Chapter 5 Genetic Analysis of Apomixis - cimmyt

(over illustration:Pictured is an interspecific crass between sexual Hieracium auricula (an leh) and apomictic Hieraciumaurantiacum (on right) that replicates acrass published by Gregor Mendel (1870)*, which is general~regarded to be the first genetic experiment on an apamiqic plant. The arrangement shows agradation in flower calor, ranging from the yellow mother to the orange-red father. Illustration fromC. H. Ostenfeld (19101, Further Studies on the Apagamy and Hybridization of the Hieracia. Zeitschrihfur Induktive Abstammungs-und Vererbungslehre.111. Band, 241-285. Plate 4.* G. Mendel 1870. Ueber einige aus kuenstlicher Belruchtung gewannene Hieraciumbastarde.Verhandlungen des naturfarschenden Vereines in Bruenn, VIII. 8d., Abhandlungen. 1869. Verlagdes Vereins, Bruenn. Pp. 26-31.Back cover illustration:The illustration shows Hieracium auricula L, one 01 the Hieracium species used for pollinationexperiments by G. Mendel in 1868. Mendel did not observe Ihe expected uniformity 01 F] progenyplants when H. auricula was used as the mother plant. This finding was thus in strict contrast to therule later lormulated as Mendel's first law. Note the stolons in the illustration, which indicate thatHieracium species also propagate vegetatively. Reference: D. F. Lvon S

The Flowering ofAPOMIXIS:From Mechanisms toGenetic EngineeringY. Savidan, J. G. Carman, and T. Dresselhaus,Editors,. CIMMYT~IsnRNATIONAl MAIZE AND WH-;AT IMPROVEMENT CeNTER•European UnionInstitut de recherchepour Ie developpement

Institut de Recherche pour Ie Developpement (IRD) is a French public research institution under theauspices of the ministers in charge of research and cooperation. For the last 50 years, it has conductedimportant research in tropical and subtropical areas. With an annual budget of US$160 million, IRDemploys approximately 750 scientists (of a total of 2,300 employees), with more than 250 of them onlong-term aSSignments in 26 different countries.The European Union (EU) is the result of a process of cooperation and integration that began in 1951between six countries and today has 15 Member States and is preparing for its fifth enlargement. TheEU's mission is to organize relations both among the Member States and their 374 million citizens in acoherent manner and on the basis of solidarity. The main objective of the European Union's Research,Technology, and Development (RTD) program FAIR (Agriculture and Fisheries) is the promotion andharmonization of research in the major European food and non-food production sectors of agriculture,horticulture, forestry, fisheries, and aquaculture. The program seeks to promote links between researchand the input and processing industries, with rural economic activities, end-users, and consumers.The International Maize and Wheat Improvement Center (CIMMYf)@ (www.cimmyt.org) is aninternationally funded, nonprofit, scientific research and training organization. Headquartered inMexico, CIMMYT works with agricultural research institutions worldwide to improve the productivity,profitability, and sustainability of maize and wheat systems for poor farmers in developing countries.It is one of 16 food and environmental organizations known as the Future Harvest Centers.© CIMMYf, fRD, EC 2001. All rights reserved. The opinions expressed in this publication are the soleresponsibility of the authors and do not necessarily reflect the views or positions of CIMMYT, IRD, orthe Commission of the EU. The designations employed in the presentation of materials in this publicationdo not imply the expressions of any opinion whatsoever on the part of CIMMYT, IRD, or the EU, ortheir contributory organizations concerning the legal status of any country, territory, city, or area, or ofits authorities, or concerning the delimitation of its frontiers or boundaries. The publishers encouragefair use of this material. Proper citation is requested.Correct citation: Savidan, Y, J.G. Carman, and T. Dresselhaus, (eds.). 2001. The Flowering of Apomixis:From Mechanisms to Genetic Engineering. Mexico, D.E: CIMMYT, IRD, European Commission DC VI(FAIR)Abstract: Apomixis, the asexual reproduction of plants through seeds, has received increasing attentionas technological advances have led to a rapid increase in knowledge about cellular biology, moleculargenetics, and the mechanisms and pathways behind plant reproduction. The fourteen chapters of thisbook address a wide range of theoretical and technical issues related to apomixis, as well as its potentialimpact on agriculture in both the developing and developed world. The technical chapters addressaspects of two complementary research paths in the ultimate quest to produce apomictic food cropplants. One path essentially seeks to either transfer the apomictic trait from a wild apomictic relativeinto a crop plant or mutagenize sexual genes into apomictic genes in the crop plant itself. This researchis currently being conducted in important food crops such as maize, wheat, and millet, as well asforages used for livestock, and model plant species such as Arabidopsis. The other path is rigorouslyexploring apomictic and sexual mechanisms and pathways in order to provide a more completeunderstanding of the overall apomixis process. This could ultimately allow scientists to target andinduce the interrelated processes of apomixis through natural or artificial means.AGROVOC Descriptors: Apomixis; Genetic engineering; Asexual reproduction; Sexual reproduction;Biotechnology; Chromosome translocation; Molecular genetics; Genetic variation; Genetic control;Genomes; Tripsacum; Zea mays; progeny forms; Plant breeding; Breeding methodsAdditional Keywords: CIMMYT, IRD, European CommissionAGRIS Category Codes: F30 Plant Genetics and BreedingDewey Decimal Classification: 631.523ISBN: 970-648-074-9

Contentsiii Contentsviii Tablesix Figuresx Acknowledgmentsxi ForewordCHAPTER 1. FEEDING THE WORLD IN THE 21sT CENTURY: PLANT BREEDING, BIOTECHNOLOGY,AND THE POTENTIAL ROLE OF ApOMIXIS(GARY H. TOENNIESSEN)1 Population Projections2 Plant Breeding3 Biotechnology6 Potential Role of Apomixis7 ReferencesCHAPTER 2.ApOMIXIS AND THE MANAGEMENT OF GENETIC DIVERSITYOULIEN BERTHAUD)8 Introduction9 Progeny of Apomictic Plants11 Diversity in Wild Apomictic Populations12 Ploidy Cycles and Organization Of Agamic Complexes12 Taraxacum and Parthenium Agamic Complexes (Asteraceae)13 Capillipedium-Dichanthium-Bothriochloa Agamic Complex (Poaceae)13 Panicum maximum Agamic Complex (Poaceae)14 Paspalum Agamic Complex (Poaceae)14 Tripsacum Agamic Complex (Poaceae)16 Cycles and Sexuality16 Management of Apomictic Varieties17 Transfer of Apomixis Gene(s) and Evolution of Landraces20 2n + n Progeny20 Relationship between Wild Relatives and Apomictic Varieties21 Promoting Genetic Diversity and Release of Apomictic Varieties22 ReferencesCHAPTER 3. CLASSIFICATION OF ApOMICTIC MECHANISMS(CHARLES F. CRANE)24 Introduction24 Types of Gametophytic Apomixis25 Nine Types of Embryo-Sac Development25 1) The Allium odorum-type25 2) The Taraxacum-type26 3) The Ixeris-type26 4) The Blumea-type26 5) The Elyrnus rectisetus-type26 6) The Antennaria-type26 7) The Hieracium-type26 8) The Eragrostis-type26 9) The Panicum-type27 Subsequent Steps of Development27 1) Embryos28 2) Endosperms28 Alternative Classifications29 Developmental Interpretation29 Meiotic Development of Megagametophytes30 Ameiotic Developments of Megagametophytes31 Subsequent Steps of Development

iv33 Outlook33 References35 Appendix: Methods to Clear Angiospenn OvulesCHAPTER 4.ULTRASTRUCTURAL ANALYSIS OF ApOMICTIC DEVELOPMENT(TAMARA N. NAUMOVA AND JEAN-PI-ITUPPE VIELLE-eALZADA)44 Introduction45 Nucellar and Integumentary Embryony46 Diplospory47 Apospory47 Differentiation of Aposporous Initials48 Aposporous Megagametogenesis48 The Cellularized Aposporous Megagametophyte57 Parthenogenesis and Fertilization58 Apogamety59 Discussion61 Future Trends62 ReferencesCHAPTER 5. GENETIC ANALYSIS OF ApOMIXIS(ROBERT T. SHERWOOD)64 Introduction64 Methods65 Chromosome Number65 Progeny Testing65 Embryo-Sac Cytology66 Sectioning or Clearing Pistils to Classify Reproductive Type66 Markers67 Biological Tests for Parthenogenesis67 Combined Cytological, Progeny, Biological, and Marker Testing68 Controlled Pollination69 Reciprocal Crossing69 Creating Tetraploid Parents70 Identification of Genomes and Chromosomes with Apomixis Genes70 Testing Inheritance70 Starting Point70 Crossing Schemes71 Classification and Grouping71 Testing Genetic Models71 Inheritance of Apomixis71 Monopolar Apospory (Gramineae-Panicoideae)7:l Bipolar Apospory75 Mitotic Diplospory7

Ii88 Applications of Molecular Genetics to Apomixis Research88 What Material?89 Molecular Mapping of Apomixis90 Cloning the Apomixis Gene(s) Using Molecular Genetics Tools93 Conclusions93 ReferencesCHAPTER 7. THE GENE EFFECT: GENOME COLLISIONS AND APOMIXISOm'lN G. CARMAN)95 Introduction95 Developmental Biology and Phylogeny of Reproductively-Anomalous Species97 Genomes of Reproductively-Anomalous Species100 The Gene Effect Hypotheses100 The Callose Hypothesis101 The Precocious Induction Hypothesis101 The Hybridization-Derived Floral Asynchrony Theory104 Testing The Gene Effect Hypotheses105 Implications of the HFA Theory105 Evolution of Apomixis and Related Anomalies106 Mendelian Analyses of Apomixis109 Making Crops Apomictic109 Acknowledgments109 ReferencesCHAPTER 8. MODEL SYSTEMS TO STUDY THE GENETICS AND DEVELOPMENTALBIOLOGY OF ApOMIXIS(Ross A. BICKNELL)111 Introduction111 Why Use a Model System for Apomixis?112 Attributes of a Model System112 Biological Attributes112 Types of Apomixis113 Genetic Attributes114 Experimental Methods114 Quantifying Apomixis115 Candidate Systems115 Modification of an Existing System117 Development of a Model System from an Existing Apomict119 Summary119 ReferencesCHAPTER 9.SCREENING PROCEDURES TO IDENTIFY AND QUANTIFY APOMIXIS(OUVJER LEBLANC AND ANDREA MAZZUCATO)121 Introduction121 Apomictic Mechanisms as Potential Screening Indic'ators122 Types of Meiotic and Apomeiotic Embryo-Sac Formation123 Embryo and Seed Formation124 Consequences of Apomictic Seed Formation124 Levels of Screening and Related Tools124 Analyses at the Plant Level124 1. Molecular markers cosegregating with apomixis125 2. Cytoembryology126 3. Egg cell parthenogenetic capacity126 Progeny Analysis128 1. Analysis of pollinated ovaries or seeds128 2. Ovule regenerated plants128 3. Analysis of progeny plants130 Choosing Suitable Procedures130 Analyses at the Plant Level versus Progeny Tests130 1. Nature of the information obtained131 2. Comparing results

vi131 Screening Procedures: Advantages and Constraints131 1. Apomixis identification and characterization133 2. Degree of apomixis expression133 Choosing a Procedure134 ReferencesCHAPTER 10. BREEDING OF ApOMICTIC SPECIES(CACTLDA BoRGES DO VALLE AND JOHN W. MILES)137 Introduction137 Prerequisites for an Effective Breeding Program139 General Structure of a Breeding Program140 Objectives140 Germplasm Acquisition and Evaluation141 Cytology, Reproductive Mode, Inheritance of Apomixis146 Breeding Plans149 Concluding Observations149 ReferencesCHAPTER 11. TRANSFER OF ApOMIXIS THROUGH WIDE CROSSES(YVES SAVIDAN)153 Introduction154 Source of Apomixis and Choice of Parental Materials154 Basic Traits to Consider154 1. Genetic resources available154 2. Chromosome number of the potential donor species154 3. Genome homoeology155 4. Pollen fertility155 5. Type of apomixis155 6. Degree of apomixis (or degree of facultativeness)155 7. Agronomic characteristics155 8. Previous knowledge155 Case History: Pennisetum157 Case History: Tripsacum158 Production of Interspecific or Intergeneric F 1Hybrids158 Crossing Techniques158 Sterility of the Fls159 Production of Apomictic Progenies through Backcrossing164 Transfer of Gene(s) for Apomixis from an Alien Chromosome to the Crop Genome166 ReferencesCHAPTER 12. FROM SEXUALITY TO ApOMIXIS: MOLECULAR AND GENETIC ApPROACHES(UEU GROSSNIKLAUS)168 Introduction169 Developmental Aspects of Sexual and Apomictic Reproduction170 Sexual Model Systems ,.171 Sexual Reproduction171 1. Megasporogenesis172 2. Megagametogenesis174 3.Double Fertilization174 Apomixis176 Interrelationship of Sexual and Apomictic Reproduction177 Models for Apomixis: Heterochronic Initiation of Development179 Genetic Control of Reproduction and Candidate Genes for the Engineering of Apomixis180 Megasporogenesis and Nonreduction183 Megagametogenesis184 Egg Activation and Parthenogenesis186 Endosperm Development and Genomic Imprinting186 1. Interrelationship of embryo and endosperm development187 2. Genomic imprinting188 3.lmprinting barriers to the introduction of apomixis into sexual species

vii189 Genetic Screens For Mutants Displaying Apomictic Traits In Sexual Model Systems189 Arabidopsis Mutants with Autonomous Seed Development191 Screen for Pseudogamous Apomixis in Cereals192 Enhancer Detection as a Powerful Tool to Study Sexual Reproduction in Arabidopsis192 Enhancer Detection and Gene Trap Systems193 Generation of Transposants and Ongoing Screens195 Identification of Developmentally Regulated Genes and Their Promoters196 Introduction of Apomixis into Sexual Species196 Introgression and Genetic Synthesis199 De novo Engineering through Biotechnology200 Field-Level Regulation of Apomictic Traits201 Conclusions and Prospects202 Acknowledgments202 ReferencesCHAPTER 13. INDUCfION OF ApOMIXIS IN SEXUAL PLANTS BY MUTAGENESIS(UTA PRAEKELT AND ROD ScOTT)212 Introduction213 Considerations213 Components of Apomixis213 1. Avoidance of meiosis213 2. Formation of aposporous embryo sacs213 3. Parthenogenesis214 4. Endosperm development214 Genetic Control of Apomixis215 How Important is Polyploidy?215 The Problem of the Endosperm216 Which Mutagen?217 Some Early Work with Mutants217 Induction of Sexuality in Apomicts218 Mutants of Sexual Plants with Apomictic Characteristics218 1. Meiotic mutants219 2. Parthenogenetic mutants219 3. Aposporous mutants220 4. Conclusions220 Current Approaches to the Isolation of Apomictic Mutants in Model Sexual Plants221 Screning for Elongated siliques in the Absence of Pollination222 Screening for Dominant Mutations in the M 1after Pollination225 Transposon Mutagenesis for the Isolation ofApomictic Mutants of Arabidopsis and Petunia225 Branching Out in the Brassicas226 Conclusions and Perspectives227 ReferencesCHAPTER 14. GENETIC ENGINEERING OF ApOMIXIS IN SEXUAL CROPS: A CRITICAL ASSESSMENTOF THE ApOMIXIS TECHNOLOGY(THOMAS DRESSELHAUS, JOHN G. CARMAN, AND YVES SAVIDAN)229 Introduction230 Transfer of the Apomixis Trait to Sexual Crops230 Breeding and Introgression from Wild Relatives231 Mutagenesis Approaches232 Known Gene Approaches236 Transformation and Inducible Promoter Systems237 Main Limitations238 Intellectual Property Rights (lPR)239 Risk Assessment Studies240 Summary241 References

VIIITables4 Table 1.15 Table 1.29 Table 2.19 Table 2.29 Table 2.310 Table 2.411 Table 2.511 Table 2.612 Table 2.714 Table 2.815 Table 2.936 Table 3.172 Table 5.1100 Table 7.1124 Table 9.1125 Table 9.2132 Table 9.3141 Table 10.1142 Table 10.2144 Table 10.3145 Table 10.4157 Table 11.1159 Table 11.2162 Table 11.3162 Table 11.4163 Table u.s163 Table 11.6233 Table 14.1234 Table 14.2International Agricultural Research CentersInstitutions facilitating the application of biotechnology to internationalagricultureGenetic constitution of progeny from apomictic plantsSize of four categories defined in Table 2.1 for two PaniClim maximum clones (fromCombes 1975)Size of four categories defined in Table 2.1 for three types of progeny involvingtwo PartlIeU/1I11l species. Adapted from Powers and Rollins (1945)Estimation of apomixis rate and categories of progeny from chromosome countsand isozyme analyses of Tripsacum populations (Berthaud et aI., unpublished data)Variation in chromosome number for progeny from wild populations of Tripsacumdactyloides mexicanllll1. (Seeds were collected in the wild.)Size of categories defined in Table 2.1 for two Pennisetum flnccidum x P mezianumcrosses. From Bashaw et al. (1992)Distribution of clones in Tripsacum wild population "La Toma"Distribution of clones according to ploidy level from the P maximum collectionestablished in Cote d'lvoire (Combes 1975)Distribution of species of Paspalltm according to their incompatibility system,ploidy level, and meiosis behavior (from studies at lEONE, Quarin, personalcomm.)Refractive index (n ) Dof common and potential clearing mediaSegregations for mode of reproduction in 10 crosses of Panicltm maximum (Savidan1981; Savidan et al. 1989)Phylogenetic, genomic, and developmental peculiarities that hypotheses for thegenetic regulation of apomixis and related reproductive anomalies must explainThe four theoretical offspring classes in progenies from facultative pseudogamousapomictsMain characteristics of megasporogenesis and megagametogenesis during bothsexual reproduction and gametophytic apomixisAdvantages and disadvantages of important procedures for the investigation ofmodes of reproduction at the plant and progeny levelsAgronomic evaluation of Brachiaria accessions in BrazilMode of reproduction of 15 species of Brachiaria, based on embryo-sac analysisSegregation for mode of reproduction in Brachiaria hybridsComparison between progeny test and embryo-sac analysis for determination ofmode of reproduction for first-generation interspecific Brachiaria hybridsCrossabilities between maize and wild Tripsacum species and presumed naturalinterspecific hybrids 0Crossabilities between pearl millet and three apomictic wild Pennisetum speciesFacultativeness of apomixis and diplospory rate in the Tripsacum accession used inthe backcross transfer of apomixis into maize and three BC Iprogenies, showingvariation for this rateChromosome numbers of BC I(2n =56) progenies as estimated by flow cytometryMaize x Tripsacum Be 3progenies, in which the BC 4s are the n + n categoryMaize x Tripsacum BC 4with known mode of reproductionExamples of isolated genes and their promoters that might be useful as tools for denovo syntb.~sisof the apomixis trait in sexual cropsExamples of patents linked with the engineering of the apomixis trait in sexualcrops

ixFigures3 Figure 1.16 Figure 1.213 Figure 2.116 Figure 2.225 Figure 3.149 Figure 4.151 Figure4253 Figure 4.354 Figure 4.455 Figure 4.573 Figure 5.196 Figure 7.1103 Figure 7.2122 Figure 9.1127 Figure 9.2127 Figure 9.3129 Figure 9.4139 Figure 10.1142 Figure 10.2144 Figure 10.3148 Figure 10.4148 Figure 10.5164 Figure 11.1164 Figure 11.2173 Figure 12.1175 Figure 12.2197 Figure 12.3197 Figure 12.4197 Figure 12.5223 Figure 13.1Biotechology tools for strengthening plant breeding.The International Agricultural Research System in the era of biotechnology.Continuos introgression and hybridization without further introgresion in anapomictic complex including three genera, Bothriochloa, Capillipedium,Dichanthium, and 18 species.Evolution ofploidy levels in Tripsacum from fertilization of female gamete (n or 211)by a male gamete (n) from 2x, 4x or 6x plants or parthenogenetic development ofegg cell (n+O).Schematic diagram of apomictic embryo sacs.Integumentary embryony in Euonymus macroptera.Apospory in Panicum maximum.Aposporous megagametophyte development in Pennistum ciliare.Organization of the mature aposporous egg apparatus in Pennisetum ciliare.Apogamety in Trillium camschatcense.Genealogical tree of the cross Ranunculus cassubicifoliu s = C, 2x = 16, meiotic (sexual)x R. megacarpus = M, 4x = 32, partially aposporous ("totally" apomictic) and thedifferent backcrosses with the sexual parent C.Developmental stages during megasporogenesis and embryo-sac development inselected sexual (monosporic, bisporic, and tetrasporic) and apomictic (Alliumodorum-type diplospory, Antennaria-type diplospory, Taraxacum-type diplospory,lxeris-type diplospory, Blumea-type diplospory, and apospory) angiosperms.Model of how asynchronously-expressed duplicate genes cause diplospory andapospory in polyploids containing two genomes divergent in the temporalexpression of female developmental schedules (floral induction, megasporeformation, gametophyte development, and embryony).Mechanisms of pseudogamous gametophytic apomixis: consequences andcomparison with sexual reproduction.Aposporous development ofHieracium type in Poo pratensis (sectioning and stainingprocedure).Clearing techniques in Tripsacum spp.Confidence limits (a=O.025) for p in binomial sampling, given a samplefraction a/noSelection and breeding scheme for apomictic forage species.Distribution of 253 acessions of Brachiaria (B = B. brizantha; R = B. ruziziensis; D = B.decumbens; and J= B. jubata in two planes (PRIN1 and PRIN2) generated by PrincipalComponent Analysis using seven morphological descriptors.Hybridization scheme for breeding Brachiaria (adapted from Gobbe et. al. 1983)Simplified diagram of recurrent mass selection employed in SEX Brachiariapopulation.Simplified diagram ofrecurrent selection scheme emplyed in APO/SEX Brachiariapopulation at ClAT.Flow-eytometric analyses on entire seeds.Backcross scheme for the transfer of apomixis from Tripsacum into maize.Diagram of megasporogenesis and megagametogenesis in Arabidopsis thaliana.The main developmental features of apomixis in relationship to the sexual pathway.The R-Navajo (R-nj) dominant maker system for embryo and endosperm.The principle of enhancer detection and gene trapping.Enhancer detector transposant with GUS expression restricted to themegagametophyte."Uncoupling" genetic screens for apomictic mutations in Arabidopsis.

xAcknowledgmentsWe are grateful to the Commission of the European Union's Research,Technology, and Development program FAIR (grant FAIR 5-CT97­3730), the International Maize and Wheat Improvement Center(CIMMYT), and the Institut de Recherche pour Ie Developpement(IRD) for the financial support they provided to produce this book.We would like to express our special gratitude to David Poland,CIMMYT Science Editor. Without his friendship and especially hisenthusiasm and commitment to the project, the book never wouldhave made it off the printer's press. Eliot Sanchez Pineda and theCIMMYT Graphic Design Unit must be lauded for the beautiful jobthey have done on this endeavor. We are grateful to all of the reviewerswho helped improve the diverse chapters of the book. Last, but notleast, we thank all of the contributing authors for their good humorand patience. They have anticipated the arrival of this book for anumber of years and have kindly reviewed and updated their workseveral times along the way.Yves Savidan,John G. Carman, and Thomas DresselhausMarch,2001

XIForewordAs implied by the title of the book-"The Flowering ofApomixis: FromMechanisms to Genetic Engineering"-this complex and mysteriousaspect of reproduction is beginning to yield its secrets to more than acentury of scientific inquiry by researchers from around the world.Building on this foundation of knowledge, and by using the rapidlyadvancing tools and techniques of biotechnology, we are probing theintricate processes of apomixis more deeply and broadly than everbefore. Consequently, our grasp of the mechanisms of both asexual andsexual reproduction has expanded tremendously in the last decade. Andthough timetables for research discoveries cannot be dictated, thepromise of applying apomixis technology to real world needs and issueshas never been brighter.One of the most urgent applications for the technology will be feedingand raising the standard of living for the burgeoning populations ofthe developing world. It is fitting that in the book's opening chapter,Gary Toenniessen, Director of Food Security for the RockefellerFoundation, succinctly sets forth the magnitude and gravity of thesituation we face, and the tremendous potential apomixis holds forhelping to meet those challenges. By producing crops that produceasexually through seeds, we can greatly hasten the development of newhigher-yielding hybrid varieties, a keystone of past productivity gainsand one that will be required to boost productivity in coming years.With costs of development coming down, seed prices to farmers mayalso decrease. Of particular import to small-holder farmers, apomixiswill allow scientists to efficiently breed varieties specifically tailored toa multitude of niche environments, many of them situated in the mostmarginal agricultural areas. Finally, because apomictic seed is selfreplicating,developing world farmers should be able to recycle seedwithout losing valuable hybrid characterrstics. Furthermore, thetechnology could be used in such a way that farmers may be able tobetter fix the traits they deem desirable within their own indigenousvarieties and landraces. Needless to say, however, there is work yet tobe done.In the following chapters, the authors follow two complementary pathsin the ultimate quest to produce apomictic food crop plants. One pathis to either transfer the apomictic trait from a wild apomictic relativeinto a crop plant or change sexual genes into apomictic genes in thecrop plant itself. This research is currently being conducted in important

xiifood crops such as maize, wheat, and millet, as well as forages used forlivestock, and model plant species such as Arabidopsis. The other path isrigorously exploring apomictic and sexual mechanisms and pathwaysin order to provide a more complete understanding of the overall apomixisprocess. This should ultimately allow scientists to target and induce theinterrelated processes of apomixis through natural or artificial means.The knowledge gained through research following both approaches hassignificantly accelerated advances in the field as a whole.It is with great pleasure that I invite those with an interest in apomixisstudents,academics, plant breeders, geneticists, and those simply with ascientifically inquisitive bent-to read and reference this book. Finally, Imust commend the authors and editors for their diligence in producingthis important and timely work.l~~Professor TImothy ReevesDirector General,The International Maize and Wheat Improvement Center (CIMMYT)EI Batan, Mexico.March,2001

Chapter 1Feeding the World in the 21st Century:Plant Breeding, Biotechnology, and thePotential Role of ApomixisGARY H. TOENNIESSENAsurplus of food in many of the world'swealthier countries has led to a certaincomplacency there about future supplies andavailability. But for the vast majority of theworld's people, who live in poorer developingcountries faced with growing populations andincreasing demand for food, concern ratherthan complacency is the order of the day. Forthe nations of the South, the task of feedingtheir future generations presents a critical andformidable challenge for agriculture over thenext half century or longer.Population ProjectionsFortunately, there are reasons to be optimisticthat an end to population growth is finally insight, albeit at some distance (Lutz et al. 1997).The rate of world population growth peakedaround 1970 and has been steadily decliningsince then. As societies have moved fromdependence on subsistence agriculture tomore intensive agriculture and more modemeconomies-in the process providingimproved nutrition and health care andexpanded educational opportunities to theirgirls and boys-desired family size hasdropped. A family planning revolution in thedeveloping world, under way now for morethan two decades, has lowered the averagenumber of children in a family from six tothree, which is reflected in a respective declinein annual population growth from 2.5 toaround 1.8 percent (United Nations 1997).Contraceptive use by women of child-bearingage in developing countries has risen fromabout 10 percent to more than 50 percentduring the last three decades; and it isestimated that there are at least an additional100 million women who wish to regulate theirfertility, but who are not now usingcontraceptives. Ifeffective family planning andreproductive health services were provided toall those wishing to use them, demographersnow predict that replacement level fertilitycould be reached as early as 2020 and that theworld's population would stabilize at 8-11billion people near the middle of the 21stcentury (Bongaarts 1994; Lutz et al. 1997).Although the task of curbing populationgrowth will be arduous, generally speakingthe agencies and institutions that providefamily planning services have the technicalknow-how required to achieve this goal; nowthey are working on mol5ilizing the necessaryfinancial resources and political commitments.To complement this effort, the agriculturalsector must provide the basic nutrition andeconomic growth needed to fuel the desire forsmaller families and the requisite familyplanning services, until the time thatreplacement level fertility is reached.These encouraging population trends will,over the long term, be good for agriculture, asthey imply that sometime during the nextcentury the ever-increasing demand forgreater food production should finallystabilize. The downside is that even giventhese positive trends, the developing worldwill need to produce two to three times as

2 Gary H. T....i.....much food as it does today. In manydeveloping countries, more than half of thepeople are just entering or are still underreproductive age. Even if these people wereto have only two children per family, a neardoubling of total population is inevitable. Inaddition, economic growth will furtherincrease the demand for food.The challenge facing agriculture in the first halfof the 21st century is formidable. It mustprovide adequate nutrition for billions morepeople and contribute to their economicdevelopment, thereby stoking the desire tolimit family size. Furthermore, agriculturemust accomplish this without jeopardizing thecapacity of the natural resource base to meetthe needs of future generations. Currently,agriculture does not have the technologies todouble or triple food production in developingcountries, and so the threat arises that farmerswill irreparably damage the natural resourcebase in their efforts to feed growingpopulations-this scenario is alreadybecoming a reality in certain locations.Meeting the food challenge will demand thediscovery of new knowledge and thedevelopment of innovative technologies,which, combined with the broader adaptationand application of existing technologies, willallow greater intensification of production ona sustainable basis.Plant BreedingMany of the institutional structures andfinancial support systems needed to addressthe food challenge are already in place and canrightly claim an impressive record ofaccomplishment. International cooperation inplant breeding has been particularly successfulin producing improved crop varieties thatbenefit the developing world. When combinedwith appropriate management practices, thesemodem varieties have substantially increasedproductivity and contributed significantly tofood self-sufficiency and economicdevelopment in many countries of Asia andLatin America.In Asia, farmers have for centuries usedirrigation, organic fertilizer, and hand weedingon their small holdings. More recently, theyhave readily adopted modem varieties and,using their traditional intensive managementpractices together with purchased inputs, havein many locations pushed yield per hectareclose to the maximum potential. Modernvarieties of rice and wheat are now grown onnearly 70 percent of the area planted to thesecrops in Asia. Because many of these varietieshave short growing seasons, farmers canobtain two or three crops per year on fertileland under irrigation. Improved varieties havealso been produced for the poorer upland andseasonally flooded regions of Asia, howevertheir performance and rates of adoption havebeen less dramatic. During the past 20 years,the proportion of the Asian populationaffected by inadequate nutrition declined from40 to 19 percent. Nevertheless, Asia still hasthe greatest number of chronicallyundernourished people, 528 million, and thelargest projected increase in population (FAO1992; Bongaarts 1994; Lutz et al. 1997).In Latin America, modem varieties have madean enormous impact, however, due to thehighly skewed and inequitable distribution ofland ~n the region, it is primarily thecommercial farmers (who control most of thefertile land) that have adopted them.Production on the larger farms has increasedsignificantly and consumers have benefitedfrom lower prices. However, the majority ofLatin American farmers, who work smallholdings on less fertile land in the highlyheterogeneous hill regions, have not gleanedthe benefits offered by modem crop varieties.Developing improved varieties for them is adifficult task and only limited progress hasbeen made. No single elite breeding line is

Feedlog rIoo World io tltt 2111 ''''.ry: Pl••IB,eedi.g. 8Io1....oIogy, IIIId Iitt Pot..tlol KIlt of Aponi.1s 3broadly applicable across such diverseagronomic and socio-economic conditions,and plant breeders are just beginning toprovide improved varieties tailored to a fewof the multitude of niche environments foundin the region.Of the major developing regions, improvedvarieties have had the least impact in sub­Saharan Africa; food production there haslagged behind rapid population growth. InAfrica as a whole, more than 168 millionpeople are chronically undernourished, and,alarmingly, nearly a fourfold increase inpopulation, from 740 million in 1996 to 2.8billion by the end of the 21st century, is nowprojected (FAa 1992, 1998; Bongaarts 1994).The defining characteristics of Africanagriculture are its complexity andheterogeneity. Most farmers have smallholdings on which they grow a variety ofcrops, often intercropped with one another. Ineach of the continent's countries, soils andclimate are highly diverse and variable.Economic realities limit the development ofirrigation and other forms of yield enhancingand risk averting infrastructure. As in muchof Latin America, no elite breeding lines arebroadly applicable and improved varietieswith specific characteristics need to bedeveloped for many different types ofagronomic and socioeconomic niches. Suchniche breeding has been successful in a fewlocations and has potential for expansion, butit is a slow process when based onconventional breeding technology. Notably,while there is no such thing as low input/highoutput agriculture, average yields in Africa areso low (often less than It/ha) that a doublingor tripling of production should be possiblewith locally well-adapted varieties using justminimal inputs. Undoubtedly, bettermanagement practices would help boostyields (the use of nutrient and SOil-enhancingcrop rota tions and associa tions looksespecially promising), but over the long term,greater use of inputs, particularly fertilizer,will be necessary.BiotechnologyModem plant breeding, which revolutionizedagriculture in the 20 th century, is now on theverge of significantly extending itstechnological potential. New geneticmonitoring and manipulation tools, inaggregate commonly referred to asbiotechnology, are becoming available as aresult of advances in molecular and cellularbiology. As indicated in Figure 1.1, these newtools are contributing to both phases of planteo:o...r:::0.....Foreign genes and modified genesI IG,"" J..l "'------+IRelated wild speciesWide hybrids Samati, hybridsI I Existing (fOp germplasmsomJclonalvatlionWider primary gene pool 14------'1Traditian,1 crossingBreeding linesc Anther culture Genotype selectionoo::>""6.1;Marker-aided assesment ofpest/pathogen papulationMarker-aided selection ofprogeny with desired genesI Improved (fOp varieties IFigure 1.1 Biote(hology tools for strengthening plant breeding.

4 Gary H. Toe"'...breeding: the evolutionary phase, in whichvariable populations are produced, and theevaluation phase, in which desirablegenotypes are selected.Variability, at the heart of the evolutionaryphase, traditionally has been created byhybridization and to a lesser extent bymutations. Wide hybridization throughembryo rescue or somatic hybridization,somaclonal variation, and genetic engineeringare biotechnology tools that can dramaticallyexpand the range of variability available tobreeders. Genetic engineering, especially,should make the process of generatingdesirable variability much more predictableand help obtain other goals that are beyondthe reach of conventional techniques.Meanwhile, the evaluation phase will becomemuch more efficient through the use of thefollowing biotechnology tools: anther cultureto produce doubled haploids and eliminatedominance variance; molecular maps andmarkers of the crop genome to tag and followthe inheritance of genes for important traits,particularly quantitative traits and those thatare difficult to score; and molecular geneticmaps and markers of pests and pathogens thatcan be used to characterize and monitorpopulation structures and dynamiCS, therebypromoting more effective selection anddeployment of resistant plants.The international agricultural research system,which has been so successful at producingimproved varieties for developing countries,is itself evolving and adding new institutions(see Tables 1.1 and 1.2) to take advantage ofthese new tools. It is drawing more on resultsgenerated by fundamental and strategicTable 1.1 International Agricultural Research CentersInternational CenterClAT(IfORClMMYT(IPICARDAIHARM'I(RAFI(RISATIFPRI*lilAIlRIIPGRIIRRIISNAR'IWMI*WARDA(entro Internocional de Agricultura Tropical(enter for Internafional Forestry Research(entro Internacional de Mejoramiento de Maiz yTrigo(entro Internacional de 10 PapaInternafional (enter for AgriculturalResearch in Dry AreasInternational (enter for Uving Aquatic ResourcesInternational (enter for Research in AgrofarestryInternational (rop Research Institutefor the Semi-Arid TropicsInternational Food Policy Reseorch InsfitufeInternational Institute for Tropical AgricultureInternational Uvestock Res. InsfitufeInternational Plant Genefic Resources InstituteInternafional Rice Research InstituteInternational Service for NafionalAgricultural ResearchInternational Water ManagementlnstituleWest Africa Rice Development Association• !CLARM, IFPRI, ISHAR, and IWMI do not directly handle plant research programs.Crops(assava, field beans, riceForestryMaize, wheat, trificalePotatoes, sweet potatoesWheat, barley, lenfils, chickpeaForestry, tree cropsSorghum, pearl millet,groundnut, pigeon pea(assava, yams, cowpea, maizeForagesRiceRiceLocation(olombiaIndonesiaMexicoPeruSyriaPhilippinesKenyaIndiaUSANigeriaKenyaIta~PhilippinesThe NetherlandsSri lanka(6te d'ivoire

Feediog the World ill the 21st Century: Pl••t Breeding. Biotedl.oIogy, .od tbe Pot••tial Role .1 ~pamixls 5research institutions and exploring new waysof gaining access to proprietary technologies.Some of the key institutions that make up thissystem in the era of biotechnology are notedin Figure 1.2. Their work includes• fundamental research conductedprimarily in advanced researchuniversities and institutes that expandsthe knowledge base on plants, insects, andmicrobes, and their interactions with oneanother and with their environment;• strategic research, conducted primarily atagricultural universities, nationalagricultural research institutes, incorporations, and increasingly at theinternational agricultural research centers(IARCs), which generates new andstrengthens existing technologies for cropgenetic improvement;• applied research (including germplasmcollection and evaluation), conductedprimarily at IARCs and national cropbreeding institutions, oftencollaboratively, which generates newbreeding lines;• adaptive research, conducted primarily atthe national level, which combines elitebreeding lines with traditional varieties toproduce improved finished varieties thatare well-suited to local needs andconditions; and• seed multiplication and delivery ofimproved varieties, usually by nationalagencies, local farmers, non-governmentalorganizations (NGOs), and increasinglythrough market mechanisms.The lARCs with plant research programs (seeTable 1.1) have the mandate and primaryTable 1.2 Institutions facilitating the application of biotechnology to international agricultureISAAA - International Service for Acquisition of Agri-Biotech Applications, Ithaca, New Yark, USA, is a not-for-profitinternational organization committed to the acquisition and transfer of proprietary agricultural biotechnologies fromthe industrial countries for the benefit of the developing world. It assists in identifying biotechnology needs andopportuni~es, evaluates the availability of proprietary lechnologies, serves as an "honest broker" that malches needswith available technology, and when necessary mobilizes the financial resources required to implement brokeredproposals.ILTAB . International Laboratory for Tropical Agricultural Biotechnology, SI. Louis, Missouri, USA, is a unit of theDan forth Plant Science (enter. Technology is transferred first from the center, where scientists are engaged inpioneering work on the development of disease resistant plants, to five ILTAB scientists and fellows from developingcountries. These scientists use the technology to produce new sources of disease resistance in tropical crops includingcassava, rice, sweet potato, and yam.CAMBIA - (enter for the Application of Molecular Biology to International Agriculture, (anberra, Australia is aresearch and technology transfer organization committed to the applica~on of biotechnology to internationalagriculture. It specializes in producing inexpensive biotechnology tools that can be effernve~ u~lized in developingcountries.ICGEB - International (enter for Genetic Engineering and Biotechnology was established by the United Na~onsIndustrial Development Organization. It has headquarters and information gathering and dissemina~on facili~es inTrieste, Italy, and agricultural biotechnology research and training facilities in New Delhi, India.ABSP . Agricultural Biotechnology for Sustainable Productivity, East Lansing, Michigan, USA, is a projectheadquartered at Michigan Slate University and funded by USAID. It is a unique bilateral program in that it supportsresearch at and technology transfer from public and corporate sector crop research insntu~ons to developing countries.IBS . Intermediary Biotechnology Service, The Hague, Netherlands, is a unit of the International Service for Na~onalAgricultural Research (ISNAR). It provides national agricultural research agencies with informa~on, advice, andassistance to help strengthen their agricultural biotechnology capacities and to enable them to establish collaborativearrangements with international biotechnology programs.BA( . Biotechnology Advisory (enter, Stockholm, Sweden, a unit of the Stockholm Environmentlns~tuIe, is anindependent resource for impartial biosafety advice. It was established to help developing countries assess the possibleenvironmental, health, and socioeconomic impacts of proposed biotechnology introductions.

6 Gary H. T....it....The potential role of apomixis in boostingyields in the developing world is considerableand varies according to region. From a plantbreeding perspective, Asia most needs newvarieties of its staple cereal crops that havesignificantly higher yield potentials tha ntoday's high-yielding varieties. Africa andLatin America most need a large number ofimproved varieties of food crops, each well­suited for production in one or more of themany ecologically and/or socioeconomicallyunique niches that can be found in thesecontinents.Asia must more than double its cerealproduction over the next fifty years and do iton the same or less area than is currently inproduction. Accomplishing this will requireresponsibility for linking the components ofthis system together and assuring that itfunctions effectively. This system has theability to take relevant scientific discoveriesfrom the "ivory towers" of academe and,through a series of technology transfers andcollaborative research projects, incorporate thenew knowledge and technology intoimproved seeds that will be sown in fieldsthroughout the developing world-and to doso in an amazingly short time frame. Theimproved cultivars and agronomic practicesgenerated by this system have helped literallybillions of people who daily consume the endproducts. If over the next century we are toachieve a stabilized world population fed bysustainable agriculture, this unique publicsector research establishment must also besustained, both financially and technologically.Potential Role of Apomixisr---1 Fundamental Research Universities andInstitutesAgriculrural UniversitiesTechnical UniversitieNat'l Biotechnology Inst.ILTAB, ICGEB, CAMBIA,ABSP---.. .­1IARCsDevelopment Assistance Agencies H r-L...+IBS I I BACNationol Crop Improvement Programsand Extension Agencies- NGOsFormersBiotechnologyCorporationsIII____ JISill JIARCs International Agriculturol Research Centers (see Table 1)ILTAB International Laboratory for Tropical Agricultural Biotechnology, La Jolla, CA, USAICGEB International Center for Genetic Engineering and Biotechnology, New Delhi, Indio and Trieste, ItoliaCAMBIA Center for the Application of Molecular Biology to International Agriculture, Canberra, AustraliaABSP Agricultural Biotechnology for Sustainable Productivity, East Lansing, MI, USAISill International Service for Acquisition of Agri-Biotech Applications, Ithaca, New York, USAIBS Intermediary Biotechnology Service, The Hogue, NetherlandsBAC Biotechnology Advisory Commission, Slockholm Environmentlnstilute, Stockholm, SwedenFigure 1.2 The International Agricultural Research System in the era of biotechnology.

F..ding Itl. World in the 211t c.n'ury: PI.., 8r.....g. 8io,.o.oIogy, and tile Po,••tlal Role of ApamiJlil 7even more double and triple annual croppingcycles and more extensive use of yieldenhancingtechnologies such as hybrid seed,by the overwhelming majority of farmers,including those with limited purchasingpower. Hybrid seed's potential to increaseproduction has been demonstrated by hybridrice in China. From 1980 to 1990, Chinaincreased its rice production by roughly 32.5million tons, or 22 percent, while decreasingthe area planted to rice by roughly 2.2 millionhectares, or six percent (FAO 1990). Yuan Longping,the "father of hybrid rice" in China,speculates that full exploitation of the heterosisavailable in rice could provide another 3D-50percent increase in yield (Yuan 1993). Newhybrid lines that are suitable for other regionsof Asia are slowly becoming available.Biotechnological tools (such as geneticallyengineered male sterility systems for elitebreeding lines) and the use of molecularmarkers to select parental lines that combinehigh levels of heterosis with other desirablecharacteristics can accelerate this process andmake the use of hybrid rice technology morebroadly applicable. And, as reported later inthis book, progress is being made on usingapomixis as the ultimate tool for fixingheterosis in cereals, thereby making thebenefits of hybrid seed available to farmers atminimal cost.African and Latin American farmers could alsobenefit from hybrid seed that self-replicatestlvough apomixis, although the application ofapomixis to niche breeding could yield evenmore consequential results. If apomixis can beintroduced into staple food crops, cultivarsthat perform well under local conditions couldbe genetically fixed early in the selection cycle.Under this scheme, variability would begenerated through traditional hybridizationor any other technique noted in theevolu tionary phase (see Figure 1.1). Theresul ting population of plants would begrown and evaluated under local conditions,and the plants that performed best could beselected and quickly developed intogenetically stable superior cultivars byincorporating the gene(s) for apomixis. Forcrops that are normally reproduced fromtubers or vegetative cuttings, apomixis wouldenable the multiplication and disseminationof improved varieties as true seed.In short, apomixis has the potential to makea significant contribution toward meetingfood production demand throughout thedeveloping world in the 21 sl century. Becauseof its limited profit potential, this technologywill probably not be fully developed in theprivate sector. Therefore, if the full potentialof apomixis as a breeding tool to help the pooris to be realized, the necessary research anddevelopment must be undertaken by thepublic sector international agriculturalresearch system-and the results mustremain freely available to public sector cropbreeding programs.ReferencesBongoorts, J. 1994. Population policy options in the developing world.Science 263: 771-76.FAD. 1990. Selected Indicators of Food and Agriwhure Development inth~ Asian·Pacific Region. Rome: RAPA Publications.FAD. 1992. The State of Food and Agriculture. Rome: FAD.FADSTAT. 1998. Populolian Statistics. Rome: FAD.luts, W, W. Sanderson, and S. Scherbov. 1997. Doubling of worldpopulation unlike~. Holure 387: 803-05.United Nations. 1997. World Population Prospeds: The 1996 Revision.New York: United Nations Population Division.Yuan, L.p. 1993. Progress of two-line system in hybrid rice breeding. InK. Muralidhoran and E.A. Siddig (eds.), Hew Frontiers in RiceResearch. Hyderabad, India: Directorate of Rice Research. Pp. 81r­93.

Chapter 2Apomixis and the Management ofGenetic DiversityJULIEN BERTHAUDIntroductionApomixis is a mode of reproduction (asexualpropagation through seeds) that exists inmany plants from different botanical families(review in Asker and Jerling 1992; Carman1997). It is most frequent in the dicots Rosaceaeand Asteraceae and in the moncot Poaceae.Some of these Poaceae genera are tropicalforages with wide colonizing ability, e.g.,Panicum maximum. From its center of origin inEast Africa, through human activities it hasexpanded to West Africa, where it can be foundcolonizing roadsides, and to tropical regionsof the Americas and Asia.Apomixis attracts considerable theoreticalinterest as it may help us better understandthe sexual mode of reproduction. It is also ofpractical interest to breeders as a means ofgenetic fixation, potentially offering thecapability of indefinite multiplication ofheterotic genetic combinations. In the case ofapomictic tropical forages (see Valle and Miles,Chap. 10), the problem faced by breeders ishow to overcome apomixis to take advantageofgenetic recombination in order to create newgenetic combinations to be maintainedthrough apomixis. Another challenge is totransfer apomixis into crops in which heterosishas been well documented. Research projectsfocused on this goal are underway for pearlmillet, Pennisetum glaucum (Hanna et al. 1993)maize (see Savidan, Chap. 11), and wheat(Carman 1992). Rice breeders are alsointerested as F] hybrids in rice show heterosis(sef~ Toenniessen, Chap. 1).Some scientists have solely pursued thesimplest model of apomixis, that with acomplete lack of sexuality, i.e., no possibilityof recombination and evolution. In this case,population genetics models show a diffusionof apomixis genes into natural populationswithout a need for some form of selectiveadvantage (Pemes 1971; Marshall and Brown1981). If this holds true, transferring apomixisto crops could ultimately decrease geneticdiversity in those crops and pose a threat tothe environment. From modern apomicticvarieties, the apomixis gene could move tolandraces and wild ancestors in their center oforigin. In a recent review of apomictic risk, vanDijk and van Damme (2000) based theirdiscussion almost entirely on this model.However, before overstating this possibility,one should know more precisely howapomixis functions, what diversity isconserved in wild populations where apomixisis the dominant mode of reproduction, andhow apomixis could be transferred tolandraces.To address these issues, this chapter discusses(i) genetic variation observed in progeny ofapomicts, (ii) diversity observed in wildapomictic populations, (iii) evolutionprocessesof agamic complexes, and (iv) thepossibility of transferring apomixis fromsynthetic apomictic crops to landraces andwild relatives.

Apamls .... ~. Maoogemtot 01 GtHlk Dlwonlty 9Progeny of Apomictic PlantsIn most apomicts, the apomictic mode ofreproduction is linked to pseudogamy,whereby the endosperm develops only afterfertilization while the embryo developsparthenogenetically (Nogler 1984; see Crane,Chap. 3). Apomixis can be split into two logicalstages, which does not necessarily imply twodifferent genetic controls (see Sherwood,Chap. 5): (i) development of an embryo sacwithout reduction and (ii) parthenogeneticdevelopment of the embryo withoutfertilization. This results in an embryo with2n + 0 chromosomes that is geneticallyidentical to the maternal plant. However, insome cases, the embryo sac is reduced whilein others fertilization occurs. It is thereforepossible to distinguish four categories in theprogeny of apomictic plants, with respectivefrequencies dependent on success rates of thedifferent stages (Table 2.1).The four categories can be identified throughthe use of chromosome counts (or flowcytometry) and isozymes or molecularmarkers. The IRD-CIMMYT Apomixis teamused isozymes to score Tripsacum progeny(Berthaud et al. 1993). In Paa, isozymes andrandom amplified polymorphic DNA(RAPDs) have been used (Huff and Bara 1993:Barcaccia et a1. 1994). It is difficult to finddetailed results of progeny analyses in theliterature. Frequently, only morphologicaldistinctions between true (maternal) and offtypeprogeny are reported.Data on Panicum maximum (Combes 1975) arepresented in Table 2.2. In the F 2generation ofa P infeslum x P maximum cross (T19 progeny),frequencies of plants produced throughsexuality and of haploid plant productionwere high. In one case (progeny from T19-36­5),40% of a 177-plant progeny were off-types,including seven haploids. F 2progeny fromother crosses involving accession T19 were lessvariable; ,only four haploids (n + 0) were foundout of 1,500 observed. In P maximum, theproportion of off-types, including 2n + nandn + 11 was3%, based on a total of 2,100 progenyobserved. We can therefore conclude apomixisin P maximum is facultative.For the Parlhenium (Asteraceae) species,Guayule and Mariola, frequencies of the fourcategories of progeny (Table 2.3) wereextracted from Powers and Rollins (1945).Haploid plants were produced at a low rate.Most plants were produced from unfertilizedunreduced female gametes, however, theTable 2.1 Genetic constitution of progeny fromapomictic plantsFertilizationFemale meiosis yes noYes II+n* 11+0No 2I1+n** 211+0a also called B II(Rulishouser 1948)** also called Bill (Rulishouser 1948)Table 2.2 Size of four categories defined in Table2.1 for two Panicum maximum clones (fromCombes -1975)Clone Progeny size n+O n+n 2n+n 2n+O256 551 0 16 6 529H267,l' 238 4** 27 2 205a hexaploid plant, from 2n + nprogeny of "267"aa 3plants with 2n =24 and 1plant wilh 2n =23Table 2.3 Size of four categories defined in Table 2.1 for three types of progeny involving twoParthenium species. Adapted from Powers and Rollins (1945)P. argentalum x P. argentatum 342 o 14 (4) 5(1.5) 323 (94.5)P. argentatum x P. inranum 888 2 48 (5.4) 78 (8.8) 760 (85.6)P. inranum x P. argentatum 567 o 76 (13.4) 66 (11.7) 425 (75.0)

10M.. Be,th.dcategories n + nand 2n + n appeared at signifi- presented in Table 2.5. According to the table,cant rates. Stebbins and Kodani (1944) showed it appears that within one species, but ina frequency of occurrence of 2n+n progeny of various populations, the rate of 2n + n5.6%, ranging from 0.14% to 49%. Thus, progeny is variable and significant, beingapomixis in Parthenillm is largely facultative. quite high in the case of population #39 "LaToma." Experiments are in progress toIn Tripsawm we found an average 2.7% (n +analyze the effect of the environment11) progeny, 8.1% (2/1 + 11), and 89.2% (2n + 0)(flowering and pollination) on the stability ofprogeny (Table 2.4). From seeds collected inthese parameters.wild populations, we analyzed the occurrenceof 2n + n progeny (it is difficult to test for n + n For Dichanthium and Bothriochloa (Poaceae),progeny in this situation because clones are Harlan et al. (1964) reported rates of 211+11distributed in small niches of land and progeny from crosses between tetraploidinterpollination occurs from identical species. These combinations, however, aregenotypes, making detection of new isozyme interspecific and therefore are difficult topatterns difficult). The frequencies for three compare with the former examples. Bashawwild populations of TripsaCllm we observed are et al. (1992) showed that in crosses betweenTable 2.4 Estimation of apomixis rate and categories of progeny from chromosome counts and isozymeanalyses of Tripsacum populations (Berthoud et 01., unpublished data)Species 2n= %Pop ID Plant ID tested Size 72 90 108 n+n 2n+n n+n 2n+n 2n+024 143 DHBV 10 10 0 0 0 a a a 10028 163,164 MZ 20 18 a 2 a 2 a 10 9029 183 MZ 12 10 a 2 a 2 a 16.7 83.337 282, 283, 772 OM 46 43 a 3 1 3 2.1 6.5 91.443 358,361 OM 14 14 a a a a a a 10047 414 BV 12 12 a a 3 a 25 a 7548 421,423 OM 39 38 a 1 a a a 5 9552 497 OM 10 5 a 5 a 5 a 50 5053 545 II 19 17 a 2 2 2 10.5 10.5 7954 588 OH 15 13 a 2 a 2 a 13.3 86. 755 608 OH 18 12 2 4 a 6 a 33.3 66. 759 641 OM 7 7 a a a a a a 10060 654, 655 OM 23 21 a 2 a 2 a 8.7 91.362 675 OH 14 12 a 2 a 2 a 14.3 85.763 689 OH 12 11 a 1 a 1 a 8.3 91.767 734 OH 11 11 a a a a a a 10071 853 BV 14 13 a 1 a 1 a 7.1 92.872 879 OM 19 18 a 1 a 1 a 5.3 94.774 898 OM 21 20 a 1 5 1 23.8 4.8 71.4B3 960 OM 8 8 a a a a a a 10087 990 OM 5 5 a a a a a a lOa96 1076 Jl 35 31 a 4 1 4 2.85 11.4 85.798 1093 OHII 23 23 a a a a a a 100100 11,201,121 IT 40 38 a 2 a 2 a 5 95Total 446 12 36 2.7 8.1 89.2Abbreviations used: TBV=T. bravum Gray, TDH=T.dactylaides var. hispidum (Hitchc.) De Wet et Harlan, TDM=T.dactylaides var. mex;canum De WeI etHarlan, TIT=T. in/ermedium De Wet et Harlan, IJl=T. jalapense De Wet et Brink, TlC=T.lancealalum Ruprecht ex Fournier, TlT=T.lalifalium Hilchc.,TMZ= T. maizar Hernandez el Randolph.

Apomixis and the MQllagement of Ge.etk Diversity 11Pen/1isetllmjlaccidllm and P. mezial1l1m, progenyof the 2/1 + 11 and /1 + /1 types are produced(Table 2.6).In summary, when progenies are producedfrom apomictic plants, we can observe plantsof the maternal type, plants with a ploidy leveldifferent from the maternal type (genomeaddition), and/or plants with the same ploidylevel that have undergone a cycle ofrecombination. With the apomictic mode ofreproduction, we have a system favoringchanges toward higher or lower ploidy levels.Changes toward higher ploidy levels are theresult of fertilization within unreduced embryosacs. Changes toward lower ploidy levels comeTable 2.5 Variation in chromosome number forprogeny from wild populations of rripsacumdactyloides mexicanum. (Seeds were collected inthe wild)Popu- Progenylation Genotype tested38 DM38-01 7839 DM39-04 17239 DM39-15 1639 DM39-16 739 DM39-20 1739 DM39-21 1039 DM39-22 1239 DM39-23 1240 DM40-0l 5640 DM40-02 20840 DM40-03 17Chromosome number211+n72 90 108 211+n (%)73 5 5 6.4111 4 57 61 35.59 7 7 43.84 2 1 3 42.911 6 6 35.39 1 1 10.010 2 2 16.77 5 5 41.755 1 I 1.8198 10 10 4.815 2 2 11.8Totols/averages per population38 78 73 5 5 6.439 246 161 79 85 34.640 281 268 13 13 4.6Totals/averages, all populations605 103 17.0from parthenogenetic development of ared uced egg cell, which is the result of meiosisand recombination. When apomixis is active,sexuality is not eliminated but ratherdistributed overseveral generations. This topicis discussed in greater detail below.Diversity in Wild ApomicticPopulationsPernes (1975) described polymorphismsobserved in wild populations of PaniclImmaximllm in East Africa, which is the center ofdiversity for this species. He identified threetypes of populations: (i) monomorphicpopulations; (ii) polymorphic, with disjointedvariation and distinct genotypes; and (iii)polymorphic, with discrete variation.The latter was discovered in zones wheresexual diploids and apomictic tetrap)oids weresympatric. The IRD-CIMMYT team'sobservations during collections of wildTripsaCllm led to the same typology. In the caseof TripsaCllm, however, different species cancoexist in the same population. Diploidpopulations are more frequent than in PaniCllm,and several ploidy levels in within species havebeen discovered in the same populations.Three different species were found to coexistin a multispecific wild TripsaCllm population("La Toma" population #39) near Tequila,jalisco, Mexico: T pi/osllm, a diploid sexualspecies, and two apomictic tetraploid species,TbmvlIm and T dacty/aides mexicamlm. Usingfingerprinting, restriction fragment lengthpolymorphisms (RFLPs) and isozymes, M.Barre et a1. (personal comm.) identified mostof the diploid plants. Plants belonging to theTable 2.6 Size of categories defined in Table 2.1 for two Pennisetum Ilaccidum x P. mezianum crosses.From Bashaw et 01. (1992)Progeny type Progeny size 11+0 II+n 211+n & II+n* 211+n 211+0 (%)PI315868xPI214061 2,505 51 20 77 2428 (96.9%)P1220606xPI214061 3,040 58 72 148 2892 (95.1 %)• Thil hybrid mlegory hOI been recognized on morphologimllroill. Nol oilihe hybrids were analyzed cytalogim1ly.

12 J.&eo BertIoaodtwo tetraploid species were distributed inclones of variable size (Table 2.7). The geneticdiversity in this population was distributedamong 54 different diploid plants, six triploidclones (11 plants), and 18 tetraploid clones (83plants). We conclude that there are almost no"widespread" genotypes in these populations.Moving from one population to another, newgenotypes of the same species are found. InMexico, populations #38 and #39 ar e about 10km apart and both contain T bravum and Tdactyloides mexicanum. Nevertheless, theirgenotypes are distinct. As a rule of thumb, theprobability of finding distinct genotypeswithin a distance of 50 to 100 m is quite high.In population #38, we analyzed 94 asexuallyreproducing triploid and tetraploid plants,distributed in 24 clones, i.e., four plants pergenotype on average. Ellstrand and Roose(1987) observed 5.9 plants per clone in aliterature survey ofstudies involving asexuallyreproducing plants. Wild populations ofdandelion (Taraxacum sp, Asteraceae) andAntennaria sp.(Asteraceae) are comparable(Lyman and Ellstrand 1984; Ford and Richards1985; Bayer 1990).Table 2.7 Distribution of dones in Tripsacum wildpopulation liLa loma llChromo-Chromo-Type* some no. Size Type* some no. SizeBVI 72 33 DM12 54BV2 72 3 DM13 54DMI 72 4 DMI4 72DM2 72 2 DM15 72DM3 72 2 DM16 72DM4 72 27 DM17 54DM5 54 5 DMIB 54DM6 90 1 DMl9 54DM7 lOB 1 DM20 72DMB 72 1 DM21 72DM9 72 I DM22 72DMIO 72 I DM23 72• BV =l bravum; DM = T. dartylaides mexiconumIn summary, studies of wild populationsdemonstrate that apomixis does not producethe uniformity that is often simplisticallysuggested. Diversity is maintained in thesepopulations. Mechanisms generating andmaintaining this diversity may involvegenetic exchanges between differentTripsacum types and genetic recombination aspreviously described.Ploidy Cycles andOrganization of AgamicComplexesIn agamic complexes, two pools exist: one issexual diploid and the other is apomicticpolyploid (very often triploids andtetraploids). Plants considered to be apomicticpresent a certain amount ofsexuality, at a ratewe will call "k." Authors of reviews onapomixis (Nogler 1984; Asker and Jerling1992) conclude that facultative apomixis is themost common. Obligate apomixis, whenfound, occurs when k = 0, and is under thesame genetic control as facultative apomixis.In many cases, apomixis and pseudogamy(endosperm produced after fertilization bypollen) are found together. Pseudogamy is therule for apomictic Poaceae, Rosaceae, andRanunculaceae. In Taraxacum (Asteraceae),fertilization is not needed for endospermdevelopment (Ford and Richards 1985), whilein Part~enium, which belongs to the samefamily, seeds are produced only afterpollination, demonstrating that fertilization isneeded for endosperm development (Powersand Rollins 1945).Taraxacum and Parthenium AgamicComplexes (Asteraceae)Taraxacum sp. is present on five continents andabout 2,000 species have been described. Thebase chromosome number is eight, anddiploid and tetraploid forms exist. Diploidforms are sexual and, depending on thespecies, self-incompatible or self-compatible.

Apo";xn ood ,~. M""ogemeo' of 60..,;, Di..,,~y 13Polyploid forms are autonomous apomicts,either facultative or obligate. Fruits(propagules) can be obtained withoutpollination, after eliminating anthers andstigmas (Mogie and Ford 1988).In Parthenillm (Asteraceae), diploid forms with2n = 2x = 36 are sexual, and polyploid formswith 211 = 54, 72, 90, or 108, are apomictic. Inthis genus, pseudogamy is prevalent andtherefore fruits are not produced in the absenceof pollen (Powers and Rollin 1945). Ploidybuildup occurs through production of 211 + 11progeny (Powers and Rollins 1945), andproduction of haplOids from hexaploids hasbeen documented (Powers 1945). In this case,a cycle exists between tetraploids, hexaploids,and triploids, with a possibility ofincorporating diploid forms into the cyclethrough their production of 211 + 11 progenywith 54 chromosomes.Capillipediunr DichanthiunrBothriochloaAgamic Complex (Poaceae)The genera of Capi/lipedium, Dichanthium, al1dBothriochloa are distributed over Europe, theMediterranean region, Asia, Australia, and theNew World, and have been studied in detailby Harlan, de Wet, and coworkers. De Wet(1968) described a possible evolution in thegenus Dichanthium based on ploidy cyclesinvolving diploids, tetraploids, and haploids.In a broader approach, de Wet and Harlan(1970) described the interrelationshipsbetween species of the three genera of thisagamic complex (Figure 2.1). The mostcommon ploidy levels are 2x, 4x, 6x, as well assome pentaploid forms. Diploids are sexual,and polyploids are apomictic. However, formsfrom the New World are sexual and polyploid.Triploid forms are not mentioned. Gene flowoccurs in several directions, but in some casesis limited by incompatibility barriers. Geneticexchanges between Capillipedium andDiclulI1thium are effective only when speciesof Bothriochloa are involved as genetic bridges.Haploid production was detectedexperimentally and haploid plants werefound to be either sexual or sterile. Tetraploidplants can be recovered from thesedihaploids through the formation of 2n + nprogeny, with n proceeding from pollen oftetraploid plants. Rates of 2n + n productionof up to 15% have been observed.Panicum maximum Agamic Complex(Poaceae)"Guineagrass" has its origin in East Africa.It has colonized West Africa as well as thetropical areas of the New World. This agamiccomplex includes three species: Panicummaximum, P. trichocladum, and P. infestum(Combes 1975). Panicum maximum is Widelydistributed and sexual diploid forms havebeen identified (Combes and Pemes 1970),though they are very rare, having only beenfound in three very limited areas in Tanzania(Combes and Pemes 1970; Nakajima et a!.1979). The other forms are tetraploid andfacultative apomicts. Occasionally, pentaandhexaploid forms have been detected.ICAPILLIPEDIUM II DICHANTHIUM Iorillolum(2x, 4x)popillolum(6x)porvinorum olsimile(2x, 4x) (2x, 4x) cori!um ontorum~ (2x,4x) 1J{2x,4x)---~:~;;~:)~-::~~==-~~~~-:~~~h;~2:~;~----,fl '17, ,,,!~,~ 'u ~' 4ewortiono inlermedio (4x) ischoemum(5x,6x) / 11 ~'4~ (4x,6x)conconenlis / insculpro (5x, 6xl waodrowii (5x)15X)'/ ~ "­/ t>

14 JoG.. B.rth""dTable 2.8 Distribution of clones according to ploidylevel from the P. maximum collection established inCote d'ivoire (Combes 1975)Total 2x 3x* 4x 5x** 6x**551 19 0 506 12 13• 3x ploidy level not found in wild populations•• 5x and 6x overrepresenled in Ihis collection (from over·collecting inthese populotions)Table 2.8 shows the distribution of clonesaccording to their ploidy level in a collectionestablished in Cote d'Ivoire.Triploid plants have been experimentallyobtained from hexaploids (n + 0 progeny) aswell as from diploid x tetraploid crosses.(Poly-) haploids also have beenexperimentally obtained from tetraploids,and the resultant plants have been eithersexual or sterile (potentially apomictic asshown by embryo sac analyses). Thesefindings led Savidan and Pernes (1982) topropose an evolutionary scheme based onploidy cycles involving di-tetra-haploid levelsas in the Dichanthium complex. The changefrom diploid to tetraploid is realized through2n + n hybridization with pollen fromtetraploid plants. In this system, sexuality ismaintained at the diploid level. Contactbetween diploid and tetraploid plants allowsgenetic exchange between these pools(compartments) and creation of sexualtetraploid plants, allowing the release of newgenetic diversity at the tetraploid level.Paspalum Agamic Complex (Poaceae)The center of diversity for the genus Paspalumis in South America. Studies conducted byQuarin (1992), Norrman et al. (1989) andcollaborators at the Instituto de Botanica delNordeste, Corrientes, Argentina (IBONE)show that many species in this genus havegenetic pools at two or more ploidy levels(Table 2.9).ln the pool with the lowest ploidylevel, plants are sexual and self-incompatible,while in pools with higher ploidy levels, theyare apomictic and self-compatible. In manycases, the two pools are at the diploid andtetraploid level (group 3 ofTable 2.9). In somespecies, however, the sexually selfincompatibleplants are tetraploid and the selfcompatibleapomictic plants are hexaploid oroctoploid (group 6 ofTable 2.9).Some species that are sexual and selfcompatibleat the tetraploid or hexaploid level(groups 5 and 7 of Table 2.9) are not apomicticat higher ploidy levels. In other species,triploids are often apomictic and are found inspecies with sexual diploids and apomictictetraploids.As with previously cited agamic complexes,sexual forms are found at the lowest ploidylevel and apomictic forms at the other levels.However, in this example, the relationshipincludes the incompatibility system.Apomictic plants are self-eompatible and thecorresponding sexual plants are selfincompatible.Experiments should beconducted to determine whether this alsooccurs in other agamic complexes.Tripsacum Agamic Complex (Poaceae)The Tripsacum genus is restricted to the NewWorld, from 42°N to 240$. Its center ofdiversity(or origin) is located in Mexico and Guatemala,and 11 of the 16 species described for the genusare found in this region. These 11 species showdifferent ploidy levels both within and amongthemselves. The collection the team assembledfrom Mexico displayed the followingdistribution (unit = one ploidy level of onespecies in one population): diploids, 16.4%;triploids, 7.9%; tetraploids, 72%; penta- andhexaploids, 3.7%.When compared to other agamic complexes,a high frequency of triploid plants in theTripsawm complex was observed. These wildtriploid plants are apomictic, produce fertilepollen, and set good seed. All of the naturalpolyploids we observed were apomictic(Leblanc et al. 1995; and unpublished data).

ApOmil~ cnd !lo. Mcacgemeot 01 Gnetk Diversity 15Diploids are sexual, and progeny with 211 + 11chromosomes from apomictic plants occur ata significant frequency (Tables 2.4 and 2.5).Through this mechanism, many hexaploidswere produced experimentally or detected inseeds collected from a wild population.Natural hexaploid plants in wild populationswere observed at a lower frequency than inthe seed progeny we analyzed.residual sexuality exists in apomicts, whichpermits production of 11 + /1 progeny. Thissexuality favors creation of new diversity atthe tetraploid level by allowing crossesbetween apomictic plants.Our model (Figure 2.2) suggests that in theTripsaCllrn agamic complex, sexuality fosterstwo stages: (i) a change from tetraploidy toTriploid plants can be obtainedin four ways: (i) from 211 + 11hybridization within diploids,(ii) from crosses betweendiploid and tetraploid plants,(iii) from haploidization ofhexaploids (/1 + aprogeny), or(iv) from asexual propagationof apomictic triploids.Evaluation of these possibilitiesis currently underway.In addition we have observedthe presence of triploids,tetraploids, and hexaploids,and absence of diploids insome wild populations, whichsuggests that some triploidscould have originated fromhaploidization of hexaploidplants. In populationscontaining diploids andtriploids, there is a possibilityof 2/1 + n hybridization, with2/1 from the triploid femaleand 11 from a diploid maleleading to the production ofnew tetraploid plants. Wehave documented such anevent in seeds from one wildpopulation. This event showsone possible route of geneexchange from the diploid tothe tetraploid genetic pool. Wedid not discover any sexualtetraploid TripsaCLlm, butTable 2.9 Distribution of species of Paspalum according to theirincompatibility system, ploidy level, and meiosis behavior (fromstudies at IBONE, Quarin, personal comm.)Species 2x 3x 4x Sx 6x 8xolmum sex, SI+ opo, SC*berlonii sex, SI+ opo, SC*brunneum sex, SI+ opo, SC*compressifolium sex, SI+ opo, SC* opo, SC*coryphoeum sex, SI+ opo, SC*cromyorrhizon sex, SI+ opo, SC*dedeccoe sex, SI+ opo, SC*denliculolum sex, SI+ opo, SC*dislichum sex, SI+ opo, SC* opo, SC*equilons sex, 51+ opo, SC*houmonii sex, 51+ opo, SC*hydrophilum sex, SI+ opo, SC* opo, SC*indecorum sex, 51+ opo, SC*inlermedium sex, 51+ opo, SC* opo, SC*moculosum sex, SI+ opo, SC*modestum sex, SI+ opo, SC*nololum sex, 51+ opo, SC* opo, SC*poluslre sex, 51+ opo,SC­procurrens sex, SI+ opo, SC*proliferum sex, SI+ opo, SC*quodrilorium sex, SI+ opo, SC* opo, SC*rulum sex, 51+ opo, SC*simplex sex, 51+ opo, SC*boscionumsex, SC+dosypleurumsex, SC+dilololum sex, SC+ opo, SC·regnelliisex, SC+virga tumsex, SC+durilolium sex, SI+ opo, SC*ionolhum sex, SI+ opo, SC*conspersumsex, SC+inoequivolvesex, SC+loxumsex, SC+romboisex, SC+lex =lexual mode of reproduction; apo =opomictic mode of reproduction; SC =lelf compotible;51 =lelf incompotible; +=meiolil regulor; • =mei~il irregulor; -=meiolis with monyunivalentl

16 M......thodhexaploidy through 2n + n hybridization, and(ii) a change from hexaploidy to triploidy bymeiosis and parthenogenetic development ofthe embryo. From triploidy to tetraploidy thepathway is as previously described (2/1 + /1hybridization) and involves diploid plants aspollinators. Complete cycles of tri-tetra-hexahaploidplants linked to diploid plants arepossible. During these cycles, recombinationand fertilization events occur, helped by theparthenogenetic development of reducedembryo sacs and by fertilization of unreducedembryo sacs. Apomixis, in this case, enhancesthe functioning of sexuality that is distributedover several generations.Cycles and SexualityIn all agamic complexes, two different ploidypools are found: a lower ploidy pool (usuallydiploid) with sexual forms and a higher ploidypool (usually several ploidy levels, the mostfrequent being the tetraploid level) withapomictic forms. Absence of apomixis at thediploid level is thought to be due to either alack of expression of this trait at this ploidylevel or to an absence of transmission throughhaploid gametes (Nogler 1984; Grimanelli etal. 1998). The sexual pool is where most of thegenetic recombination occurs and is thereforethe pool where most of the selection on newcombinations is acting.n+n (6x)2n+n (4x) 2n+n (4xl 2n+n (4x) 2n+n (4xlFigure 2.2 Evolution of ploidy levels in Tripsacumfrom fertilization of female gamete (n or 2n) by amale gamete (n) from 2x, 4x or 6x plants orparthenogenetic development of egg cell (n+O).Gene flow from the diploid to the polyploidpool is realized in several ways. Diploidsexual plants, in some cases, can produce 2/1female gametes (Harlan and de Wet 1975). Ifthese gametes are pollinated by pollen fromtetraploid plants, tetraploid progeny will beproduced that will be sexual to a certainextent, providing an opportunity for a newburst of diversity to be tested at the tetraploidlevel. Another flow, as discussed earlier,comes from the pollination of unreduced eggsfrom triploid plants by normal pollen grainsfrom diploid plants. The triploid plants canresult from crosses between diploid andtetraploid plants. As can be seen, manyopportunities exist for the diploid pool tocontribute to the genetic diversity of theapomictic tetraploid pool. In the A/1tl?l1/1ariacomplex, several genomes from diploidspecies can be accumulated in polyploidspecies (Bayer 1987).In the polyploid apomictic pool, new geneticcombinations may also arise through residualsexuality (n + /1 progeny). We have also seenevidence that sexuality is distributed overseveral generations by creation of 2n + nprogeny in one generation, followed by n + 0progeny in the next generation. By permittingsome perenniality for each stage of the sexualcycle, this wealth of genetic recombination isfavored by apomixis, and it may becharacteristic of the apomictic mode. Moreexperimental data and modeling are requiredto isolate all of the factors involved in thegenetic recombination of apomicts.Management ofApomictic VarietiesTwo types of apomictic varieties can bedistinguished: forage varieties, which arealready released as apomictic varieties, andapomictic varieties ofcrops such as maize andpearl millet, which may be released in thenear future.

Apooni.~ aod I~. Maoogo....'.f Geoeti< D1ve"ity 17In the breeding of apomictic forage grasses,sexuality is involved at different steps andpermits genetic recombination (Valle and Miles1992; see Valle and Miles, Chap. 10). Releasedvarieties are apomictic and have beendistributed mainly outside their centers ofdiversity. In this instance, breeding activity isgenerating new genetic diversity.Because projects are now underway to transferapomixis to pearl millet, maize, wheat, andrice, we must consider the consequences ofapomixis on the diversity management oflandraces and that apomixis drasticallyreduces the recombination rate. It is importantto remember that these landraces and theirwild ancestors represent our current reservoirof genetic diversity. Thought should also begiven to conserving the diversity of wildancestors that grow near fields planted withapomictic varieties, which could be recipientsof apomixis genes through naturally occurringgene flow.Projects to transfer apomixis to pearl millet andmaize have reached an intermediate stage:advanced generations of interspecific hybridsbetween apomictic forms and cultivatedspecies have been produced that retain theapomictic trait. In the case of rice, possiblesources of apomixis are yet to be identified. Forwheat, F} and BC lhybrids between Triticumand Elymus have been produced (Peel et al.1997; Savidan et al., Chap. 11). Pearl millet andmaize are allogamous crops and so methodsmust be developed to maintain geneticallyadaptative processes once this new mode ofreproduction is introduced. In its currentdeSign, the Penl1isetum project considers thecreation of tetraploid apomictic varieties ofpearl millet (Dujardin and Hanna 1989). Uponrelease, the distinct ploidy levels of currentlycultivated millet and the tetraploid apomicticnew varieties will act as a genetic barrierbetween them. Dissemination of apomixisgene(s) from the tetraploid to the diploid levelwould involve production of triploid plants,which are usually male sterile; sodissemination through triploids should benegligible. However, in agamic complexes,apomixis seldom occurs at the diploid level.Some mechanism may suppress the expressionof apomixis or impeach transmission to thediploid level. In the pearl millet program, thereis no clear evidence that apomixis can beexpressed at the diploid level. In contrast, afew BC 2diploid-like hybrids in the maize­Tripsacum program were found to expressapomixis (Leblanc et al. 1996). These plants are211 =28 with x =10 from maize and x =18 fromTripsacum. Furthermore, triploid Tripsacum aremale and female fertile. Thus, tetraploidapomictic varieties of maize will probably notrestrict diffusion of apomixis gene(s) to othermaize lines or its wild ancestor, teosinte.Therefore, the models of diffusion of apomixisdiscussed below are based on diploidy.Apomixis fixes heterosis, thereby presentingtwo options for its use: (i) to produce apomicticF 1hybrids through breeding programs andrelease them to farmers as end products; and(ii) to release to farmers apomictic varieties thatwould be used to transfer (diffuse) gene(s) tolandraces, which would eventually becomeapomictic. In the latter case, breeding forapomixis would be a local activi ty. Infact, thesetwo options are complementary and relatedas they pertain to the diffusion of apomixisgene(s). F1apomictic hybrids could be releasedin an area where landraces and wild relativesstill exist. The transfer of the gene to theselandraces and wild relatives will depend onthe parameters cited above in option 2.Transfer of Apomixis Gene(s) andEvolution of landracesWe deduce from Sherwood (see Chap. 5), thatapomixis is probably initiated by onedominant gene (see also Valle and Savidan1996). The active A allele of this "apomixisgene" would be found mostly in the

18 JI&.. lertIoaodheterozygous condition (Aa). Thehomozygous stage (AA) has been consideredlethal in some cases (Nogler 1984).Nevertheless, in discussing apomixis transfer,we will consider three models: (i) apomixis isactive as a dominant trait, either heterozygousor homozygous (Aa or AA) with the recessivehomozygote (aa) being sexual; (ii) apomixis isactive only as a heterozygote (Aa), with therecessive homozygote (an) being sexual; and(iii) apomixis is only expressed as a recessivehomozygote (ss), while sS and SS are sexual.We will also consider a residual rate ofsexuality, k, in apomictic plants, with 0 < k < 1.Simple models of population genetics predict,in the absence of selection, the diffusion of theapomixis gene (Pemes 1971; Marshall andBrown 1981). According to the models, it ispossible for the apomixis gene to transfer tolandraces, such as maize or pearl millet, andto inadvertently move to wild relatives (Pemes1971; van Dijk and van Damme 2000).In model 1, there is one dominant allele forapomixis and three categories of genotypes atgeneration n: AA (apomictic) at a frequency ofP n . Aa (apomictic) at a frequency of 2Qn, andaa (sexual) at a frequency of R n . Gametes forgeneration n+1 are distributed according to thefollowing frequencies: male gametes A have afrequency of P n + Q n and gametes a have afrequency of Q n + R ; female gametes A havena frequency of 0, gametes a, a frequency of R n ,gametes AA, a frequency of P n , and gametesAa a frequency of 2Qn'Three genotypes will appear at generation n +1 with the following frequencies (randommating of gametes): AA at a frequency ofP n+ I = P n , Aa at a frequency of2Q n+ 1 = 2Q n +Rn(Pn + Qn)' and nn at a frequency of R n+ 1=R (R + Q )·I1 n nWith Pn+ 2Q n + R n =1, we obtain Q =1/2(1­ nPn- R n ) and the recurrence relation:Equilibrium is reached for R = 1, thepopulation being entirely sexual, or for R = 0,the population being completely apomictic.This model is identical to the model proposedby Fisher (1941) for autogamy. In fact,apomictic plantsself-reproduce, however theysimultaneously release pollen with thedominant allele to the sexual plant forms;consequently, a portion of the progeny ofsexual forms becomes apomictic.If we take into account a rate of residualsexuality, k, the variation in frequency for Aallele becomes Pn + 1 + Q n + I = (Pn + Qn)(1 +1/2(l-k)R n ) (pemes 1971).The change in frequency of allele A fromgeneration n to generation n + 1 is a functionof Rn, the frequency of the recessive allele, anda function of k. A zero value for k (obligateapomixis) maximizes the frequency of A, whilehigher values of k reduce the frequency of A.This variation would be zero if k = 1, i.e., whenall plantsare sexual with either the A ora allele.In this model, we assume random mating ofgametes. Transfer would be favored if anapomictic variety, homozygous for A, wereinterplanted with the variety (landrace) to bemodified. In the case of maize, by detassellingand harvesting only the landrace, onlyheterozygous progeny would be produced.These new plants would be apomictic andgenetically fixed. Their ability to evolve wouldrely on'the rate of residual sexuality, k. Aproportion k of the apomictic forms can beferti lized by pollen from other sources.Moreover, pollen from the first generation ofapomictic forms can be used to pollinate thelandrace. After several cycles of suc!"1backcrossing, the new variety will be identicalto the land race except that it carries theapomixis gene. Evolution in these "new"landraces will depend on the rate of residualsexuality that is retained at the end of thetransfer process.

Apomi.is aod rio. Maoage...., of Ge..'k Dlnnlty 19In model 2, apomixis is active in plants withthe Aa association of alleles. The aa genotypesare sexual. If Rn is the frequency of aagenotypes (sexual) and Qn is the frequency ofAa genotypes (apomictic), frequencies in thenext generation (n + 1) will be R n+ 1 = R (1-1 / n2Qn)' and Q n+ 1 = Q n (1 + 1/2R n )· In this case,the apomixis allele, A, diffuses in thepopulation as1 + 1/2Rn >1 and Qn + 1>Qn.We can use this model to define conditions ofequilibrium between sexual and apomicticforms if a differential fitness exists between thetwo forms. With a fitness of 1 + S for the an and1 for the Aa, the frequency changes fromgeneration n to generation n + 1 are as follows:R n+ 1= R n (1-1/2Qn)·(1 + s)/(l + sR n + sR n 2)Q n+ 1 =Qn(1 + 1/2R n ).1/(1 + sR n + sR/).In this case, equilibrium between sexual andapomictic forms will be reached for s = 1/1 +R. Initially, when apomixis starts to beestablished in a population, R is close to 1, andequilibrium can be reached with s values closeto 0.5. The fitness advantage of the sexual formsin relation to the apomictic forms has to be atleast 1.5:1 to reach the equilibrium. Onceapomixis is widely established, R is lower, andequilibrium will be reached only with highers values. In the extreme case of Q close to 1,equilibrium will be reached with s values closeto 1. In this instance, sexual forms will have toproduce twice as many seeds as apomicticforms to survive in the successive generations.If model 2 applies to apomictic varieties,transfer of apomixis to landraces could beaccomplished according to the process foundin model 1; but the transfer will take longer (atleast one more generation) because the firstgeneration will be made from Aa x aa crossesproducing Aa and aa genotypes, not from AAx aa crosses, which produce only Aa progeny.Conservation of diversity in the apomicticlandraces will depend, as in the former model,on the rate of residual sexuality, k.In model 3, apomixis is active only in plantsthat are homozygous for the recessive allele s.In this case, 55 (sexual) has a frequency of Pn,5s (sexual) has a frequency of 2Qn, and ss(apomictic) has a frequency of Rn. Using thismodel, it can be shown (Pemes 1971) that thefrequency of 5 behaves as follows:Pn + 1 + Qn + 1 = (Pn + Qn)(l - 1/2Rn)The frequency of 5 is reduced from onegeneration to the next, as 1 - 1/2Rn is alwayslower than 1.If the genetic control of apomixis follows thismodel, then transfer of apomixis will requireat least two generations. The pathway totransfer can be imagined as follows:1st generation: 55 [sexual] female x ss[apomictic] male =5s [sexual]2nd generation: 5s [sexual] female x ss[apomictic] male = 5s [sexual] + ss [apomictic]3rd generation: 5s [sexual] + ss [apomictic] xss [apomictic] or 5s [sexual] + ss [apomictic] =5s [sexual] + ss [apomictic] or 55[sexual] + 5s[sexual] + ss [apomictic]The apomixis gene can diffuse within thepopulation through backcrossing betweenplant& from the first generation and the donorvariety as male parent. In order to haveapomixis transferred within a reasonabletimeframe, the donor must be used as the malevariety of each generation. After severalbackcrosses, the local variety will betransformed to an apomictic variety, but it willbe almost identical to the donor variety.Therefore, if apomixis is active only whenrecessive alleles are present, it will be difficultto transfer apomixis to landraces while at thesame time maintaining the original traits ofthese landraces. It would require (i) the use of

20 JM Iertltaodmarkers to retain the a allele, (ii) the productionof near isogenic lines through backcrossingwith the landrace, and (iii) the selfing ofisogenic, heterozygous (Aa) lines to produceaa apomicts.2n + n ProgenyIn Tripsacum, we saw an average of 10% ofprogeny come from 2n + n hybridization; insome samples, this rate rises to 35%. Crossesbetween apomictic species of Pennisetum alsoproduced this type of progeny (Bashaw et al.1992). These forms are less frequent in otherspecies, such as Panicum maximum. If this traitis inherited during the transfer of apomixis,what behavior can be expected from cultivatedapomictic forms?The transfer projects now underway considera type of apomixis linked to pseudogamy.Once apomictic varieties are produced, mostprobably they will be also pseudogamous. Inthis case, we are concerned with the ratiobetween embryo ploidy and endospermploidy, as it has been often reported that a ratiodifferent from 2:3 (or 2:5) would introducesome developmental incompatibility at theseed level and a loss in productivity(endosperm development also depends onmaternal:paternal genome ratio; see Chap. 6,11, 12, and 13). However, for the Tripsacum, weobserved that triploid plants produce seedseven when their pollen environment comesmostly from tetraploid plants. In this case, theploidy ratio between embryo and endospermis 3:8. The 2n + n progeny we detected werefrom normal seeds with normally developedendosperm. In Tripsacl/m, the 2:3 ratio (or 2:5)between embryo ploidy and endospermploidy does not appear to be necessary for seedfilling. In general terms, we have twohypotheses to consider:1. Endosperm development is deficientwhen the ratio of embryo ploidy toendosperm ploidy is different from 2:3(or 2:5). In this case, ears display poorlyfilled kernels (with 2n + n embryo) atharvest time. There is a potential loss ofproduction due to the presence of these2n + n embryos, but these kernels wouldnot be selected as seed for the nextgeneration.2. Endosperm development is not affectedby a ratio of embryo ploidy to endospermploidy different from 2:3 (or 2:5). In thiscase, kernels with 2n + n embryos wouldgo undetected and could be used as seedfor the next generation. Apomictic plantsobtained from such embryos are triploid;they may produce normal seeds but thepollen could be sterile, which could limitfield production. If the pollen is stillfertile, as noted with triploid Tripsac!lm,no loss in production should be detected.However, ploidy buildup will occur, andmany different ploidy levels will bestored in the same variety. This ploidybuildup could raise chromosomenumbers to levels far above the optimumfor productivity, potentially resulting inlower production.In nature, 211 + n progeny production is astrategy that takes advantage of geneticrecombination, as these plants would give rise,after meiosis, to some haploid progeny byparthenogenetic development of reducedembryo sacs. In the case of an apomictic crop,it is a trai t tha t should be reduced oreliminated.Relationship between Wild Relatives andApomictic VarietiesFor the purpose of discussing the relationshipbetween wild relatives and apomictic varieties,we will use the maize-teosinte model,however, it is our belief that it can beextrapolated to pearl millet in instances wherewild relatives are still in contact with cultivatedplants. Teosinte is only found in Mexico andGuatemala. Relationships between wildrelatives and maize are not identical over thedistribution area of teosinte. The variety

Apooohis aod Ih......,.1 Gnelk DivenIty 21paruiglumis may be found in southwest Mexicoand is considered to be a very wild form, withalmost no link to modem maize. In the statesof Michoacan and Mexico, teosinte should beconsidered a weed. An incompatibility systemexhibited by these weedy teosintes, whichefficiently controls gene flow from maize toteosinte, has been detected and analyzed(Kermicle and Allen 1990). Moreover, asdescribed by Wilkes (1967), teosintes generallyhave a flowering period that is distinct frommaize. These mechanisms limit gene flowbetween this wild relative and maize.If we use model 1 to explain the transfer ofapomixis from apomictic plants to landraces,we can envisage the following process. Thefirst generation hybrid between teosinte(sexual, aa) and apomictic maize (AA) wouldbe apomictic (Aa), and BC Iplants with teosinteas female would produce Aa (apomictic) andaa (sexual) progeny. At each generation, theapomictic forms are fixed but they stillparticipate in the next generation from sexualplants through their pollen, which can transferthe apomixis allele to sexual plants. Therefore,a portion of each generation's progenybecomes apomictic. We can then deduce thatthe apomictic allele will diffuse into the wildpopulation. However, the assumptions madeto simplify the model may not prove accuratewhen applied to the relationship betweencultivated plants and wild relatives.Cultivated maize and its wild teosinte relativesare, morphologically, widely distinct.Apomictic maize x teosinte F] hybrids will beapomictic and will breed true. Sexual maize xteosinte Fls are known to have a low fitnessdue to their intermediate morphology andadaptation, and they are easily recognizedmorphologically. When they grow in a field,they are not harvested. However, if the hybridis apomictic, its pollen will transmit the A alleleat a rate of50%. Pollination efficiency dependson synchronization between flowering of thesehybrids and the wild relatives. As a lack ofsynchronization between the two types ofplants is anticipated, the gene flow betweenthem should be minimal. These observationsdeviate considerably from the assumptionsposited in the model in which apomictic plantsare expected to engage in pollination inproportion to their frequency in thepopulation. Moreover, in the long run, theapomictic intermediate forms should have alower fitness than the sexual forms, becausethe latter can take advantage of more newrecombinations and adapt faster toenvironmental changes. As noted earlier, astable polymorphism between sexual andapomictic forms is possible when fitness valuesof the two forms reach a certain ratio. We havealso observed that the speed of apomixisdiffusion is a function of the rate of residualsexuality-a high level of residual sexualitywill slow apomixis diffusion.Promoting Genetic Diversity and theRelease of Apomictic VarietiesWe base our models for apomixis diffusion onthe hypothesis that this mode of reproductionis under a simple genetic control. Currentknowledge about the mechanisms underlyingapomixis, however, is very incomplete,especially regarding the expression of anapomixis gene in a new genetic background,as would be the case with a Tripsacum apomixisgene transferred into a maize background. Ifgenetic control of apomixis in landraces andnew varieties involves several genes or a majorgene and modifiers, the dynamics of diffusionwill be more difficult to describe andtransformation ofcurrent varieties to apomicticvarieties would have to be carried out byprofessional breeders. In this instance,apomixis could be used as a genetic fixationtool and new varieties with a complex geneticstructure could be created and released. Suchvarieties would contribute to the maintenanceof diversity at the farmer's field level.

22 JoA•• BerthudFurthermore, if apomixis is controlled bymultiple genes, the probability of diffusing thistrait to wild relatives is extremely low. A wildplant would need to receive several genes(probably on several different chromosomes)from the cultivated plant to become apomictic.This transfer would certainly lower its fitnessto a value unacceptable for survival in the wild.If apomixis is under a simple genetic control,diffusion of apomixis to landraces and wildrelatives is possible. Apomixis reducesrecombination rates and could be perceivedas a danger for conservation of geneticdiversity of wild relatives and landraces. Inactuality, current genetic diversity is the resultof a long process of domestication, which isstill underway in some regions of the world,especially where wild and cultivated plantscontinue to exchange genes, often within atraditional agricultural system. Somewhatsurprisingly, it is in regions where traditionalagriculture prevails that apomixis could be themost helpful. We know that obligate apomixisis an exception and facultative apomixis ispredominant (Asker 1979). If during thetransfer of apomixis to crops, residual sexualityis also transmitted and expressed in the newapomictic crop, we could rely·on the rate ofrecombination inherent in this process togenerate new genetic combinations. Even atlow rates, new combinations may beinteresting to farmers who could select andpropagate them easily. As long as apomixis isnot obligate, landraces can still evolve. It mayalso be possible to introduce new genes from"exotic" and modem sexual varieties. Crosseswill occur only in the proportion k (rate ofresidual sexuality). But if these new productscan be detected by markers or by their hybridvigor, following selection, they could serve asan important source of seed for the nextgeneration. The possibility and rate ofevolution of these apomictic varieties willeventually depend upon the rate of residualsexuality; therefore, it will be important toconsider this paramet-er when transferringapomixis from wild apomixis donor plants tofirst apomictic varieties. This rate of residualsexuality may depend on genetic factors.Controlling these factors, in order to adapt thevalue of this parameter in new apomicticvarieties, could be extremely useful as we seekto conserve the genetic diversity of landracesand allow for their continual evolution.Areas of traditional agriculture are repositoriesfor most of the genetic diversity of crops. Theconservation of this diversity is threatened,however, by changes in technical practices thatcan suppress current gene flow and by theintroduction of new modern varieties withlimited genetic diversity (e.g., F jhybrids).Producing new varieties from localgermplasm may be advantageous to farmers,and it could be more easily accomplished ifapomixis is incorporated into the breedingscheme (see Toenniessen, Chap.1). In thisscenario, landraces with high genetic diversitywould be maintained in these farmingsystems, thus limiting the diffusion of varietieswith low genetic diversity. This diversitywould serve as a reservoir for future evolution.ReferencesAsker, S. 1979. Progress in opomixis resellrch. Heredffos91: 231-40.Asker, S., and l. Jerling. 1992. Apomixis in Plonll. Boco Roton, Florida:CRC Press.Bashaw, E.e., M.A. Hussey, and KW. Hignight. 1992. Hybridizotion (n+nand 211+n) of farultafive apomidtic species in the Pennise/um agamiccomplex. In/emotionol Journal of PIont x;ence 153: 466-70.Barca((ia, G., A. Mazzucato, M. Pezzotti, and M. Fokinelli. 1994.Comporison between isozyme ond RAPD ono~ses to meen aberrantplants in Poo pra/ensis Lprogeny. Apomixis News/elfer 7: 29-30.Boyer, RJ. 1987. Evolution and ph~ogenelic relationships of theAn/ennorio (Asteraceoe: Inuleoel polyplOid agamic com~exes. BioI.Zen/. bl. 106: 683-98.--.1990. Patterns of donal diversity in the An/ennorio raseo(Asteroceoe) po~ploid agamic complex. American Journal of B%ny77: 1313-19.Ber1haud, J., M. Borre, and Y. Savidon. '993. Managing genetic resourcesof the agamic genus, Tripsowm exploration, conservation,disfrib~tion. ASA meeting, Gndnnoti, Ohio, Nov. 7-12, 1993.abstract: Pp. 186.Corman, J.G., 1992. Unifying our effOllsla create opom:ctic crops.Apomixis News/elfer. 5: 47-50.

Apomixis .,d Ih. M..og.....t .f G...tic Dly...~y 23--. 1997. Asynchronous expression of duplicate genes inongiosperms moy couse opomixis, bispory, tetraspory, ondpo~embryony. Biological Journol of the Unneon Soriety 61: 51-94.Combes, D. 1975. Po~morphisme el modes de reproduction dons 10section des Moximae du genre Ponicum (Graminees) en Afrique.Memoires ORSTOM (Paris) #77.Combes, D., ond J Pernes 1970. Voriations dons les nombreschramosomiques de Panicum maximum en relation avec Ie mode dereproduction. CR. Arad. Sri. Paris 270: 782-85.De Wet, J.MJ. 1968. Diploid·tetroplaid-hoploid eyries and the origin ofvoriobility in Dirhanthium. Evolution 22: 394-397.De Wet, JMJ., ond JR. Harlan. 1970. Apomixis, po~ploidy, ondspeciotion in Dirhanlhium. Evolution 24: 270--77.Dujardin, M., and W,w. Hanna 1989. Developing apomictic pearl milletrharocterizotionof 0 BC:J plant. Journol of Genetics ond Breeding 43:14S-50.Ellstrand, N.e., ond M. Roose. 1987. Pa"erns of genotypic diversity inrlonal plant species. American Journal of Botany 74: 123-31.Fisher, R.A. 1941. Average excess ond overage effect of a genesubstirurion. Ann. Eugen. 11: 53-61Ford, H., and AJ. Richards. 1985. Isozyme variation within and betweenTaroxacum agamospecies in a single locality. Heredity 55: 289-91.Grimanelli, D., O. leblanc, E. Espinosa, E. Peroni, D. Gonzalez de lean,and Y. Savidon. 1998. Non-Mendelian transmission of apomixis inmaize- Tripsacum hybrids caused by a transmission ralio distorlion.Heredity 80: 40--47.Hanno, W., M. Dujordin, P. Ozios-Akins, E. Lubbers, and L. Arlhur. 1991Reproduction, cytology, and ferlility of pearl millet x Penniselumsquamulatum BC 4plants. Journal of Heredity. B4: 213-16.Harlan, JR., M.H. Brooks, D.S. 8argaonkar, and JMJ. de Wet. 1964.Nature and inheritance of apomixis in Bathriarhloa and Dirhan/hium.Botanical Gazette 125: 41-46.Harlan, J.R., and JMJ. De Wet. 1975. On 0. Winge and a prayer: theorigins of po~laidy. Bot. Rev.41: 361-69.Huff, D.R., and JM. Bara. 1991 Determining genetic origin of aberrantprogeny from facultative apomictic Kentucky bluegrass using 0combination of flow cytometry and ~Iver·stained RAPD markers.Theare/iral and Applied Genetics 87: 201-208.Kermirle, J.L., and J.O. Allen. 1990. Cross·incompatibility between maizeand teosinte. Maydiro 35: 399-408.Leblanc, D., M.D. Peel, JG. Carman, and Y. Sovidan. 1995.Megasporogenesis and megagametogenesis in several Tripsacumspecies (Poaceae). Americon Journal of Botany B2: 57-61Leblanc, 0., D. Grimonelli, H.1. Faridi, J. Berlhaud, and Y. Sovidan. 1996.Reproductive behavior on moize- Tripsacum pa~haploid plants:implications for the transfer of apomixis inlo moize. Journal ofHeredity 87: 108-11.lyman, J.C., and H.e. Elistrand. 1984. donal diver~ty in TaraxacumaHirinale (Asteraceael, an apomict. Heredity 53: 1-10.Marshall, D.R., and A.H.D. Brown. 1981. The evolution of apomixis.Heredity47: 1-15.Magie, M., and H. Ford. 19B8. Sexual and asexual Taraxacum species.Biologiral Journal of the Unneas Society 35: 16s-68.Nakajima, K., 1. Komatsu, H. Mochizuki, and S. Suzuki. 1979. Isolation ofdiploid and tetraploid sexual plants in Guinea gross (Panicummaximum). Japan Journal of Breeding. 29: 228-38.Hogler, G.A. 19B4. Gametophytic apomixis. In B.M. Johri (ed.),&nbryology of Angiosperms. New York: Springer-Verlag. Pp. 47S­518.Horrmonn, G.A., CL. Quarin, and B.L. Burson. 1989. Cytogenetics andreproductive behavior of different chromosome roces in six Paspalumspecies. Journol of Heredity 80: 24-28.Peel, M.D., J.G. Cormon, Z,w. Uu, and R.R.C Wang 1997. Meioricanomalies in hybrids between wheat and apomictic f}ymus rertise/us(Nees in Lehm) A. love and Connor. Crop Srienre 37: 717-21Pernes, J 1971. Etude du mode de reproduction: apomixie facultative dupoinr de vue de 10 genetique des populations. Travaux el documenlsde I'ORSTOM (Paris) #9.--. 1975. Organisation evolutive d'un groupe ogomique: 10 sec1iondes Moximae du genre Panicum. Mtimaires ORSTOM (Paris) #75.Powers, L1945. Ferlilizatian wilhout redution in guayule (Ponheniumargentotum Gray) and a hypothesis oslo the evolution of apomixisand polyploidy. Genetirs 30: 323-46.Powers, L, and R.C Rollins. 1945. Reproduction and pollination sludies onguayule, Panhenium argentatum Gray and P. inranum H.B.K. Journalof the Ameriran Society of Agronomy. 91r-112.Quarin, e.L 1992. The nature of apomixis and its origin in Panicoidgrasses. Apomixis Newsletter 5: 7-15.Rurishouser, A. 1948. ~eudogamie und po~morphie in der GaNungPotentilla. Arrh. Julius Klaus-Stih Vererbungsforsrh 23: 267-424.Sovidan, Y., and J. Pernes. 1982. Diploid-Ietraploid-diho~oid eyries andthe evolution of Panicum maximum Jacq. Evolution 36: 596--600.Stebbins, G.L, and M. Kodani. 1944. Chromosomal varia lion in Guayuleand Moriola. Journal of Heredity. Pp. 163-72.Valle, e.B. do, and J'w. Miles. 1992. Breeding of opomictic species.Apomixis Newsletter 5: 37-47.Volle, C.B. do, and Y. Sovidan 1996. Genetics, cytogenetics, andreproductive biology of Brachiaria. In J'w. Miles, B.L Ma~s, ond CB.do Valle (eds.), Brarhiaria: Biology, Agronomy, and Improvement.Coli, Colombia: C1AT-EMBRAPA. Pp.147-61von Dijk, P., ond J. von Damme. 2000. Apomixislechnology and rheparadox of sex. Trends in Plant Scienre 5(2): 81-84.Wilkes, H.G. 1967. Tea~nte: the dosest relative of moize. Ph.D_disserlation. Bussey Institute, Harvard University, Cambridge,MassachuseNs.

Chapter 3Classification of Apomictic MechanismsCHARLES F. CRANEIntroductionApomixis has traditionally beenseparated intoasexual seed production (agamospermy) andreplacement of flowers by vegetativepropagules (vivipary). In practice, manyresearchers define apomixis as agamospermy,which in turn is divided into adventitiousembryony and gametophytic apomixis.Adventitious embryony is the formation ofsomatic embryos from ovular tissues outsidethe embryo sac, although endosperm in theembryo sac usually is necessary to supporttheir maturation, and the resulting somaticembryos sometimes compete with a zygoticembryo within the same ovule. Gametophyticapomixis is, at the least, a two-step processinvolving the production of a 2nmegagametophyte, whose egg developswithout fusion of egg and sperm nuclei; otheraspects of fertilization can be completelynormal or completely missing, depending onthe type.Adventitious embryony can be divided intonucellar and integumentary embryony,depending on where the embryos arise.According to Naumova (1993), adventitiousembryos in nature always originate fromsingle cells, termed embryocytes, which candifferentiate in the nucellus or integuments,before or after fertilization of the embryo sac.As summarized by Naumova (1993), the earlyorganization of adventitious embryoscharacteristically differs from that of zygoticembryos, possibly because space isconstrained and the mechanical stress field inthe nucellus differs from that in the embryosac. Multiple adventitious embryos candevelop asynchronously within the sameovule (Gustafsson 1946), whereuponcompetition and packing further affect theirmorphology. Polyembryony is very frequentin most cases of adventitious embryony, andfertilization of the central cell is generallynecessary for seed set. The developmentalinterpretation of adventitious embryony issimple: the embryocytes are induced to act likezygotes. The number of ways in which thisinduction can occur is not known, but they allmust cause repeated mitosis.Types of GametophyticApomixisIn gametophytic apomixis, 2n embryosacs canarise in at least nine different ways, dependingon the species. The embryo and endospermin such embryo sacs can develop in at leastfive ways. Therefore, at least 45 types ofgametophytic apomixis are theoreticallypossible, but only half of these account fornearly all cases of naturally occurringgametophytic apomixis. The nine types areschematically diagrammed in Figure 3.1,which lays out their proposed developmentalbasis. The types are described below(references are given afterward as part of ahistorical perspective). Note that "strike"(nondivision or precocious degeneration ofnuclei in the chalazal end of developing

'r·\(1a"lfi

26 (llarIesf.(r••3) The Ixeris-type. The Ixeris-type follows theTaraxacum-type through the formation of ameiotic restitution nucleus. The next nucleardivision is not followed by formation of a cellwall, nor is the division after it. Cell walls formonly as the embryo sac matures after the thirdround of divisions. All daughter nuclei fromthe restitution nucleus contribute to themature, 8-nucleate embryo sac.4) The Blumea-type. In the Blumea-type, amore or less mitotic division of themegasporocyte yields a dyad of 2nmegaspores. The chalaza} megaspore gives riseto the mature, 8-nucleate embryo sac. Thedevelopment is identical to the Taraxacumtypeafter the restitution nucleus has formed.Therefore, existence of the Blumea-type mustbe demonstrated by the absence of therestitutional stages of the Taraxacum-type; thisproof requires careful, thorough sampling andordering of all the meiotic stages.5) The Elymus rectisetus-type. In Elymusrectisetus, three interrelated developmentsoccur that superficially resemble the Blumeatype.The megasporocyte nucleus enlarges andis deformed in all three types. The nucleusappears to be under tension, as evidenced byparallel alignment of its chromatin, and by itschalazally pointed or occasionally waspwaistedshape. Triads and dyads with unequalnuclei indicate amitotic division in a very lowpercentage of megasporocytes. The first visibleprophase follows the nuclear deformationepisode and leads to what could be considered,more or less, a mitotic division. Here the threesubtypes diverge: this division can result in adyad of completely separated cells, ahemidyad of incompletely separated cells, ora binucleate embryo sac. All three behaviorscan occur on the same individual. Thechalazal-end nucleus in all three cases oftenundergoes a milder form of the deformationseen in the megasporocyte. The fate ofhemidyads and directly binucleate embryosacs has not been thoroughly investigated, butthe chalazal member of dyads typicallyundergoes three rounds of mitosis and formsthe mature, 8-nucleate embryo sac. Themicropylar nucleus of hemidyads sometimesdegenerates as if it had been completely cutoff.6) The Antennaria-type. In the Antennariatype,there are no meiotic stages. Themegasporocyte undergoes three rounds ofmitosis, typically after an initial period ofenlargement and vacuolation. The matureembryo sac has eight nuclei, which arearranged as in the conventional Polygonumtype.7) The Hieracium-type. In the Hieraciumtype,one to several nucellar cells enlarge,become vacuolate, and go through threerounds of mitosis, resulting ideally in aconventionally organized 8-nucleate embryosac. The megasporocyte undergoes meiosis insome species and degenerates in others. Insome cases, reduced and unreduced embryosacs can coexist in the same nucellus. Theability to form multiple embryo sacs in thesame ovule increases the frequency ofpolyembryony.8) The Eragrostis-type. In the Eragrostis-type,there are no meiotic stages. Themegasporocyte undergoes only two roundsof mitotic division, leading to a 4-nucleateembryo sac with an egg, two synergids, andone polar nucleus, or alternatively, to an egg,one synergid, and two polar nuclei. Theoriginal description of the type also includedbipolar sacs with more than four nuclei, whichat that time were interpreted as modificationsof the Antennaria-type (Streetman 1963). Suchsacs may instead represent facultativeoccurrence of the Polygonum-type.9) The Panicum-type. In the Panicum-type,one to 25 nucellar cells enlarge, increase invacuolation, and ideally divide twicemitotically, although some divide only once

Ool$ifit.tIo. of Apoonktk MedIa.lsms 27and others not at all. The mature embryo sachas an egg, two synergids, and a polar nucleus,or else, an egg, one synergid, and two polarnuclei. The megasporocyte can undergomeiosis or degenerate, depending on thespecies. The potential for polyembryony ishigh in this type.Most of these developments were discussedby Battaglia (1963), who also discussed variousaberrations, and later by Nogler (1984) andAsker and Jerling (1992). Descriptions werefirst published for the Antennaria-type in 1898(JueI1898), the Taraxacum-type in 1904 Quel1904), the Hieracium-type in 1907 (Rosenberg1907), the Ixeris-type in 1919 (Holmgren 1919;Okabe 1932), the Allium odorum-type in 1951(Hakansson and Levan 1951, 1957), thePanicum-type in 1954 (Warmke 1954), theEragrostis-type in 1963 (Streetman 1963; Voigtand Bashaw 1972), and the Elymus-syndromein 1956 (Hair 1956), with significantamendments in 1987 (Crane and Carman1987). The Blumea-type, described for Erigeronramostls by Holmgren (1919), was latergenerically refuted (Fagerlind 1947) and finallyrehabilitated for Blumea eriantha(Chennaveeraiah and PatiI1971).A number of previous classifications, e.g.,Asker and Jerling (1992), have recognizedautomixis, which is the fusion of nuclei in thesame gametophyte to produce a 2n,homozygous egg cell. Automixis has beenpurported in several species of Axonoptls(Gledhill 1967), but the evidence can easily bereinterpreted as early degeneration of one ofthe synergids in the Polygonum-type beforefertilization. Like gametophytic selfing inferns, the mechanism would produceindividuals that are completely homozygouswithin genomes. There is no genetic evidencefor or against apomixis in the affectedAxonopus species. Automixis has also beenclaimed for Rtlbus caesitls, as discussed byAsker and Jerling (1992), but the case isscarcely closed. The evolution of automixis isimplausible in angiosperms, becauseautogamy evolves easily in most groups andhas the same genetic result over the course ofgenerations.Subsequent Steps of Development1) Embryos. Apomictic embryo developmentfrom egg cells has been summarized in earlierresearch, e.g., Asker and Jerling (1 "92) andBattaglia (1963). There are three major types:pseudogamy, semigamy (= hemigamy), andautonomous parthenogenesis. There is notradition for naming these last threedevelopments for the type genera in whichthey occur. The developments differ in theorder in which egg and primaryendospermatic nuclei divide, the ability of thecentral-cell nuclei to divide withoutfertilization, the number of polar nuclei thatcontribute to the endosperm, and the abilityof the egg cell to undergo plasmogamy.Pseudogamy is traditionally defined asasexual seed set that requires pollination anddoes not involve adventitious embryos.However, a more restricted definition seemspreferable: pseudogamy is seed set throughfertilization of the central cell, but not the egg,in the absence of adventitious embryos.Semigamy is seed set with full fertilization ofthe central cell and only plasmogamy in theegg, ~here the sperm nucleus can be walledoff, or divide one to several times, or contributeequally to chimeric or twinned embryos. Thelast of these behaviors is not found routinelyin recurrent apomicts, but is typical of the Semutant in cotton (Turcotte and Feaster 1969),in which the sectors can be easily visualizedwith leaf-color mutations. In Se cotton,complex chimeras with maternal haploid,paternal haploid, and hybrid sectors arefrequent; they probably arise from coalescenceof one pair of adjacent spindle poles upon

simultaneous division of the proximate eggand sperm nuclei in the zygote (Turcotte andFeaster 1973,1974).Autonomous parthenogenesis is the formationof embryo and endosperm withoutfertilization ofeither the egg or the central cell;such apomicts require no pollination for seedset and frequently are male-sterile.Autonomous endosperm formation has alsobeen reported in a case of adventitiousembryony, namely Euphorbia dulcis(Gustafsson 1946).2) Endosperm and embryo development.Pseudogamy and autonomous parthenogenesiscan be subdivided on the basis of theircapacity to form embryos in the absence ofendosperm. The apomictic egg dividesautonomously in Potentilla (pseudogamy) andWikstroemia (autonomous parthenogenesis)(Gustafsson 1946). The egg divides after thefirst endosperm nucleus in Themeda(pseudogamy) and Crepis (autonomousparthenogenesis), although the potential forits eventual division in the absence ofendosperm has not been critically evaluatedin most cases. In semigamy, the unfertilizedegg degenerates without dividing (Crane1978), contrary to an earlier report of eventualdivision (Cae 1953). There is also variation inthe fusion of the polar nuclei with each otherin apomicts. In autonomous parthenogenesis,the polar nuclei can fuse before dividing (e.g.,in Taraxacum, Elatostema, and most Alchemillaspecies), facultatively fuse (e.g., in Crepis), ordivide without fusing (e.g., in Antennaria)(Gustafsson 1946). In pseudogamy, the polarnuclei may (Ranunculus auricomus; Nogler1972) or may not (Amelanchier; Campbell et al.1987) fuse with each other when the centralcell is fertilized. Both polar nuclei contributeto the endosperm in semigamous Cooperin andin the Se mutant of cotton (Crane 1978; G. L.Hodnett, unpublished).Alternative ClassificationsThe standard classification of apomicticembryo sacs (e.g., Asker and ]erling 1992) hasfollowed Edman (1931) and Gustafsson (1946)in recognizing embryo sacs that arise fromdiploid megaspores (diplospory) and embryosacs that do not arise from megaspores (apospory).Under this scheme, the Hieracium- andPanicum-types are aposporous, and theremaining seven types are diplosporous (Note:Types that were undescribed at the time ofthese old papers (Eragrostis, Panicum, Elymus,and Allium odorum) have been assigned on thebasis of the classifying paper's underlyingconcepts). Fagerlind (1940, 1944) defineddiplospory to be the production of diploidmegaspores from meiosis, in effect restrictingit to the Taraxacum-, Ixeris-, and Alliumodorum-types. He termed purely mitoticdevelopments apospory, which could begenerative (Antennaria, Eragrostis) or somatic(Hieracium, Panicum). The Blumea- and dyadformingElymus-types felI into a halfwaycategory, semiapospory. Battaglia (1963) alsorecognized apospory for purely mitoticdevelopments and differentiated gonial andsomatic subtypes. However, he grouped allmodifications of meiosis under aneuspory.Thus he considered the AlIium odorum-,Taraxacum-, Ixeris-, Blumea-, and dyadformingElymus-types to be fundamentallysimilar in that they involved modified meiosis.Conseq'uently, one issue that has emerged isthe definition of a megaspore: is it morefundamentalIy the product of amegasporocyte or the product of femalemeiosis? Many other terms have been definedand redefined, leading to terminological andconceptualconfusion. Forexample,]ohri et al.(1992) speak of "apospory of the Taraxacumtype"in Balanophora globosa. The majorproblem has been an absence ofhard evidenceas to what genetic changes are necessary toestablish apomixis and what the relevant geneproducts do. Furthermore, the traditional

(lasslfl

30 (\aries f. Crao.differentiating the egg and micropylar-endpolar nucleus. The egg cannot form if the thirdmitotic division does not occur; its precursordefaults to being a polar nucleus. Thus meiosisII and the first embryo-sac mitosis can beskipped without affecting female fertility oroffspring ploidy, as in the Adoxa type of sexualembryo sac, however the last divisions arecritical for female fertility. The details of thenumerical abnormalities observed in Cooperia,and their implications for the evolution ofsexual embryo sacs, are reported elsewhere.Following are the author's interpretations ofthe nine apomictic types of embryo sacs.Ameiotic Developments ofMgagametophytesEndomitosis in the Allium odorum-typewould result from relaxation of thechromosomal double-strandedness checkpointpostulated above for DNA synthesis.Endomitosis can occur in a wide variety ofplant cells, including, most relevantly, theanther tapetum (Crane et al. 1993), which ishomologous to cells in the nucellus.Megasporocyte endomitosis does not affect themorphological course of meiosis; it canprecede any of the sexual types of embryo-sacdevelopment, and it is probablyunderreported. One consideration is how thechromosomes behave after chromatidseparation and secondary replication in thepremeiotic G z nucleus. If they scarcely move,the former sister chromatids would lie in closeproximity as pairing begins in leptotene, and--mttltiva!en!? rarely, if ever, would form.The Taraxacum-type results from induction ofthe meiotic interphase during the prophase ormetaphase of meiosis 1. The contractedchromosomes are thus induced to decondenseand await the second meiotic division.Another possible explanation for theTaraxacum-type is induction of an Elymusstyle"waiting state" (see under Elymus below)during prophase I, with recovery in time toundertake meiosis II, although this inductionwould not necessarily result in a restitutionnucleus. Direct induction of meioticinterphase would likely acceleratemegasporogenesis relative to sexual ovules,whereas entering a "waiting state" wouldcause a significant delay.Induction of meiotic interphase also explainsthe Ixeris-type, but in this case the inductionis superimposed.on a bisporic or tetrasporictype of reduced embryo sac, e.g., the Fritillqriatypein Rudbeckia (Ba ttaglia 1946, 1963).Assuming a checkpoint for chromosomaldouble-strandedness, bisporic and tetrasporictypes lack the second meiotic division, andthus have their own induction of megasporegermination after the first meiotic division(bisporic types) or during the first meioticdivision (tetrasporic types). Accordingly, therestitution nucleus awaits the first embryo-sacmitosis rather than meiosis II in the Ixeris-type,and there is no cross-wall to separate 2n"megaspores."In the Blumea-type, meiotic interphase isinduced in the premeiotic megasporocyte. Itsgenetic basis could be the same as in theTaraxacum-type, and the two types couldcoexist in the same species or individual if thetiming of induction varied. PrOVing theexistence of the Blumea-type is intrinsicallydifficult because it must be shown that therestitutional stage of the Taraxacum-type isabsent. Therefore, measuring the frequenciesof these two types in a polymorphic individualmight not be feasible with microscopictechniques that kill the ovule before it isexamined. The outcome of superimposing theBlumea-type on a bisporic or tetrasporicsexual type remains unknown. It mightsuperficially conform to the Antennaria-type.The Elymus-type appears to result fromincomplete differentiation of themegasporocyte from the nucellus. Not only is

00,,11k.1Io1 of ApIIIlktl< M.d'.II,.., 31its callosic wall deficient, but the pattern ofvacuolation and the occasional presence of anoxalate crystal in the megasporocyte conformto typical behaviors of the adjacent nucellarcells. Theex ten t to which this similari ty causesthe wall to be deficient is not clear, nor is thedegree to which the deficiency prevents anecessary physiological isolation for fulldifferentiation of the megasporocyte. Thecytoskeleton of the megasporocyte isdisturbed, as evidenced by frequently obliquedivision planes and the deformation of thenucleus. Perhaps the meiotic spindleapparatus is not completely suppressed, andso interacts with the preapomeiotic nucleus.But this does not explain the seconddeformation in the chalazal 211 megaspore.Another feature of the E1ymus-type is the needfor parallel control of development. Themegasporocyte enters a "waiting" state andvariably emerges in time to undertake meiosisII, the first embryo-sac mitosis, or anintermediate condition, respectivelyaccounting for dyads, directly binucleate sacs,or hemidyads. During the "wait," a master"clock" continues to run, albeit abnormallyslowly in comparison to the Polygonum-typein related sexual genotypes; this clock allowsrecovery from the wait.The Antennaria-type can be explained in twoways: by induction of megaspore germinationin the megasporocyte or as a much belatedrecovery from an Elymus-style waitingcq'"ndition. The frequently abnormalehlargement and vacuolation of themegasporocyte are consistent with the latterexplanation, but intraspecific polymorphismfor the Antennaria- and Hieracium-types inPatentilla and Paa is consistent with the formerexplanation. Details of gene expression ingerminating Polygonum-type megasporesshould provide clues as to which examples ofthe Antennaria-type result from megasporegermination. It should be noted that thetetrasporic sexual types also involve inductionof megaspore germination during meiosis I,and that having tetrasporic sacs does not seemto predispose individual taxa to evolve theAntennaria-type or vice versa. Thus different,separately inducible aspects of megasporegermination might impact meiosis in differentways, or the event that calls up megasporegermination might fun

32 CWIts F. C,..haploids from normal cotton plants pollinatedwith semigamous pollen have been reported(Turcotte and Feaster 1963). In the absence ofsuch rare haploids, one is tempted to presumethat semigamy results from induction of asubset of zygotic behavior that preventskaryogamy but not plasmogamy. Ifsemigamywere found to function in both eggs andsperms, one would instead suspect a generaldefect in fertilization capacity.The variation from autonomous to pollinationdependentembryo induction in pseudogamysuggests that embryos can be induced in morethan one way. Perhaps one way triggers fullembryonic development, while anothermerely primes the egg (or sexual zygote) torespond to a stimulus from the dividingendosperm. In the latter case, the apomicticegg of an unfertilized ovule would degeneratewithout ever dividing, whereas in the formercase, the apomictic egg would eventuallydivide, even if it does not regularly do sobefore the primary endosperm nucleusdivides or before unfertilized embryo sacsdegenerate. Savidan (1989) has proposed thatbecause pseudogamous Panicum maximumundergoes such early induction andmaturation of the embryo sac, the egg cellcompletes its cell wall before pollination.While the formation of the egg wall isprogressive in grasses (Cass et al. 1986), it isnot clear that old eggs inevitably wouldcomplete their wall. In vitro fusion of egg andsperm protoplasts of maize leads to rapid cellwall formation (Breton et al. 1995, andreferences cited therein), suggesting thatinduction of zygotic behavior is responsiblefor the physical barrier to fertilization ofpseudogamous eggs.Autonomous parthenogenesis would seem torequire separate inductions of endospermdivision and either the "triggered"(Wikstroemia) or merely the "primed" (Crepis)condition. Yet separate inductions wouldimply that the genes for each could beseparated and that genotypes lacking theautonomous endosperm induction would befully pseudogamous.This is possible in generawhere instances of pseudogamy andautonomous parthenogenesis are known, e.g.,Poa and Crataegus, but it is inconsistent withunreduced autonomous parthenogenesis inthe absence of pseudogamy in apomicticCichoreae (Taraxacum, Chondrilla, Crepis, andIxeris).Genomic imprinting of gametic nuclei isanother consideration in the developmentalinterpretation ofapomictic embryogenesis. Inthe Polygonum-type, the endosperm has a 2:1maternal:paternal genome ratio; deviationsfrom this ratio frequently cause endospermabortion in interploidy crosses. InsemigamousCooperia, the reduced sperm nucleus fuseswith both unreduced polar nuclei (Crane1978), resulting in a 4:1 maternal:paternal ratioin the endosperm, which functions adequatelyfor seed germination in this genus. Inpseudogamous Rtmunculus, both sperm nucleican fertilize the central cell, resulting in a 4:2maternal:paternal ratio and normalendosperm function (Nogler 1972). Inpseudogamous Crataegus, the polar nuclei donot fuse with each other, and a sperm fuseswith only one polar nucleus, resulting in thenormal 2:1 maternal:paternal ratio (Campbellet al. 1987). The situation in autonomousparthenogenesis is not clear, because thenumber and biochemical effect of imprintedloci is not obvious. Imprinting of a tubulinlocus (Lund et al. 1995) and the dzrl locus(Chaudhuri and Messin 1994) has beendocumented in developing maize endosperm,and it is possible that the number and sexspeCificityof imprinted loci vary greatlyamong plant species. What matters here is thenumbe~ of loci that orchestrate the imprintingpattern. Since the evolution of apomixis iseasier if fewer loci control it, one might expect

Clanllkatlo. of Apamictlc ~is .. 33one or a few key loci to affect this pattern.Ehlenfeldt and Hanneman (1988) analyzed thecrossing behavior of diploid hybrids betweentwo diploid, sexual Solanum species that differin endosperm balance number, and theyconcluded that three unlinked loci controlledit. Obviously, comparable evidence is neededfor the immediate relatives of autonomouslyparthenogenetic apomicts.OutlookThe speculative nature of the precedingdiscussion emphasizes how little isexperimentally known about meiosis,megasporogenesis, gametogenesis, andfertilization in sexual plants, and how little ofthe diversity of apomictic behaviors has beenstudied with modern methods ofgenetics, cellbiology, and molecular biology. Most apomictsare polyploid, perennial herbs with longgeneration times, and the effort to maintainsuitably large, uncontaminated populations ofadvanced generation hybrids has repeatedlydefeated even rudimentary attempts tounderstand the inheritance of apomixis. Thepattern of genomic affinity is not known formost pseudogamous or semigamousapomicts; recombination between thecentromere and the causative loci can furtherchange the expected segregation. The generalproblem is that more than one equally bestfittingmodel can be found for the samesegregation data, especially when multipleallelism and complex dosage requirements areconsidered and advanced generation selfs andbackcrosses are not available. Since thenumber of developmental inductions isconstrained by the number and interaction ofcausative loci, the difficulty in completelyunderstanding the genetics of apomixis hashamstrung the more direct developmentalapproaches. This is particularly true for therelationship among imprintable loci,endosperm function, and embryo inductionin autonomous apomicts. Endospermimprinting relationships might well becomethe biggest obstacle to the utilization ofapomixis in sexual crop species.At the level of cell and molecular biology,critical questions remain regarding thesequential and parallel nature of clocks thatgovern the relative and absolute duration ofevents in the Polygonum-type, and thus thenumber ofcontrol points and interconnectionsthat suffice to push the whole program along.Diagnostic mRNA or protein suites have notyet been recognized for the steps in thePolygonum-type and applied to recognizeparallel steps in apomictic developments. Thepromoters, reading frame sequences, andaction of genes at control points remain amystery. Thus the basis of one of the mainassumptions in the preceding developmentalinterpretations, namely that the induction oflate steps cancels intervening steps within thePolygonum-type, has yet to be justified.Natural apomicts are difficult experimentalsubjects, but they provide embryologicalbehaviors that might be even more difficult toscreen from mutagenized populations ofArabidopsis thaliana or annual crop species,especially if the appropriate mutations requiregains in function. Apomictic studies mightcontribute significantly to our futureunderstanding of sexual plant reproductionand to the successful utilization of apomixisin currently sexual crops.ReferencesAsker, S. E., and LJerling. 1992. Apomixis in Planls. , Baca Ratan,Rorida: CRC Press.BoMaglia, E. 1946. Ricerche cariolagiche ed embriologiche sui genereRudbeckio (Asleroceoe). I-V. 11 gametaflto femminile e mos

34 cales F. (,...Campbell, C. S., C. W. Greene, and Scon E.Bergquist. 19B7. Apomixis and sexuality inthree species of Amelanchier, shadbush(Rosaceae, Maloideael. Amer. 1. Bot. 74:321-2B.Coss, D. D., D. J. Peteya, and B. l Robertson.19B6. Megagametaphyte development inHordeum vulgare. 2. Later stages of wolldevelopment and morphological aspects ofmegogamelophyte cell differentialion. (an.J. Bat. 64: 2327-36.Chaudhuri, S., and J. Messing 1994. Allele·specific parentol imprinting of dzrT, aposnranscriptional regulator of zeinaccumulation. Proe. Natl. Acad. Sci. (USA)91: 4B67-7l.Ch~nnaveeraiah, M. S., and R. M. Patil. 1971.Apomixis in Blumea. Phytomorphology 21 :71-76.Coe, G. E. 1953. Cytology of reproduction in(ooperia pedunculata. Arner. 1. Bot. 40:335-43.Crone, C. F. 197B. Apomixis ond cr~singincompotibilities in some Zephyrontheoe.Ph.D. dissertation, the University of Texas otAustin, Austin, Texos.Crone, C. E, and J. G. Carman. 19B7.Mechanisms of apomixis in Elymusredisetus from eastern Australia and NewZealand. Amer. 1. Bot. 74: 477-96.Crane, C. E, H. J. Price, D. M. Stelly, D. G.Czeschin, and 1. D. McKnight. 1993.Identification of a homeologouschromosome poir by in s~u DNAhybridization to rib~omal RNA loci inmeiotic chromosomes of conon IGossypiumhirsutum Ll. Genome 36: 1015-22.Edman, G. 1931. Apomei~is und Apomixis beiAtraphaxis frutescens C. Koch. Ado Hort.Berg. 11: 13-66.Ehlenfeldt, M. Il, and R. E. Hanneman. 1988.Genetic control of endosperm balancenumber (EBN): three additive loci in 0threshold-like system. Theor. Appl. Genet.75: 825-32.Fogerlind, E1940. Die Terminologie derApomixis-Prozesse. Hereditas 26: 1-22­--.1944. Is my terminology of theopomictic phenomeno of 1940 incorrectond inappropriate? Hereditas 30: 59~96.--.1947. Macrogametophyte formationin two agamospermous Erigeran species.Ada Hort. Berg. 14: 223-47.Gledhill, D. 1967. Embryo soc formotion inAfrican Axanapus spe

aa"llkatio. 01 ApOIIIidic Medla.lsm, 35Appendix: Methods to ClearAngiosperm OvulesApomictic development can be studied withany of the tools used to study other plantanatomical problems. Particularly useful toolsinclude thick and thin sections, clearings, flowcytometry, and Feulgen microspectrophotometry.Clearings have played aprominent role in recent studies because theyallow characterization of large, reproductivelyheterogeneous samples. Sections and clearingsallow the positions and divisions of nuclei andprotoplasts to be observed directly; sequencesof nuclear divisions and movements can beinferred by observing the set of still images. Inthis respect, both are inferior tocinematography of living embryo sacs inovules suspended in silicone oil (Erdelska etal. 1971, 1979), but the latter technique has notbeen widely applied because of the usuallyinsufficient transparency of the nucellus andinteguments. Flow cytometry providesinformation on the DNA content (ploidy) ofembryonic and endospermatic nuclei and thuson the frequency of apomixis and the role ofthe sperm nuclei in apomictic endospermdevelopment. The older technique of Feulgenmicrospectrophotometry provides this sameinformation for species or developmentalstages in which there are too few nuclei forflow cytometry or for which positiveidentification of all nuclei is desired. Exceptfor cinematography, these methods requirefixation of the specimen to stop all divisionsand preserve the cells of interest in asufficiently "lifelike" condition. The type offixation limits the quality of the resulting data.The researcher's objectives (both scholarly andmicroscopic) determine the image quality thatis needed and therefore the time and labor thatare required; screening a segregating F 2population for the Panicum-type does notrequire ultrastructural preservation orfreedom from plasmolysis.Clearings usually require less effort thanembedding and sectioning, but the overallquality is limited by the amount of visualinformation that can be accrued in a singleoptical section (plane of focus) before the ovuleas a whole is rendered opaque. Objects can beseen because they differ from theirsurroundings in color or refractive index; thelatter causes light to bend at the surface of theobject. Clearing media are designed to closely,but not perfectly, match the refractive index ofcell walls or organelles; an object becomesinvisible when immersed in a medium ofequalrefractive index. Variation in refractive indexamong cellular components assures that someaspect of the specimen will always be visiblewhen the refractive index of the medium isclose to that of the bulk specimen.Furthermore, many structures, such as cellwalls, starch grains, and oxalate crystals, varyinternally in refractive index as the light pathmoves relative to their crystal axes. Thisproperty is termed birefringence, and it can bea severe nuisance in examining cleared ovules.Specimen thickness, refractive index, andinformation content (organelle or nuclearfrequency per unit thickness) determine theoptimal refractive index of a clearing medium.Single cells can be successfully examined inwater, and a movie has been made of nuclearmovements and divisions within viable ovulesofJasione montana (Campanulaceae) immersedin silicone oil (Erdelska et al. 1971). "Normal"ovules and grass ovaries require a considerablycloser matching of refractive index, i.e., greaterloss of visual information from individualplanes of focus, for successful visualization oftheir interiors. Information content reflectsorganellar preservation and thus depends onthe type and duration of fixation. Theorganellar refractive index appears to respondto hydration, and thus the optimum refractiveindex differs between hydrous and anhydrous

36 GaOO F. er...media. There may also be a statisticalcorrelation of optimum refractive index toDNA content per chromosome, with largegenomes tending to require a higher no'For example, Zephyranthes can be clearedin methyl salicylate (no = 1.537), butNothoscordum (which has biggerchromosomes) requires a higher no' asprovided by 2:1 benzyl benzoate: dibutylphthalate (no = 1.542). Possibly thiscorrelates more directly to G-C content,pitch of the double helix, or degree ofcytosine-methylation in heterochromatin,than to total DNA content; the smallergenomes simply carry lessheterochromatin.Table 3.1 presents refractive indices ofvarious published and candidate clearingmedia as measured with an Abberefractometer at ca. 20°e. Most ovuleclearingstudies have used one of twoclasses of media: modified lactophenol(Herr's 4-1/2 and BB 4-1/2, Herr 1971 and1974) and aromatic esters (Crane 1978;Young et al. 1979; Crane and Carman 1987).The latter have been combined successfullywith staining in hemalum (Stelly et aI.1984) or azure dyes (G. L. Hodnett,personal comm.) for brightfieldobservation. The Herr media do not fullydehydrate the ovule because of the 15%water in commercial lactic acid, andorganelles and cell walls therein seem toclear optimally around no 1.51, whereasthe optimum in aromatic esters is aroundno 1.53 or 1.54. A third class, nearlysaturated sugar solutions, has been usedto document callose deposition (Peel et al.1997), with lesser success in displayingnuclear and cellular locations. A fourthclass, salts in glycerol or sugar solutions,remains to be tested, but appropriaterefractive indices have been obtained forhigh-quality clearing (Table 3.1).Both of Herr's 4-1/2 media contain chloralhydrate, a federally controlled substance thatsometimes presents legal problems for institutionsthat possess it. They also contain eugenol, whichturns yellow upon exposure to light, and phenol,which likewise discolors and is a chemical contacthazard. On the other hand, excellent photographshave been obtained with Herr's media inappropriate species, and their refractive indexprobably can be adjusted upward withoutTable 3.1 Refractive index (n D) of (ommon andpotential dearing mediaSubstance20% sucrose 1.36340% sucrose 1.39960% sucrose 1.44170% sucrose (soturated) 1.465saturoted sucrose in 50% glycerol 1.46880% fructose (saturated) 1.48380% fructose in 10M Kacetate, 10 mM KOH 1.49480% fructose in saturated (ca. 8.6 M) KI, 10 mM KOH 1.51795% glycerol saturated with KI, 1mM KOH 1.506saturated CaCl 2in H0 1.4692saturated CaCI 2in glycerol (glassy solid at room temperature) 1.534saturated ZnCI 2in H021.550saturated ZnCI 2in 80% (v:v) glycerol 1.5321:1:4 (v:v:w) water, glycerol, ZnCl 21.518FeCI 36H 20 liquefied with a trace of glycerol 1.56050% polyvinylpyrrolidone (MW 10000) 1.42975% polyvinylpyrrolidone (MW 10000) 1.487saturated naphthalene in glycerol 1.472Herr's 41/2 medium 1.503Herr's BB 41/2 medium 1.510methyl salicylaJe 1.537dibutyl phthalate (DBP) 1.492benzyl benzoate (BB) 1.569BB:DBP mixes, v:v:50:50 1.53055:45 1.53360:40 1.53765:35 1.54170:30 1.54575:25 1.54980:20 1.552saturated KI + 2.5 mg/mllris base in 50% DMSO 1.475saturated KI + 2.5 mg/mllris base in 70% DMSO 1.496saturated KI + 2.5 mg/ml Iris base in 80% DMSO 1.508saturated KI + 2.5 mg/mltris base in 90% DMSO 1.519saturated KI + 2.5 mg/mltris base in DMSO 1.525n D

Oo,silk.tlo••f Apamiclic MotH.isll' 37affecting their general properties by partiallysubstituting isoeugenol (n D= 1.574; Windholzet al. 1983) for eugenol (n D = 1.541; Windholzet al. 1983) in the recipes given below.Aromatic oils require complete or nearlycomplete dehydration of the ovule, whichtypically lengthens the clearing procedure andleads to shrinkage and possibly distortion. Theovules normally become hard and brittle,which is good on the microscope slide but noton the dissecting stage. While the oils that havebeen used are only mildly toxic, this is notuniformly true, and some oils (polynucleararomatic hydrocarbons) must be rejected ascomponents ofclearing media because of theircarcinogenicity.Both the lactophenol and aromatic-oil clearingagents offer limited opportunities forcytochemistry. Most of the establishedcytochemical procedures are based onreactions in water and may behave abnormallyin less polar solvents. Opaque reactionproducts can obscure unstained regionsbehind them, and accessibility to reactants isalways more difficult in intact ovules than insections thereof. Reaction products can bemobilized or lost upon dehydration andinfiltration with clearing agents. Nevertheless,there is interest in distinguishing various typesof cell walls (especially callose) and cellularinclusions (e.g., starch grains) in clearedovules, and a medium that can be used to thisend is introduced here for furtherinvestigation.Several inorganic salts, including those listedin Table 3.1, dissolve to a considerable degreein OMSO, glycerol, concentrated sugarsolutions, or polyethylene glycol. None ofthem has proven completely satisfactory as analternative to the aromatic oils or 4-1/2 media,but they might permit additionalcytochemistry to be performed on wholemounts. Some media with KI equal the 4-1/2media in refractive index (Table 3.1), but donot clear the tissue well. This is attributed toiodination of the macromolecules in the cell,which raises its refractive index. The problemwith CaCl 2is the high viscosity of its solutionsin polyols. Its hygroscopicity also causes therefractive index of the preparation to vary withthe relative humidity in the room. Ferric andzinc chlorides are extremely corrosive tometals and human flesh, and they destroynuclei relatively rapidly. Potassiumthiocyanate gives a high refractive index insolution, but is chaotropic (destroys nucleiagain) and hazardous to the mental health ofthe slide user (Handbook of Chemistry andPhysics). Finally, the problem with DMSO forembryological clearings is its denaturantaction on DNA and proteins, resulting in poorquality of nuclei or no nuclei at all. This actionis the basis of a very effective recipe (providedbelow) for destructively clearing leaves forvascular and epidermal study.The following protocols are mostly based onfixation in organic solvents that kill the cellquickly and preserve the chromosomes well.The choices include FAA (37% aqueousformaldehyde, acetic acid, and 50% or 70%ethanol, 1:1:18 by volume); FPA (the samesolution with propionic acid substituted foracetic acid), 3:1 etbanol:acetic acid;, Carnoy'ssolution (6:3:1 ethanol:chloroform:acetic acid,by volume); and 2:1 acetone:acetic acid. Crafand Allen-Bouin-type fixatives can also beused, but require longer dehydrations.Formaldehyde and glutaraldehyde, alone orin combination, also require longdehydrationsand typically leave the ovules yellow. Theyalso contribute to autofluorescence in thedetection of callose with aniline blue. The lesspolar fixatives generally cause severeplasmolysis of embryo sacs with two or morenuclei, but they also extract chlorophyll andcarotenoid pigments more completely fromovary walls. Acetone/acetic acid isparticularly effective at this and often leavestissues snow-white.

38 (IIarIe,F.e-.The following recipes include examples inwhich the fixative and clearing agent areimmiscible, requiring an intermediarydehydrating agent, and examples in whichthe fixative and clearing agent mix freely. Inboth cases, one must pay close attention tothe solubility of air in each liquid in theprocec;lure. Air appears to be much lesssoluble in aromatic oils and concentrated saltsolutions than in ethanol or acetone. Overrapidinfiltration typically results inexsolution of bubbles or films of air that donot disappear with time. Affected ovulesgenerally float and remain opaque during theinfiltration series. Ifsuch bubbles appear, onemust backtrack, possibly all the way to thefixative, and start again. Adding extra stepsat closer gradations in concentrationeliminates bubbling more effectively thandoes lengthening individual steps.The dehydration and infiltration steps can beadvantageously recombined in many of thefollowing recipes. When the clearing agentand fixative are miscible, a general principleis to progress through a graded series ofmixtures of the two, including enough stepsto avoid exsolution, as discussed above. Agood starting point is 10 gradations fromfixative to the microscope slide. Users shouldfeel free to experiment, especially whenadapting a recipe to novel species or stages.The requirements for dissection must also beconsidered; 70% ethanol is much less noxiousand corrosive than FAA, 2:1 acetone: aceticacid, or salt solutions in dimethyl sulfoxide(DMSO). In many cases, one does not havetime to dissect out ovules before fixation ordoes not want to lose small, nearly invisibleovules when solutions are changed.Microscopy can make or break any clearingprocedure. Confocal laser scanningmicroscopy has been used impressively onthin orchid ovules (Fredrikson 1990), but itsusefulness on thicker specimens remainsuntested, and it is not widely accessiblebecause of its expense. Nomarski interferencecontrasthas been applied successfully toapomictic amaryllids (Crane 1978) andpanicoid grasses (Young et al. 1979). Both ofthese methods create a very shallow depth offocus that helps to eliminate information fromoutside the focal plane. Hoffman modulationcontrastis a less expensive alternative tointerference-eontrast. Phase-contrast provides"optical staining" that can help distinguishnuclei from similarly sized vacuoles, but itrequires a closer match between the sampleand the medium in refractive index. Stainclearedspecimens (and many unstainedspecimens) can be examined satisfactorilyunder brightfield optics once the medium isproperly matched to the specimen in refractiveindex. Brightfield is also insensitive tobirefringent cell walls or crystals in thespecimen, which greatly degrade aninterference-contrast or phase-contrast image.Specimen orientation is frequently critical toeasy and correct interpretation of clearedembryo sacs. Ovaries of pooid grassestypically are laterally compressed and presenta nearly end-on view of the embryo sac whenthey are allowed to lie flat. Standing them onedge in a sawtooth cut in index card oraluminum foil yields the desired sagittaloptical section. Stelly et al. (1984) observed thatsqua~hing cleared ovules was generallycounterproductive; the image got worseinstead of better as more visual informationwas crammed into nearly the same plane.Thus clearings are often examined with "Raj"slides (Herr 1974), with which the cover slipis supported by underlying adjacent coverslips that are conveniently held in place with"Superglue" (G. Hodnett, personal comm.).

(Io"il'ollo. of Apomi

40 CWIes F.

Oa"imatlo. of -,."';

42 C"Ie

(lanilkoll.. of Apomktlt MeUotb... 434. The method works also for meristems,floral primordia, and other generalsubjects where one wants to see only thecell walls. Oxalate crystals and opalinesilica survive the procedure intact. If theepidermis separates from the mesophyll,use less heat next time during fixation.AcknowledgmentsThe author would like to thank Drs. David M.Stelly and H. J. Price for their help during thewriting of this chapter. This research wassupported by Texas Advanced Technology andResearch Program Grant 999902090.ReferencesBer~n, G. P, and 1 PMiksche. 1976. Botonical Mierotechnique ondCytochemistry. Ames, Iowa: The lowo State Unive~ify Press.Crone, C. F. 1978. Apomixis and crossing incompatibilities in someZephyrantheoe. Ph.D. dissertation, Ihe Unive~ify of Texas ot Austin,Austin, Texos.Crane, C. F., and J. G. Carman. 1987. Mechanisms of apamixis in Bymusredisetus fram eastern Australia and New Zealand. Americon. Journolof Botony. 74: 477-96.&delska, 0., H. K. Galle, H. H. Heunert, and K. Philipp. 1971.Somenonloge und fruhe Endospermenlwicklung von Josione mantonoIComponumceae). Insmuf fUr den Wissenschahlichen Film.Erde~ka, 0., H. H. Heunert, T. Hard, J. Kaeding, and H. WllImonn. 1979.Be!ruchtung und fruhe Enlwidclung von Embryo und Endosperm beimSchneeglikkchen.lnstiluf fiir den Wissensmahlichen Film, no. 1465.Fredrikson, M. 1990. Embryological Sludy of Herminium monorchis(Orchidaceae) using conlocal scanning loser mi

Chapter 4Ultrastructural Analysis of ApomicticDevelopmentTAMARA N. NALIMOVA AND JEAN-PHILIPPE VIELLE-CALZADAIntroductionSince the first applications of electronmicroscopy to the observation of plant tissues,numerous studies have focused on theultrastructural characteristics of sexualreproduction in flowering plants (Jensen1965; Jensen and Fischer 1968). In manyspecies, a wealth of information has beenpresented on the morphology of the femalegametophyte (megagametophyte or embryosac) before and after pollen tube arrival(Russell 1985; Mogensen 1988; Russell 1992;Huang and Russell 1992). Valuablecontributions to the understanding of doublefertilization have also emerged from studiesthat combine light and electron microscopywith technological advances on the in vitroisolation and fusion of gametes (Kranz et al.1991; Faure et al. 1994). Nevertheless, thefundamental mysteries involving cell to cellsignals and interactions duringme,gagametogenesis and fertilization insexual and apomictic species are stillunsolved. These enigmas include the natureof the movement of sperm cells within theegg apparatus, the specific recognition ofmale and female gametes, the molecularnature of plasma membrane adherence, andthe association, fusion, and activation ofparental nuclei (Gerassimova-Navashina1957; Russell 1992).Despite the increasing scientific interest inapomixis (Asker and Jerling 1992; Vie lle­Calzada et al. 1996a), few studies havecharacterized the fine structure ofmegagametophytes involved in apomicticreproduction. Cytological studies based onlight microscopy have identified thefundamental differences existing between thegeneral organization of sexual and apomicticfemale gametophyes in different genera(Bashaw and Holt 1958; Voigt and Bashaw1972; Philipson 1978; Campbell et al. 1987;Crane and Carman 1987; Burson et al. 1990).The fine structure of nucellar andintegumentary embryony has beeninvestigated in some detail for a limitednumber of species (Naumova and Willemse1982; Wilms et al. 1983; Naumova 1993), andultrastructural studies of diplospory have beeninitiated in only a few apomicitic species ofPoa (Naumova et al. 1999). In addition, a fewultrastructural characterizations ofmegagametophyte development andfertilization have been conducted in someaposporous members of the Poaceae (Chapmanand Busri 1994; Naumova and Willemse 1995;Vielle et al. 1995),The occurrence of apomixis in species thathave conserved the ability to reproducesexually provides unique opportunities forcomparative ultrastructural studies. Theformation of sexual and apomicticmegagametophytes can simultaneously occurin the same genotype or in different genotypesof the same population. Ultrastructuralcomparisons of megagametogenesis and earlyfertilization events in sexual and apomictic

Ultra,t..""aI AIlaly,l, af Apamktk O...Ia,...., 45ovules of the same species can be used toobtain mechanistic information on specificcellular differences that distinguish thesedevelopmental processes. Although thenumber of species in which ultrastructuralstudies have been conducted remains limited,recent electron microscopy studies in someaposporous grasses have provided newinformation on megasporogenesis,aposporous initiation, the organization of thedifferentiated megagametophyte, and theautonomous division of unreduced egg cells.Nucellar and IntegumentaryEmbryonyDuring integumentary or nucellar embryony(also called adventive embryony), asexualembryos are formed from inner integumentaryor nucellar cells that aredifferentiating in tissues external to themeiotically derived megagametophyte.Nucellar or integumentary cells that give riseto adventive embryos are called embryocytes.A comprehensive review of adventitiveembryony provides significant ultrastructuralinformation about this process (Naumova1993) by summarizing light and electronmicroscopy observations and reviewingembryological information on representativesof more than 250 species of flowering plants.The first morphological evidence ofembryocyte differentiation is usually observedafter the initiation of megagametogenesis. Asthe nucellar cells divide, they invade thecentral cell of the sexually functional embryosac. The formation of viable seed from nucellarembryos usually requires fertilization of thepolar nuclei and subsequent development ofthe endosperm; however, autonomousendosperm formation sporadically occurs insome species. Nucellar embryos are usuallyinitiated independently of pollination.Polyembryony, the formation of severalembryos in a single ovule, is characteristic ofadventive embryony. Several ultrastructuralinvestigations have improved ourunderstanding of embryocyte differentiationand early adventive embryogenesis (Naumova1978; Naumova and Wi1Iemse 1982; Wilms etal. 1983).Embryocytes are generally characterized by adense cytoplasm, an irregularly shapednucleus, and a larger volume than most of thenucellar cells present in the ovule. Theabundance of polysomes, free ribosomes,mitochondria, and plastids suggests highphYSiological activity in embryocytes. Theircell walls are significantly thickened andlacking plasmodesmata. In the genusSarcococca, the number of plasmodesmatagradually decreases during embryocytedifferentiation (Naumova and Willemse 1982).Using light microscopy, Kol tunow et al (1995)found that the nucellar initials form a thick cellwall in Citrus sinensis. They also found alocalized degeneration of the nucellar tissueat the chalazaI pole, a region where nucellaradventive initials are confined. In Sarcoccoca,the general organization of the embryocytecytoplasm is modified during consecutivestages of mitosis, undergoing events similarto those characterizing the dedifferentiation ofpre-meiotic megaspore mother cells (MMC) insexual species (Dickinson and Potter 1978). InEuonymus macroptera, integumentary embryoformation occurs in tenuinucellate ovules.Asexual embryos originate from two or threecell layers enveloping the micropylar regionof the sexually functional megagametophyte(Figure 4.1a,b,c; Naumova 1990). Integumentaryembryos coexisting in a single ovuleusually develop asynchronously (Figure4.1d,e). Plasmodesmata are not present in theembryocyte cell wall or in the integumentarywall that is in direct contact with the centralcell (Figure4.1f,g). The transversal cell wall thatseparates sister integumentary cells varies inthickness (Figure 4.1h,j).

46 T.""". N. Naomo•• ..dJ...·P~iipp. Y~l..."'''.d.The ultrastructure of both cells contained in atwo-cellular adventive embryo is extremelysimilar: their external cell wall remains thickand lacks plasmodesmata. Young adventiveembryos also lack a distinct polarity, and noregular cell divisions can be discerned; asuspensor is frequently missing. Nucellarembryos do not arise synchronously, and theirearly development can be severely delayed.Whereas some embryos can appear to becomposed of only a few cells, others in thesame ovule may already be undergoingorganogenesis. Adventive embryos are able topenetrate into the adjacent central cell of asexually derived megagametophyte.DiplosporyIn diplosporous species, the megagametophyteis formed from an aberrant meiotic cyclethat prevents reduction and recombination. Anul trastructuraIcharac-terization of d iplosporyhas only been initiated in two apomicts: Poapa!lIslris and Poa nemoratis (Naumova et al.1999). In these two species, diplospory ischaracterized by a complete omission ofmeio~is; the unreduced megasporocytedevelops into an embryo sac after gradualvacuolation and three consecutive mitoticdivisions. In the young developing ovule, thearchesporial cell progressively differentiatesinto a diplosporous precursor or diplosporousembryosac megaspore mother cell (DMC).Subsequently, after three mitotic divisions, theDMC develops into an 8-nucleate embryo sacof the Antennaria-type.In contrast to nucellar cells present within thedeveloping ovule, the archesporial cell of bothspecies is characterized by a large nucleus withdecreased chromatin contraction in regionshaving close contact with the surface of thenucleolus. Numerous poorly differentiatedmitochondria and plastids are uniformlydistributed in the enlMged cell. Interestingly,isolilted enclil\"es of cytoplasm containingmultiple membranes are often observed,whereas the endoplasmic reticulum (ER) ispoorly developed. Compared to the DMC,vacuoles are also abundant but small andlocated mainly at the micropylar and chalazalpoles of the cell.Significant cellular elongation marks thetransition from archesporial cell to DMC.Whereas the archespore is 160 J.lm on average,the DMC is about 350-380 J.lm long, with littleor no changes in width. This increase in size isassociated with a reduction in chromatincondensation and nuclear volume. In theDMC, the nucleus is positioned at themicropylar end of the cell and becomeselongated and irregular in shape. There is asubstantial increase in the population ofribosomes, polysomes, dictyosomes. At thisstage, the vacuoles fuse into two large oneslocated at the micropylar and chalazal poles,and there is an obvious increase in thethickness of the cell wall. The transition fromDMC to a noncellularized embryo sac is alsocharacterized by a gradual increase in celllength to about 500 J.lm. The nucleus becomesirregular and lobbed, with numerousprotuberances in the nuclear envelope. Severalnucleoli differing in size and shape can besimultaneously observed. There is also amarked increase in the population ofmitochondria and plastids that often formclusters heterogeneously distributedthroughout the cytoplasm. Contacts betweenthe external membrane of the nuclear envelopeand granular ER cisternae are often observed.Additionally, series of concentric membranousstructures were also observed in the cytoplasm.The cell wall became slightly thick during thechange of the DMC into a one nucleate embryosac. Nucellar cells adjacent to the developingdiplosporous embryo sac showed signs ofdegeneration, with an electron-opaquecytoplasm and highly condensed nuclei.

Ultr••IA,""aI A.aI,.is .1 ApamKtk D...lop..... 47AposporyApospory is a form of apomixis in whichsporophytic cells in the ovule give rise tounreduced female gametophytes (Gustafsson1947). The autonomous division of aposporousegg cells generates viable embryos withoutfertilization; however, the majority ofaposporous species are pseudogamous andrequire fertilization of the polar nuclei forendosperm development (Nogler 1984).An ultrastructural characterization ofaposporous megagametophytes has beenconducted in PaniCllm and Penniselum. In thesetwo grass genera, the mechanism ofaposporous female gametophyte formation isvery similar. Early during megasporogenesis,the orientation of the ovule changes within theovary, and the integuments progressivelyenclose the nucellus. The ovule becomesanatropous as it rotates toward the base of thepistil, leaving the micropyle facing away fromthe style. In PaniCllm maximllm and PenlJiselumciliare (syn = Cel1c}zYlls ciliaris L.), sexualmegagametophyte development ismonosporic; a single meiotically derivedmegaspore gives rise to the embryo sac. Thefunctional megaspore enlarges, and its nucleusdivides mitotically three times to form themegagametophyte. Buth species develop asexual megagametophyte of the PolygolJlImtype: two synergids, the egg cell, a binucleatecen tra I ce II whose n uc lei fuse prior tofertilization, and three antipodals thatproliferate to give a cluster of cells at thechalazal pole. After fertilizatio:1 of both the eggcell and the central cell, the ovuIe develops intoa seed.Whereas sexual ovules only develop ameiotically derived megagametophyte, inaposporous plants one or several nucellar cellsacquire a reproductive fate and are able topursue growth and differentiation whilemegasporogenesis proceeds. The spatial andtemporal patterns of megaspore mother cell(MMC) differentiation and meiosis are thesame during sexual and aposporousdevelopment. In obligately apomicticgenotypes, all meiotically derived megasporesdie and reduced mt'gagametophytes are notformed. Active nucellar cells undergo twomitotic division~. After cellularization, 4­nucleated female gametophytes differentiateinto an egg apparatus and a central cellcontaining usually one or two polar nuclei. Inboth species, the egg cell is able to divideparthenogenetically and give rise to viableembryos. Fertilization of the polar nucleus (ornuclei) is necessary for endosperm formationand seed set. In facultative plants, bothdevelopmental processes are viable; sexual andaposporous female gametophytes can coexistwithin the same ovule, or in different ovulesof the same inflorescence.Differentiation of Aposporous InitialsAposporous development requires thedifferentiation of nucellar cells into unreducedorganized female gametophytes. In mostaposporous species, the differentiation of theMMC appears to take place in a subepidermalla yer of the n ucell us, and followsmorphological characteristics that have beenpreviously described in many sexuallyreproducing angiosperms (Huang and Russell1992). In Panicum maxim11m, the MMC ischaracterized by a centrally located nucleuscontaining a conspicuous nucleolus and densecytoplasm. In contrast, the nucellar cellssurrounding the MMC have thinner cell wallsand numerous plasmodesmata that shareultrastructural similarities with youngmeristematic cells. In aposporous genotypes,meiotically derived dyads and linear tetradsoften degenerate during megasporogenesis.Aposporous initials usually differentiateadjacent to degenerating megaspores or onenucleatedsexual embryo sacs. Chalazallylocatednucellar cells assume vacuolizedspherical shapes and increase in volume

(Figure 4.2a,b,c). Their cytoplasm containsnumerous plastids and mitochondria, andplasmodemata are scarce or appear to be filledwith a cellulose-like matrix. The degenerationof meiotically derived megaspores and earlystages of aposporous initiation has recentlybeen investigated in Brachiaria brizantha. In thisgrass, aposporous initials appear to containdedifferentiated organelles reminiscent oforganellar populations found in pre-meioticMMCs (A.c. Guerra de Araujo and V. Carneiro,personal communication).Aposporous MegagametogenesisSeveral aposporous embryo sacs can developin a single ovule (Figure 4.2d). Aposporousfemale gametophytes show a differentorientation with respect to the micropylarchalazalaxis, and usually differentiateheterochronically with respect to each other.The transition from aposporous initial to onenucleatedembryo sac is characterized by anincrease in cell size (Figure 4.3a). The first signsof cellular polarity are the consequence ofvesicular fusion; two large vacuoles are formedat opposite sides of a centrally located nucleus.Cell wall thickness also increases, and noplasmodesmata can be discerned (Figure4.2d,e). In Pennisetum ciliare, the nucleusmigrates to the periphery of the cell beforedividing (Figure 4.3b). The first mitotic divisionis perpendicular to the long axis of the cell andusually close to the micropylar region of theone-nucleated embryo sac (Figure 4.3c). Thesecond mitotic division is synchronous in bothsister nuclei, giving rise to a four-nucleatedtype of embryo sac that lacks antipodals at thechalazal pole (Figure 4.3c,d).The Cellularized AposporousMegagametophyteThe second mitotic division is followed by thecellularization of individual nuclei. Little isknown about the processes that regulatecelll;llarization and differentiation ofaposporous megagametophytes.ln Pennisetllnlciliare, as in most reported aposporous species,female gametophytes usually contain twosynergids, an egg cell, and a single polarnucleus in the central cell; however, a variablenumber of female gametophytes (up to 20%in certain genotypes) may be composed of asingle synergid, an egg cell, and two polarnuclei. On rare occasions, embryo sacscontaining three polar nuclei and no synergidshave been observed. This variable organizationappears to be associated with the localizationof nuclei prior to cellularization, and suggeststhat positional information plays a role ingametophytic cell specification.In all aposporous grasses ultrastructurallyexamined to date, the egg apparatusdifferentiates with morphologicalcharacteristics similar to those found insexually functional synergids and egg cells.These cells are attached at the micropylar apex,but the triangular organization of the eggapparatus is not necessarily conserved, as theunreduced egg cell may appear in a morelateral position with respect to one of thesynergids. Except for the presence of thefiliform apparatus, few differences are foundin the ultrastructural constitution of thesynergids and the egg cell in unpollinatedpistils of Panicum maximum. The three cellsappear vacuolated, with a centrally locatednucleus, and organelles preferentially locatedin the n:icropylar pole (Figure 4.2f,g,h).Naumov'a and Willemse (1995) characterize

Ultra.t..ct.raI baty.;. of Apamldk D...lop....' 49Figure 4.1 Integumentary embryony in Euonymus macroprera.(a) JunctiDn between meristematic integumentary tissue and endosperm, x5.000; (b) and (e) Micropylar region of ovules containingmeristematic integumentary tissue embryocytes and endosperm (bl, x570, and integumentary embryo (c), x850; (dl Seed with threeintegumentary embryos at different developmental stages, x 130; (e) Initial cells of integumentary embryo at micropylar region, x850; (f) Integumentary cell adjacent to the endosperm, x8.500; (g) Cell wall junction between an integumentary cell and endosperm,x42.500; (h) to (j) Cell walls tangenlialto the meristematic integumentary cells; (hl, x34.000; (il, x 10.200; li)' x42.500.Abbreviations on following page.

50 Tomaro N. Naurnovo ond Jeon-Philippe V"telle·(olzadofigure 4.1 (cont'd)AbbreviationsAE - aposporous embryo sac; CC -central cell; E-egg cell; EC -initial cell of nucellar or integumentary embryo =embryocyte;ES -embryo sac; 1- inner integument; IA - initial cell of oposporous embryo sac; IE - integumentary embryo; M- micropyle;MC - meristematic cell; MIT - meristematic inner integument tissue; EN -endosperm; PN - polar nuclei; 5-synergid;T-tetrade of megaspores; Z. zygote; ZE . zygotic embryo; SE -synergid embryo; FA - filiform apparatus; CW - cell wall;C- cuticle; M- mitochondrio; N- nucleus; P. plastid; PD -plasmodesmata; V- vacuole.

Ultrastru

Figure 4.2 (cont'd)

Uhrastm'.raJ A.alysis al Apomi

54 Tamara N. Nauma.a and J...-Phaipp. VIelle-CabadaFigure 4.4 Organization of the mature aposporous egg apparatus in Pennisetum ciliare.(0) The egg celllhree hours aher pollination. The chalalal end is completely covered by a cell wall (CW)" vacuole (V), egg cell (EO;(b) Detail of the cell wall (CW) separating the chalalal region of the egg cell (Ee) from the central cell (CO cytoplasm; M,mitochondria; (c) Micropylar region of the apasparaus egg apparatus, filiform apparatus (FA), egg cell (Ee); (d) Numerous Galgi(arrowheads) are present in fhe apical pocket, a region located between the central cell (CO wall and the egg apparatus, egg cell (EO,plastid (P), vacuole (V), and cell wall (CW).

UIITas'",,'ural Analysis of Apomilli< Develop....l 55Figure 4.5. Apogamety in Trillium camschatcense.(a) A2-celled embryo derived from a synergid adjacent to a zygotic embryo, x850; (b) Embryo sac wilh 3-celled synergid embryoand a zygotic embryo, x850; (c) Fine structure of a synergid, x6.000; (d) Cell wall separoting the egg celilrom the central cell, x72.000; (e) cell walls between synergid and central cell 01 the mature embryo sac, x68.000; (f) zygotic embryo, terminal port, x1.600; (g) outer cell wall of zygotic embryo, x 24.000; (h) 2-cellular synergid embryo, x 1.500; (i) cell wall and cytoplasm of thesynergid embryo, x 12.00.

56 Tamara H. Haumaya ...d J...-Phmpp. V1e1.·CalzadaFigure 4.5 (cont'd)

Ullra.'nd"aI Analy.i. of Apolllidic D...lop....' 57aposporous embryo sacs in the progeny of aPel1l1isetl/ni glal/cum (sexual) x Pennisetllmsqllaml/latl/m (obligate apomict) interspecificcross. In facultative apomictic genotypes, theycompared sexual and aposporous embryo sacsin the same ovule. Particular attention wasgiven to the distribution of internal cell wallingrowths (transfer walls) in the central cell.Many wall projections were observed in themicropylar region, but few ingrowths werefound in the chalazal region of both sexual andapomictic megagametophytes. Plasmodesmatawere present in the cell wall that separatesthe central cell from the antipodals, butappeared to be absent at the chalazal pole ofaposporous embryo sacs.Parthenogenesis and FertilizationIn Pennisetl/m ci/iare, most genotypes reproduceby obligate apospory. Breeding efforts arebased on the identification of rare genotypesthat have completely lost the ability todifferentiate aposporous megagametophytesand only form sexually functional, reduced(haploid) embryo sacs of the Polygonum-type.The comparison of the egg apparatus in sexualand aposporous megagametophytes of P. ci/iareoffers an opportunity to analyze the cellulardynamics of fertilization in apomictic plants.Such a comparison is particularly valuable ifconducted where independent genotypes ofapomictic and sexual germplasm are availablewithin a population segregating for methodof reproduction.In P. ci/iare, fertilization occurs 3-4 hours afterpollination. The examination of thE' eggapparatus of buffelgrass at several timeintervals after pollination has provided someinformation on the structural and functionalfeatures that distinguish sexual and apomicticdevelopment (Vielle et a!. 1995). Compared tothe synergids of the sexually derived eggapparatus, the degenerative process in one orboth aposporous synergids appears to beaccelerated. Prior to pollination, the cytoplasmof both cells contains small vacuoles, and feworganelles can be identified. In general, thecells are similar in appearance to the sexualsynergids during the first two hours followingpollination, but before pollen tube arrival.During the first two hours followingpollination, the cytoplasm of one of thesynergids becomes electron-dense, the plasmamembrane appears disrupted, organellescannot be identified, and the nucleus isirregularly shaped and pressed to the plasmamembrane. Increased amounts of vesiculartraffic are observed. Later, the cytoplasm of thesecond synergid also shows signs ofdegeneration. Two to three hours afterpollination, the degenerated synergid hasentirely collapsed and its remnants appear asan electron-dense fringe closely associatedwith the egg cell. The time of initiation of thedegenerative process is variable in both sexualand aposporous embryo sacs, and in somecases, highly degenerated synergids arepresent in unpollinated ovules during or justafter embryo-sac cellularization.Significant differences are also observedbetween the aposporous and the sexual eggcell. After cellularization but before pollination,the aposporous egg cell is characterized by aconspicuous centrally located nucleus thatcontains a single nucleolus (Figure 4.3e). Thenucleus of the sexually derived egg cell isgener-ally centrally located, but has a smallernucleolus. The chalazal vacuole present in thesexually derived egg cell is replaced by severalsmaller vacuoles that restrict the cytoplasm toa region located around the nucleus. In contrastto the sexual egg cell, the cytoplasm containsabundant mitochondria and polysomes, butfew Golgi bodies and endoplasmic reticulum(ER) can be observed. At the micropylar region,both sexual and aposporous eggs sharecommon cell walls with both synergids, butthese walls disappear in the middle portions

58 Tamara N. N....... Old JOGO-PUIp,. VIeIle-CAbadaof the cell. The plasma membranes of the eggand synergids are in direct contact at thechalazal pole.Three to four hours after pollination, strikingchanges are detected in the ultrastructure ofthe aposporous egg cell. A cell wall without amiddle lamella has covered the chalazal regionof the egg plasma membrane, separating thecell from the degenerated cytoplasm of acollapsed synergid (Figure 4.4a,b). No cell wallcovers the chalazaI end of the sexual egg cell,even after pollen tube discharge. Theaposporous egg cell cytoplasm appearsvacuolated and contains numerousundifferentiated plastids preferentiallyorganized in clusters at the periphery of thenuclear membrane. A thin layer of central cellcytoplasm is associated with the externalsurface of this de novo formed cell wall (Figure4.4c,d). The central cell cytoplasm contains alarge number of Colgi cisternae andmitochondria, particularly in the so-calledapical pocket, a region of the central cell formedby the proximity of the egg apparatus to thecentral cell wall (Figure 4.4d). The unreducedpolar nucleus usually contains more than onenucleolus. In some rare occasions, amulticellular embryo can be present beforepollen tube arrival into the micropyle.ApogametyApogamety designates the formation ofembryos from a cell of the megagametophyteother than the egg. Even if this phenomenonhas rarely been reported in sexual andapomictic species (Asker and Jerling 1992), itimplies the autonomous activation ofreproductive cells and can be considered anonrecurrent form of apomixis.The fine structure of synergids undergoingautonomous activation has been described inTrillium camschatcense Ker. Caw!., a species thatis endemic to the extreme eastern region of theformer USSR. Synergid embryos develop inmore than 80% of the embryo sacs investigatedin this species (Naumova 1978, 1990). Lightmicroscopy studies have shown that the eggcell and the synergids undergo limiteddifferentiation. They are characterized by acentrally located nucleus and no centralvacuole. The three cells appear similar in sizeand shape, but the presence of a conspicuousfiliform apparatus is characteristic of thesynergids. Differences between theircytoplasmic constitution can only be discernedat the ultrastructural level. In contrast to theegg cell, synergids are rich in endoplasmicreticulum and Golgi cisterneae. Plastids,mitochondria, and polysomes are abundantin the egg cell. Before pollen tube arrival intoone of the synergids, the three cells have anincomplete cell wall at their chalazal pole(Figure 4.5a,b).Following fertilization of the egg and centralcell, drastic ultrastructural changes occur inthe persistent synergid. An increase in thenuclear and nucleolar size is followed by acomplete reorganization of the organellarpopulation, which becomes similar to the onepresent in the egg cell before sperm celldelivery. Mitochondria with numerous cristaeand plastids dense in stroma are abundant.Whereas the sexually derived embryo isalready surrounded by a thick cell wall devoidof plasmodesmata, the cell wall of the 2­cellular embryo derived from autonomoussynergid activation thickens (Figure 4.5c,d)and becomes progressively isolated throughthe loss of all plasmodesmata. In summary, theultrastructural transformations taking place inembryogenic synergids of T. camschatcense arecomparable to the changes occurring in the eggcell during the gametophytic to sporophytictransition of sexual flowering plants.....

Ultras'",',," AoaIy.1s of Apomk'k O•••lop....' 59DiscussionComparisons between sexual and apomicticmegagametophytes represent an ideal systemto start dissecting the cellular and moleculardifferences that distinguish sexuality fromapomixis. Sexual reproduction is strictlydependent on the production and fusion ofhaploid male and female gametes. In contrast,apomictic reproduction is dependent on theformation of unreduced female gametophytes,the autonomous activation of the egg cell, andthe eventual fertilization of the polar nucleus(or nuclei). Regardless of the reproductivemethod, the megagametophyte forms entirelywithin the nucellar tissue and is dependent onthe sporophyte for its development andfunction. UI trastructural analysis of apomicticdevelopment has provided new, but limited,information on the developmentalparticularities of adventive embryony andgametophytic apomixis. So far, results havemainly described specific aspects ofdifferentiated embryocytes and aposporousinitials, and the fine structure of cellularizedmegagametophytic cells before and afterparthenogenetic activation.The similarity of ultrastructural characteristicsshared by adventive embryocytes andaposporous initials deserves special attention.The presence of a thick cell wall appears to beprevalent around embryocytes andaposporous initials. An extensive survey of thegametophytic-sporophytic junction in landplants reveals that the two generations of thesexual angiosperm life cycle are almostinvariably separated by thickened cell wallslacking plasmodesmata (Ligrone et aJ. 1993;Bell 1995). The presence of a conspicuousboundary surrounding the megagametophyteof some mosses and ferns of the earlyDevonian (Remy et aJ. 1993) suggests thatgametophytic cell isolation may have 3 crucialfunction of fundamental importance for theevolution of the angiosperms. Some alsosuggest that a combination of hyd ratedpolysaccharides and callose in the cell wall ofmeiotically derived megaspores may act as amolecular filter that impedes the transport oflow molecular weight peptides and/or nucleicacids expressed in nucellar cells, since theirpresence may represent a threat to the initiationof gametophytic development (Knox andHeslop-Harrison 1970). In sexual plants, thepresence of callose surrounding degeneratingmegaspores (but not viable ones) suggests thatcallose may be suppressing the nonfunctionalmegaspores by isolation. In lower plants(mosses and ferns), selective permeability ofthick wall boundaries allows only the transportof minerals and simple sugars (Bell 1988, 1992);plasmodesmata appear to be completelyabsent. Callose deposition was detected inadventive embryocytes in young Citrus seeds(Wakana and Uemoto 1987). Carman et al.(1991) compared the distribution of callose insexual and apomictic megaspores of Elymusrecticetus and found that, contrary to normalmeiosis, the cell wall of diplosporous MMCsinvariably lacks callose. These results havebeen confirmed in several diplosporous grassgenera (Leblanc et at. 1995; Peel et aJ. 1997; seealso Leblanc and Mazzucato, Chap. 9). Thedrastic cytoplasmic transformations occurringin young embryocytes and aposporous initialsmay depend on a developmental program thatis only initiated in the absence of informativemolecules originating in the nucellus. A recenthypothesis postulated by Carman (1997)suggests that the role of isolation duringmegasporogenesis may depend on expressionof duplicate genes, especially in polyploids inwhich the time and duration of specificcytological events could be the cause ofanomalies during megasporogenesis. Needlessto say, the role of cell wall thickening andcallose deposition during megasporo-genesis,diplospory, aposporous initiation, or nucellarembryony remains an unsolved problem.

60 T.mar. N. N••may•••dJ•••·Phmppe y"He·(.llOd.The examination of the apomicticmegagametophyte in some aposporousgrasses illustrates many ultrastructuralcharacteristics that are conserved in theangiosperms. Ultrastructural comparisons tosexually functional megagametophytes do notalways distinguish facultative apomicticgenotypes (with sexual and aposporousembryo sacs in the same ovule orinflorescence) from obligate apomictic ones(aposporous embryo sacs exclusively present).In addition, the distinction between theultrastructural observations of unpollinated orpollinated pistils has rarely been reported. InPalliCllm maximllm (Naumova and Willemse1995) and Pellniseillm interspecific progeny(Chapman and Busri 1994) the generalorganization of the egg apparatus appears tobe very similar to that of corresponding sexualmegagametophytes.The ultrastructural comparison of the sexualand aposporous egg apparatus in Pell/lise/llmciliare raises questions regarding the role ofsynergids and the nature of the signal thattriggers egg cell parthenogenesis. The severesigns ofsynergid degeneration present in someunpollinated pistils indicate that a certainflexibility characterizes the timing andduration of megagametophytic developmentin sexual and apomictic ovules, and suggestthat the fate of the degenerating synergidmight be independent of pollen germinationor pollen tube growth. The spatial associationbetween the egg cell and the degeneratingsynergid is particularly close and suggests apossible involvement of the degenerativeevents in the activation or repression of theegg cell. In the sexually derived egg cell, thelow frequency of endoplasmic reticulum,Golgi bodies, and polysomes, together withthe small number of cristae present in themitochondrial population, suggests that thecell is in a rather quiescent physiological state,presumably prolonged until pollen tubearrival into the micropylar vicinity. A quiescentegg cell prior to fertilization has also beendescribed in several sexual species (Diboll1968; Schultz and Jensen 1968). In contrast, thelarge amount of ribosomes and polysomes andlarge number of cristae in mitochondriasuggest that the mature aposporous egg cellis in a highly active metabolic state even beforepollination.The changes that occur in the egg cell severalhours after pollination are likely to be the resultof an important reorganization of thecytoplasm that occurs during cell wallformation. The chalazal portion of the egg cellwall is synthesized de novo and does notdepend on cell division or the formation of acell plate. The completion cf the egg cell wall_ lacking plasmodesmata is presumably theresult of a previous activation process;however, a role for cell wall completion in theinduction of egg cell division can not bediscarded. Some researchers suggest that a cellwall might impede fusion of the sperm andegg plasma membranes (Savidan 1982; Askerand Jerling 1992). The ultrastructuralobservations in Pennise//lm ciliare represent thefirst direct evidence demonstrating thepresence of a complete egg cell wall beforepollen tube arrival (Vielle et at. 1995). Theabundance ofGolgi bodies in the apical pocketsuggests that the central cell may beresponsible for the synthesis and transport ofpolysaccharides necessary for the formationof this unique portion of the egg cell walt.Whereas it is unlikely that a chalazally-locatedwall is the only factor preventing egg cellfertilization, it is worth mentioning thatfertilization of apomictic egg cells occurs atvariable frequencies in the grasses, and thatprecocious pollination increases the frequencyof this phenomenon (Martinez et at. 1994). Thepresence of multicellular embryos inunpollinated pistils suggests that autonomousegg cell activation may also be independent

UItr••hwd.ra1 Allalysi••f Apoonktk O...lop....' 61of pollination (Naumova et al. 1992; Naumovaand Matzk 1998). In the case of apogameticdevelopment, synergid activation appears todepend on cytological modifications thatmirror events occurring in fertilized egg cells.Future TrendsThe number of studies dealing withultrastructural analyses of apomicticdevelopment is extremely limited. Even ifsome insights have been gained on the finestructure of adventive embryocytes andaposporous initials, the characterization of themechanistic events that distinguish apomicticinitiation from sexual megasporogenesis is farfrom complete. Additional descriptions ofearly megasporogenesis in members ofdicotyledonous apomicts would be veryvaluable. Low-magnification electronmicroscopy could provide a betterunderstanding of the sporophyticgametophyticjunction, and of the generaldistribution of aposporous initialsdifferentiating in the nucellus. Whereas thedeposition ofcallose has provided an excellentcytological marker to distinguish normalmeiosis from diplosporous differentiation(Carman et al. 1991), a detailed ultrastructuralanalysis of diplospory and all its variants isurgently needed. A description of apomeiosisat the ultrastructural level could provideunique insights about the specifc frequency ofchiasma formation for different species; thistype of analysis has been extensively used tocharacterize mutants affecting female meiosisin maize (Golubovskaya 1979; Golubovskayaet al. 1992).Electron microscopy investigations can now becombined with immunocytochemicalapproaches that take advantage of a widerange of monoclonal antibodies raised againstspecific components of the plant cell surface.The use of monoclonal antibodies representsa valuable alternative to the scarce eDNAprobes and gene expression studies that relateto regulatory genes involved inmegagametophyte development (Vielle­Calzada et al. 1996b; see also Chapter 12). Ofparticular interest are probes that identifyspecific glycoproteins involved in events suchas positional sensing and cell determination(Knox 1992; Pennell 1992). The establishmentof a reproductive lineage can be associatedwith changes in the distribution ofglycoprotein epitopes present at the outer faceof the plant cell plasma membrane (Pennelland Roberts 1990; Pennell et al. 1991). Theseand other discoveries (Knox et al. 1990) suggestthat cell surface arabinogalactan proteins(AGPs) participate in the local control ofreproductive transitions, from a sporophyticto a gametophytic development duringgametogenesis, and from a gametophytic to asporophytic condition after fertilization. In thisregard, plasma membrane AGPs bearcomparison with components of animal cellglycocalyx and suggest a functional similaritybetween plant and apimal cell surfaces(Pennell 1992). These/particular patterns ofAGP distribution seem to be the consequenceof a reproductive cellular commitmentassociated with gametophytic gene expression.The investigation of these proteins in apomicticovules could provide valuable information onthe mechanisms that regulate megagametophytedevelopment in the angiosperms.In Pennisetum ciliare for example, AGP epitopesrecognized by JIM13, a monoclonal antibodyimplicated in embryogenic cell specification(Pennell and Roberts 1990), are. abundantlylocalized in the plasma membrane of sexualand aposporous synergids. This localization isconserved in aposporous megagametophytesaberrantly positioned with respect to themicropylar-chalazal axis, suggesting that thesporophytic tissue is unlikely to playafundamental role on the specification of theegg apparatus (J-P Vielle-Calzada and K.VandenBosch, unpublished results).

62 Tamara N. Nouma.o and Jeon-Phmppe '(..lIe-Cal.adoThe comparison of early fertilization events inPellllisetllnJ ciliare has identified ultrastructuraldifferences related to the control of egg cellparthenogenesis, an essential event in theapomictic life cycle that is absent from sexualreprod uction; however, the dynamics of pollentube arrival and sperm cell delivery have notbeen investigated. Further ultrastructuralstudies will be necessary to determine if spermcell delivery and movement are equivalent insexual and aposporous female gametophytes.The nature of the signal that activates theaposporous egg cell and the fate of the spermcells after pollen tube delivery remain to beelucidated. Levels of calcium and otherelements have never been measured insynergids of apomictic embryo sacs (Chaubaland Reger 1992). Additional studies shouldinclude three-dimensional reconstruction ofegg cells during the first hours followingpollination and quantitative information onthe pattern of synergid degeneration and onthe specific fate of sperm cells within themegagametophyte. Attempts to follow themovement of fluorochromatically-stainedsperm cells in sexual embryo sacs have beenlimited, and the reliability of such studies usingconventional clearing techniques isquestionable; however, in vitro analysis ofsperm nuclei movement within isolatedembryo sacs has been accomplished (Faure etal. 1994). Isolation and fusion of gametes cannow be used to compare the interaction ofsperm cells with sexual or aposporous eggcells, in order to determine if cell to cellinteractions impede the fusion of sperm nucleiwith an ilpomictically-derived egg cell. Finally,some \'ery enlightening findings may bel'btilined in the near future by following onthe pioneering work of Olga Erdelska(Skwakian Academy of Sciences, Bratislilvil).Recent investigations in Torellia jOllmieri haveshown that, at least in certain species in whichthe megagametophyte can be readily dissectedor is partially uncovered by the integuments,some unique discoveries can be made byobserving the process of double fertilizationin vivo (Higashiyama et at. 2000). Microinjectionand cell ablation technology maysoon be combined with microcinerr:atographyto investigate sexual and apomictic megagametophytedevelopment in vivo.AcknowledgmentsWe thank A. Vassilyev and U. Grossniklausfor their critical reading of the manuscript andstimulating discussions, and J. Osadtchiy fortechnical assistance. Work by TamaraN'lUmova. was partially supported by a grantfrom the lAC (the Netherlands) and byfinancial support from the InternationalScience Foundation (USA).ReferencesAsker, 5.E., and l. Jerling. 1992. Apomixis in Plonts. Boca Rotan, Florida:CRC Press.Bashaw, H., and E.C. Holt. 1958. Megasporogenesis, embryo socdevelopment and embryogenesis in dallisgrass, Paspalum dilatatum.Agran.1. 50: 753-56.Bell, P.R. 1988. Control points in generation cycle of ferns. 1. Plant Physiol.132: 378.--. 1992. Apospory and apogamy: implications for understandingthe plant life cycle. Int. 1. Plant Sci. 153: 123-36.---.1995. Incompatibility in flowering plants: adaptation of onancient response. The Plont (ell 7: S--16.Burson, B.l., PW. Voigt, B.A. Sherman, and c.L. Dewald. 1990. Apomixisand sexuality in eastern gamagrass. (rap. Sci. 30: 81>-89.Campbell, 5.C, WG. Craig, and E.B. ScaN. 1987. Apomixis ond sexuality inthree species of Amelanchier, shodbush (Rosaceae, Moloideael. Am.1. Bal. 74: 321-28.Corman, J.G., C.F. Crone, and O. Riera-lizarazu. 1991. Comparativehislology of cell walls during meiosis and apomicticmegasporogenesis in !wo hexaploid Australasian Elymus species.(rop Sri. 31: '527-32.Corman, J.G. 1997. Asynchronous expression of duplicate genes inangiosperms may couse apomixis, bispory, tetraspory, andpolyembryony. Bioi. 1. linn. Soc 61: 51-94.Chapman, G.P, and N. 8usri. 1994. Apomixis in Pennisefum: onultrastructural study. International 1. of Plonf Sciences 155: 492-97.Chaubal, R., and BJ Reger. 1992. Calcium in the synergid cells and otherregions of pearl millet ovaries. Sex. Plant. Reprod. 5: 34-46.Crone, CF., and J.G. Carmon 1987. Mechanisms 01 apomixis in ElymusreC1isefus from eastern Australia and New Zealand. Ame~ 1. Bot. 74:477-96.Diboll, A.G 1968. Fine structural development of the megagametophyte01 lea mays follOWing fertilization. Am 1. Bal. 55: 787-806.

Ultraslrvdural Analysis of Apomktit Deveiopmellt 63Dickinson, H.G., ond U. Potter. 1978. Cytoplosmkchanges accompanying the female mei~isin li/ium longiflorum Thunb. 1. (ell Sc;. 29:147-69.Foure, J·E, e. Digonnet, ond e. Dumos. 1994. Anin vitro system for adhesion and fusion ofmaize gometes. 5cience 263: 1598-1600.Gerassimovo·Novoshina, H. 1957. On somecytologicol principles under~ingfertilization. Phytomorphology 7: 1582-98.Golubovskaya, I.N. 1979. Genetk control ofmeiosis. Int. Rev. (yrol. 58: 247-90.Golubovskoya, I.N., N.A. Avalkin, ond W.F.Sheridan. 1992. EIIect of several meioticmutants on female meiosis in maize. Dev.Genet. 13: 411-24­Gustafsson, A. 1947. Apomixis in higher plantsII. lunds Univ. Arsskr NF1I43: 71-179.Huang, B.Q., and S.D. Russell. 1992. The femolegerm unit. In S.D. Russell and e. Dumas(eds.!, Sexual Reproduction in FloweringPlants. Int. Rev. (yrol. 140: 233-93.Higashiyamo 1, H. Kuroiwo, S. Kawano, and 1Kuraiwo. 2000. Explosive discharge afpollen tube contents in Torenio Fournier;.Plant Phys. 122: 11-14.Jensen, W.A. 1965. The ultrastructure andhistochemistry of the synergids of catton.Amer. 1. 80t. 52: 238-56.Jensen, W.A., and D.B. Rsher. 1968. Coltonembryogenesis: the enlrance and dischargeof the pollen tube in the embryo sac. Plonto78: 158-83Knox, J.P. 1992. Molecular prabes for the plontcell surface. Protoplosmo 167: 1-9.Knox, R.B., and J. Heslop·Harrisan. 1990. Directdemonslratbn of the low permeability ofthe angiosperm meiotk tetrad using aflu orogenic filter. 1. Pflonzenphysiol. 62:451-59.Knox, J.P., P.J.linslead, J. King, e. Cooper, and K.Roberts. 1990. Pectin esterification isspotially reguloted both within cell walls andbetween developing tissues of root apices.Plonto 181: 512-21.Koltunow, A.M., K. Soltys, N. Nobumasa, and s.Mcclure. 1995. Anther, ovule, and nucellarembryo development in Gtrus sinensis cvVolencia. (on. 1. 80t. 73: 1567-82.Kronz, E., J. Bautor, and H.larz. 1991.ln vitrofertilizalian of single, isolated gametes ofmaize mediated by electro fusion. 5ex. PlantReprod. 4: 12-16.leblanc, 0., M.D. Peel, J.G. Corman, and Y.Savidan. 1995. Megasporogenesis andmegagametogenesis in several Tripsocumspecies (Paoceae) Amer. J. 80t. 82: 57-63ligrane, R., J.G. Duckett, and K.S. Renzaglia.1993. The gametophytk·sporophytkjunction in land plants. Advanced 80tonicolResearch 19: 231-317.Martinez, E.J., F. Espinoza, and e.L Quarin.1994. Bill progeny (2/ltn) from apomicticPospolum nototum obtained through ear~pollination. Journal of Heredity. 85: 29>­97.Mogensen, H.L 1988. Exclusion of malemitochondria and plastids during syngamyin barley as a basis for mote rnaIinheritance. Proe. Notl. Acod. Sri. (USA) 85:2594-97.Naumova, IN. 1978. Peculiarity ofmauagametogenesis and p~t·fertilizationdevelopment of Trillium camS(hatcenseKer.·Gawi. 5oc.80t. Fr. Aduol;tes80toniques 1·2: 183-86.--.1990. Fami~ Trilliaceae. In M.S.Yakolev and lB. Batygina (eds.),(omporotive Embryology of FloweringPlants (Butomaceae·lemnaceae).leningrad: Nauka, (in Russian). Pp. 151­59.---. 1993 Apomixis in Angiosperms:Nucellor and Integumentary Embryony.Boca Raton, Florida: CRC Press.Naumova, IN., A.PM. den Niis, and M.T.M.Willemse. 1992. Quantitative ona~is ofaposporous parthenogenesis in Pooprotens;s genotypes. Ado 801. Neerlondica43: 1-14.Naumovo, IN., ond F. Matzk. 1998. Differencesin the initiotion of the zygotic andparthenogenetic pothwoy in the Salmonlines of wheot: ultrostructurolstudies. 5exPlant Reprodudion 11: 121-30.Naumava, IN., J.V. Osadtchiy, V.K. Shorma, PDijkhuis, and K.S. Ramulu. 1999. Apomixisin plants: structural and functional ~pectsof of dipl~pry in Poo nemorolis ond P.polustris. Protoplosmo 208: 186-95.Naumova, IN., and M.T.M. Willemse. 1982.Nucellar po~embryany in 5orrococcohumilir. ultrastructural aspects.Phytomorphology 32: 94-108.--.1995. Ultrastructuralchorocterization of ap~pory in Ponicummaximum. 5ex. Plant Repr. 8: 197-204.Nogler, G.A. 1984. Gametophytic opomixis. InB.M. Johri (ed.), Embryology ofAngiosperms. Berlin: Springer·Veriag. Pp.47>-518.Peel, M.D., J.G. Carman, and o. leblonc. 1997.Megasporocyte callose in opomicticbuHelgr~s, Kentucky bluegrass,Pennisetum squomulotum Fresen,Tripsocum Lond weeping lovegrass. (rap5c;. 37: 717-23.Pennell, R.I. 1992. Cell surface orabinogoloctonproteins, arobinogalactons ond plontdevelopment. In J.A. Callow ond J.R. Green(eds.1, Perspedives in Plant (ell Recognition.Society of Exp. BioI. Series 48. CambridgeU.K.: Cambridge Univ. Press.Pennell, R.I., l. Jonniche, P. Kjelbom, G.N. Scofield,J.M. Peart, ond K.A. Roberts. 1991.Developmental regulation of a plasmamembrane arabinagalactan protein epitope inoilseed rape flowers. Plant (ell 3: 1317-26.Pennell, R.I., and K.A. Roberts. 1990. Sexualdevelopment in the pea is presaged byaltered exprmion 01 arabinogolactanpratein. Nature 344: 547-49.Philipson, M.N. 1978. Apomixis in Cortoderioiuboto (Gramineoel. New Zeal. 1. 801. 16:4>-59.Remy, w., P.G. Gensel, and H. Hoss. 1993. Thegametophyte generation of some earlyDevonian land plants. Int. 1. Plont5ci. 154:3>-58.Russell, S.D. 1985. Preferential fertilization inPlumbago; Ultrastructural evidence forgamete·level recognition in an angiosperm.Proe. Notl. Acad. 5ri. (USA) 82: 6129-32.--. 1992. Dauble fertilization. In S.D.Russell and e. Dumas (eds.!, SexualRepoduction in Flowering Plants. Int. Rev.(ylol. 140: 357-88.Savidan, Y. 1982. Nalure et heredite de I'apomixiechez Poniwm maximum Jacq. Trov et DocOR5TOM 153:1-159.Schulz, S.R., and W.A. Jensen. 1968. (opselloembryogenesis: the egg, zygote and yaungembrya. Amer. 1. 80t. 55: 807-19.Vielle, J.p, B.l. Burson, E.e. Bashaw, and M.A.Hussey. 1995. Eorly fertilizotion events in thesexual and aposporous egg apparotus ofPenn;setum ciliore ILl link. Plant 1. 8: 309­16.Vielle·Colzada, J.p, C.F. Crane, ond D.M. Stel~.1996a. Apomixis: the asexual revolution.5cience 274: 1322-23.Vielle·Calzada, J.p, M.l. Nuccio, M.A. Budimon, 1l.Thomas, B.L Burson, M.A. Hussey, and R.A.Wing. 1996b. Comparative gene expressionin sexual and apamictic ovaries ofPennisetum ciliore. Plant Mol. 8iol. 32:1085-92.Voigt, PW., and E.e. 8ashaw. 1972. Apomixis andsexuality in &ogr05tis curvulo. (rop Sci. 12:843-47.Wakana, A., ond S. Uemoto. 1987. Advenliveembryogenesis in Glrus (Rutoceae) II. Theoccurrence of adventive embryos withoutpollination or fertilizatian Arner. 1. 801. 74:517-30.Wilms, H.J., J.l. van Wenl, M. Cresli, and f.Gampolini. 1983. Advenlive embryagenesisin Citrus. (oryolagio 36: 6>-78.

Chapter 5Genetic Analysis of ApomixisROBERT T. SHERWOODIntroductionGeneticists interested in analyzing theinheritance of apomixis face challengingproblems. Apomixis overrides certainprocesses essential to the analysis ofinheritance, i.e., it usurps meiotic megasporogenesisand megagametogenesis. Obligatelyapomictic plants cannot serve as maternalparents in hybridization. However, sinceapomixis normally does not prevent meioticpollen formation, apomicts can be used asmale parents in crosses with sexual orfacultatively apomictic female parents. Whenfacultative apomicts function as maternalparents, three types ofprogeny may be formed(see Berthaud, Chap.2), and each type mustbe distinguished when testing genetic models,i.e., normal B IIhybrids from fertilization ofreduced embryo sacs, apomictic progeny fromparthenogenetic embryogenesis of umed ucedeggs, and B m progeny from fertilization ofunreduced eggs by reduced pollen yieldingnonmaternal types at an increased ploidylevel. Sexual reproduction is unknown insome apomictic species, accordingly, sexualplants must be found or created before geneticmanipulation is possible. Sexual members ofagamic complexes usually are at a differentploidy level than the apomicts. The polyploid,highly heterozygous nature of most apomicticplants complicates genetic analysis (Stebbins1950; Nogler 1984a).Early reports indicated that apomixis isheritable, but did not point to specific genesor genetic systems (Gustaffson 1946--47; Nogler1984a). Successful hybridization was difficult,and methods for classifying progeny weretedious and unreliable. Discovery of thecytologically distinctive Panicum-type ofapospory in the 1950s fostered the beginningofcreditable inheritance studies. Pistil clearingtechniques were introduced in the 1970s (seeCrane, Chap. 3) that permitted the classificationof large numbers of progeny. More recently, theapplication of molecular technology tocharacterizing, locating, and isolating apomixissequences has augmented our understandingof the regulation of apomixis (see Grimanelliet aI., Chap. 6). Presently, it appears that theexpression of aposporous apomixis requires adominantly acting master gene or linkage unit;roles have been indicated for dosage, additivity,recessive lethality, and modifying genes. Thelimited data available for diplosporous taxaindicate that diplospory also may be regulatedby a dominant linkat with modifiers. Applieddiligently, the methods suited to hybridizationand classification of apomictic plants describedbelow can lead eventually to a resolution ofthese difficult problems.MethodsThis section discusses methods for selectingparents, characterizing parents and progeny,and making crosses of apomictic species.Chapters 2, 6, 9, and 10 should also beconsulted. Chromosomal constitution,reproductive behavior, and phenotype ofbothparents must be completely known.

Getoetk AooIysls ., Ap••obls 65Chromosome NumberChromosome counts in progeny are essentialwhen the parents are at different ploidy levels,whenever aneuploidy is suspected, and fordetecting BIll hybrids. When facultativeapomicts are used as the maternal parent, thenonmaternal progeny, or a sample thereof,should be examined for their chromosomenumbers to determine whether they are BnorBIll hybrids. Counts are made from anaphasefigures of root tips using standard techniques(Sherwood et al. 1980; Dujardin and Hanna1989; Hignight et al. 1991). Some 2n + 1aneuploid plants have a tendency to eliminateone chromosome in root tips, makingdetection ofaneuploidy difficult (Nogler 1989).Flow cytometry is useful in determiningapproximate ploidy levels (den Nijs 1990; Huffand Bara 1993; Naumova et al. 1993;Mazzucato et al. 1994; Leblanc et al. 1995a;Leblanc and Mazzucato, Chap.9). Ploidy levelof individual reproductive nuclei can bedetermined with image cytometry of Feulgenstained sections (Sherwood 1995) orphotocytometry of DAPI-stained isolatedembryo sacs (Naumova et al. 1993). Cell cyclestages must be accounted for when applyingflow cytometry or photocytometry.Chromosome pairing and disjunction usuallycan be determined by examiningmicrosporogenesis (Hignight et al. 1991;Burson 1992). Pollen viability is determinedusing 1 2KI stain, fluorescein diacetate stain, orgermination tests (Dujardin and Hanna 1989;Hill et al. 1989). Female fertility is estimatedby determining percentage of seed set per 100florets in open or self-pollinated inflorescences(Hignight et al. 1991).Progeny TestingProgeny testing originally was practiced aswhole plant morphological comparison ofprogeny with the maternal parent. Broadlyspeaking, comparative analysis of any traitincluding cytological and molecular markersconstitutes progeny testing. Identity ofoffspring with the maternal type in all respectsindicates possible apomictic origin (Bashaw1980). However, maternal appearance can alsostem from matrocliny and from selffertilizationof highly homozygous parentalmaterial, as found in advanced generationsof naturally inbred plants (Asker 1980).Embryo-Sac CytologyIt is necessary to recognize all stages of normalmeiotic megasporogenesis and megagemetogenesis,as well as the stages of apomeioticmegasporogenesis and gametogenesis.Luckily, the number of different types foundamong species for which inheritance studiesare feasible is small compared with the totalrange displayed by angiosperms. For about90% of genera known to reproduce bothsexually and apomictically, normal meioticreproduction is based exclusively onformation of the Polygonum-type of embryosac. About 10% of the genera commonly havebisporic or tetrasporic sexual embryo sacdevelopment (Carman 1997). Fortunately,most apomictic species exhibit only one typeof apomictic sac from among the four typesdescribed below (Nogler 1984a).The distinction between meiotic (= sexual) andapomeiotic (= apomictic) events becomescytologically discernable after differentiationof the, megaspore mother cell (MMC) in allovules (see Leblanc and Mazzucato, Chap. 9).In the normal monosporic Polygonum-typemeiosis, walls of megasporocytes andmegaspores (tetrad cells) become investedwith callose. It is a simple matter to visualizethis cage-like indicator of meiotic activityusing fluorescence m"icroscopy of intact,aniline blue treated pistils (see Leblanc andMazzucato, Chap. 9). The fully differentiatedPolygonum-type sac is based on an 8-nucleatescheme, with an egg apparatus, polar nuclei,and antipodal cells.

66 Robe" T. SherwoodIn mitotic diplospory and restitutionaldiplospory, the MMC is diverted intofunctioning as an unreduced, apomicticembryo sac initial. Its size and shape may varyfrom those of reduced megasporocytes andprovide fleeting evidence for diplospory thatis tedious to acquire. Illustrations are given byGustafsson (1946-1947), Hair (1956), Battaglia(1963), Rutishauser (1969), Nogler (1984a), andCrane and Carman (1987). A much moreconvenient test for diplospory is based on theabsence of callose deposition in pistils duringdiplosporous megasporogenesis (Carman etal. 1991; Leblanc et al. 1995c).Monopolar aposporous embryo sac (=Panicum-type embryo sac) formation followsa 4-nucleate scheme. Panicum-type sacs areknown only in the Panicoideae andArundinaceae (Brown and Emery 1958,Nogler 1984a). Illustrations are given in Fisheret al. (1954), Snyder et al. (1955), Bashaw andHolt (1956), and Bashaw (1962). Each embryosac develops from a somatic cell (usually anucellar cell) about the time that meiosis beginsin the MMC. Meiotic and apomeiotic features,including callose deposition around tetrads,occur in the same ovule. The mature Panicumtypesacs lack antipodals and usually can bedistinguished from sexual sacs by this feature.Occasionally, large antipodals of sexual sacscan be mistaken for multiple aposporous sacs.Aposporous activity usually crowds outsexualsac formation. However, some pistils offacultative plants eventually may contain afully differentiated sexual sac, or even sacs ofboth types (Sherwood et al. 1980). Bipolaraposporous sacs and diplosporous sacs formantipodals and at maturity closely resemblethe sexual sacs (Nogler 1984a).Sectioning or Clearing Pistils to ClassifyReproductive TypeThe reproductive mode can be classified byexamining diagnostic stages of megasporogenesisand / or megagametogenesis in clearedor sectioned pistils. In monopolar (Panicumtype)aposporous species, the pistils can beexamined at any stage after embryo sacenlargement has begun. In otherspecies, stagesof megasporogenesis are examined. Toaccurately determine whether embryo sacs ofa plant are exclusively or predominantlysexual or apomictic, at least 20-100 pistils fromthe plant should be viewed. A few pistils willbe unclassifiable.Sectioning paraffin embedded pistils istechnically demanding and time-consuming.Early studies were based on sectioning andonly small samples of ovules were examined,consequently, facultativeness was not readilydetected. However, sectioning remains usefulfor classifying certain species (Burson 1992).Spikelets of Erngroslis have multiple floretswith a range of developmental stages. Carefulsectioning of a spikelet permits observation ofdevelopmental sequences and assists ininterpretation in facultative apomicts in whichdiplospory affects timing of meiotic eventsrelative to those of normal sexualdevelopment.Clearing methods are much faster and easierthan sectioning. Nuclei and walls ofreproductive cells can be visualized in theirproper relations in intact pistils or ovules.Several c1earants and protocols are available(Herr 1971, 1982; Crane 1978; Young et al. 1979;Crane '1.nd Carman 1987; Nogler 1990, Savidan1990a; Crane, appendix of Chap. 3; Leblancand Mazzucato, Chap, 9), Users mustdetermine which clearant provides the bestresults for their material; they should also befamiliar with sectioned material of the species,MarkersAside from their effects on embryo-sacformation, parthenogenesis, and conservationof the maternal genotype, apomixis genes haveno known effect on plant characteristics-norhave transcripts or other direct products of the

Ge••ti< Alaly.;'.f Apamil;' 67genes been located. No conventionalmorphological, agronomic, or physiologicaltraits are specifically associated with apomixis.The lack of linkage information is hardlyunexpected given the obstacles to traditionalmapping in species that have the barrier tocrossing imposed by apomixis and for the mostpart are alloploids with indistinguishablechromosomes, irregular chromosomeduplication, and secondary economic status.Conventional, unlinked, monogenicallyinherited traits have been used as markers todistinguish maternal from hybrid progeny. Ifthe maternal parent is homozygous for arecessive trait, and the pollen parent ishomozygous dominant, uniformity of progenyfor the maternal marker suggests maternalinheritance. Homozygous or heterozygousdominant markers in the pollen parent havebeen employed to reveal hybrid Fls (Hanna etal. 1970; Hanna and Powell 1973; Dujardin andHanna 1989; Hignight et al. 1991).Isozyme polymorphism has been used tocharacterize variability in apomictic parentsand progeny (Marshall and Downes 1977;Hacker 1988; Cruz et al. 1989; Roy andRieseberg 1989; Bayer et al. 1990; Kojima et al.1991; Poverene and Voigt 1995; Gustine et al.1996; Berthaud, Chap. 2; Leblanc andMazzucato, Chap. 9).Several molecular markers that apparently arelinked with apomixis genes have been found(Leblanc and Mazzucato, Chap. 9; Grimanelliet aI., Chap. 6). Ozias-Akins et al. (1993, 1998)and Lubbers et al. (1994) described a randomamplified polymorphic DNA (RAPD) markerand a sequence-tagged site (STS) markertightly linked with apomixis in Pellniseillmspecies. Gustine et al. (1997) described twoadditional linked RAPD markers in P cilinreand derived a preliminary linkage map of threemarkers with the apospory locus. Leblanc etal. (1995b) prepared three restriction fragmentlength probes (RFLP) from a maize-TripsnCllmdnctylaides F\ population that cosegregatedwith diplospory. They also were linked on thelong arm of chromosome 6 of maize.Biological Tests for ParthenogenesisMatzk (1991a) devised an auxin test fordetecting parthenogenetic capacity.Unpollinated plants are treated with DIC;2,4-0; 2,4,5-T; or CPAA. Parthenogeneticindividuals form grains with a mature embryobut no endosperm. Results are positive forparthenogenetic mutants of nonapomicticspecies (barley, wheat) and for apomictic plantsof apomictic species. The test can be used toscreen for parthenogenetic plants in sexualspecies and to detect sexual plants in apomicticpopulations. It has proven useful incharacterizing Pan pmlensis lines that vary indegree of facultative apomixis (Matzk 1991b;Mazzucato et al. 1996).Naumova et al. (1993) described a cytologicaltest for quantitative analysis ofparthenogenesis in Pan pmlensis. Embryo sacswere isolated mechanically and examined forspontaneous embryogenesis.An ovule culture medium facilitatedidentification of apomixis in diplosporousAllillm II/beros/lm (Kojima and Kawaguchi1989). Up to 80% of apomictic embryos, butno sexual embryos, showed development onthe medium.•Combined Cytological, Progeny,Biological, and Marker TestingWhen used alone, none of the progeny testingmethods discussed above can unequivocallyestablish the reproductive status of every plant.Two or more approaches applied together aremore informative (Naumova et al. 1993;Mazzucato et al. 1996).Whole plant progeny testing views the endproduct of seed formation and is the ultimatetest of whether apomixis is functional.Cytological examination of ovules during

68 ••It T. Sllorwoodmegasporogenesis and megagametogenesistests whether the plant has a capacity forapomictic embryo-sac formation and howstrong that capacity may be relative to thesexual alternative, but it does not indicatewhether the unreduced sacs will formfunctional seeds. Cytological and whole plantmethods must be used in tandem tocharacterize parents (Savidan 1992).Comparisons of methods of classificationgenerally show good agreement. Cytologyand whole plant progeny testing give similarresults for most lines of Eragrostis curvula(Voigt and Burson 1983) and Panicummaximum (Nakajima and Mochizuki 1983). InBrachiaria, Miles and Valle (1991) foundcorrespondence between the two methods inclaSSifying 54 Fls as sexual and 37 Fls asapomictic, however, 10 plants that appearedsexual in progeny tests were facultativelyapomictic in embryo-sac analysis, i.e., progenytesting underestimated the genetic potentialfor apomixis. Cytological analyses generallyreveal higher sexual potential than is indicatedby whole plant progeny testing (Savidan1982a; Mazzucato et al. 1996).When apomixis is essentially obligate,progeny tests are considered as reliable ascytological analyses (Savidan 1982a). Infacultative lines with high levels ofapomeiosis, progeny testing in conjunctionwith a determination of chromosome levelsof off-type progeny may be efficient indetecting sexuality (Sherwood et al. 1980;Savidan 1982a). Progeny tests are unreliablein detecting low levels of apomixis in apredominantly sexual line (Savidan 1982a;Voigt and Burson 1992). Heterogeneity withina progeny cannot be considered proof of theabsence of apomixis (Yudin 1994).Early identification of nonmatemal plants inprogeny of facultatively apomictic Poapratellsis has been facilitated by isozyme andRAPD markers (Huff and Bara 1993;Mazzucato et al. 1995). Estimates of the degreeof apomixis or parthenogenesis in P pratensiswere higher with progeny testing than withembryo sac analysis (Mazzucato et al. 1996).The auxin test gave similar or higher estimatesthan embryo-sac analysis. It is necessary toexamine a large sample of ovules or seed(upwards of 100 individuals) from eachpotential parent to detect any tendenciestoward facultativeness.Controlled PollinationAccidental self- or cross-pollination candramatically influence genetic inferences,especially when wide ratios are being tested.Unintentional self-fertilization will skew ratiosin favor of the phenotype of the maternalparent. Markers to recognize hybrid or selfedprogeny should be used when available.Several approaches have been used to reduceunwanted fertilization in apomixis research:1) Protogyny. If the inflorescence exsertsreceptive stigmas before anthers are exsertedand dehisce, stigmas can be pollinated beforethe maternal floret sheds pollen (Voigt andBashaw 1972; Hanna and Powell 1973; Voigtand Burson 1983; Bashaw et al. 1992; Valle andMiles, Chap. 10). Extraneous pollen is excludedby covering the head with a paper or glassinebag prior to stigma exsertion and continuinguntil seeds are set.2) Suppressed anther dehiscence. Dehiscenceof exserted anthers can be suppressed bymaintaining high humidity. Humiditychambers (Taliaferro and Bashaw 1966),glassine bags (Hanna et al. 1973), and glassbottles lined with moist filter paper (Sherwoodet al. 1994) have been used for that purpose.3) Hand emasculation. Emasculation prior toopening of flowers, followed by bagging, hasbeen practiced for Potentilla (Asker 1970a) andRalllll1CltIlls (Nogler 1984b). Valle et al. (1991)showed that emasculation and bagging of

Goottk AIaIy,Is .f Apolllilis 69Brachiaria did not totally prevent selfing.Techniques have been published foremasculating small flowered grasses (Burson1985,1992; Richardson 1958).4) Male gametocide. Young inflorescences ofPennisetum ciliare were sprayed with a malegametocide (Bashaw and Hignigh 1990;Hignight et al. 1991).5) Self incompatibility. Self incompatibility ofsexual female lines has been used to advantagein Taraxacum (Richards 1970), Paspalum notatum(Burton and Forbes 1960), and Hieracium(Gadella 1987). Dujardin and Hanna (1989)used male sterile pearl millet (Pennisetumglallcum) as a female parent. Selfincompatibilityis incomplete in guineagrass(Panicum maximum); the degree of crossfertilization depends upon the procedurepracticed for isolation (Savidan et al. 1989;Savidan 1990b). Because genotypes can varyin degree of spontaneous selfing, it may benecessary to use control tests to establishreliability of each female parent (Matzk 1989;Valle et al. 1991).Reciprocal CrossingReciprocal crossing detects nuclear andcytoplasmic maternal effects (includingmatrociiny). The apomixis gene(s) has such apowerful maternal effect that geneticists havebeen discouraged from using this technique,but apomixis need not be a deterrent to its use.It is feasible to use one or two facultativeparents in the crossing scheme (Savidan 1981).Nogler (1984b) conducted reciprocalbackcrosses with a sexual diploid genotype asthe male or female in pairings with facultativelines. Jassem (1990) used reciprocals in beets.Creating Tetraploid ParentsIt is best if both parents are at the same ploidylevel. The tetraploid level seems to be thenatural milieu for expression of apomixis,whereas apomixis is rarely confirmed indiploids. If fully sexual tetraploids are notavailable, they may be produced by variousstrategies. Colchicine treatment of sexualdiploids has created sexual tetraploids usedin hybridizations (Burton and Forbes 1960;Richards 1970; Savidan 1981; Miles and Valle1991; Valle et al. 1991). Leblanc et al. (1995a)treated embryogenic calli of sexual diploidTripsacum with colchicine to inducechromosome doubling; the regeneratedtetraploid plants reproduced sexually.However, when a sexual line of PaspalumheXilstachium was doubled, the tetraploid wasfacultatively aposporous (Quarin and Hanna1980). Asker (1967) started with an apomicticdiploid (possibly a dihaploid?) biotype ofPotentilla argentea and obtained a partiallysexual tetraploid. Thus colchicine doublingmay reveal latent capacities for apomixis orsexuality.Tetraploids can be created by hybridization.Harlan et al. (1964) selected a completelysexual tetraploid of Bothriochloa grahamii froma cross between two facultatively apomictictetraploids. Savidan's (1981) crosses offacultative tetraploids of genotype Aaaayielded sexual:aposporous (S:A) progeny inthe ratio of 1:3, indicating that the sexualprogeny were genotype aaaa. The geneticanalyses of Harlan et al. (1964) also used asexual tetraploid Dichanthium annulatumaccession that originated from a BIllhybridization ofan emasculated diploid sexualplant with a tetraploid male. Burton andHanna (1986) grew diploid sexual Pensacolabahiagrass (Paspa/um notatum) in isolation withan apomictic tetraploid to produce a triploidBmhybrid. Open pollination of the triploidwith Pensacola bahiagrass yielded Bill hybridsat the tetraploid level. These facultativelyapomictic lines could be used as maternalparents in crosses and selfs to create sexualtetraploids. Quarin (1992) successfully used asimilar scheme.

70 Robert T. SherwoodIdentification of Genomes andChromosomes with Apomixis GenesInterspecific or intergeneric crosses haveprovided evidence for apomixis genes in agenome or chromosome unique to one of theparents. Sexual tetraploid vaseygrass(Paspalllnl lIrvillei) with the genomic formulaIIJJ was hybridized with hexaploid P.dilatatllnl biotypes with genomic formulas ofIIJJXX and IlJJXX 2(Burson 1992). Thepentaploid hybrids were facultativeapomicts, indicating a gene(s) controllingapomixis was located in the X genome.Examinations of aneuploids indicate thatgenetic control for apomixis may reside inone chromosome (Sorenson 1958; Nogler1984b; Mogie 1988). This also appeared toexplain introgressive transfer of aposporyinto a sexual Pennisetllnl glallCllnl background(Dujardin and Hanna 1989). It appeared thatthe apomictic 29 chromosome BC) plantreceived 28 chromosomes from pearl milletand one, bearing an apomixis gene, fromPe/wisetllnJ sqllanJlIlatllnl; however, moleculardata of Ozias-Akins et al. (1993) did notsupport this assumption.RFLP markers were used successfully tolocate a portion of a sexual maizechromosome homologous to the TripsaCllnlchromosome bearing the diplospory gene(Leblanc et al. 1995b; Grimanelli et aI.,Chap.6). Three RFLP probes cosegregatedwith diplospory in a maize-TripsaCllnJ F)population that was segregating for mode ofreproduction. The markers also wereassociated with the long arm of maizechromosome 6.Testing InheritanceStarting PointGenetic relatedness, chromosomal balance,and sexuality of the maternal parent are theprimary considerations in selecting parents(Savidan 1990a). Generally speaking, the bestsituation is to use intraspecific crosses;interspecific crosses are a second choice, andintergeneric crosses are a poor third alternative.Interspecific and intergeneric crosses of distantparents usually create abnormalities thatinterfere with reproduction. Selection ofgenetically close parents, however, must betempered with consideration of naturalcrossability rather than arbitrary taxonomicdistinctions. Interspecific crossing has beenrelatively successful among some species of theDichanthillnl / Bothrioclrloa / CapillipedillnJcomplex (de Wet and Harlan 1970) and withinBrachiaria (Valle and Miles, Chap. 10). On theother hand, a composite species such asPotentilla argentea may constitute a complexaggregation of sexual and apomictic biotypesthat do not readily interbreed (Asker 1970a, b).Crosses between plants at the same ploidy levelare preferred in order to reduce meioticdisturbances and sterility. Hybridization acrossploidy levels leads to intermediate, and oftenunstable, chromosome numbers. Triploidhybrids resulting from crossing diploid andtetraploid parents can express apomixis, butthey will not be desirable for further crossingor selfing except for developing aneuploids.Sexual parents should be devoid of apomixisalleles. Savidan (1990a) warns of hiddenapomixis within naturally occurringtetraploids that appear to be sexual. Hesugges,ts that sexuality in natural tetraploidscould be facultative, even if it looks obligate ina progeny test, because of modifying factorsthat influence expression of the apomixis gene.Savidan goes on to recommend creation oftotally sexual tetraploids by colchicinedoubling of sexual diploids.Crossing SchemesAll apomixis inheritance studies published todate have included crosses between sexualmaternal parents and apomictic pollen parentsfollowed by characterization of the Fl' Usually,severaI crosses were made, and fi rst selfed

G...1i< loaly,i, .f Apomixi, 71generation SI plants of sexual parents wereanalyzed. Information from Fls and SIS hasonly limited power for testing alternativegenetic models. Interpretations can bestrengthened using advanced generations, asin studies by Burton and Forbes (1960); Nogler(1984b); Savidan (1981); and Valle and Miles(Chap. 10).Classification and GroupingHow should facultative progeny displayingvarious degrees of apomixis be grouped whentesting segregation ratios? Current practiceplaces all plants showing any apomixis intoone group (deemed apomictic) and all plantsdevoid of apomixis into a second group(deemed sexual) (Savidan 1981; Voigt andBurson 1983; Miles and Valle 1991; Sherwoodet al. 1994). Most facultative plants produce farmore apomictic sacs and progeny than sexualsacs or nonmatemal progeny, and classificationis relatively straightforward. With only twogroups being recognized, it is inevitable thatthe genetic models are for Simple gene actionwith Mendelian interpretations. Unimodaldistributions of reproductive types indicativeof quantitative or polygenic inheritance havenot been tested.Testing Genetic ModelsAll reasonable genetic models andpermutations of genotypes should be tested.Predicted ratios should consider the effects ofnumber of loci, dominance, epistasis, andlethali ty. If the degree of alloploidy versusautoploidy of tetraploids is unknown, bothdisomic and tetrasomic inheritance and bothpartial and complete random assortment ofchromosomes or chromatids should beconsidered (Sherwood et al. 1994).Inheritance of ApomixisLiterature on inheritance of apomixis has beenre\'iewed by Stebbins (1941), Gustafsson (1946­1947), Asker (1980), Nogler (1984a), Bashawand Hanna (1990), and Asker and Jerling(1992). Reports issued since 1950 aresummarized here, with a reinterpretation ofsome of the results. Symbol "AU is used todenote the dominant allele of the putativeapomixis gene regardless of the symbols usedin the original reports.Monopolar Apospory (Gramineae­Panicoideae)Paspaillm notatum (bahiagrass). Burton andForbes (1960) crossed colchicine-ind ucedsexual autotetraploid lines of Pensacolabahiagrass with an apomictic tetraploid line.Progeny were classified only by whole plantprogeny testing. The crosses producedsexual:apomictic (S:A) segregation ratios near3:1. Selfing the sexual progeny gave an F 2witha ratio near 35: 1. Selfing the apomictic progenygave only apomictic F 2progeny. Burton (1992)postulated an A gene dominant for apomixisand an independent 5 gene dominant forsexuaJity. He aSSigned genotype Aaaassss toMH, and genotype 111111115555 to the sexualparents. Some progeny from selfing sexualsappeared apomictic based on their uniformitybut actually may have been sexual; this couldgive rise to the 35:1 ratios. In the crosses giving3:1 ratios, some facultatively apomictic hybridF I progeny may have been inadvertentlyclassified sexual, skewing the ratios in favorof high numbers of sexual progeny. Morerecent data support the view that apomixis iscoded for by dominant genes (Burton andHanna '1992). The bahiagrass system deservesreexamination using modern classificationmethods.Dichanthium-Bothriochloa (bl uestem).Harlan et al. (1964) crossed sexual tetraploidsof D. 1I11111t1l1tl/m and D. grahnmii with apomictictetraploid pollen parents. Crosses of sexual xsexual provided only sexual Fls. Crosses ofsexuals x apomicts gave a S:A of 1:4.1. Theypostu lated random assortment of two disomicgenes and assigned genotype A ll A z II tol 1 Ztetraploid apomicts, and II 11 11 11 1 1 Z 2to sexuals. If

72 R....11 T. SlMrwoodwe postulate that the tetraploids weretetrasomic with genotype AAoo, the observedratios fit ratios expected for the assumptionof random chromatid assortment (1:3.67) orthe assumption of a recessive lethal effect ofthe A allele (1:4) (Sherwood et al. 1994).Pennisetum ciliaTe (buffelgrass). A naturallyoccurring tetraploid sexual plant, designatedB-1S, was selfed and crossed with twoaposporous biotypes by Taliaferro andBashaw (1966). The S:A ratios (near 13:3 forSIS and 5:3 for Fls) suggest two disomicindependent genes with the dominant alleleof gene A being required for apospory, andthe dominant allele of gene B being epistaticto A and conferring sexuality. Sexual parentB-1S was assigned genotype AaBa. Furtherevaluation of Fls, F 2s, and a BC I from B-1S,identified true breeding sexual progeny ofapparent genotype aabb and apomictic plantsof putative genotypes Aabb and AAbb (Readand Bashaw 1969; Bashaw et al. 1970).Crane (1992) proposed a tetraploid Single genemodel to explain the Taliaferro and Bashawsegregations. Three alleles were postulated: a(wild type sexual), A (aposporous), and A+(super sexual). Only genotypes AAAA, AAAa,and AAaa would be apomictic. Chromosomalsegregation patterns were postulated to bepreferential and to differ duringmicrosporogenesis and megasporogenesis.Sherwood et al. (1994) studied inheritance ofembryo sac type of sexual tetraploid plant B­2S. From open pollinated B-25, five siblingsexual lines and five sibling aposporous lineswere selected as parents. Segregations weredetermined for crosses ofsexual x aposporouslines and sexual x sexual lines, and selfs ofsexual lines. Selfs and crosses ofsexual plantsgave only sexual progeny. F 1 s from sexual xaposporous combinations segregated for S:Aat ratios near 15:13 for four aposporous linesand 1:2.8 for the other line. Segregations didnot fit any one- or two-disomic gene models,nor any recessive gene models. Data werecompatible with a one-tetrasomic-gene modelwith apospory regulated by dominant allele A,under either of two assumptions:(i) randomassortment of chromatids, or (ii) A acting as arecessive lethal in gametes. Sexual plants wereassigned genotype aooa; apomicts were Aaaaand AAaa. Data on linkage of apospory inPennisetum with molecular markers (Gustine etal. 1997; see Grimanelli et al., Chap. 6) providesadditional evidence that a single major locusregulates apospory in Pennisetum.Panicum maximum (guineagrass). Hanna et al.(1973) reported that four naturally sexualtetraploid accessions produced 51 progeniessegregating in a combined ratio of 116S:54A.Crosses of sexuals x apomicts gave 21S:28A.They proposed a digenic, disomic additivemodel using the assignment of AaBb for thesexual plants and Aabb, aaBb, or aabb forapomicts (two dominant doses required forsexuality).Savidan (1981) crossed a colchicine-inducedautotetraploid sexual plant and a naturalapomictic tetraploid. Ten kinds of crosses weretested (Table 5.1). All the data fit perfectly withTable 5.1 Segregations for mode of reproduction in10 crosses of Panicum maximum (Savidan 1981;Savidan et a!. 1989)sexual x apomictic crosses sum apo sexFl hybrids,; SIx Al 133 71 623-way hybrids: IS1xAIlsex xA2 279 135 144Backcross: (S1xAIlsex xA1 26 14 12Backcross: sexual 3-way hybrid xA2 170 73 97sexual 3-way xapomictic 3-way 60 26 34Backcross: S1x(S1xAllapa 23 13 10total sex xapo crosses 691 332 359sexual x sexual aosses (or selfed)sexual Fl hybrids selfed 126 0 126sexual 3-way hybrids selfed 57 0 57sexual 3-way xsexual 3-way B2 0 B2total sex xsex crosses 265 0 265apomictic x apomictic crossesapomictic 3·way xapomictic 3-way' 71 53 18.. analysis made of off-types (maternal types nal counted)

Ge..tk Aoaly,ls of Apomixi, 73the hypothesis of one tetraploid genedominant for apomixis, with all sexual parentsassigned genotype aaaa and all apomictsassigned genotype Aaaa.Brachiaria (Gramineae). Valle and coworkers(1991, 1992, 1993, and Chap. 10) conductedextensive studies of Brachiaria along the linesof Savidan's (1981) guineagrass program. Theresults pointed to a single dominant genedetermining apospory with genotypes of aanafor the colchicine-induced tetraploid sexualparents and Aaaa for the aposporoustetraploid parents.Bipolar AposporyRanunculus (buttercup, Ranunculaceae).Nogler (1984b) crossed diploid sexual R.cassubicifolius with tetraploid apomictic R.megacarpus (Figure 5.1). Four fertile triploidfacultatively aposporous progeny wereobtained and used to initiate three generationsof reciprocal backcrossing to the sexual diploid.Nogler deduced that apospory is caused by adominant factor A (designated A- in Nagler1984b), the wild allele of which (a) (designatedA+ in Nogler 1984b) does not enhanceapospory. Dominance of A is incomplete andadditive. Furthermore, A, when homozygousrnRanunculus cassuhicifolius = ex M = R. mcgacarpuslevel of fl deS"'" ofp polyploidy 2x aposory4xd' A A ••Iplant;;,obtainedBC,la l'JDB II hybrid: n-ndihaploid: n+O8 m hybrid: ::n + nmaternal: 2" .... 0aa [Aaa] aaa Aa41 aa aaa Aaa (mal.).....• trisomic hybrid: 1n=179'~9~9Q~~~~r "aa A a (mat.) a a A a a DIHAPlOIOS a a a A a a (m~t.) A a a a ao rn·~~·.·.·.·.. Da 0 Aa D U E"~13 0 , 5 D D 2 ~ 21 24 9 ·.... cc·7217" 1S 16 , ..... 1 I / " / 2942J& 23 + 5 veg. 66 7~~­~~II!c-3 0 c-=l ~ 0EJ c........=::! L....3 ., 61 0 , 16 0 J& 208 a A a a II A a (maL) A a aFigure 5.1 Genealogical tree of the cross Ranuncu!us cassubicilo!iu s =C, 2x =16, meiotic (sexual) x R.megacarpus =M. 4x =32, partially aposporous (Utotally" apomictic) and the different backcrosses withthe sexual parent C. The number of plants obtained, the level of polyploidy, the approximate degree ofapospory, and the genotype are indicated for each offspring.Reproduced with permission of the publisher, Birkhouser·Yerlog AG, Basel, Switzerland. From Nagler, G.A. 19B4. Genetics of opospory in apomicticRanunculus auricomus. Y. Conclusion. Botanica Helvetica 94: 411-422. Updated by G.A. Nagler (personal comm.) and Nogler (1995). See Nogler(I 984b) for details.

74 Rob.rt T. Sherwoodin the gametophyte, is lethal to thegametophyte; there are no functional A, AA, orAAA gametes because of recessive lethality.Gametes must carry the wild type a allele to beviable. Several lines of evidence point tolethality, including failure to find aposporousdiploid hybrid progeny. A highly aposporous211 + 1 aneuploid, line T, was believed to haveoriginated from an a egg and an Aa sperm andhave genotype Aaa. Line T transmitted a andAa gametes, but not A gametes. The data, forthe most part, did not permit testing ofsegregation ratios, but cross aa x Aaan and itsreciprocal did yield 1:1 ratios as expected. [ntrisomic ilpomictic lines assigned genotype Aaa,there was declining strength of apomixis withadvance from BC lto BC,. Nogler believed thismight be due to modifying genes for sexualitythat increased with each generation ofbackcrossing. Within generations there WilS adosage effect; for example genotype Aaashowed greater apospory than Aaaa. Someplants were identified as dihaploids withgenotype Aa and were highly aposporous.Dihaploids are formed by parthenogeneticdevelopment of reduced Aa eggs offacultatively apomictic AAaa or Aaaatetraploids; in other species, dihaploids usuallyare sexual (Nogler 1984a).Line T expressed low parthenogenicity. Nogler(1984b) postulated that parthenogenicity wascoded by a separate gene closely linked to A.Later tests revealed that apomictic line Ttransmitted genes for parthenogenicity.Therefore, it was not necessary to insist thatapomixis and parthenogenesis were coded byseparate genes (Nogler 1989, 1995). Nogler's(1984b) study encountered all the problemsanticipated from interspecific crossing atdifferent ploidy levels-poor fertility, poor seedset, facultative expression, aneuploidy, andambiguous segregation ratios-but succeededbecause of exceptional effort and insight.Paa pmtensis (Kentucky bluegrass, Poaceae).There are several inconclusive andcontradictory reports on inheritance ofapomixis in Kentucky bluegrass. Bluegrasspopulations have complex arrays ofpolyploidy and aneuploidy, with chromosomenumbers ranging continuously from 24 to 124.In addition, somatic cells within plants mayvary by as many as 30 chromosomes (Huff1992). Plants vary widely in the degree ofapomixis. Almgard (1966) concluded that thepresence of aposporous embryo sacs showeddominant inheritance, but retention of thematernal phenotype (functional apomixis)was recessive.Matzk (1991b) explored regulation of theparthenogenetic capacity of P pmtclIsis usingthe auxin test. He concluded that the apomicticparents were heterozygous for one or moredominant alleles for parthenogenesis. Theresults were consistent with expectations fora single major gene.Study of the genetic regulation of facultativeexpression (Nogler's modifying genes forsexuality) may be feasible using molecular andbiological probes developed for P. pmtel1sis(Huff and Bara 1993; Naumova et al. 1993;Mazzucato et al. 1995; Mazzucato et al. 1996).There is also an interesting report ofHieracium-type embryo-sac formation indiploid specimens of three Tribalill11l species(Poaceae) (Visser and Spies 1994).Hiemcil/m (hawkweed, Asteraceae). From across of diploid sexual H. allriCll/a x tetraploidapomictic H. allrl1lltillCIIIII. Christoff (19-l2)found 32 sexual and 27 apomictic progeny, aratio that fits the expected 1:1 if the male has asingle dominant allele for apomixis. Gadella(1987) believed results from a cross oftetraploid sexual H. pi/ascI/a with a pentaploidapomict could be explained by monogenicdominant inheritance with the pentaploid

Ge••ti< loal"is 01 Apomixi. 75having genotype Annnn. However, theobserved ra tio 36:annn:8Annn:5Annaa:2nannn didnot fit the expected ratio 3:2:3:2.Bicknell and Borst (1996) observedsegregation for sexuality and apomixis amongtetraploid regenerants of H. pilosel/n derivedfrom reduced calli of an apomictic biotype.They considered this evidence for dominantinheritance of apomixis. Bicknell (1994, andChap. 8); Koltunow et al. (1995) believeHierncilllli can be a model system for studyingmolecular genetics of apomixis.Mitotic DiplosporyMost reports on diplospory, summarizedbelow, were handicapped by lack of suitablesexual parents and by unavailability ofconvenient classification techniques.Eragrostis curvlIla (weeping Iovegrass,Poaceae). Crosses of naturally occurringtetraploid sexual plants with tetraploid andhexaploid apomicts gave F Iprogeny testsegrega tions ind ica ti ng tha t a pom ix is ismonogenic and dominant (Voigt and Burson1983). Results of a cross with an aneuploidindicated possible dosage effects, but mayhave been confounded by Bill hybridizationor chromosome elimination.Tripsacllm dactyloides (Eastern gamagrass,Poaceae). Sherman et al. (1991) crossed asexual diploid female parent with a triploidapomict. Forty-six hyperdiplOid progeny wereidentified as hybrids. All but two of theseshowed cytological indiCations of apomixis;the degree of diplospory ranged frompredominantly sexual to highly apomictic.The authors believed this indicated thatapomixis is incompletely dominant, or thatminor additive genes on variouschromosomes affect penetrance of apomixis.Recent production of sexual tetraploids bycolchicine doubling should facilitate futurestudy of inheritilnce. The tetraplOid T.dnctyloides parent genotype is simplex for thediplospory allele. Maize-TripsnclIlIl Fjssegregated 1:1 for mode of reproduction(Leblanc et al. 1995b).Partllcllillln argelltatllln (guayule,Asteracae). Gerstel and Mishanec (1950)reciprocilily crossed a sexual diploid (2/1 = 36)with il facultiltively diplospofOushyperdiploid (2/1 = 37). With the sexual plant,~s the milternal parent, all F] diploid hybridprogeny (21 plants) were sexual. In thereciprocal, most progeny, as expected, werematernal or polyploid apomicts, but fourdiploid sexual F1s were formed (Gerstel et al.1953). They concluded that apomixis genesacted recessively but additively andpostulated that polyploids with two ilpomixisgenomes and one sexual genome wereapomictic.There is an alternate interpretation. Thestarting materials and results with Pnrti/(:/Iilllllresemble those of NogleI' (1984b) foraneuploid plant T of Rail I IIICII Ills (Figure 5.1).The Parilic/lilllll results may indicate the samecontrol as in Ralill/lwllls, i.e., a single factordominant for apomixis, that acts as a recessivelethal, with the polyhaploid parent beinggenotype Ann and the diploids being nn.Restitutional DiplosporyTaraxacum (dandelion, Asteracae).Eutriploids (211 = 24) and many hypotriploids(2/1 = 23) have facultative diplospory. Certain2/1 = "23 aberrants are primarily sexual(Sorensen 1958). Mogie (1988) offered thefollowing interpretation of earlier studies.Expression of apomictic phenotype inTI1rt1XnCfllll depends on one or more geneslocated on one chromosome and on dos,lge.At least two copies of the mutant apomixisallele are required to obtain apomixis; theallele prevents meiosis in diplosporousapomicts. The dominance relationshipbetween the wild type and mut,mt allele isdetermined by balance and environmenl.

76 Rokrt T. SHtwoodPerhaps the most significant aspect of thehypothesis is.that Mogie deduced that the wildtype (a) allele of the apomixis locus has anessential function in the plant. He suggests thatit codes for meiotic reduction and that it is alsoinvolved in the control of mitosis, whichwould be disrupted by the expression of themutant allele in somatic cells.Multicellular ArchesporiaBeta (beet, Chenopodiaceae). At least twospecies in the section Corollinae formmulticellular archesporia that show bothdiplosporous and aposporous development.Jassem (1990) conducted extensive analyses ofcrosses involving sexual and apomicticspecies, across ploidy levels, including F 2andBC lgenerations. Bill hybridization andaneuploidy complicated the results. Althoughunable to draw unequivocal conclusions, shebelieved that apomixis genes were partlydominant and acted in a complementaryfashion.Sorbus (mountain ash, Rosaceae). Liljefors(1955) used leaf morphology and chromosomepairing to deduce genomic formulas ofaposporous polyploids in the agamic complexSorbllS. He assigned the genomic formula BBto totally sexual S. aucuparea. Species assignedgenomic formulae AAAA, AAAB, and AABwere fully aposporous, and AABB wasfacultative. He postulated that a gene or genesfor apomixis were associated with genome A,and that expression was dosage/balancerelated. However, species with genomicformulae ABBB and ABB were also highlyapomictic, and he had to postulate exchangeof the gene into the B genome. Liljefors'observations may be equally well explainedusing a one major-gene model with facultativeexpression, and postulating that genotypesAAnn, Annn, and Aan are aposporous.Towards a ComprehensiveModel of InheritanceInheritance of apomixis has been explored inrelatively few species, yet several geneticmodels have been put forward that differwidely in postulated number of loci andnature of gene action (Bashaw and Hanna1990; Asker and JerJing 1992; den Nijs andvan Dijk 1993). Is the seemingly capriciousoccurrence and regulation of apomixis indifferent taxa to be attributed to independent,random mutations at various reproductiveloci leading to similar phenotypicconsequences? Is there a single apomixislocus or linkage group shared by allapomicts? The reality lies somewherebetween these extremes.Regulation of Monopolar AposporyMonopolar (Panicurn-type) apospory occurscommonly throughout Panicoideae andArundinaceae and is found nowhere else.This indicates a common genetic basis. Brownand Emery (1958) postulated that coding forthe monopolar pattern arose early in theevolution of the group. Inheritance data forall the Panicoid species studied have beeninterpreted as indicating that expression ofapomixis requires a major locus, withapomixis behaving as a dominant trait.Common gene action and phenotype inrelated species indicate the same linkagegrouP. may be involved. The molecularevidence for one linkage group in Pennisetumis impressive. However, C. F. Crane (personalcomm.) cautions that this does not necessarilyimply that the monopolar type evolved onlyonce and spread laterally among relatedgenera that are well populated with sexualspecies. He considers it more likely that theancestor of the A-a locus hi,ls becomeWidespread in the panicoids and chloridoids,and that apospory has emerged repeatedlyby mutation of the wild type locus.

G...tk holy,1s OfApoonllls 77Savidan (1991a, 1992), Peacock (1993), andothers suggest that a single master gene isresponsible for induction of embryo-sacformation. They view induction as triggeringa cascade of events that requires direction bymany genes with a potential for modifying theend result. Savidan (1989) suggests that severalgenes controlling apomixis may be linked onone small chromosome segment or Iinkat.Jefferson (1993) suggests tha t apomixisinvolves phenotypic mutations at several lociacting together as a non-recombining unit andhaving the appearance of a single gene.Accordingly, classical genetic observations ofsegregations might be less informative thanexpected (Grimanelli et al. 1995, and Chap. 6).Some early studies on Panicoideae interpretedthe data as indicating not only a locus forapospory, but also a second, independent locusthat affected sexuality (Burton and Forbes1960, reinterpreted by Burton 1992; Taliaferroand Bashaw 1966). In the Taliaferro andBashaw (1966) report, the gene was postulatedto be epistatic to the apomixis allele. Hanna etal. (1973) proposed two loci with genes actingadditively to confer sexuality. Theseinterpretations followed observations thatselfing of apparently sexual parents gaveprogeny segregating for mode ofreproduction. The data sets from these threestudies could not be fitted to the singletetraploid apomixis gene model or to any othercited model (Sherwood et al. 1994). Savidan(1982b) noted some rare facultative genotypesof Pal1icum maximum with nearly 90%sexuality. Later, Savidan (1991a, b) proposeda technical explanation for the observation ofapomictic progeny when naturally occurringtetraploid sexual parents were selfed in theearlier studies. He believes the tetraploidparents were facultative apomicts ofgenotypeAaaa with a high frequency of sexualreproduction and that the parents weremistakenly classified sexual because of thelimitations of the sectioning and progenytesting techniques used at the time. However,selfing or crossing Aaaa parents should give51 progeny of 1:3, S:A (Figure 5.1), not thereported ratios of 2.5:1 or 13:3, so the matter isnot yet resolved.Alternatively, the data of Hanna et a!.,Taliaferro and Bashaw, and those of all otherstudies on Panicoideae can be accommodatedin one genetic model that postulates twotetrasomically inherited loci-the A locus witha gene dominant for apomixis (and recessivelylethal), and a Blocus with the dominant alleleepistatic to A (Sherwood et al. 1994).Regulation of DiplosporyRecent studies indicate that segregation fordiplospory in maize-Tripsacum progenyinvolves a single Mendelian factor, withperhaps some modifying factors (Leblanc etal. 1995b; Grimanelli et al. 1995; Savidan et al.1995). The factor appears to be an apomixislinkat. Savidan et at. (1995) stated they "maynow have a series of concrete reasons to believethat apomixis is indeed controlled bysomething more complex than this dominantgene, including at least one recessive factorwhich prevents apomixis expression indiploids." See also Grimanelli et al. (Chap. 6).Regulation of Facultative ExpressionAll apomictic species for which inheritancedata are available show facultative expression.DegreE!'of apomixis could be due to dosage orpenetrance effects of a major gene and / or tomodifying genes. Data for Ramll1culus (Nogler1984b) and Paspalum (Quarin 1986, 1992)indicate that penetrance of the A allele isincomplete; degree of apomixis may increasewith increased number ofA alleles. Quarin andHanna (1980) suggested that a certain geneticthreshold must be reached for apomixis to beexpressed in some Paspalttm. A single dose ofthe A allele is sufficient to support a high levelof apomixis in Pel111isetum (Sherwood et al.1994) and Pal1icum (Savidan 1981).

78 Robe" T. S~erwoodEnvironment plays a role in expression(Nogler 1984a; den Nijs and van Dijk 1993).A short photoperiod increases the frequencyof aposporous vs. sexual embryo sacs inDichal1thil/m aristattlm (Knox 1967) andPaspall/ni chromyorrhizol1 (Quarin 1986). Saltstress affects facultative expression in P ciliare(Gounaris et a!. 1991).Harlan et a!. (1964) accurately assessed therelation between sexual and aposporousreproduction. "Apomixis (read apospory)and sexual reproduction are not alternativemodes of reproduction, either genetically oroperationally, but are simultaneous andindependent phenomena. The genescontrolling normal sexual reproduction arenot allelic to those controlling apomixis in theconventional sense." This accounts forfacultative expression of apospory. Inaposporous lines, meiotic reduction of thearchesporial nucleus and apomeioticinduction of apospory in nucellar cells goforward at about the same time. Aposporousini tials and umed uced embryo sacs normallycrowd out the red uced megaspores and sacs.Facultatively displosporous plants alsopossess all of the genetic information requiredfor completion of both meiotic andapomeiotic embryo sacs. However, in contrastto facultative apospory, the two eventscannotproceed simultaneously in a facultativelydiplosporous ovule, for they compete for thesame site in the ovule. Events beginning inthe megaspore mother cell can proceed onlytowards normal meiosis or apomeiosis, butnot both. Variability within facultativediplosporous or aposporous types indicatesthat the entire apomictic developmentalprocess cannot be explained on the basis of asingle gene (Grimanelli et a!. 1995; Savidanet a!. 1995).The lethal Gene as the Basis forHeterozygosityNogler (1984b) concluded that functionalgametes of Rammcl/lus contain a copy of thewild type a allele. Noirot (1993) reviewedevidence that the A allele may act adverselyin Pal1iCllm maximum; his report focused onmale and female sterility. Male and femalesterility is encountered in spontaneousdihaploids and trihaploids of Pal1icl/n1 ofputative genotypes Aa and Aaa, respectively(Combes 1975). Mogie (1988) proposed that thewild typea allele has a function that is essentialto normal plant processes. If gametophytesand gametes bearing dominant allele A mustalso bear the wild type a allele to remainfunctional, this would account for theobservation (H arlan et aI. 1964) tha taposporous apomicts invariably areheterozygous at the apomixis locus. Noapomict has ever transmitted an exclusivecapacity for apomixis to the offspring; acapacity for sexual reproduction is always alsotransmitted, al though it may not surface untillater generations. Heterozygous Aa gametescan be formed by dihaploids, triploids, ortetraploids. In the case of the dihaploid, theAa gamete'is from an unreduced (apomictic)sac, and the progeny are either maternal (nofertilization) or Bill hybrids (fertilization), asshown for the dihaploids of Ralll/I1Cl1Il/s(Figure 5.1). In the case of triploids andtetraploids, the Aa gamete occurs in sexuallyreduced (meiotic) sacs, and the egg is usuallyfertilized (B IIhybridization), but occasionallymay parthenogenetically form an Aadihaploid as part of a diploidtetraploid-dihaploidcycle (de Wet and Harlan1970; Savidan and Pernes 1982). Insofar assexual transmission of A is concerned, theparent must be polyploid and heterozygous;for asexual transmission, the parent must beheterozygous.

Geoelk Ao.lysis of Apomi.is 79The results of the maize-TripsoCllnJ Summarybackcrossing effort conform with the idea of a The results of all published studies onrecessive lethal factor linked to a dominant inheritance of apospory are compatible withdiplospory gene (Savidan et al. 1995). No the hypothesis that expression of aposporyapomictic representatives were found in Be 2requires the dominant allele of a major geneprogeny with 211 = 38= 20M+18T chro­ or Iinkat. There is no species for which themosomes produced from F of 211 = 56 hypothesis of a major dominant apospory1=20M+36T chromosomes. See also Grimanelli factor can be ruled out. Limited data availableet al. (Chap. 6) for discussion.for diplosporous species also suggest a singlemaster gene or linkat (Savidan 1989); itsRichards (1996) postulates that lethal recessivepOSSible correspondence to the aposporymutants might occur at loci linked with thelinkat is entirely unknown. Some reportsdominant apomixis locus and accumulate inindicate that degree of expression is regulatedheterozygotic apomictic lines without beingby modifying genes that promote sexuality;expressed. These harmful recessive genes maythe apomixis gene(s) may show dosage effects,be expressed in haploid gametophytes (i.e.,and environment may influence expression.pollen and embryo sacs) resulting in theirSeveral lines of evidence indicate that the wildabortion. The implied corollary is that thetype allele of the apomixis gene plays anapomixis allele is not, of itself, the source ofessential role in cell function and that itsrecessive lethality. The evidence from studiespresence is required for survival of theby Nogler (1984b), Mogie (1988), Sherwood etgametophyte. There is no clear evidence for aal. (1994), and others still points to themajor gene for parthenogenesis independentapomixis gene as being recessively lethal.of the major gene(s) for apospory orSegregation ratios from the facultativediplospory (Nogler 1984a, 1995; Leblanc et al.populations cited above show an absolute lack1995b; Mazzucato 1996). Recombinationof recombination of recessive lethality vis a viswithin the linkat is rare or nonexistent. Thethe apomixis linkat on the male side, but notlinkat encodes sufficient information to directon the female side.the complex sequence of developmentalevents in embryo sac initiation, differentiation,maturation, and function.--.1980. Gametaphytic apomixis: Bashaw, E.C., and KW. Hignight. 1990. GeneReferenceselements and genetic regulation. Hereditas transfer in apomiclic buffelgrass throughAlmgard, G. 1966. Experiments with Poa. III. 93: 277-93. , fertilization of an unreduced egg. Crop Sci.Further studies of Poa longifolia Trin. withAsker, S.E., and LJeding. 1992. Apomixis in 30: 571-75.special reference to its crass with Poa Plan/s. Boca Raton, Florida: CRC Press. Bashaw, E.C., and E.C. Halt. 1956.pratensis L Lantbrukshogskolan Annales 8ashaw, E.C. 1962. Apomixis and sexuality in Megasporogenesis, embryo soc~2: l--64. buffelgrass. (rap Sci. 2: 412-15. development and embryogenesis inAsker, S.E. 1967. Induced sexuality aher--.1980. Apomixis and its application dallisgrass, Paspalum dilatatum, Poir.chronmome doubling in an apomiclic in crap improvement.. In W.R. Fehr and Agran.1. 50: 75l--56.Polentilla argenteo-biatype. Hereditas 57: H.H. Hadley leds.), Hybridization of (rop Bashaw, E.C., A.W. Havin, and E.C. Halt. 1970.339-42. Plants. Madison, Wisconsin: ASA and CSSA. Apomixis, ils evolutionary signi~cance and--.19700. Apomixis and sexuality in thePp.45-63utilization in plant breeding. Proe. I'th Int.Potentilla argentea complex. I. Crosses withBashaw, E.c., and w.w. Hanna. 1990. Apomiclic Grassl. (ongr. Pp. 245-48.ather spe

80 Robert T. SherwoodBanaglio, E. 1963. Apomixis. In P. MlIheshwari(ed.), Recent Advances in Ihe Embryologyof Angiosperms. Delhi, Indio: Int. Soc. ofPlant Morphologists, Univ. of Delhi. Pp.221-64.Bayer, RJ., K. Rilland, and B.G. Purdy. 1990.Evidence af portial apamixis in Anlennariamedia IAsteraceoe:lnuleoel detected by thesegregation of genetic markers). Arner. 1.801. 77: 107S-B1Bicknell, R. 1994. Hieraciunr. Amodel systemfor studying the molecular genetics afopomixis. Apomixis Newsletter 7: 8-10.Bicknell, R.A., and N.K. Bo~t. 1996. Isalatian ofreduced genatypes of Hieracium pi/osellausing anather culture. P10nl Cell, Tissue andOrgan Culture 45: 37-41.Brown, W.V., and W.H.P. Emery. 1958. Apomixisin the Gramineae: Panicaideae. Arner. J.801.45: 253--63.Burson, B.L 1985. Cytalogy of Paspolumchacoence and P. durifolium and theirrelationship ta P. dilalalum. Bal. Gal. 146:124-29.--.1992. Cytalogy and reproductivebehaviar of hybrids between Paspolumurvillei and two hexaploid P. dilalalumbiotypes. Genome 35: ]002~6.Burton, G.w. 1992. Manipulating apomixis inPaspa/um. In J.H.E1gin and lP. Miksche(eds.), Proc. Apomixis Workshop, February11-12,1992, Atlanta, GA. USDA, ARS.ARS-] 04. Pp.I6-19.Burton, G.w., and I. Farbes. 1960. The geneticsand manipulatian of obligate apamixis incomman Bahia grass (Paspolum nolalumAuggel. PrOf. 8th Inl. Grassl. Congr.Pp.66-71.Burton, G.w., and W.W. Hanna. 1986. Bahiagrass (Paspalum notalum) tetra plaidsproduced by making (apamictic tetraploidxdiploid) xdiploid hybrids. Crop xi. 26:1254-56.--.1992. Using apomictic tetraploids tomake aself-incompatible diploid PensocalaBahiagrass dane set seed. 1. Hered. 83:30~6.Carman, l 1997. Asynchranous expression afduplicote genes in angiosperms may couseapomixis, bispary, tetraspory, andpolyembryony. Bioi. 1. of Ihe Unn. Soc. 61:51-94.Carman, lG., c.F. Crane, and O. Riera-lizarazu.1991. Comparative histology of cell wallsduring meotic and apomeioticmegasporogenesis in twa hexapklid.Australian Bymus species. Crop 5ci. 31:1527-32.Christoff, M. 1942. Die genetische grundlage derapomiktischen Fortpflanzung bei Hierociumauranliacum Zeitschrih fur IndudiveAbstammungs-und Vererbungslehre 80:101Combes, D. 1975. Po~morphisme et modes dereproduction dons la section des Maximaedu genre Ponicum (Gramineesl en Afrique.Memaires ORSTOM 77: 1-99.Crane, CF. 1978. Apomixis and crassinginrompatibilities in some Zephyronlhaceoe.Ph.D. thesis, Univ. of Texos, Austin, Texas.--.1992. Arationale for theinvestigation of certain wild apomicts. InElgin, lH., and lP. Miksche (eds.) Proc.Apomixis Workshop, February 11-2, 1992,Atlanta, Georgia. USDA, ARS. ARS-l 04. Pp.5S-61.Crane, CJ., and lG. Carman. 1987. Mechanismsof apomixis in Elymlis rectisetus fromeastern Australia and New Zealand. Amer.1. 80t. 74: 477-96.Cruz, R., lW. Miles, W. Roca, and H. Ramirez.1989. Apomixis and sexuality in Brochiorio.I. Biochemical studies. Cuban 1. Agric. Sci.23: 317-21.den Nijs, A.P.M. 1990. Experimenting withapomixis and sexuality in Poa prolensis.Apomixis Newsletter 2: 52-54.den Nijs, A.P.M. and G.E. van Dijk 1991Apomixis.. In M.D. Hayward, N.O.Basemark, and I. Romagasa (eds.), P10nlBreeding: Principles and Prospedl. landan:Chapman and Hall. Pp. 229-45de Wet, lMJ., and JR. Harlan. 1970. Apomixis,po~plaidy, and speciation in Dichonlhium.Evolution 24: 270-77.Dujardin, M., and W.W. Hanna. 1989.Developing apomictic peorl milletcharacterizationof a B~ plant. J. Genel.Breed. 43: 145-51.fISher, W.O., E.c. Bashaw, and E.C. Holt. 1954.Evidence for apamixis in Penniselum ciliareand Cenchrus seligerus. Agron. J. 46: 401­04. GGadella,IWJ. 1987. Sexual tetra plaid andapomictic pentaploid populatians ofHieracium pi/Ofelia (Compositae). PI. Syst.Evol. 157: 219-46.Gerstel, D.U., B.l. Hammond, and C. KiM 1953.An additional note on the inheritance alapomixis in guayule. Bal. Gal. 1J5: 89­93.Gerstel. D.U., and W. Mishanec. 1950. On theinherirance af apomixis in Pof1heniumargenlolum. Bol. Gal. 112: %-106.Gaunaris, E.K., R.T. Sherwood, I. Gounaris, R.H.Hamiltan, and D.l. Gustine. 1991.Inorganic salts modify embrya socdevelapment in sexual and aposporausCenchrus ci/iaris. Sex. Plant ReprorJ. 4:18S-92.Grimanelli, D., a.leblanc, D. Gonzalez deleon,and Y. Savidan. 1995. Mapping apomixisin tetraploid Tripsocum, preliminary results.Apomixis Newsletter B: 37-39.Gustafssan, A. 1946-1947. Apomixis in higherplants. I. The mechanism of apomixis.lands Univ., Msskr. NJ. Mv. 242(3):1­66. II. The causal aspect af apomixis. Mv.243(2):69-182. III. Biotype and speciesformation. Mv. 243(12): 183-370.Gustine, D.L, R.T. Sherwood, and D.R. Huff.1997. Apaspory-linked molecular markersin buffelgrass. Crop Sci 37: 947-51.Gustine, D.L, R.T. Sherwood, Y. Gounaris, andD.R. Huff. 1996. Isozyme, protein, andRAPD markers within ahalf-sib lami~ ofbuffelgrass segregating far apo

G...H, Analy.i. of Apamlxl. 81Huff, R.D., and 1M. Bera. 1991 Determininggenetic origins of aberrant progeny fromfacultative apomictic Kentucky bluegrassusing acombination of flow cytometry andsilver-stained RAPD markers. Theor. Appl.Genet. 87: 201~B.Jassem, B. 1990. Apomixis in the genus Beta.Apomixis Newsletter 2: 7-21Jefferson, R.A. 1991 Strategic development ofapomixis as a general tool for agriculture..In KJ. Wilson led.), Prae.lnternationolWorkshop on Apomixis in Rice, Jan. 13­15, 1992, Changsha, China. Canberra,Australia: CAMBIA. pp. 206-17.Knox, R.B. 1967. Apomixis: seasonal andpopulation differences in a gras~. Science157: 325-26.Koltunow, A.M., R.A. Bicknell, and A.M.Choudhury. 1995. Apomixi~: molecularstrategies for the generation of genetical~identical seed~ withoul fertilization. PlontPhysiol. lOB: 1345-52.Kojima, A., and T. Kawaguchi. 1989. Apomicticnature of Chinese chive (Allium tuberosumRollI.) detected by unpollinated ovuleculture. lapJ.Breeding 39: 449-56.Kojima, A., Y. Nagato, and K. Hinala. 1991.Degree of apomixis in Chinese chive(Allium tuberasum) estimated by esteraseisozyme ana~i~. lop. 1. Breeding 41: 73­B4.leblanc, 0., M. Dueiias, M. Hernandez, S. Bello,V. Garcia, 1 Berthoud, and Y. Savidan.19950. Chromosome doubling inTripsacum: the production of artifidal,sexual tetraploid plants. Plont Breeding114: 226-30.leblanc, 0., D. Grimonelli, D. Gonzalez de lean,ond Y. Savidan. 1995b. Detection of theapomictic mode of reproduction in maize­Tripsocum hybrids using maize RFlPmarkers. Theor. Appl. Genel. 90: 1198­1201leblanc, 0., M.D. Peel, 1G. Carman, and Y.Savidan. 1995e. Megasporogenesis andmegagamelogenesis in several Tripsocumspecies (Poaceael. Amer. 1. Bot. B2: 57­61Liljefors, A. 1955. Cytological studies in Sorbus.Ada Holti Bergianil7: 46-113.lubbers, E.L, l. Arthur, w.w. Hanna, and P.Ozias-Akins. 1994. Molecular markersshared by diverse apomictic Pennisetumspecies. Thear. Appl. Genet. 89: 636-42.Marshall, D.R., and R.w. Downes. 1977. Atestfor obligate apomixis in grain sorghumR471 Euphytica26: 661-64.Matzk, F. 19B9. Genetic studies onparthenogenesis in Poo pro/ensis LApomixis Newsletter 1: 32-34.--.1991 a. Anovel approoch 10differentiated embryos in the absence ofendosperm. Sex. Plant Reprod. 4: B8-94.---. 1991 b. New efforts 10 overcomeapomixis in Poa pratensis L Euphytica 55:65-72.Mazzucato, A. 1996. Which genets) are welooking for? Apomixis Newslefter 9: 7.Mazzucato, A., G. Beracda, M. Pezzotti, and M.Falcinelli. 1995. Biochemical and molewlarmarkers for investigating the mode ofreproduction in the facultative apomict Poapratensis LSex. Plant Reprod. B: 133-38.Mazzucato, A., A.P.M. den Nij~, and M. Falcinelli1996. Eslimation of parthenogensisfrequency in Kentucky bluegrass withauxin·induced parthenocarpic seeds. ClopSci. 36: 9-16.Mazzucato, A., M. Wagenvoort, and A.P.M. denNijs 1994. Flow cytometric ana~ses toestimate the mode of reproduction in Poaprotensis LApomixis Newslefter 7: 22-24.Miles, T.W., and C.B. do Valle. 1991. Assessmentof reproductive behavior of interspecificBrachiaria hybrids. Apomixis Newslefter 3:9-10.Mogie, M. 1988. Amodel for the evolution andcantrol of generative apomixis. Bioi. 1.linn. Soc. 35: 127-51Nakajima, K., and N. Mochizuki. 1983. Degreesof sexuality in sexual plants of guineagrass by the simplified embryo socana~is. lap. l. Breeding 33: 45-54.Naumova, T., A.P.M. den Nij~, nnd M.T.M.Willemse. 1993. Quantitative anal~is ofaposporous parthenogenesis in Poapratensis genotypes. Ada Bot. Neer/ond.42: 299-312.Nogler, G.A. 19840. Gametophytic apomixis.. InB.M. Jahri led.), Embryology ofAngiosperms. Berlin: Springer-Verlag.Pp.475-51 B.---. 19B4b. Genetics of apospory inapomictic Ranuncu/us auricamus: V.Conclusion. Bot. Helvetica 94: 41 1-22.--.1989.Cytogenetics ofparthenogensis -/irst results onRanunculus auricomus. ApomixisNewsletter 1:44-47.--.1990 Simplified methods forembryologicol studies. Apomixis Newslefter2: 56-58.---. 1995. Genetics of apomixis inRanuncu/us auricomus. VI. Epilogue. Bot.Helvetica 105: 111-15.Noirot, M. 1993. Allelic ratios and sterility in theagamic complex of the MaximaeIPanicoideae): evolutionory role of theresidual sexuality. 1. Evol. Bioi. 6: 95-101.Ozias-Akins, P., E.L lubbe~, W.W. Hanna, and1W. McNay. 1991 Transmission of theapomictic mode of reproduction inPennisetum: co-inheritance of the trait andmolecular markers. Theor. Appl. Genel. 85:632-38.Peacock, WJ. 1993. Genetic engineering andmulagenesis for apomixis in rice. In KJ.Wilson led.), Pro

82 Robe" 1. Sherwood---. 1989. Apomixis in plont breeding:tronsfer vs. synthesis. Apomixis News/eNer1: 22-24.---.19900. The genetic control ofopomixis. Apomixis NewsleNer 2: 24-27.---. 1990b. Commentory on thepreceding contribution. ApomixisNewsletter 2: 51.---.19910. La critique est oisee mois I'artest difficile. Apomixis NewsleNer 1 26-27.--. 1991 bDo we need a c1assificotion ofapomixis? Apomixis NewsleNerl 27-29.---. 1992. Genetic control and screeningtools for apomixis. Apomixis NewsleNer 5:16-19.Savidan, Y.H., D. Grimanelli, and O. leblanc.1995. Apomixis expression in maize­Tripsawm hybrid derivatives and theimplicotions regarding its control andpotential for manipulation. ApomixisNewsletter 8: 35-37.Savidan, Y.H, l. Jank, and C8. do Valle. 1989.Breeding Panicum maximum in Brazil. 1.Genetic resources, modes of reproductionand breeding procedures. Euphytica 41:107-12.Sovidan, Y.H., and J. Pernes. 19B2.Diploid·tetraploid-dihaploid cycles and theevolution of Poniwm moximum Jacq.Evo/u/ion 36: 596-600.Sherman, R.A., PW. Voigt, B.L Burson, and CLDewald. 1991. Apomixis in diploid xtriploidTripsowm daelyloides hybrids. Genome 34:528--32.Sherwood, R.T. 1995. Nuclear DNA amountduring sporogenesis and gametogenesis insexual and apasporous buffelgrass. Sex.Plan/ Reprod. 8: 85-90.Sherwood, R.T., CC Berg, and B.A. Young. 1994.Inheritonce of apospory in buffelgross. CropSci. 34: 1490-94.Sherwood, R.T., 8.A. Young, and E.C Bashaw.1980. Facuhalive apomixis in buffelgrass.Crop Sci. 20: 375-79.Snyder, LA., A.R. Hernandez, and H.E. Warmke.1955. The mechanism of apomixis inPennise/um ciliare. Bo/. G01. 116: 209-21.Sorensen, T. 1958. Sexual chromasomeaOOrranlsin Iriploid opomictic Taroxaca.Bo/anisk TidsskriFt 54: 1-22.Stebbins, G.L 1941. Apomixis in theangiosperms. Bo/. Rev. 7: 507-42.--. 1950. Varia/ion and Evo/u/ion inPlants. New York: Columbia Univ. Press.Taliaferro, CM., ond E.C Bashaw. 1966Inheritonce and conlrol 01 obligateapomixis in breeding buflelgross,Pennise/um ciliare. Crop Sci. 6: 473-76.Volle, CB. do, and CGlienke. 1993. Towardsdefining Ihe inheritance of opomixis inBrachiaria. Apamixis NewsleNer 6: 24-25.Valle, CB. do, G. leguizamon, and N.R. Guedes.1991. Interspecific hybrids of Brochiorio(Gramineael. Apomixis NewsleNerl lOll.Valle, CB. do, and J.W. Miles. 1992. Breeding ofapomictic species. Apomixis NewsleNer 5:37-47.Visser, N.C, and J.J. Spies. 1994. Cytogeneticstudies in the genus Tribolium (Poaceae:Danthonieael. II. Areport on embryo \ocdevelopment, with speciol reference to theoccurrence of opomixis in diploidspecimens. South Afrir. 1. Bo/. 60: 22-26.Voigt, Pw., and E.C Bashaw. 1972. Apomixisand sexuality in Eragros/is wrvulo. CropSci. 12: 843-47Voigt, PW., and B.L Burson. 1983. Breeding ofapomictic &agros/is wrvula. Pror. 14/h In/.Grassl. Congr. Pp.160-63.--. '992. Apomixis in Erogros/is.. In lH.8gin and lP Miksche leds.), Prar.Apomixis Workshop, February 11-12,1992, Atlonto, Georgio. USDA, ARS. ARS­104. Pp. 8--11.Young, BA, R.T. Sherwood, and E.C Bashaw1979. Geored·pistil ond thick·sectioningtechniques for detecting opasporousapomixis in grosses. Can. 1. Bo/. 57:1668--72.Yudin, BJ 1994. Towords progeny·tests osrecognition tool of apomicts. ApomixisNewsletter 7: 10-12.

Chapter 6Applications of Molecular Genetics inApomixis ResearchDANIEL GRIMANELlI, JOE TOHME, AND DIEGO GONzALEZ-DE-LEONIntroductionApomixis in higher plants refers to a widerange of mechanisms of asexual reproductionthrough seeds (Nogler 1984a). It is found in atleast 400 wild species belonging to 35 higherplant families (Richards 1986; Asker andJerling 1992; Carman 1997). The modalities ofapomictic development in the wild are nearlyas diverse as the number of species studied,but in most cases, apomictic processescompletely bypass meiosis and egg cellfertilization, and produce offspring that areexact genetic replicas of the mother plant.Two major types of gametophytic apomixishave been described, namely diplosporousapomixis and aposporous apomixis, based onthe origin of the megagametophytes. Inaposporous apomicts, one or more unreducedfemale gametophytes form mitotically fromsomatic nucellar cells while the legitimatesexual line generally aborts. Diplospory resultsfrom meiotic failure in megasporocytes thatdirectly develop into mature unreducedfemale gametophytes through three or moremitoses. Typically, apomixis is a facultativephenomenon, and an apomictic plant usuallyproduces both asexually (apome;otic) andsexually derived embryos.Apomixis, for the most part, is found in wildspecies. In contrast, major crop plants aresexual, with only rare exceptions such as someprominent tropical forages. This could beconceived partly as a consequence of cropdomestication because the process necessarilyimplies that early farmers had access tovariability and segregation among the wildtypes. In modern agriculture, however, theability to fix superior genotypes throughgenerations would offer numerousadvantages. Recognition of these advantageshas led to a growing interest in apomixisresearch, and indeed, many scientists haveextolled the tremendous potential thatapomixis holds for plant improvement (thisvolume; Jefferson and Bicknell 1996;Grossniklauss et al. 1999; Savidan 2000).Various strategies are being considered by agrowing number of research groups aroundthe world to introduce apomixis into majorfood crops. The oldest efforts were directedtoward the introgression of the genes forapomixis from wild species into cultivatedrelatives (see review by Savidan 2000). As analternative approach, the de novo synthesis ofapomixis in sexual plants through geneticengineering is now underway through anumber of intiatives Oefferson and Bicknell1996; Grossniklaus et al. 1999; Luo et al. 2000).Despite this growing interest, surprisingly littleis known about the biology of apomictic plants.This is certainly the primary reason whyattempts to manipulate apomixis have failedto yield useful products to date, and it is clearthat harnessing the potential of apomixis willstrongly depend on our ability to develop areliable understanding of the basic features ofthe biological processes of apomixis and itsgenetic control. The emergence of powerful

84 Dcaitl ~11i-, lot To...., .od Diego Goo,"',-

Ap,IicaIiH••f Molo

86 Do.lel Griome/I-, Joe To~ ...., ood Diego GOllcil.,-

AppW.liol. 01 Mole

88 Oaoitl G

ApfI

90 o.lel ~II-, Jo. lob.., aod Diego Gaucile,......le••Cloning the Apomixis Gene(s) UsingMolecular Genetics ToolsA major difficulty encountered by thoseinterested in cloning "apomixis genes" issimply defining what they are. Introducingapomixis into crops implies that specific genesare transferred or altered and expressed in thetarget crops. Most likely, not all of the genesinvolved in the apomictic process should betargeted: most, if not all, of them shouldalready be present and playing a role insexually reproducing plants. The issue then iswhich alleles of pertinent genes must betransmitted or manipulated for the inductionand successful development of apomicticembryos and seeds. To date, all efforts to tagapomixis genes, including those presented inthis paper, have focused on the mechanism ofnonreduction, mainly because it is an excellentindicator of apomictic development and it isprobably the easiest one to score. Nevertheless,it should be remembered that apomixis isprobably more complex than the Simpleprocess of nonreduction. The importance ofthis constraint will likely emerge whenattempts are made to synthesize de novoapomicts in sexual organisms"Map-based" cloning in apomictic species.Once a gene has been located on a genetic map,subsequent efforts to specify its position canultimately lead to its isolation (for the firstsu~cessful efforts in plants, see Giraudat et al.1992; Martin et al. 1994). The recentdevelopment of powerful new approaches forphysically mapping chromosome segmentscombined with the ability to clone large DNAfragments (Burke et al. 1987; Shizuya et al.1992), and progress in genome sequencingtechniques have created new and higherstandards for positional cloning in plants. It isstill a laborious and risky task outside of a fewwell-characterized model genomes, but thenumber of genes cloned in this manner arerapidly increasing. However, positionalcloning for apomixis is not very promisingbecause most, if not all, of the candidatespecies for a map-based cloning project arehighly heterozygous tetraploids, for whichlittle genomic characterization exists.Furthermore, when attempting positionalcloning, the first step is to identify achromosomal region, defined by two or moremolecular markers, that flanks the gene understudy. The precision of the estimated positionof the gene is therefore limited by the smallestmeasurable recombination unit, meaning onerecombinant in a given mapping population.Hence, the recombination level around theapomixis gene(s) presents another significantchallenge: positional cloning will proveefficient only insofar as recombination can beobserved near the locus of interest. Asmentioned earlier, recombination near theapomictic alleles is very likely restricted, atleast in Pennisetum and Tripsacum.Consequently, the smallest recombination unitdefined by two markers that encompasses theapomixis locus might well be a relatively largeamount of DNA.Transposon tagging of apomixis genes. Somemodel plants, such as maize, rice, tomato,Arabidopsis, and Petunia have undergoneextensive genome characterization. Specificapproaches are available for gene taggingthese plants that might be considered fortagging apomixis gene(s), provided thatcomponents of apomixis occur in one of theseorganisms.A very promising approach is that oftransposon tagging. Transposable elementsare short DNA sequences that have theproperty to transpose to more or less randomlocations in the genome (see Walbot 1992, fora review). They were discovered in maize, buthave since been identified or introduced invery diverse organisms. They have been usedin a wide range of genetic studies, and havebeen found to be highly effective for genetagging and cloning.

App/kaliold of Mol...lor Ge..tks I. Aporixls R....... 91Transposon tagging in apomicts presentssomeconstraints, including access to transposableelements and the genetic control of the trait.To the best of our knowledge, transposonactivity has not been demonstrated inapomictic species. This might be overcome byintroducing functional transposable elementsinto apomicts, either through transformation(as in Hieracium, Bicknell, Chap. 8) or throughhybridization with a close relative (as withmaize and Tripsacum, Grimanelli 1997). In bothcases, maize transposable elements weresuccessfully introduced into an apomicticbackground, and transposable activity wasdemonstrated.In our view, the main issue concerningtransposon tagging of apomixis is geneticcontrol of the trait. While this approach isefficient for phenotypes controlled by singlegenes, it might yield no, or disappointing,results if apomixis is genetically more complex.But taken further, it would at least provide anelegant method to determine whetherapomixis is controlled by one or several genes:if a single allele controls the trait, then a singlemutation should allow complete reversion tosexuality; if a more complex system isinvolved, then individual mutations shouldlead to abnormal or only partial expression ofthe trait.Candidate gene approaches. Althoughapomixis is unknown in major crop plants orother genetically well-characterizedorganisms, useful information can be derivedfrom detailed analyses of the reproductionprocesses of select sexual organisms. Forexample, genes involved in the control ofovuledevelopment, the initiation of meiosis,embryogenesis, and endosperm developmenthave been described in various organisms, anda close look at these genes might provideuseful information about the regulation ofapomixis. Such genes, but not necessarily theirrespective alleles, might represent prospective"candidates" for the apomixis gene(s), i.e., thegene(s) that would code for identicalfunctions as their apomicitic counterparts.The best, though not the only, candidates arethe yeast genes responsible for the inductionof meiosis and the meiotic mutants identifiedin higher plants.Major biochemical pathways involved in theregulation of the cell cycle and meiosis appearto be relatively well conserved betweendistant organisms such as yeast and higherplants, and the advance of whole-genomesequencing puts provides complete catalogsof putative candidate genes. This progressoffers great promise, but it is tempered by thefact that it is usually difficult to verifywhether a yeast gene of known function playsa similar role in plants. One powerful way tocorroborate such gene functions is the socalled "reversegenetics" strategy, using eitherinsertional mutagenesis or homologousrecombination. When based on transposon orT-DNA insertions, reverse genetics (or sitespecifictransposon mutagenesis) implies thattransposon tagging is performed to identifyindividuals carrying a transposon insertionin a gene of known sequence. The expectedfunction of that given gene can then becorroborated by confirming that its disruptionleads to the loss or alteration of the expectedfunction. Powerful reverse-genetic systemsare av'ailable in various plant species,including maize, Arabidapsis, and tomato.A specific candidate gene strategy based oncomparative mapping can also be undertakenwithin the grass family. The identification oforthologous genes between species (i.e., genesthat diverged from a common gene at the timethat the species harboring them diverged)could be used to understand the relationshipsbetween the genes responsible for variousc0mponents of apomixis in apomictic plants,

and meiotic or developmental mutants that arewell characterized in sexual plants. Numerousmutants are known in grasses, especially inmaize (Neuffer et a!. 1997), for various aspectsof sexual reproduction. Furthermore, largenumbers of such mutants can be generatedthrough classical (e.g., chemical) or transposonmutagenesis. Recent results of comparativemapping among grasses (Bennetzen andFreeling 1993; Ahn and Tanksley 1993; Mooreel a!. 1995) demonstrate that most grassesprobably share the same basic set of genes, andthat the obvious differences separating thespecies are based on allelic variations and noton their relative gene combinations..Therefore, we suggest that the genes whoseactions produce an apomictic phenotype insome grasses almost certainly can be found insexual species. In this instance, comparativemapping could be used to identify genes inmaize or some other sexual grass that areorthologous to the apomixis genes, and thenuse them to isolate their counterparts in theapomictic species.The process of identifying maize orthologs ofgenes responsible for apomixis involves threesuccessive steps: (i) candidate genes areidentified through phenotypic characterizationand genetic mapping; (iJ) promisingcandidates are then isolated in maize; oncecloned, lhe isolated genes are sequenced, andlhe sequence information is used to cloneorthologous genes in the apomicts; (iii) therelationship between the alleles isolated in theprevious steps and the expression of apomixisis confirmed using a reverse genetic strategyin apomictic plants. For step iii, theconstruction of transposon tagging populationsin apomicls are of great interest to R.Bicknell and the CIMMYT apomixis team.Three criteria can be employed to selectcandidate genes: (i) because apomixis oftenaffects only the female function, we proposethat the gene(s) responsible for the failure ofmeiosis have a megasporogenesis-specificphenotype, meaning that mutants of interestshould affect only the female function; (ii) asin diplosporous plants, the candidates shouldaffect early stages of meiosis, ideally, theinduction of meiosis; meiotic mutations actingat later stages in meiosis are probably notdirectly related to apomixis; and (iii)interesting candidates should be able toproduce unreduced gametes, (thus, as inapomictic plants, the completion of unreducedgamete formation implies that the checkpoints(Hartwell and Weinert 1989), which usuallyact during the meiotic cell-cycle to ensure theproduction of normally reduced haploidgametes, failed to override abnormal behavior.With aposporous-Iike mutants, obviousphenotypes relate to the induction ofmegagametogenesis in somatic cell. Sheridanet al. (1996) describe a remarkable example ofthis type of mutant.Manipulation of gene expression in modelspecies: To date, this is probably the mostwidely used approach for developingapomictic cultivars, (details are discussedelsewhere in this volume). Current workcenters on large-scale mutant screening inArabidopsis and Petunia Gefferson and Bicknell1996; Ohad et a!. 1996; Chaudurhy et al. 1997;Grossniklaus et al 1999; Luo el al. 2000}. Thebest prospecl from these approaches would bethe engineering of a mode of apomixis thatbetterCmeets the requirements of agriculturalproduction than the apomixis mechanismsfound in the wild (see Jefferson and Bicknell1996, and Chap. 8). The remarkable resultsobtained recently with a sel of mutations inPolycomb-related genes in Arabidopsis(Grossniklaus et a!. 1998; Luo et a!. 2000) arevery encouraging. They demonstrate thatphenotypes related to apomixis mayeventually be obtained by manipulating theexpression of genes involved in sexualreproduction, without reference to apomixisas seen in the wild.

Applicallo.,.f MoItc.Ia, Ge••10 i. Apomixi1 R....,do 93ConclusionsOur understanding of the genetics of apomixisis changing rapidly, from the idea that a simplegenetic system might control the wholedevelopmental process, to a more integratedconception and sophisticated models. Part ofthat evolution stems from the application ofmolecular genetic technologies to the study ofapomixis. Still, many important questions andproblems remain unresolved; there is noshortage of challenges in the field of apomixisresearch. Many serious research efforts mayonly serve as preliminary and somewhatacademic steps toward the long-term goal ofintroducing apomixis into farmers' fields. Toreach the distant goal of deployment tofarmers, future research should include anassessment of the social and economic impactof apomixis, and a definition of adequatedeployment strategies. These critical elementswill strongly influence the biological aspectsof apomixis research and what "kind" ofapomixis should be targeted for developmentand deployment.ReferencesAhn, S.N., and S.D. Tanksley. 1993.Comparative linkage maps of the rice andmaize genomes. Prac Natl Acad Sci (USA)90: 7980-84.Asker, S., and l. Jerling. 1992. Apomixis inPlants. Boco Raton, Florida: CRC Press.Bennetzen, J.L., and M. Freeling. 1993. Grassesas a single genetic system: genomecompasilion, collinearity and compo~bility.Trends in Genetics 9:259-61.Bicknell, R.A., N.K. Borstk, and A.M. Koltunow.2000. Monogenic inherilance of apomixisin two Hieracium species with distinddevelopmental me

94 D..iel Grimallelli-, Jo. Tohme, .od Di.go GN,OIt,-d...I.t••--. 19840. Gametophytic apomixis. In8.M. Johri (ed.), Embryology ofAngiosperms. Berlin: Springer-Verlag. Pp_47>-518.--. 1984b. Genetics 01 apospory inapomictic Ranunculus ouricamus V.Conclusions. Botanica Helvetica 94: 411­22.Noirot, M. 1993. Allelic ratios and steri~ty in theagamic complex 01 the Maximeoe(Panicoideae): evolutionary role 01 theresidual sexuality.J. Evo/. BioI. 6:9S-101.Nayes, R.O., and R.H. Rieseberg. 2000. Twoindependent loci control agamospermy(apomixis) in the triploid flowering plantfIigeran annuus. Genetics 155: 379-90.Ohad, N., l. Margassian, Y. -C Hsu, CWilliams,p.Repe"i, and R_L Fisher. 1996. Amutationthat allows endosperm developmentdevelopment without fertiliza1ian. Proe.Natl. Acad. Sci. (USA) 93: 5319-24.Ozias-Akins, P. 1998. Tight clustering andhemizygosity of apomixis-linked molecularmarkers in Pennisetum squamulatumimplies genetic control of apospory by adivergent locus that may have no allelic Iform in sexual genotypes. Proe. Natl. Acdd.Sci. (USA) 95: 5127-32.Pessin a, S.C 1997.ldenliflcation of a maizelinkage group related 10 apomixis inBracharia. Theor. Appl. Genet. 94: 439­44.Richards, AJ. 1986. Plant breeding systems.london: George Allen and Unwin.Savidan, Y. 1982. Nature et herMite de"apomixie chez Panicum maximum }acq.Ph.D_ dissertation Universite of Paris XI.---. 2000. Apomixis: Genelics and8reeding. Plant Breeding Reviews 18: 13­86.Sheridan, W.F., A.A. Nadezhda, 1.1. Shamrov, lB.Batygina, and LN. Golunovskayo. 1996.The mac gene: controlling the commitment10 Ihe meiotic pathway in maize. Genetics142: 1009-20.Shizuya, H., B. 8irren, U.-1. Kim, V. Manano, 1Slepak, Y. Tachiiri, and M. Simon. 1992.Cloning ond stable maintenonce 01300­kilobase-poir fragmenls 01 human DNA inEscherichia coli using a Hoctor-basedvector. Pro. Natl. Acod. Sci. (USA) 89:8794-97.van Dijk, PJ. 1999. Crosses between sexual andapomictic dandelions (Taraxacum). II. Thebreakdown of apomixis. Heredity 83:71S-21.Vielle-Calzoda J.P., J. Thomos, L Spillane, A.Coluccio, M.A. Hoeppner, and U.Grossniklaus. 1999. Maintenance ofgenomic imprinting at the Arabidopsismedea locus requires zygotic ODM1activity. Genes Dev. 13(22): 2971-82.Walbot, V. 1992. Strategies lor mutagenesis andgene cloning using transposon lagging andlDNA mutagenesis. Ann Rev. Planf Physiol.Plant Mol. BioI. 43: 49-82.

Chapter 7The Gene Effect:Genome Collisions and ApomixisJOHN G. CARMANIntroductionIn the vast majority of angiosperms, femalemeiosis results in the formation of a tetrad ofmononucleate megaspores, of which threedegenerate and one forms the geneticallyreduced8-nucleate female gametophyte(Polygonum-type embryo sac). Consistentlyexpressedcytological deviations from thisnorm occur in certain species in 506 of the13,479 genera of angiosperms recognized bythe Kew Botanical Gardens (Carman 1997).However, most species in most of these 506genera reproduce normally. Thus, thepercentage of species consistently expressingreproductive anomalies (probably < 0.3 %) isfar less than the percentage of genera (3.8 %)currently known to contain anomalous species.Reproductively-anomalous species occur in atleast 184 families, which is 53 % of those inwhich some embryological analyses have beenreported, and are much more abundant insome families than in others (Carman 1997).The reproductive anomalies considered in thischapter generally belong to three categories:gametophytic apomixis, polyspory, andpolyembryony. Gametophytic apomictsproduce unreduced embryo sacs that containparthenogenetic eggs, are generally polyploid,and occur in at least 33 of 460 families ofangiosperms. Diplospory and apospory aretwo major subdivisions of gametophyticapomixis (referred to as apomixis hereafter)and occur when unreduced embryo sacs formprecociously from ameiotic megaspore mothercells (MMCs) or nearby somatic cells,respectively (Asker and Jerling 1992; Carman1997; Peel et al. 1997a, b; Crane, Chap. 3).Polysporic species (bisporic or tetrasporic) aresexual and occur in at least 88 families. As indiplospory, embryo sacs of polysporic speciesform precociously from MMCs, but onlyportions of meiosis not critical to geneticreduction are affected (Battaglia 1989; Johri etal. 1992; Carman 1997). Polyembryonyinvolves the formation of embryos from cellsof other embryos, synergids, antipodals,nucelli, integuments, and even leaves (Tisseratet al. 1979; Johri et al. 1992). Like parthenogenesisin apomicts, polyembryony oftenbegins before pollination (Naumova 1993).This chapter summarizes these anomalies interms of developmental similarities,phylogenetic associations, and gene effecthypotheses, and it discusses implications of thegene effect hypotheses for future research andplant improvement. It concludes that a majorpurging of some widely accepted dogmaconcerning the evol ution and geneticregulation of apomixis will probably occur asthe mechanisms involved give way to moreaccurate evolutionary, developmental, andmolecular characterizations.Developmental Biology andPhylogeny of Reproductively­Anomalous SpeciesAt the developmental level, some apomicticmechanisms resemble sexual polysporicmechanisms more than other apomictic

96 10k. G.(o,.,..mechanisms (Figure 7.1). For example,Antennaria-type diplospory is identical totetraspory (sexual), except that the nucleardivisions leading to a tetranucleate embryo sacare meiotic in tetraspory but mitotic inAntennaria-type diplospory. Both Antennariatypediplospory and tetraspory occur inAntennaria, Erigeron, Limonium, and Rudbeckia(Carman 1997). In both anomalies, MMC lackcallose (Peel et a1. 1997a). Ixeris-typediplospory is even more similar to tetrasporybecause, as in tetraspory, a meiotic prophaseoccurs. However, in the diplosporousmechanism, a first division restitution ensues.Ixeris-type diplospory is identical to bispory(sexual) except that meiosis I fails in the former(Figure 7.1). Both Ixeris-type diplospory andbispory (and apospory and tetraspory) occurin Erigeron. Allium odorum-type diplosporyand bispory differ only in that a chromosomalendoreduplication occurs in the former. Bothare found in Allium. The recognition of thesedevelopmental similarities and phylogeneticassociations led Carman (1997) to analyzetaxonomic data for all known species thatexpress these anomalies. It was found thatapomictic, polysporic, and polyembryonicspecies are polyphyletic and tend to bephylogenetically related. Many highlysignificant associations were discovered.The variation represented in Figure 7.1 issomewhat continuous. For example,Antennaria-type diplospory is similar to bothtetraspory and apospory. In some families,both species and genera span this continuum.For example, apomixis and polyspory occurtogether in 13 of 127 apomixis-containinggenera (Allium, Antennaria, Burmannia, Cordia,Cynoglossum, Erigeron, Eurybiopsis, Leontodon,Limonium, Rubus, Rudbeckia, Sambucus, andTridax) and 18 of 33 apomixis-containingAbundantcollole IynlhelilUni-nudeote embryo lOCIFigure 7.1 Developmental stages during megasporogenesis and embryo-sac development in selected sexual(monosporic, bisporic, and tetrasporic) and apomictic (Allium odorum-type diplospory, Antennaria-typediplospory, Taraxacum-type diplospory, Ixeris-type diplospory, Blumea-type diplospory, and apospory)angiosperms.

Tilt Goo. IH.

98 Jolo.G.(_explain the existence of apomixis must alsoaddress these highly significant peculiaritiesin genome composition, phylogeneticrelatedness, and developmental (ormechanistic) affinities.Other genome-related factors may abnormallyaffect female development in polyploids orpaleopolyploids. For example, meiotic rates arelinearly correlated with DNA content, but theregression line is much steeper in polyploids.And, meiosis in tetraploids usually requires thesame period of time as in related diploidscontaining half the DNA. This occurs becausegenes for meiosis in polyploids are duplicated(Bennett 1977). That the meiotic rate to DNAcontent regression slopes in paleopolyploidsreflect either a diploid or a polyploid conditionmay be critical to the evolutionofapomixis andrelated anomalies, Examples include Scillanonscriptus, 2/1 = 2x = 16, which belongs to ananeuploid series with x = 6 to 9, 15, and 17 asstabilized base numbers, and Convallariamaja/is, 2/1 = 2x = 38, with a sole base numberofx = 19. Both arepaleopolyploid diploids withlarge quantities of DNA, but their meiosesoccur in only50% of the time predicted for nonpaleopolyploiddiploids with similar amountsof DNA, i.e., they behave as polyploids. Incontrast, Omit/roga/um virens (2/1 = 6), which isat the bottom of a descending aneuploid seriesin which x = 3 to 5 and 7, Allium cepa (2/1 = 16),which is at the middle of an aneuploid serieswith x = 7 to 9, and Fritillaria me/eagris (2n =24), which is at the top of an ascendinganeuploid series with x = 7, 9, and 12, arepaleopolyploid diploids with slow meiosesthat is indicative ofdiploids with considerableDNA. Duplicate genes for meiosis in thesespecies have either been lost throughaneuploidy or genetically silenced. Hexaploidnulli 5B tetra 5D wheat is another example.Meiotic rates in this line reflect a tetraploid, nota hexaploid, probably because of imbalancedsets of meiotic genes (Bennett 1977).Several questions relevant to the evolution offemale developmental anomalies can beformulated from this information. Forexample, how is the synchrony of femaledevelopment affected when some of theduplicated genes responsible for megasporogenesis,embryo-sac development, andembryony from one of two genomes aresilenced or lost during diploidization(paleopolyploid formation)? What happenswhen there are duplicate doses of genes forcertain stages of meiosis or embryo sacdevelopment and not other stages, as isanticipated in highly aneuploid paleopolyploidpolysporic species? Could suchimbalances cause some of the anomalies ofembryo sac development observed inpolysporic species, such as a precociousgametophytization of the MMC or theformation of 4 to 32 nucleate embryo sacs?Total quantities of DNA also influence thetypes of life cycles angiosperms assume. Forexample, species of Fritillaria (many of whichare polysporic paleopolyploids) have largeamounts of nuclear DNA and their meiosesmay require 3-4 weeks to complete. Incontrast, annuals have little DNA and veryshort meioses (Bennett 1977), and apomixisand related anomalies are rare among them(Asker and Jerling 1992). Hence, a minimumthreshold in duration of meiosis may be aprerequisite for the evolution of certainreprod~ctiveanomalies.Reproductive anomalies in angiosperms mightalso be influenced by differences in meioticdurations between genders. In cereals, femaleand male meioses are generally synchronousand similar in duration. However, in speciesin which female meiosis occurs later than malemeiosis, differences in duration may be asgreat as 50 % (Bennett 1977). Such differencesmight encourage anomalous development inone gender but not the other.

Tlte G... flfe

100 JolI.G.C....parameters affecting whether meiosis and that diplospory might be similar toembryo-sac formation occur early or late tetraspory (Rodkievitz 1970) with regard toduring floral bud development. Hybridization an absence of MMC callose. The molecularamong ecotypes expressing such divergent sieve properties of callose (Heslop-Harrisonfloral development schedules may cause and Mackenzie 1967) led Crane and Carmanapomixis in the absence of mutation (Carman to hypothesize that the absence of callose1997,2000). may allow regulatory factors necessary forThe Gene Effect Hypothesesmeiosis to diffuse away from the MMC and/or allow regulatory factors responsible forInsights concerning the molecular control of mitosis to diffuse into the MMC from thefemale development can be gleaned by nucellus or integuments. This callosestudying the many developmental, hypothesis was formally discussed in 1986phylogenetic, and genomic peculiarities that (Carman 1986; Carman and Crane 1986).apomicts share with polysporic and Subsequent reports (Crane and Carmanpolyembryonic species (Table 7.1). The gene 1987; Carman et al. 1991) suggested that theeffects that ultimately explain apomixis will deficiency in callose deposition aroundprobably also largely explain these MMCs of apomictic E. rectisetus may only bepeculiarities. Several gene effect hypotheses coincidental to a more fundamental geneticare presented below and are judged by how lesion that causes apomixis. The latter reportwell they explain the data of Table 7.1. (Carman et al. 1991) documented that sexualThe Callose HypothesisMMCs from diplosporous lines of E.In 1984, Charles Crane, working in the rectisetHs, which usually constitute less thanauthor's lab, noted that MMC walls of 5% of all MMCs in this facultative apomict,diplosporous lines of Elymlls rectisetlls were are richly enveloped by callose, as in normalabnormally thin. This was the first indication sexual species.Table 7.1 Phylogenetic, genomic, and developmental peculiarities that hypotheses for the genetic regulationof apomixis and related reproductive anomalies must explainPhylogeneticApomixis, po~spory, and po~embryony are rare yet polyphyleticApomixis, pa~spory, and po~embryony tend to ocwr in the same species/genera/familiesIAMC degeneration/replacement acwrs in apomictic or polysporic genera "Preleptotene chromosomal condensations tend to ocwr in polysporic genera The majority of apomicts evolved during the postthree million years, i.e., during the last 2% of the evolutionary existence of angiosperms (see Carman, 2000).Genomic/developmentalChromosome bose numbers are low in apomicts and high in polysporic/po~embryonic speciesApomicts tend to be genome-balanced, polysporic/polyembryonic species tend not to beApomixis, palyspory, and polyembryony are general~ absent in annuals (low amounts of DNA/rapid meioses)Paleopolyploids may behove as diploids or polyploids with respect to meiotic durationPolygonum-type reproduction is fawltative in apomictic, polysporic, and polyembryonic speciesTendencies to apomixis are observed in some wide hybrids including Raphanabrassica (consistent aposporous embryo socformation)Male and female meioses in some species start at different times and have different durationsApomixis, polyspory, and polyembrony tend to ocwr in genera capable of strong adaptations to photoperiodApomixis, polyspory, and polyembryony are characterized by asynchronausly·expressed substitutions, replacements, orduplications of discrete reproductive phases Most apomicts display relaxed endosperm balance number requirements (seeGrimanelli et 01., Chop 6; Grossniklaus, Chop 12)

The G... EHo

102 1010. G.(arma.was developed in 1994 by the author whileattempting to reconcile simple inheritance forapomixis (the prevailing opinion at the time;Nogler 1984; Asker and Jerling 1992; Mogie1992) with (i) the many apparent asynchronousreplacements, competitions, andduplications of discrete developmentalsegments in reproductiveIy-anomalousspecies (Figure 7.1; many phenomena inaddition to embryo-sac induction), and (ii) thefact that nearly all apomicts are polyploid. Theauthor concluded that such a reconciliation isunreasonable. According to HFA theory,apomixis occurs when hybrids are producedfrom ecotypes that are distinctly divergentwith respect to their start times and rates ofMMC formation, meiosis, embryo sacformation, and embryogenesis relative to grossovule development. Such "genome collisions"(terminology suggested by Sven Asker,personal comm., 1997) explaill the abundantduplicity and asynchrony of developmentdepicted in Figure 7.1.In 1994, the author conducted a preliminarysearch of the literature to determine ifpolysporic and polyembryonic species containmultiple genomes, i.e., whether they arepolyploid. A negative result was soonobtained, which seemed to deal a fatal blowto this fledgling multi-genome "asynchronyhypothesis." However, in studying thegenome composition of the polysporic andpolyembryonic diploids, it was found thatthese "diploids" generally have highchromosome base numbers indicative ofpaleopolyploidy. The author concluded thatif the HFA theory is correct, the base numbertrends observed in the preliminary 1994 studyshould hold in a large-scale study of all knownapomictic, polysporic, and polyembryonicspecies. The theory survived the large-scaleexamination, was refined, and additionalhypotheses concerning the origins of apomixisand its role in the evolution of somereproductively-novel polysporic andpolyembryonic species and genera weredeveloped (Carman 1997; Peel et al. 1997a, b).The HFA theory states (i) duplicate sets ofgenes encoding female developmentalpathways exist in interracial or interspecifichybrids, polyploids, mesopolyploids, andpaleopolyploids; (ii) polygenic "heterozygosity"for photoperiodic floral induction andstart times and durations of MMC formation,megasporogenesis, embryo-sac formation,endosperm formation, and embryony, is theprimary cause of apomixis, pOlyspory,polyembryony, and related anomalies; (iii)allopolyploidy or segmental allopolyploidy isoften required for apomixis because it preventsor greatly reduces the incidence of geneticrecombination between genomically-isolatedsets of parental genes, which otherwise wouldlead to recombination among the many genesrequired for apomixis, resulting in reversionto sexuality (Carman, in preparation); (iv)polyploidy also influences apomixis byinfluencing the timing and duration of meiosis(Bennett 1977) and because divergent genomesare probably more prone to be physicallypartitioned in the nucleus (Leitch et al. 1990)when present as homologous pairs and thusmore functionally independent (Carman1997); and (v) mutations are of secondaryimportance and may improve reproductivefitness through null-allele formation in one orboth genomes. This theory is consistent withcurrent models of developmental geneexpression, including (i) the ABC model, inwhich floral genes from a B cassette areexpressed only when genes from an A cassetteare expressed (the expression of C genesrequires expression of B genes, etc.) (Theissenet al. 2000), and (ii) checkpoint models, inwhich precocious expression of checkpointgenes causes developmental phases to beskipped, e.g., fusing G]-phase yeast cells withM-phase cells causes G] nuclei of

Th. Go•• [H.,,: Go..... Colli.io,,, ..dApomixi. 103heterokaryons to skip Sand G2 and proceedprecociously to mitosis (Lewin 1994).Many examples of checkpoint phenomena canbe hypothesized from an examination of Figure7.1. For example, if embryo-sac developmentsignals from one genome are superimposed onmegasporogenesis signals from anothergenome, meiosis may be skipped (diplospory),similar to the heterokaryon examplesdescribed above, or embryo-sac developmentmay be ectopic (apospory). Accordingly,apomictic-like tendencies would occur inpolyploids only if major differences in timingof megasporogenesis and embryo-sacdevelopment (relative to other ovule and ovarytissues) exist among the ancestral ecotypes orspecies (Figure 7.2). Such natural variation maybe infrequently found in highly cosmopolitangenera, i.e., genera with broad latitudinal andecological distributions, but possibly absent inless cosmopolitan genera.The HFA theory was refined using criteriatabulated in Table 7.1. Hence, it explains thesecriteria as well as many inconsistencies in theapomixis literature. For example, apomixis,polyspory, and polyembryony are rare buttend to occur together in cosmopolitanfamilies, such as Poaceae, Asteraceae, andRosaceae, because sufficient ecotypic variationin reproductive start-times, etc., is rare in mostfamilies but high in these cosmopolitanfamilies. Sexual reproduction of themonosporic Polygonum-type occursfacultatively in apomictic and polysporicspecies because, barring deletions ormutations, each parental genome containsgenes required for normal reproduction, andgrowing conditions may occasionally favor theexpression of one genome over the other,causing sexual development to occur.Facultativeness may be influenced by (i)differential silencing of genomes (epigeneticFigure 7.2 Model of how asynchronously-expressed duplicate genes cause diplospory and apospory inpolyploids containing two genomes divergent in the temporal expression of female developmental schedules(floral induction, megaspore formation, gametophyte development, and embryony). Bolded developmentalphases are skipped as described below.Developmentally-critical stages'Genome 2 3 4Genome I Archespore Meiosis Embryo sac Double fertilization/(unmodified)early embryonyGenome I Embryo sac Double fertilization/ Fertilization of(modified) : early embryony central cell onlyGenome II Meiosis Embryo sac Double fertilization! Fertilization 01early embryony

104 J... G.(.....effects), which could be caused by differencesin genetic background, or (ii) envirorunentalfactors that reduce the degree of asynchronyby accelerating or decelerating geneexpression from one genome relative to thatof another (photoperiod or temperatureresponses, e.g., as occurs in Dicanthium,Themedtl), thus allowing sexual developmentto occur facultatively (Carman 2000).According to the HFA theory, polyspory andpolyembryony result from the competitiveexpression of grossly imbalanced genomes(incompletely duplicated sets of reproductivegenes) in which some checkpoint systems aremissing. In contrast, competitive expressionamong genomes is terminated by checkpointgenes in apomicts, which generally containbalanced sets of reproductive genes (Carman1997), thus allowing a somewhat smoothtransition to apomixis (Figure 7.2).Apomixis is much more prevalent amongoutcrossing species than inbreeding species(Asker and Jerling 1992). This is consistentwith the HFA theory because outcrossingspecies are much more prone to forminterecotypic or interspecific polyploids whensecondary contacts occur, e.g., during thenumerous major. climatic shifts associated withthe Pleistocene glaciations (Frakes et al. 1992;Carman 2000). Likewise, more apomicts arealiopolyploid than autopolyploid becausepolyploidization by Bill hybrid formation isexpected to occur more frequently in naturein interspecific hybrids than interracialhybrids. Similarly, the chances of Bill hybridformation occurring in interecotypic orinterspecific F 1hybrids that are sterile andannual are low compared to their formationin sterile perennials, which may flowerannually for many years. This factor limits thechances of annuals becoming apomictic andexplains their low frequency in nature.The HFA theory also predicts ambiguousoutcomes regarding the sexuality of progenywhen an apomict is crossed with a sexual orwith another apomict, regardless of thecloseness or wideness of the cross. The modeof reproduction expressed in the progeny willdepend on how the added or removedgenome(s) affect asynchrony, and this cannotbe predicted without some a priori knowledgeof the female developmental schedulesencoded by the involved genomes (Carman1997). That these many inconsistencies in theapomixis literature are explained by the HFAtheory is strong evidence for its validity.Testing the Gene EffectHypothesesIf apomixis is the result of one or a fewmutations, similar artificially inducedmutations might produce apomicts fromsexual species. Research programs currentlyexploring this possibility are reviewed byBicknell (Chap. 8), Grossniklaus (Chap. 12),and Praekelt and Scott (Chap. 13). Likewise,simple inheritance should permit transfer ofapomixis gene(s) to sexual species. To date,introgression projects have failed to conferapomixis upon sexual species by addinganything less than at least one complete alienchromosome (Savidan, Chap 11). Kindiger etal. (1996) reported a condition that might leadto an exception. They isolated, from theirmaize-Tripsacum backcross program, a line (30Mz + 9 Tr chromosomes) that appears tocontain a maize Tripsacum translocationpossessing gene(s) for apomixis. However,Blakey et al. (1997, reviewed below)determined that the genes required forapomixis occur in five distinct Tripsacumlinkage groups that are syntenic to regions onthree maize chromosomes. These data castdoubt on Kindiger's simple inheritance model.

If apomixis is caused by hybridization, theparental phenotypes (divergent floralinduction stimuli and meiotic start times inovules, etc.) may be identified by acombination of phenological and cytologicalstudies. Because many developmentalpathways occur asynchronously in apomicts(Figure 7.1), it may be possible to use molecularprobes to determine if the asynchronoussignals originate from the same genomes ordifferent genomes. For example, do bothgenomes in a tetraploid apomict produce bothmeiotic and embryo-sac development signalsat the same time, or does one genome producemeiotic signals (and not embryo-sacdevelopment signals) at the same time theother genome is producing embryo-sacdevelopment signals (see Figure 7.2)?Many criteria should be considered in testingfor such asynchrony. First, it would bedesirable for the apomict to (i) be allotetraploidwith known sexual diploid progenitors, (ii)contain well-mapped genomes, (iii) beamenable to transposon tagging, (iv) be easilygrown with a short vegetative phase, and (v)have ovules readily accessible to cytologicalanalyses. At present, no apomict meets all ofthese criteria. Second, molecular probes thatrecognize mRNAs specific to differentdevelopmental stages would need to beproduced, and the genes from which they aredeveloped would need to contain adequateintergenomic sequence divergence such thatprobes unique to each genome could beproduced. Such probes may currently be underdevelopment.Portions of meiotic prophase and earlyembryo-sac development occur simultaneouslyin most aposporous apomicts and allTaraxacum- and Ixeris-type diplosporousapomicts (Carman 1997; Peel et al. 1997a). TheHFA theory would be confirmed if thefollowing two conditions are observed: (i) aprobe unique to meiotic prophase of genomeA plus a probe unique to early embryo-sacdevelopment from genome B detect theirrespective genome-specific mRNAs in ovulesfixed and sectioned during meiotic prophase,and (ii) the probe for meiotic prophase fromgenome B does not detect its respectivegenome-specific mRNA product (or viceversa). This would confirm that one genomecodes for meiosis (but not embryo-sacdevelopment) at the same time the othergenome is coding for embryo-sac development.This test would not be valid if the mRNAsynthesis identified by the respective probesis produced in response to transactingregulatory genes.Implications of the HFA TheoryIf the HFA theory is correct, much of theapomixis literature will need to bereinterpreted. This includes interpretations ofhow apomixis evolved, the role of apomixisin evolution, the genetic control of apomixis,and how apomixis might be transferred to (orinduced to occur in) sexual species.Evolution of Apomixis andRelated AnomaliesAccording to HFA theory, many secondarycontacts must have occurred between ecotypes(or closely related species) that had beenisolated from each other for many years alonglatitudinal or other ecological gradients. Majorclimatic shifts could account for suchseconQary contacts (Carman 2000). Thedistribution patterns of most apomicts indicatea Pleistocene origin (Stebbins 1971; Asker andJerling 1992), i.e., the geographic distributionsand centers of diversity of many apomicts arecentered near the margins of the Pleistoceneglaciations, but their ranges often encompassthe ecological ranges of the putative sexualprogenitors, which lie north and south of theglacial margins. Eight major glaciations, whichcovered as much as 20% of the earth's surface,occurred during the Pleistocene. These wereseparated by warm interglacial periods lasting

106 J.hG.eforseveral thousand to a hundred thousandyears each. Likewise, most of the major glacialevents consisted of glacial advancesinterrupted by minor interglacial periods thatlasted for a few thousand years (Frakes et al.1992). Hence, during the Pleistocene, the highlatitudes of both hemispheres were repeatedlydeglaciated and revegetated by cosmopolitantaxa capable of adapting to cool climates, longdays, and short growing seasons.A long-day flowering response and aprecocious meiosis and embryo-sacdevelopment in young ovules of sexualAnlennaria, and probably many other taxa, areadaptations to short summers in high latitudesor altitudes (Carman, unpublished). Glacialadvances, which followed the numerousinterglacial periods, cooled the midlatitudes,permitting higher latitude flora to invademidlatitude flora. This provided opportunitiesfor ecotypes with a putatively-precociousfemale development (from higher latitudes orelevations, i.e., temperate to arctic climates) toform polyploids with ecotypes (or relatedspecies) with a putatively-delayed femaledevelopment (from lower latitudes, i.e,. tropicto full-sun temperate climates). Suchpolyploids may have given rise to modernapomicts (Carman 1997, 2000).This interpretation is consistent with theobserved effects of climatic factors onfacultativeness in certain apomicts. Forexample, exposing Dicnnlhillm a11l1ll11111lm, D.illiermedillm, and Themedn lrumdm to short daysand low temperatures increases the frequencyof apospory. The opposite conditions increasethe frequency of sexuality in a partially-sexualDicnllihilim hybrid (reviewed by Asker andJerling 1992; Carman 2000). Such shifts infacultativeness are expected if signaltransduction pathways vary among genomesin sensitivity to morphogens (hormones, etc.)that accumulate in response to changingseasons and photoperiods.Once asynchrony is initiated, further shifts infacultativeness might occur in response togrowing conditions. G. Ledyard Stebbins(personal communication, 1997) suggests thatconditions favoring rapid growth (hightemperatures, moisture, and light) mightenforce competition between asynchronousgenomes causing an increased frequency ofapomixis. This prediction was observed inclones of F 1hybrids obtained between wheatand diplosporous Elymlls recliselus. Clonesgrown under favorable conditions (hightemperatures and light intensities) producedmore apomeiotic-like MMCs (Peel et al. 1997b).Additional research should be conducted toelucidate such effects on facultativeness (Askerand Jerling 1992). Such research could provideclues concerning the nature of the divergentparental phenotypes that may have producedapomicts upon hybridization and polyploidizationduring the Pleistocene.The HFA theory also suggests that the role ofapomixis in evolution is more prominent thanpreviously supposed. What happens if someof the many checkpoint g-enes permitting areasonably smooth replacement ofdevelopmental segments (resulting inapomixis) are mutated or lost duringdiploidization? Phylogenetic and genomicevidence suggests that such confusions ofdevelopment may manifest themselves aspolysp~ry or polyembryony. Thus, apomixis,instead of being an evolutionary dead end,may occasionally serve as an evolutionaryspringboard in the evolution of normal ordevelopmentally-novel paleopolyploid sexualspecies and genera (Carman 1997).Mendelian Analyses of ApomixisEssentially all Mendelian analyses of apomixisface reinterpretation if the HFA theory iscorrect. Data from a variety of apomicts haveon one or more occasions (or when grown incertain environments) tended to fit atetrasomic inheritance model with apomixis

Th. Go•• U1e

108 Jelo.G.(_large linkage group in which recombinationis suppressed (Grimanelli et al. 1998), and asimilar linkage group appears to exist inapomictic Pennisetum (Grimanelli et al. 2000,Chap 6), Cenchrus (Roche et al. 1999), andBrachiaria (Pessino et al. 1999). Such linkagegroups may contain multiple genes requiredfor apomeiosis (Grimanelli et al. 1998,2000).Two groups are introgressing apomixis intomaize from Tripsacum, and neither has reportedits expression in addition lines with less thannine Tripsacum chromosomes. In one group,two apomictic maize triploids containing nineTripsacum chromosomes (3x + 9) wereproduced. Cytogenetic and molecular studiesstrongly suggested that the nine Tripsacumchromosomes in each line were the same(Kindiger et al. 1996). A third triploid additionline, again with nine Tripsacum chromosomes(3x + 9), was produced by the same group.However, many of the nine chromosomes inthis line differed from the nine chromosomesof the two former lines. The maizechromosomes were the same for all three lines.The latter 3x + 9 plant was also apomictic, butthe frequency of apomixis was only 10-15%,compared to 95-100% for the two former lines(Victor Sokolov, personal comm., 1997). Thesedata and unpublished findings from the othergroup attempting to transfer apomixis to maize(Grimanelli et al. 2000, Chap. 6) suggest a muchmore complex mode of inheritance forapomixis.In another study, sexual T dactyloides diploidswere crossed with highly apomictic T.dactyloides triploids (%) to produce aneuploids.All but three of 46 Fls showed tendencies forapomeiosis (determined cytologically).However, the highly apomeiotic FIS containedseven or more chromosomes above the diploidlevel, and all Fls with chromosome numbersnear the diploid level were sexual (Shermanet al. 1991), which also suggests complexinheritance. Finally, genes essential to theexpression of apomiXis in artificially producedTripsacum triploids cosegregated with fiveTripsacum linkage groups that are syntenic withregions from three maize chromosomes(Blakey et al. 1997).These latter studies infer (i) the interaction ofmultiple genes from multiple chromosomes (atleast in Tripsacum) is required for the expressionof apomixis; (ii) many genes affectfacultativeness and behave additively; (iii)some Tripsacum chromosomes affectfacultativeness much more than others; and (iv)alleles from at least three maize chromosomesfail to substitute for their homeologous(syntenic) counterparts from Tripsacum inconferring apomixis. These inferences evokethe following questions: How many genes arerequired to efficiently express apomixis in analien genetic background? What selectivepressures cause the accumulation ofappropriate combinations of alleles that conferapomixis and regulate facultativeness? Cansuch selective pressures predispose sexualplants to apomixis upon inter-ecotypichybridization? What role does polyploidyplay? Does mutation play any role at all? TheHFA theory answers these questions.That Mendelian analyses of some apomictsconsistently result in expected ratios does notprove apomixis genes exist. Apomixis in suchspecies may be largely controlled by onerecombi{lationally-isolated linkage group withmany minor unlinked genes affectingfacultativeness (Carman 2000). According toHFA theory, such linkage groups do notcontain apomixis genes per se and will notconfer apomixis when syntenic regions of asexual species are thereby replaced. Suchlinkage groups mimic heterozygous dominantgene action because of genomic configurationsthat isolate them from recombination. Thisisolation maintains the intergenomic (orintersegmental) polygenic heterozygosityrequired for apomixis and is critical to the

stabilization of apomixis, i.e., the ability ofapomicts to t" .:ltatively produce progenysexually without such progeny being sexualrevertants (Carman, in preparation). Themapping of genes or linkage groups withstrong effects on apomixis penetrance isimportant, and, as in many other breedingstudies, segregation ratios and levels offacultativeness should be verified by testingputative sexual segregants for apomixis(cytology of 10D-200 ovules and/or progenytests involving molecular markers) overmultiple years and multiple locations.Putative segregants should be consideredunreliable for mapping studies until such testsare performed. Finally, because levels ofapomixis expression (facultativeness) oftenare highly variable in Mendelian analyses(reviewed by Carman 2000), QTL analysesshould be conducted to estimate the numberof genes involved.Making Crops ApomicticAccording to HFA theory, "apomixis genes"that confer apomixis when transferred tosexual species do not exist. In contrast,heterozygosity at many loci is required; thisheterozygosity involves traits such as floralinduction stimuli and stages of ovuledevelopment in which meiosis, embryo-sacdevelopment, and embryony occur.Transferring such linkage groups to sexualspecies will not confer apomixis unless thetiming of female reproductive developmentencoded by the recipient plant isappropriately asynchronous with thatencoded by the alien segment. Hence, HFAtheory predicts that programs designed tointrogress apomixis into crops will experiencedifficulties producing apomictic lines thatpossess anything less than one to a few alienchromosomes.It should be possible to produce apomicts inat least some crops by (i) selection forappropriate parental phenotypes (divergencein female developmental schedules), (ii)crossing the appropriately-divergentphenotypes, and (iii) stabilizing apomixisthrough cytogenetic modifications that isolatethe responsible heterozygous loci fromrecombination (polyplOidy or inversion ortranslocation heterozygosity) or throughtransgenic modifications that obligately abortfemale meiosis. The success of this approachwill depend on the existence in primary genepools ofsufficient genetic variability for femaledevelopmental schedules and on correctlyidentifying the appropriate cytological parentalphenotypes. It may be pOSSible to enhanceinsufficient variabili ty in some crops byoutcrossing to the secondary gene pool.According to HFA theory, the required genesare not apomixis genes per se, but consist ofnormal genes with multiple ecotype-specificalleles, which when found in specificcombinations confer temporally-divergentschedules of sexual female development tonatural ecotypes.AcknowledgmentsThis study was supported by a Centers ofExcellence grant from the State of Utah and bythe Utah Agricultural Experiment Station,Logan, UT 84322-4845; approved as journalpaper no. 4891.ReferencesAsker, 51, and LJerling. 1992. Apomixis in Plants. Boro Rolon, Florida:0( Pr~s,Bo"aglia, E. 1989, The evolulion of the female gametophyte ofongiosperms: on interpretative key, Annali di Batanica 47: 7-144,Bennen, M.D. 1977. The time and duration of meiosis. Phi/osophiwlTranslations of the Royol Society of london, Series B. 277: 201-26.Bennen, M.D., and H. Stern. 1975. The time and duration of preleptotenechromosome condensation stage in U1ium hybrid cv. block beauty.Proceedings of the Royol Society of london, Series B. 188: 477-91Blakey, LA., C.L Dewold, and S.L Goldman. 1997. Cl>-segregation of DNAmarkers with Tripsocumferlilily. Moydiw42: 363--69.Brummill, R.K. 1992. Vaswlor PIont Families and Genera. Whintable, Kent,U,K.: Royal Botanic Gardens, Kew, Whitstable Litho, IJd.Carmon, J.G. 1986. Genetic engineering: potentials for underslanding andusing apomixis. Presentation at the workshop "The Polenliol Use ofApomiXis in Crap Improvement," April 21-26, 1986, RockefellerFoundation, Bellogio Study and Conference Center, Belogio, Ita~.

11 0 Joh. G.(orma.---. 1997. Asynchronous exprellion 01duplicote genes in angiosperms may causeapomixis, bispary, tetraspory, andpo~embryony. Biologicol Journol of theLinneon Society 61: 51-94.--.2000. The evolution of gametophyticapomixis. In I Batygina (ed.l, Embryologyof Flowering Plants, Terminology andConcepts, Vol. 3, The Systems of!leproduclion., SI. Petersburg: RussionAcademy of Sciences, World and Fami~Press. (in press).Carman, J.G., and CE Crane. 19B6.Comparative cytochemical analyses ofmegasporogenesis between apomictic andsexual Australasian Elymus species.Agronomy Abs/racls. Pp. 60.Carman, J.G., CE Crane, and O. Riera-Lizarazu.1991. Comparative histology 01 cell WIllisduring meiotic and apomeioticmegasporogenesis in two auslralasianElymus Lspecies. (rop Science 31: 1527­32.Crane, CE, and J.G. Carman. 19B7. Mechanismsof apomixis in £1ymus reclise/us fromeastern A~trolio and New Zealand.American Journol of B%ny 74: 477-96.Czapik, R. 1985. Apomic1ic embryo sacs indiploid Waldsteinia geoides Willd.(Rosaceae). Acla Biologico (racoVleOlia 27:29-37.Oujardin, M., and w.w. Hanna. 1983. Apamic1icand sexual pearl millet XPennise/umsquomularum hybrids. Journal of Heredity74: 277-79.Davis, G.L. 1966. Systematic Embryology of theAngiosperms. New York: Wiley and Sans.Frakes, LA.,!.E. Francis, and J.I. Syktus. 1992.Climate Modes of the Phanerozoic.Cambridge: Cambridge University Press.Goldblall, P19BO. Polyploidy in angiosperms:monocotyledons. In. WH. lewis (ed.),Polyploidy. New York: Plenum Press. Pp219-39Grimonelli, D., O. leblanc, E. Espinoso, E. Perolli,D. Gonzlilez·de-leOn, and Y. Sovidan.1998. Mapping diplosporo~ apomixis intetraploid Tripsacum one gene or severalgenes? Heredity BD: 33-39Grinanelli, D., P. Ozias-Akins, J. Tohme, and D.Gonzlilez-de·le6n. 2000. Moleculargenetics in apomixis research. In Y. Savidanand J.G. Carman (eds.), Advances inApomixis !leseorch. Rome: FAO.Hanna, WW., M. Oujardin, POzias·Akins, and L.AUlhur. 1992. Transfer 01 apomixis inPennise/um.. In J.H. Bgin and !..P Mikscheleds.l, Proceedings of the apomixisworkshop, February 11·12, 1992, Atlanta,Georgia. USDA·ARS, ARS-1 04. Pp 30-33.Heslop-Horrison, J., end A. Mackenzie. 1967.Autoradiography of soluble (2.1 4 ().thymidine derivatives during meiosis andmicrosporogenesis in Lilium anthers.Journal of Cell Science 2: 387-400.Johri, B.M., K.B. Ambegaokar, and PS. Srivastro.1992. Compora/ive Embryology ofAngiosperms. Volume 1and 2. New York:Springer·Verlag.Kindiger, B.K., B. Dopeng, and V. Sakolov. 1996.Cytological and molecular studies ofapomic1ic moize· Tripsocum hybrids.Agronomy Am/rocls. pp. 132.Koltunow, A.M. 1993. Apomixis: embryo sacsand embryos formed without meiosis orfertilization in ovules. The Plont Cell S:1425-37.leblanc, 0., D. Grimanelli, D. Gonllilez·de·le6n,and Y. Sovidon. 1995b. Detection of theapomic1ic mode of reproduc1ion in maize·Tripsacum hybrids using maize RFlPmarkers. Theore/ical and Applied Gene/in90: 1198-1203.leblanc, 0., M.D. Peel, J.G. Carman, and Y.Savidan. 19950. Megasporogenesis andmegagametogenesis in several Tripsocumspecies (Pooceae). American Journal ofBotany B2: 57-63.leitch, A.R., W. Mosgoller, I Schworzacher, M.O.Bennell, and J.S. Heslap·Harrison. 1990.Genomic in situ hybridization to sectionednuclei shoWl chromosome domains in grasshybrids. Journol of Cell Science 95: 335­4l.lewin, B. 1994. Genes V. New York: OxfordUniversity Press.lewis, WH. 1980. Polyploidy in angiosperms:dicotyledons.. In W.H. lewis (ed.l,Polyploidy. New York: Plenum Press. Pp.241-6B.Mogie, M. 1992. The Evolu/ion of Asexual!leproduclion in Plon/s. london: Chapmonand Hall.Naumova, IN. 1993. Apomixis in Angiosperms:Nucellor and In/egumentary Embiyony.Boca Raton, Florida: CRC Press.Noumova, IN., A.PM. den Nijs, and M.T.M.Willemse. 1993. Quantitative ona~sis ofaposporous parthenogenesis in Poapratensis genotypes. Acla Bo/anicaNeerlandica 42: 299-312.Naumova, IN., M.T.M. Willemse. 1995.Ultraslruc1ural charac1erizotion of aposporyin Panicum maximum. Sexuol Plan/!leproduclion B: 197-204.Nogler, G.A. 19B4. Gametophytic apomixis. InB.M. Johri {ed.1. Embryology ofAngiosperms. New York: Springer·Verlag.Pp. 475-51 B.Patil, B.D., M.W. Hardas, and A.B. Joshi. 1961.Auto-alloploid nature of Pennisetumsquamula/um Fresen. Narure 1B9: 419-20.Peacock, J. 1993. Genetic engineering andmutagenesis for apomixis in rice. In KJ.Wilson (ed.l, Proceeding of theInterno/ionol Workshop on Apomixis in !lice,Chongsha, China. New York: RockefellerFoundmion. Pp. l1-nPeel, M.D., J.G. Carmon, and O. leblanc. 19970.Megasporocyte callose in apomicticbuffelgrall, Kentucky bluegrass,Pennise/um squamula/um Fresen, TripsocumLand weeping lavegrass. Crop Science 37:724-32Peel, M.D., J.G. Carman, Z.w. Liu, and R.R.CWang. 1997b. Meiotic anomalies in hybridsbetween wheal and apomictic £1ymusreclisetus (Nees in lehm.l A. love ondConnor. Crop Science 37: 717-23.Pellino, S.C, J.PA. Ortiz, M.D. Hayward, and CLQuarin. 1999. The molecular genetics ofgametophytic apomixis. Hereditas 130: I­ll.Roche, D., PCong, Z. Chen, WW. Hanna, D.LGustine, R.T. Sherwood, and POzias·Akins.1999. An apospory·specilic genomic regionis conserved between bulfelgrass {Cenmrusciliaris U ond Pennisetum squomulotumFresn. The Plont Jouma/19: 20~B.Rodkiewia, B. 1970. Callose in cell WIllis duringmegasporogenesis in angiosperms. PIon/a93: 39-47.Salisbury, EB., and CW. Ross. 1992. Plan/Physiology. Belmont, California: Wadsworth,Inc.Sovidon, Y. 1982. Nature et heredile deI'apomixie chez Panicum maximum Jocq.Ph.D. thesis, Universite Paris XI, France.Sherman, R.A., PW. Voigt, B.L Burson, and Cl.Dewald. 1991. Apomixis in diploid HtriploidTripsacum daelyloides hybrids. Genome 34:528-32.Stebbins, G.L 1971. Chramosomol Evolution inHigher Plan/s. london: Addison-Wesley.Theissen, G., A. Becker, A.O. Roso, A. Kanno, J.T.Kim, I Munster, K.U. Winter, and H. Saedler.2000. Ashort history of MADS-box genes inplants. Plonf Mo/elUlar Biology 42: 115­49.Tisserat, B., E.B. Eson, and I Murashige. 1979.Somatic embryogenesis in angiosperms.HOr1icul/ural!leviews 1: 1-77.Valle, CB. Do,and J.w. Miles. 2000. Breeding ofapomic1ic species. In Y. Savidan and J.G.Carman leds.l, Advances in Apomixis!leseorch. Rome: FAO.Wallace, B. 1991. Coodaptation revisited. Journalof Heredity. B2: B9-95.

Chapter 8Model Systems to Study the Geneticsand Developmental Biology ofApomixisRoss A. BICKNellIntroductionApomixis is a diverse topic of study,encompassing the cellular and sub-cellularevents associated with the trai t, its inheritanceacross taxa, the taxonomic, ecological,biogeographica I and evol utionary consequencesof its expression, and the potentialeconomic impacts of its incorporation intocrop species. For each intended purpose,different characteristics are required in thestudy material, so the choice of anexperimental system will vary betweenresearch groups. Historical reasons, access tomaterial and expertise, and fundingopportunities are clearly also critical factorsin such decisions. For this brief chapter, Iintentionally restricted my focus to the choiceof model systems for studying thedevelopmental biology of apomixis and thegenetics underlying its expression andinheritance. Furthermore, characteristics ofexperimental merit have been ranked abovethose with more immediate commercial value.It is appreciated that this introduces a bias thatis not appropriate for all researchers, however,it is hoped that the considerations discussedremain relevant across different program aims.Why Use a ModelSystem for Apomixis?Until very recently, most published studiesregarding apomixis focused on eitherdescribing the cellular mechanismsunderlying asexual seed formation or on theecological implications of the trait for manydifferent species. Much of the available dataon apomixis are therefore quite diverse. Thishas been valuable for identifying elementscommon to different types of apomixis,however, it has also led to difficulties whencomparing results from widely disparatesystems. Consequently, it now seemsappropriate to concentrate research effort ona smaller number of systems that may serveas models for the trait. The benefits of usingmodel systems are clearly recognized in mostfields of biology. Concentrating on a singlesystem makes it possible to establish researchnetworks; accumulate information on theexperimental manipulation of the organism;develop genetic maps; maintain repositoriesfor associated materials (such as geneticallycharacterized plants, DNA libraries andprobes); and to devote limited resources todeveloping a comprehensive understandingof a representative system(s).As other chapters in this book make clear,many critical questions in apomixis researchremain unanswered, such as the number andnature of the major genes involved and therole of epistasis, with regard to modifiers andthe interaction with polyploidy. Temporal andspatial gene regulation clearly occur within theovule to determine tissue fate, and timingappears to be critical in determining therelative importance of the meiotic andapomeJotic pathways of seed formation infacultative systems (Savidan 1989; Koltunow

112 a... A. 8icUeI1993). Before the trait can be successfullycommercialized, more information isrequired about the impact of environmentalvariables on the expression of apomixis andtheir interaction with sexuality, particularlywith regard to fertility and resourceallocation. Ultrastructural studies are alsorequired to more clearly elucidate thecytological and histological events involvedin apomixis, such as the role of differentialcallose deposition (Carman et al. 1991;Leblanc et al. 1995a; Peel et al. 1997).Attributes of A Model SystemThe innate advantages and limitations of anysystem will dictate the available researchopportunities and impact on the rate ofpossible progress. Therefore, beforediscussing candidates, it is helpful to considerthe features that would facilitate their use,specifically in a study of the developmentalbiology and molecular genetics of apomixis.Biological AttributesTo facilitate experimental progress, there area number of cultural characteristics toconsider in choosing a model system. Forsimplicity, the plant should be easy tocultivate, both in vivo and in vitro. Ideally itshould also be a perennial that can be easilypropagated vegetatively to permit themaintenance of sterile or self-incompatiblesexual biotypes. A small, compact plantstature, which does not require training, willfacilitate the manipulation of largepopulations, such as those used duringmutant screens and inheritance studies. Forthe rapid turnover of experimentalpopulations it is convenient to use a specieswith a short generation time, abundant seedset, and adequate seed fertility. This isparticularly important for studying apomixisbecause plants are assessed at the time offlowering or seed formation. For theevaluation of apomixis, it is advantageous touse a species in which apomixis can be easilyassessed, preferably in a format that can bequantified, to facilitate the study of allelicdifferences and epistasis.The inheritance of apomixis is typicallyassessed by crossing sexual and apomicticbiotypes, often using the sexual plant as thematernal parent. Although it is usually alsopossible, and sometimes necessary, to performthe reciprocal cross, it requires the separationof recombinant (B IIand Bmhybrids) from non­recombinant (maternal and polyhaploid)progeny. Inheritance studies, therefore, requirethat cross-compatible sexual and apomicticbiotypes are available, and if a sexual recipientis used, that microsporogenesis andmicrogametogenesis are functional in theapomictic biotypes used. Sexual biotypes arealso useful in molecular studies for assessingthe developmental importance of reintroduced,putative control sequences.Types of ApomixisTwo prinCipal mechanisms of apomixis havebeen reported: (i) adventitious embryony, inwhich the maternal embryo arises directlyfrom a somatic cell, and (ii) gametophyticapomixis, in which the maternal embryo isderived from an egg cell within an unreducedembryo sac. Gametophytic apomixis is furtherdivided into diplosporous and aposporousmechanisms, depending on whether theunred~ced embryo sac arises from amodification of the meiotic apparatus or froma separate cell, respectively. Intermediatesbetween the two forms of gametophyticapomixis have been reported (Gustafsson1946), indicating that they possibly representmodifications of a single developmentalmechanism. For more detailed descriptions ofmechanisms, see Nogler (1984a) andKol tunow (1993). Most current research effortson native apomixis focus on gametophyticmechanisms, including studies of bothaposporous and diplosporous species.

Model Sy" ...... s..dy til. Geoetk, aN Dew...p......1llo!ovY o. Apaonllls 113For an experimental system using a nativeapomict, it is convenient to use a facultativespecies in which both maternal andnonmaternal ("aberrant") progeny can beharvested from the same plant. Rutishauser(1948) identified three aberrant types amongthe progeny of these plants: B lIhybrids,derived from the fusion of a reduced egg celland sperm nucleus; Bill hybrids, derived fromthe fusion of an unreduced egg cell and spermnucleus; and polyhaploids, which arisethrough parthenogenesis from a red uced eggcell. Aberrant progeny are often very usefulfor studying the genetics of apomixis. Hybridprogeny permit the evaluation of cytoplasmicinheritanceand the hybridization of pollensterilebiotypes such as interspecific hybrids(Savidan et al. 1994). Similarly polyhaploidshave been used to study the expression ofapomixis in diploids (Nogler 1982; Bicknell1997).Competence to form both meiotic andapomeiotic seed is also invaluable for mutationscreening. Dominant inheritance (see below)suggests that the selective inactivation ofeitherdevelopmental pathway by a mutation willlead to the exclusive expression of the other.This dual competency provides a usefulinternal control. The continued formation ofat least one class of seed indicates that themutation(s) is specific to an event(s) associatedwith apomixis and that related requirementsfor floral development, megagametogenesis,and embryogenesis remain intact. Similarly, asfacultative mechanisms incorporate a balancebetween the utilization of paralleldevelopmental pathways, they can be used tostudy factors that influence that balance, suchas the impact of physiological stress and ofinteractions between genetic modifiers.Finally, as one clear goal of this research is toincorporate apomixis into crop populations, afacultative mechanism is likely to provide themaximum flexibility for farmers and plantbreeders alike. Fortunately, most gametophyticapomicts of all types appear to be facultative,although they differ in the relative importanceof the meiotic and apomeiotic pathways ofseed formation.The formation of the endosperm may also bean important consideration in the choice of anexperimental system. The endosperm mayeither form spontaneously, as in"autonomous" apomicts, or require thefertilization of the polar nuclei by a spermnucleus (pseudogamy). As pseudogamy onlyrequires spontaneous embryogenesis and notspontaneous endosperm formation,pseudogamous species may be consideredpotentially simpler models to study. Thispossible advantage, however, is offset byexperimental constraints associated withpseudogamy. In these plants, pollination isrequired for seed formation, so it becomesnecessary to demonstrate that the appliedpollen led only to the fertilization of the polarnuclei and not of the egg. This difficulty hasbeen largely overcome, however, in speciesthat can be assessed for apomixis using acorrelated cytological character, such as callosedeposition in diplosporous species ofTripsacum (Leblanc et al. 1995a) and four-

114 ROil A. 8idlooll(Sherwood et al. 1994); Brachiaria sp. (Valle andSavidan 1996); Hi.:racium (Gadella 1991;Bicknell et al. 2000); and diplospory inTripsaClim dactylaides (Leblanc et al. 1995b;Grimanelli et al. 1998); TaraxaClim (van Dijk etal. 1999); and Erigeron (Noyes and Rieseberg2000). Earlier work by Dopp (1939) indicatesthat it may even be true for the fern Dryapteris.Simple inheritance would be particularlyvaluable for analyzing the molecular biologyof the trait.For the advancement of a molecular researchprogram, it is particularly advantageous if amodel system can be genetically transformed.This permits the introduction of marker genesto follow segregation and recombination andmutagenic sequences such as T-DNA tags andtransposons to assist in the cloning ofsequences associated with the expression ofapomixis. Furthermore, transformationpermits expression studies about putativeregulatory elements by their fusion to markersequences and the functional testing ofputative control genes by their introductioninto sexual or mutant genotypes.It would also be preferable to use a specieswith a small genome size, such as Arabidapsis,to facilitate the identification of critical loci.Similarly, an ideal system would already becharacterized with respect to genomicsequence and mapped morphological andmolecular markers (deletions, translocations,RFLPs, RAPDs, transposons, etc.) for thelocalization of loci. Finally, it would beadvantageous, although not essential, to usea crop species to assist in the transfer ofresearch findings into practical outcomes.Experimental MethodsThe success of any program on apomixis willdepend on both the natural attributes of thespecies used and on the power of thetechniques available to manipulate itexperimentally. For some species, techniqueshave already been documented, while forothers they would need to be established.Tissue culture and transformation techniquesare particularly valuable. They permit therapid micropropagation of selected types, theretention of somatic genotypes through invitro shoot regeneration, and the maintenanceof valuable genotypes through long-term coldstorage. It is also possible to recoverinterspecific hybrids using embryo rescue,and to manipulate ploidy in culture throughthe use of anther or microspore culture toisolate reduced genotypes and by colchicineapplication for chromosome doubling.As the initiation of the embryo sac is the firstdecisive event of the female reproductivephase, a thorough embryological analysis isan indispensable base for an investigation ofapomixis, while embryology continues to benecessary at all stages of the research. Reliablecytological techniques are therefore essentialfor this work. Fortunately, this is one area ofapomixis research that is well documented,particularly with respect to the use ofversatile, routine clearing techniques (Leblancet al. 1995a) that enable the use of thicksections to visualize the complex internalstructure of the ovule.Quantifying ApomixisMost studies of apomixis require a methodfor determining the presence of apomixis, andwhen possible, for quantifying its extent,either with respect to prevalence in apopulation or level of expression in a singlegenotype. Measuring apomixis, however, isa difficult task. When considered as a geneticevent, apomixis is reproduction through seedwithout either allelic segregation orrecombination. Assessment, therefore, shouldbe strictly based on whether allelicrearrangement occurs. This is seldompractical experimentally and most systems forassessing apomixis use a correlated character.The reliability of these correlations is clearly

Model Sy.tems to Study the Geneti" ond Developmental Biology .f Apomixi. 11 5an important concern in interpreting datacollected from any system. For example, inmost apomicts demonstrating apospory of thePanicum-type, maternal embryos typicallyarise from four-celled, unreduced embryo sacs,while reduced embryo sacs are usually eightcelled.Apomixis can therefore be readilyscored in these plants by determining therelative abundance of the different structures.The correlation between the form of theembryo sac and nuclear state appears to bevery good, but the technique is also based onthe assumption that the number of unreducedembryo sacs is a direct reflection of the numberof maternal embryos that will reach maturity.As the assessment is conducted before therequisite events involved in apomixis haveoccurred, an overestimate of intact apomixisis possible. In contrast, asexual seed formswithout pollination in autonomous apomicts,therefore, apomixis can be quantified in thesespecies from the seed that sets withoutpollination. This is attractive experimentally,but it presents a conservative bias becauseindividuals with only partial apomixis areeasily scored as either sexual or sterile. It isclear in every case that results should beverified by either using more than one methodfor quantifying apomixis or by establishing thecredibility of the results with either anembryological or a genetic study of the system.Candidate SystemsTwo approaches have been taken indeveloping model systems for studyingapomixis: (i) the modification of an establishedmodel plant species, and (ii) the experimentalmaniplliation of a known apomict.The first approach has the advantage ofimmediate access to extensive records on thegenetics, biochemistry, and developmentalbiology of the system. Experimental methodshave been devised, genotype colle(tion~ art'available, DNAlibraries, probes, map markers,and sequenced genomic fragments arcaccessible, and existing laboratory collaborationsare present for cross-referencingfindings. Conversely, existing model plantsystems have been chosen for reasons otherthan apomixis. This approach, therefore,dictates targeting the mechanism ofapomixis,depending on the nature of the availablemutations. It is also based on the assumptionthat apomixis is a derivative of sexualityand / or tha t it can be derived artificially fromsexuality. While this currently appearsintuitive, it should be noted that thisassumption remains unproven.In contrast, the use of a naturally apomicticspecies enables the selection of both theapomictic mechanism for its applicability andthe plant type for its amenability. The modelplant systems currently used to studydevelopmental biology, however, are obligatesexuals. Consequently, this approach requiresthe establishment of experimental methods forthe selected species, the collection andcharacterization of critical genotypes and otherexperimental tools, the maintenance of thosecollections, and the establishment of newcollaborative networks.Modification of an Existing SystemSeveral genera have been used as models tostudy plant reproductive biology includingAlltirrlzillllnl, Arnbidopsis, Capsella, Dauctts,Hordellm, Lentna, Lo/illlll, Lycopersicum,Nicotwna, Perilla, Petllnia, Pharbitis, Sinapis,Xantlzilllll, and Zea. Recently, with an increasedemphasis on molecular genetics, the mostimportant model system has become thecruciferous species Arabidopsis tlwliana, withless emphasis on the solanaceous generaLycopersiclInI, Nicotiana, and Petllllia, and themonocotyledonous genus Lea. These plantsreprod uce through seed by obligate sexuality.Two approdches have been taken to useobligate sexual species as models for apomixisresearch.

116 R... A. IkbeIlinduced mutation or by the accumulation ofmutant alleles. A second approach has beento attempt the transfer of the trait byintrogression from an apomictic relative.Two important advantages of mutagenesis arethat it can be conducted on a species withoutany known close apomictic relative and it hasthe potential to rapidly provide theexperimental material required in acharacterized genetic background. Theprincipal difficulty with this approach is thatmethods must be developed for screeningmutant populations for apomixis without anyforeknowledge of the developmentalmechanism(s) that will arise. In particular, itis not known whether a single mutagenicevent can cause an unambiguous phenotypicchange that can be identified as either an intactapomixis or a recognizable component of thetrait. The screening technique must thereforebe based both on the appearance ofdevelopmental anomalies that arecharacteristic of apomixis and, ultimately, onthe retention of the maternal genotype. Onegood example of this approach is the mutantscreening of Arabidopsis thaliana for mutationsleading to the formation of seed withoutfertilization (fie andfis mutations) (Ohad et al.19%; Chaudhury et al. 1997). The screens werebased on the identification of plants thatdeveloped elongated siliques without priorfertilization. The advantages of Arabidopsis forthis approach are highlighted by the authors'use of male sterile mutants (popl and pistillata)to avoid self-fertilization, and the use of GUSas a paternal marker in test crosses with themutants. The description of these mutants isa particularly exciting outcome, providingevidence for the involvement of chromatinremodeling factors in the control of celldivision at the time of fertilization(Grossniklaus et al. 1998; Kiyosue et al. 1999).Furthermore, recently reported data indicatesthat methylation is critical to the regulation ofthese genes, specifically with their expressionduring gametogenesis and earlyembryogenesis and that this may be associatedwith mechanisms of maternal inheritance(Vielle-Calzada et al. 1999).When mutations that alter megasporogenesis,megagametogenesis, and embryogenesis arealready known, it may be possible to createan apomictic mechanism by the accumulationofappropriate alleles within a single genotype.In potato, the homozygous representation ofthe ds-l allele Significantly reduces chiasmatafrequencies on all chromosomes during bothmegasporogenesis and microsporogenesis,leading to high levels of 2n gametes throughfirst division restitution (FDR) Oongedijk et al.1991). If this could be combined withparthenogenesis, true potato seed could begenerated through a synthetic diplosporicmechanism (Hermsen et al. 1985).When an apomictic relative is available, it maybe possible to transfer the trait into acharacterized system by introgression. Thistypically involves a backcrossing programusing the sexual species as the recurrentparent. Introgression has been used to attemptthe introduction of apomixis into several cropspecies, normally as part of a cropimprovement program (Asker and ]erling1992). Examples of this approach includeattempted transfers from PennisetumsquamHlatum to cultivated pearl millet(Dujardin and Hanna 1985), Elymus rectisetusto Triticum (wheat) (Torabinejad and Mueller1993), and Tripsacum dactyloides to Zea (maize)(Leblanc et al. 1995b). Similarly, it may bepossible to transfer the trait from apomicticcrucifers, such as Arabis holboellii (Roy 1995)or Draba oligosperma (Mulligan and Findlay1969), to Arabidopsis thnliana to take advantageof the versatility of this model system. UnlikesyntheSiS, transfer has the advantage that theprogram is based on a known, functionalapomictic mechanism. Furthermore, the

Model 5'11." to Stody til. Gett.,kl _ Dn........tal Biology of ApotIIIII.117inheritance of that trait can be monitoredthrough each generation, providinginformation on its genetic basis in the systemunder study. Finally, as indicated above, whenthe recipient species is a crop, the product maybe of immediate commercial value. Theprincipal disadvantage of transfer is that theavailability of apomictic relatives typicallydictates the mechanism used. It has alsoproven difficult to incorporate the trait intoobligate sexual crops with a significant levelof expression. The reason for this difficulty isunclear. It may be associated with theinheritance of modifiers or an importantassociation with polyploidy, which is lostduring the backcrossing program. Alternatively,the difficulties encountered may bemore a problem of experimental scale. Mostcrop species do not have a close apomicticrelative so introgression requires widecrossing. This often results in poor fertilityamong the progeny and little, if any, crossingover during meiosis. Large populations needto be formed and assessed for ploidy andapomixis, but traditional methods ofchromosome counting and thin sectioning aretoo labor-intensive to be practical for mostresearch teams. Recent advances in DNAquantification through flow cytometry andanalytical cytology using clearing techniquesare overcoming these difficulties. Over the pasttwo years, researchers in the maizeI Tripsacumprogram have screened more than 200,000plants for chromosome number using flowcytometry (Savidan, personal comm.) and theexperimental populations have beenadvanced to the BC s generation.Development of a Model System from anExisting ApomictApomixis occurs throughout the plantkingdom. Species utilizing various forms ofasexual reproduction that involvegametopnytic structures have been recordedamong the algae, pteridophtyes, and in morethan 400 species of flowering plants from morethan 35 families (Asker and Jerling 1992;Carman 1997).Which is the best to study? Different speciesclearly have different advantages. The uniquebiology of ferns, for example, presents someunusual opportunities to study apomixis(Sheffield and Bell 1981). The events ofmegasporogenesis and megagametogenesisare physically separated in these plants,permitting the study of the individualcomponent processes of apomixis in isolation.Unlike the angiosperm embryo sac, the freeliving fern gametophyte is isolated fromparental influence, and, in some systems,parthenogenetic development of thesporophyte can be induced in vitro (Sheffieldand Bell 1987). Furthermore, the sporogenictissue is relatively exposed in ferns,simplifying the study ofevents associated withthe avoidance of meiosis. It is interesting tonote that Manton (1950) reported thatunreduced spores of Crytomium arose fromtissue immediately adjacent to meiotic tissuein the sporangium, a situation closelyanalogous to the development of aposporousinitials in the angiosperm ovule. Finally, as thefern sporophyte develops without the need foran endosperm, difficulties associated with thedevelopment of that tissue do not arise. Thereare, ofcourse, several limitations in using fernsas model systems, particularly for a molecularstudy of development. They are often largeslow-growing plants that can be difficult tocultivate, and they present some realchallenges when conducting controlledfertilizations with motile sperm cells. Finally,very little is known of the molecular biologyof this group, which would impede progressin any molecular study of apomixis.Apomixis occurs throughout the angiospermsincluding representatives of bothmonocotyledonous and dicotyledonousgenera. Many of the most comprehensive

118 Ron A. Mlltllstudies of apomixis used monocotyledonousspecies, principally relatives of the cereals, suchas ELymlls (Torabinejad and Mueller 1993),Tripsacum (Leblanc et al. 1995a), Panicum(Savidan 1982), and important forage grasses,such as Pennise/um (Dujardin and Hanna 1985;Ozias-Akins et al. 1998)), Brachiaria (Valle et al.1994), and Paspalum (Bonilla and Quarin 1997).The inheritance of apomixis has beenparticularly well characterized for Panicummaximum. Savidan (1990, 2000) partiallyascribed his succe~~with this system to the useof apomicJic-aria~exual forms from within thesqmespecies and at the same plOidy level. Thisimportant advantage is shared with only a verysmall number of studied apomictic taxa. Incontrast, most known apomictic species appearto have either evolved away from theirimmediate sexual progenitor(s) or theirprogenitor(s) no longer exists. As a result,apomicts have often been crossed wi th a relatedbut distinctly different sexual species. In suchstudies, the subsequent analysis of the progenymust consider the inheritance of the breedingsystem while also allowing for unrelated effectsresulting from interspecific hybridisation.Despite this caveat, however, mappingstrategies have led to the isolation of molecularmarkers linked to apospory in the grass genusPennise/um (Ozais-Akins et al. 1993, 1998). In asimilar approach, colinearity between grassgenomes has also been used to propose alinkage to apomeiosis in Tripsacum (Leblanc etal. 1995b) and Brachiaria (Pessino et al. 1997;see Grimanelli et a!., Chap. 6).Some dicotyledonous species have been usedpreviously as model systems, most notablyPo/en/ilia (Rutishauser 1948; Asker 1970, 1971);Taraxacum (Richards 1973; van Dijk et al. 1999);Hypericum (Noack 1939); Rammculus (Nogler1984a); and Hieracium (Bicknell et al. 2000). Theinheritance of apospory in Ramll1cultls has beenparticularly well studied (Nogler 1984b)through four generations of crosses andbackcrosses. The results indicate that aposporyis inherited as a simple dominant Mendeliantrait in this system, however, the alleleconferring apomixis could only be transferredin a diploid or polyploid gamete. Nogler notedthat this mechanism could explain the veryclose association observed betweenpolyploidy and apomixis in native systems.In contrast, recent results with thediplosporous genus Taraxacum (van Dijk et al.1999) indicate that a two loci model better fitsthe data obtained from controlled crosseswithin this taxon. Taraxacum is an attractiveexperimental system because it is a wellstudied ecological model (Richards 1986) andit has also been successfully transformed (Songand Chua 1991).One dicotyledonous taxon that appears to bewell suited for use as a model system isHieracium subgenus Pilosella (Koltunow et al.1995). The plants are small herbaceousperennial daisies that are easily propagatedand maintained in the greenhouse. Hieraciumis a long-day plant, flowering in response toextended photoperiods (Yeung 1989).Photoperiodicity is a very useful experimentaltool as it enables the programmed productionof flowers at any time during the year usingday-length extension lighting. Both H. pilosellaand H. auran/iacum set seed within 3-4 monthsof germination, allowing 3-4 generations perannum,(Bicknell 1994a).Apomictic biotypes of Hieracium subgenusPilosella develop seed by facultative aposporycoupled to autonomous endospermdevelopment. Pollination is therefore notrequired for the formation of maternal seed,and apomixis can be scored by quantifying theseeds that set after the excl usion of pollen. Aswith Taraxacum, the simplest method ofexcluding pollen is to decapitate the immaturebud, removing both the anthers and stigmas(Ostenfeld 1906; Richards 1986).

Model Sy.l.ms 10 Study I~. Goo.li

120 I... A. BkbollLeblanc, 0., M.D. Peel. J.G. Cormon, and Y. Sovidan. 19950.Megasporogenesis and megagometogenesis in several Tripsacumspecies (Pooceae). American Journal of Botany 81: 57-63.Manton, I. 1950. Apogamous ferns: The general phenomenon. Problemsof Cytology and Evolution in the Pteridophyta. Cambridge, U.K.:Cambridge University Pr~.Mulligan, G.A., and J.N. Findloy. 1969. Sexual reprodlKlion andagamosperrny in the genus Droba. Canadian. Journal of Botany 48:269-70.Noock, K.L 1939. aber Hypericum-Kreuzungen VI.Fortphflanzungsverhiiltnisse und Bastarde von Hypericum perforatumLZ. Indukt. Abstamn Vererbungslehre. 76: 569-601.Nogler, GA 1982. How to obtain diploid apomictic Ranunculus auricorousplants not found in the wild state. Bot. HeIvet. 92: 13-22­--.19840. Gametophytic apomixis. In 8.1.1. Jahri (edJ,Embryology ofAngiosperms. New York: Springer-Verlag..Pp.475-518.--.1984b. Genetics of apospary in apomictic Ranunculusauricomus. V. Conclusion. Botan. Helvet. 94 (21: 411-22.Noyes, R,D., and L.H. Rieseberg_ 2000.lwo independent loci cootrolagamospermy (apomixisl in the Iriploid flowering plant Erigeronannuus. Genetics 155(1 ): 379-90.Ohad, N., L. Morgossion, Y. -c. Hsu, C. Williams, P. Repeni, and R.L flS(her.1996. Amutation that allows endosperm development wiThoutfertilization. Proceedings of the National Academy of Science (USA)93: 5319-24.Ostenfeld, C.H. 1906. Experimental and cylologicalstudies in the Hieraoo.I. Cos/ration and hybridization experiments with some species ofHieracia. Bo/. Tidsskr. 27: 225-48.Ozias-Akins, P., E.L Lubbers, W.W. Hanna, and J.W. McNay. 1993.Transmission of the apomictic mode of reproduction in Pennisetumco-inher~ance of the tra~ and molecular morkers. Theoretical AppliedGenetics 1993 (8S): 632-38.Ozias-Akins, P., D. Roche, and W.W. Hanna. 1998. TIght clustering andhemizygosity of apomixis-linked molecular markers in Pennisetumsquamulatum impnes genetic ronlral af apospory by a divergenllocusthat may have na allelic form in sexual genotypes. Proceedings of theNational Academy of Science (USA) 95: 5127-32.Peel. M.D., J.G. Corman, and O. Leblanc 1997. Megaspore callose inapomictic buffelgrass, Kentucky bluegrass, Pennisetum squamulatumFresen, Tripsacum L, and weeping lovegrass. Crop Science 37: 724­32.Pessina, S.c., J.P. OJ1iz, O. Leblanc, C.8. do Volle, C. Evans, and M.D.Hayward. 1997. Identification of a maize linkage group related 10opomixis in Brachiaria. Theoretical Applied Genetics 94: 439-44_Richards, AJ. 1973. The origin 01 Taraxacum ogomospecies. BiologicolJournal of the Linneon Society 66: 189-211.--.1986. Plont Breeding Systems. London: George Allen andUnwin.Roy, B.A. 1995. The breeding systems of six species 01 Arabis(Brassicaceae). American Journal 01 Botany 82(7): 869-77.Rutishauser, A. 1948. Pseudogamie und Po~morphie in der GanungPo/en/ilIa. Arch. Julius-klaus-Stihung f. Vererb. -FoTSCh. 23: 267-424.Sovidan, Y. 1980. Chromosomol and embryological analyses in sexual xapomictic hybrids 01 Panicum maximum Jocq. Theoretical AppliedGenetics. 57: 153-56.--.1982. Nature et herooile de I'apomixie mez Panicummaximum Jacq. Travaux &Documents ORSTOM 153: 1-159.--.1989. Another 'working hypothesis" for the control 01parthenogenesis in Panicum Apomixis News/effer 1: 47-51.--.1990. The genetic conlrol 01 apomixis. Apomixis Newsleffer2:24.--.2000. Apomixis: Genetics and Breeding. In J. Janick (ed.1,Plant Breeding Reviews. New York: John Wiley ond Sons. 18: 13-86.Savidon, Y., O. Leblanc, and J. 8eJ1haud. 1994. The moize-Tripsacum BC2production -gening late to win the bet? Apomixis Newsleffer 7: 27­28.Sheffield, E., and P.R. 8ell. 1981. Experimental studies 01 apospary inferns. Ann. Bot. 47: 187-95.--.1987. Current Studies of the Pleridophyte life Cycle. TheBotonical Review 53: 442-90.Sherwood, RI, c.c. Berg, and BA Young. 1994. Inheritance al aposparyin buffelgross. Crop Science 34: 149~94.Song, Y.C.N., and N.-H Chua. 1991. TISSue wlture and genetictransformatian of Dandelion. Ada. Holt. 289: 261-62.Torabinejad, J., and RJ. Mueller 1993. Genome ana~is 01 intergenerichybrids of apamictic and sexual Australian Bymus species wiTh wheat,barley and rye: implication for the transfer 01 apomixis to cereals.Theore/icol Applied Gene/ics 86: 283-94.Valle, CB. da, CGlienke, and G.O.c. Leguizamon. 1994. Inheritance ofapomixis in Brachiaria, atrapical lorage gross. Apomixis News/effer7: 42-43.Valle, C.B. da, and Y. Sovidan. 1996. Genetics, cylogenetics andreproductive biology 01 Srachiaria. In J.W. Miles, B.L Maass and C. B.da Valle (eds.1, Brachiaria: Biology, Agronomy, and Improvement.(oli, Colombia: C1AT!EMBRAPA. Pp. 147-163.van Dijk, PJ., I.ca. Tas, M. Falque, and T. Bakx-Schotman, 1999. Crossesbetween sexual and apomictic dondelions (Toraxacum).11. Thebreakdown 01 opomixis. Heredity 8316): 715-21.Viella-Calzada, J-p, J. Thomas, et al. 1999. Maintenance 01 genomicimprinting at the Arabidopsis medea Lows requires zygatic 001.11activity.oGenes and Oevelopment 13(22): 2971-82.Yeung, U. (1989). Hieracium. CRC Handbook of Flowering. 80ca Ratan,Aarida: CRC Press.

Chapter 9Screening Procedures to Identify andQuantify ApomixisOLIVIER LEBLANC AND ANDREA MAZZUCATOIntroductionMendel was no doubt puzzled when hestudied the progenies from crosses in theHieracillm species to confirm his Pisumexperiments: F 1families were highly variableand some F] hybrids produced homogeneousprogenies (Mendel 1870). So what about theworld famous laws ofinheritance worked outon Pisum? In fact, the laws still held. Whatthe great geneticist did not know was that hehad made the first progeny tests withapomictic species.The term "apomixis" initially covered all ofthe mechanisms of asexual reproduction(Winkler 1908), but today it is applied strictlyto asexual reproduction through seeds(NogleI' 1984). There are two main types ofapomixis, based on the origin of the embryo:adventitious embryony, in which the embryoforms directly from the sporophyte (thegametophyte phase is bypassed), andgametophytic apomixis, in which the embryodevelops parthenogenetically from anunreduced female gametophyte (Gustafsson1947; Stebbins 1950). Apomicticdevelopments bypass meiosis andfertilization, the bases of sexual reproductionand genetic recombination, and therefore,offspring are genetically identical to themother plant.Apomixis has been widely identified in theplant kingdom (Asker and Jerling 1992;Carman 1997), and occurs in families ofeconomic importance (Rutaceae, Poaceae,Rosaceae). Moreover, it appears to be a verycommon mode of reproduction in thePanicoideae subfamily of Poaceae (Brown andEmery 1958), which includes several majorgrain crops. Apomictic processes are stillpoorly understood, but the potential impactof apomixis on agriculture appears great,provided that it is proven to be ecologicallysafe (Vielie-Calzada et a1. 1996; van Dijk andvan Damme 1999; Toenniessen, Chap. 1).Identifying sources of apomixis, understandingits inheritance, and breeding andmanipulating apomictic species will requirereliable and efficient procedures to screen formode of reproduction.This chapter concentrates on identifying andquantifying gametophytic apomixis, but forthe most part, the procedures are the same foradventitious embryony. After presenting basicfeatures of apomixis, screening procedures forthe reproductive mode are described and thevarious challenges encountered by scientistsworking with apomixis are discussed.Apomictic Mechanisms asPotential Screening IndicatorsSeed production through gametophyticapomixis requires production of embryo sacswith unreduced nuclei unreduced femalegamete (no red uction of chromosome numberor apomeiosis), followed by embryogenesiswithout fusion of nuclei of the male and femalegametes (parthenogenesis). The regulatoryand quantitative aspects of parthenogenesisin unreduced egg cells have been poorly

122 Olivier L.bla.,.ncI A.d,ea M.nu,ol.documented (Asker 1980; NogleI' 1984; Mogie1988), but fertilization-independant mutantsfor both seed and endosperm development,recently described in Arabidopsis fhaliana(Ohad et at. 1996; Chaudhury et aJ. 1997;Grossniklaus et at. 1998), might provide newinsights into embryogenesis in apomicts.Pseudogamy is the most common path of seeddevelopment, but autonomous apomixisoccurs in some cases (NogleI' 1984). Bycontrast, apomeiosis is well documented andmay follow different pathways. Severalreviews (NogleI' 1984; Asker and Jerling 1992;Koltunow 1993; Crane, Chap. 3) providedetailed descriptions of most types ofapomixis that occur in the wild.The two types of apomeiosis-apospory anddiplospory-and their characteristics arebriefly described in this chapter in order tohighlightdifferences with sexual reproductionthat are pertinent for the development ofscreening tools. In adventitious embryony,both megasporogenesis and megagame­~20... or ...oral .omatk n PankuM-typeIIIKOIIar ,01. d...lop °2n Part"--- tk_Inltlat. IS. ~ •....."...._C""'Y .tiIdov.loP"'"t 01 omb

Sue.oiog Pro

124 Olivier lt~1ao< aod Aock.. Mau"atothe ovule, unreduced egg cells may not bereceptive. This loss of receptivity is not yet wellunderstood, but several hypotheses have beenproposed, including chemical or mechanicalbarriers (e.g., a complete cell wall around theegg) and a temporal window of receptivity,among others.Consequences of Apomictic SeedFormationIn sexual reproduction, the two gametes thatfuse are produced through meiosis. Sexualdevelopment allows genetic recombinationand segregation, thereby enhancing geneticdiversity. Aside from strict autogamy and fromthe very specific case of permanenttranslocation heterozygosity (El1strand andLevin 1982), offspring from sexual plants arenew genotypes. Apomictic pathways arecharacterized by unreduced egg cel1parthenogenesis, resulting in offspring that areexact genotypic replicas of the mother plant.However, genetic recombination may occurduring apomictic reproduction in plants thatshow partially synaptic and restitutionalmeiosis or somatic DNA rearrangements(Richards 1997).Complete (100%) maternal progenies arerecovered when the mother plant reproducesthrough obligate apomixis. But generally,apomixis is facultative and progenies comprisevarious types, each resulting from a differentTable 9.1 The four theoretical offspring dasses in progeniesfrom facultative pseudogamous apomicts. Note thatapomeiotic mechanisms can induce chromosome losses andresult in unbalanced unreduced female gametes.Egg cell originReduced megasporealter meiosis: ngameteApomeiosis (aposporyor diplospory):2n gameteEgg cell development afterFusion with asperm cell (+n)Sexualityn+n offspringNew genotypes"Genomic accumulation"2n+n offspringNew genotypesParthenogenesis(+0)(Poly)haploid productionn+O offspringNew genotypesApomixis2n+0 offspringMaternal genotypescombination of failure or success in meiosisand fertilization (Table 9.1). A fairly strictgenetic control for both the formation ofunreduced ES (reviewed by Sherwood, Chap.5) and the degree of apomixis (Asker 1980) hasbeen reported in most taxa studied.Levels of Screening andRelated ToolsThere are several indicators of apomixis,including high frequency of multipleseedlings, high seed fertility in plants expectedto be sterile (e.g., wide hybrids, triploids,autopolyploids, and aneuploids), homogeneousprogenies, etc. (Bashaw 1980; Hannaand Bashaw 1987; den Nijsand van Dijk 1993).They are sometimes difficult to use in the caseof wild materials and, in all cases, furtherinvestigation is required to assess apomixistype and level of expression. The relativeadvantages or disadvantages of the screeningprocedures presented here are discussedfurther in "Choosing Suitable Procedures."Analyses at the Plant Level1. Molecular markers cosegregating withapomixis. To date, the identification ofisozymic or molecular markers strongly linkedwi th apomixis is the only procedure fordetecting apomixis prior to flowering.Molecular marker-based analyses inapomicts were conducted either to.investigate the molecular basis ofapomixis, to assist in transferringapomixis into crops, or to ultimatelyisolate the gene(s) responsible for itsregulation. Segregating progenies orbulk segregant analyses were usedafter determining the reproductivebehavior on the basis ofcytoembryological observations orprogeny testing. Because ofconflicting results, debate continuesover whether apomixis is controlledby a single locus or by multiple loci.

s"...1og Pro

126 OIlrier Ltblaooc ood AocI...Mo,,_Paraffin sectioning methods (Figure 9.2)combined with staining (e.g., safranin-fastgreen stain, Johansen 1940; or aniline blue,Russel 1978) have been used over the lastcentury for cytoembryological studies ofreproductive development and in apomixisresearch (e.g., Snyder 1957; Voigt and Bashaw1972; Burson et al. 1990). However, preparingparaffin sections is arduous and timeconsuming, and interpretations may bedifficult. Clearing procedures (Figure 9.3) weredescribed more than 90 years ago (Strasburgerand Hillhouse 1900), but have been recentlyrediscovered and greatly improved (see Crane,appendix of Chap.3). They do not requiresectioning or squashing and thus allow ovulesto be observed in situ in three dimensions,making interpretations easier than from aseries of sections. Squashing techniques,generally combined with staining, weredeveloped for studies of megagametogenesisor megasporogenesis in various species(Hillary 1940; Bradley 1948; Saran and de Wet1966; Darlington and La Cour 1966), but haveproven only moderately successful.Nevertheless, improved squashing techniquescombined with clearing procedures providegood results when analyzing female meiosisOongedijk 1987; Kojima et al. 1991a; Kojimaand Nagato 1992b).Clearing techniques using non-aqueous fluids(Herr 1971; Young et a1.1979; Crane andCarman 1987) now represent the best tool forobserving ovule details during bothmegasporogenesis and megagametogenesis inaposporous and diplosporous materials.Procedures combining Mayer's hemalumstaining with methyl-salicilate clearing havebeen successfully used for observations withinwhole ovules of Solanum (Stelly et al. 1984) andMedicago (Tavoletti et al. 1991). Thesetechniques are of great interest forembryological analyses in apomicts becausethey do not require the use of special optics.Clearing procedures combining aqueoussolution (sucrose, KI) and aniline blue haverecently been developed for observation ofcallose deposition during megasporogenesis(Carman et al. 1991; Leblanc et al. 1995b; Peelet al. 1997).3. Egg cell parthenogenetic capacity. Egg cellsproduced through apospory or diplosporyshould be better able to differentiateparthenogenetically than those producedthrough sexual development, because of theapparent linkage between the two steps ofapomictic development. Matzk (1991) recentlyproposed a new procedure to identify andquantify parthenogenesis for a wide range ofcool season grasses. The technique, known asthe auxin test, involves applying a syntheticauxin compound a few days before anthesisto induce parthenocarpic development inunpollinated ovaries. Auxin induced grainswill lack endosperms, because the fusion ofthe sperm and polar nuclei is no longerpossible, but egg cells with parthenogeneticcapacities will develop into embryos. Studiesin Poa pratensis using the auxin test to estimatethe degree of parthenogenesis in variousgenotypes showed good reliability and lowvariation across years and environments(Mazzucato et al. 1996).Progeny analysisIn classical progeny testing, one compares themother plant with its offspring and/orevaluates heterogeneity within progeny.Offspring from apomictic plants are expectedto be genetically identical to the mother plant;therefore phenotypic identity with thematernal type suggests apomicticreproduction, whereas variations indicatesexuality, recombination, and/or fertilization.Traditionally, progeny tests based on grossmorphology have been used in apomixisresearch because they are easy to perform (e.g.,Duich and Musser 1959; Burton et al. 1973;Cadella 1983), but many other descriptors may

5treenlng P'o

128 or..Ie, leblal( allll Aodrea Mam"',"be useful for progeny tests and shouldtherefore be considered, Progeny tests areusually performed on seedlings or fully-grownplants, but other tissues from earlierdevelopmental stages, such as ovaries,endosperms or seeds, can also be used.1. Analysis of pollinated ovaries or seeds.Determining ploidy levels in pollinated ovariesor seeds (albuminated) provides informationon both reduction (meiosis) and fertilizationevents. Ratios between endosperm andembryos and between female and malecontributions to the endosperm in apomictsoften differs from those in sexual plants exceptfor the Panicurn-type a posporousdevelopment (Figure. 9.1). For many otherapomictic pathways, these ratios differ. Forexample, endosperms found in tetraploiddiplosporous apomicts are higher than in theirsexual tetraploid counterparts for identicalpollen donors (i.e., lOx versus 6x if the pollenis 2x); endosperm/embryo ratio forautonomous apomixis is 2:1 and 5:2 [(4x + 4x)+ 2x / 4x + OJ for tetraploid pseudogamousapomicts (tetraploid pollen donor).Fertilization by unreduced pollen (Chao 1980;Huff and Bara 1993) and endopolyploidization,which sometimes occurs during endospermdevelopment, is also possible and may furthercomplicate analyses. However, endospermploidy level(s) may suggest apomicticreproduction or allow the quantification offacultative apomixis. Nevertheless, it cannotreveal the precise nature of the apomicticmechanisms involved.Ploidy level in fertilized ovaries or immatureseeds cannot easily be determined usingclassical chromosome counting methods, butflow cytometry now permits rapidmeasurement of DNA content in a variety ofplant tissues, including single embryos, youngendosperms, or seeds (Galbraith et at. 1983;Kowles et al. 1990; Hignight et al. 1991).Analyses in numerous apomictic species haveproven flow cytometry to be a rapid andreliable procedure for determining the modeof reproduction (Mazzucato et al. 1994;Brautigam and Brautigam 1996; Grimanelli etal. 1997b; Gupta et al. 1998; Naumova et al.1999; Matzk et al. 2000). Another option forDNA content estimation of the endospermnuclei is to combine staining with 4'-6­diamidino-2-phenylindole (DAPI), Huoresencemicroscopy, and image analysis(Naumova et al. 1993; Sherwood 1995; Cacereset al. 1999).2. Ovule regenerated plants. In tetraploidaccessions of AIlium tuberosum, Kojima andKawaguchi (1989) reported a high frequencyof tetraploid regenerated plants fromunpollinated cultured ovules, suggestingapomixis expression. This indicator could beapplied in screening because, in similarculture media, sexual plants would generatefew (poly)haploids, whereas apomeioticovules would grow mostly into plantlets withthe same number of chromosomes as themother plant.3. Analysis of progeny plants. Progeny testsmust clearly identify either hybrid offspring(n + 0 types are generally poorly represented)or seed production in absence of pollinationwhen pseudogamous apomixis orautonomous apomixis, respectively, aresuspected. Hybrids can be identified using (i)morphological descriptors, (ii) cytologicaldata, and/or (iii) marker analyses, if the originof the progeny is appropriate.Remarks on progeny size. The use ofprogenies from controlled crosses isrecommended. Male parents bearingdiscriminating traits (dominant traits,different chromosome numbers, etc.) shouldbe chosen when available, limiting possibleconfusions between selfed and hybridprogenies. However, open pollinatedprogenies can be used when mother plantsare sufficiently heterozygous to detect

S"...iog Pro...,.. '0 Idnllfy ood a.aotlfy Apomixis 129segregation after selfing and when there issignificant diversity in the surrounding fieldcollection, as is the case for most apomictic species,which are generally polyploid, polymorphic, andhighly heterozygous.Identifying or quantifying apomixis does notrequire the same number of progeny. To detectapomixis, a relatively small number ofprogeny (15­25) can be analyzed. Aberrant rates typically are a:nratios with 'n' the progeny size and 'a' the numberof aberrants observed in the progeny. Statistically,such samplings are binomial; 'p' (aberrant rate) isthe ratio to be estimated for a given value of n(progeny size) on the basis of an observed value fora (number of aberrants detected within theprogeny). Confidence limits for p in a binomialsampling are given in Figure 9.4 for various valuesof n (a = 0.025). Note that for n>30, confidence limitscan be estimated using formulas for the normalFigure 9.4 (onfidence limits (a=0.025) for pin binomialsampling, given a sample fraction a/n. The numbersprinted along the curves indicate the sample size, n. For agiven value of aln (abscissa), limits for p (PA and PB) arethe ordinates read from the appropriate lower and uppercurves (Pr{PA ~ p ~ PB} ~ 1-2a).distribution. Curves shown in Figure 9.4clearly indicate that, up to n = 100, theinformation obtained is poorly signjficantregardless of the value of a. Finally, toobtain good estimations of aberrant rates(i.e., less than 10% confidence limits), itappears that a high number ofindividuals is required.Chromosome number determinationwithin progenies. The sexual or ;:lsexualorigin of offspring is not detectable fromcrosses made at the same level of ploidy,but 2n + nand n + 0 off-types are easilydetected even at the seedling stage.lnterploidy-Ievel crosses could be usedto detect all classes of offspring, butinformation can be biased bydisturbances caused by unstable ploidylevels or ploidy barriers. Chromosomecounts can be made from root tips,microspores, or any somatic tissue usingflow cytometry (Hignight et al. 1991).Detection of seed production in absenceof pollination. Species carrying nonhermaphroditeflowers obviouslyrepresent the easiest situation, the onlyprecaution required being to avoid pollencontamination. Contrarily, emasculationwill be required unless an appropriategenetic system that ensures male sterilitycan be developed. Such systems requirethorough knowledge of the genetics andgenetic stocks of the material understudy, making their application verylimited in natural populations. They havebeen exclusively developed Inexperimental mutagenic populations ofsexual model species (Aradidopsis tha/ulna,Petllnia hybrida), with the aim ofidentifying mutants that reproducethrough autonomous apomixis(Koltunow et al. 1995b; Chaudhury et al.1997; Ramulu et al. 1997).

130 Ofjyier LH......AodrH Mom","Markers lor hybrid dete(tion. Traits undersimple genetic control are ideal for progenytesting by crossing recessive maternalgenotypes with homozygous dominant testers(Hanna et al. 1970; Bashaw and Hanna 1990).Models for estimating levels of apomixis byfollowing marker segregation have beendeveloped (Marshall and Brown 1974),however, recombination can occur withoutfertilization, and the presence of dominanttraits in progeny tells nothing about the originof the off-types (n + n or 2n + n) in the absenceof cytological data. Moreover, identification ofsuch "ideal" markers in apomictic species oragamic complexes is not necessarily easy,because traits in polyploid apomicts aredifficult to analyze genetically.Morphological descriptors are the easiestmeans for conducting progeny tests. If thetester (pollen donor) differs significantly fromthe progeny-tested plant, hybrids will varysufficiently from the maternal type to allowdetection. In the case of selfing, becauseapomicts are generally highly heterozygous,offspring arising through sexuality will varysufficiently from the mother plant to be scoredas off-types. In most species, (poly)haploidsare easily detected because of their pa,:ticularphenotypes and the low vigor they exhibit(Asker and Jerling 1992). However, whenusing morphological descriptors, it is often notpo.ssible to distinguish between sexuality(n+n) and genomic accumulation (2n + n). Butwhen morphological and cytological(chromosome number) data are combined, theidentification of all classes is theoreticallypossible. Analysis of seedlings has the majoradvantages of timeliness and saving space, butthe most informative descriptors for screeningpurposes are usually expressed at maturity.There are few reports of successful progenytesting for morphology on seedlings afterinterspecific crosses (Williamson 1981).Isozymes or molecular markers can be usedto assess variation in progenies (fingerprintinganalyses; Nybom 1996). Finding goodpolymorphic isozyme systems, RFLP probes,or primers for peR as candidates forfingerprint experiments is not a major obstacle.Although genetic analysis is still hindered bypolyploidy, any variation in isozyme or DNApatterns indicates off-type production,provided that somatic recombination does notoccur frequently in the material under study.Esterase and peroxydase were the first systemsused to isolate sexual plants from Panicummaximum (Smith 1972). Apomixis expressionwas also confirmed or quantified usingisozymes in Taraxacum (Ford and Richard1985), Arabis holboellii (Roy and Rieseberg1989), Allium tuberosum (Kojima et al. 1991b),Poa pratensis (Wu et al. 1984; Barcaccia et al.1994), Tripsaatm spp. (Leblanc 1995), and Malussp. (Ur-Rahman et al. 1997).Mazzucato et a1. (1995) showed a slightlyhigher capacity of RAPD markers indiscriminating off-types in progenies from thesame species, when compared with threepolymorphic isozyme systems or with analysisof traditional morphological traits. Althoughstill seldom used, molecular markers havebeen successfully used for progenyfingerprinting (e.g., Poa pratensis: Huff andBara 1993; Barcaccia et al. 1997; Paspalumnotatum: Ortiz et al. 1997).rChoosing Suitable ProceduresAnalyses at the Plant Level versusProgeny Tests1. Nature of the information obtained.Apomixis results from apomeiotic processes(apospory or diplospory) that produceunreduced ESs, and parthenogenetic embryodevelopment from unreduced eggs. Althoughnonreduction and parthenogenesis arethought to be closely linked in apomicts,observations and/or analyses of the plant itself

Sueeolog P........,.. r. Idootily ood Qoaotily ApMixIs 131obviously provides insights only aboutapomeiotic or meiotic events, not about thecomplete process of apomixis. Data on the nextgeneration (progeny test) must be collected tostudy fertilization and parthenogenesis eventsas well as the degree of apomixis. The choiceof the level of analysis (apomeiosis /parthenogenesis / apomixis) depends on theobjectives of the research, i.e., whether onewishes to determine only cytologicalprocesses, study parthenogenesis, orinvestigate apomixis in its entirety.2. Comparing results. Limited information isavailable on diplosporous development, butcytological analyses ofparent plantscomparedwith progeny tests generally show goodagreement between apomeiosis and apomixisscreenings in Eragrostis curvula (Voigt andBurson 1981), Allium tuberosum (Kojima andNagato 1992b), and Tripsacum spp. (Leblanc1995). By contrast, the situation in aposporousspecies appears more complex:cytoembryological analyses generallyrevealed higher sexual potential than didmorphology-based progeny tests in Panicummaximum (Savidan 1982b), Poa pratensis(Nygren 1951), and Bothriochloa-Diclumthium(Harlan et al. 1964). The same tendency wasalso observed by Mazzucato et al. (1996) in Poapratensis, when auxin tests and field data werecompared. However, using progeny tests onmore than 100 Brachiaria F1s, Miles and doValle (1991) classified ten plants that werehighly facultative apomicts as sexual,according to cytoembryological tests. Sexualpotential in aposporous tropical grasses hasgenerally been scored according to theformation of 8-nucleate ESs that may developconcomitantly with several apomeiotic (4­nucleate) ESs. The competition among ESsmorefavorable to apomeiotics (Savidan1982a)--and the possible weakness of certainhybrids that are eliminated early, may explainthe overestimation of sexuality in facultativeapospore as measured using cytoembryology(Clausen et al. 1947; Kojima and Nagato 1992b).This was confirmed by Savidan (1982a) in onePanicum maximum accession: sexual potentialwas estimated using a clearing procedure at22.5%, but only3% of the open pollinated adultprogeny, were off-type5. Elimination of hybridoffspring occurred at germination (-7%) oraftertransferring plants to the field (-12.5%), becauseof their inbred nature (resulting from selfingor hydridization with genetically closegenotypes in the collection). On the other hand,after self- or sib-pollination, the lack ofheterozygous loci in segregation may cause anoverestimation of apomixis, with progeny testsshowing the presence of "apparent apomixis"(Bayer et al. 1990).Screening Procedures:Advantages and ConstraintsUntil recently, screening tools for mode ofreproduction were limited to easy-but-latemorphological progeny tests or skilldemandingand time-eonsuming cytologicalsectioning methods (see Table 9.3). During thepast 15 years, new tools in molecular and cellbiology have made screening for mode ofreproduction more efficient, rapid, and reliable.These techniques include ovary progenytesting, flow cytometry for determining ploidylevel, auxin test, and molecular markers thatcosegregate with reproductive mode. Themajor disadvantage of the new methods is theirexpense. In addition, though the methods seemto agree with cytological and/or fieldobservations, additional data are needed toconfirm their reliability.1. Apomixis identification andcharacterization. As mentioned, apomixis maybe detected in various ways, butcytoembryological observations are ultimatelyneeded to confirm the origin of the ES and todetermine the type of apomixis. Clearingtechniques are now quick and easy but requirethe use of phase-contrast or differential

132 Olivier leblaoc and Aldr.. Mall,,"'ainterference contrast optics, both entailing However, these tests do not require muchconsiderable expense. Stain clearing equipment or technical skill, and can thus betechniques that allow observation of ovule managed everywhere. Their main drawbackdetails under traditional optics are less is that they produce frequent errors becauseexpensive. Molecular markers that cosegregate facultative apomixis occurs more often thanwith apomixis, which enable analysis at earlier previously thought. Moreover, progeny withgrowth stages than cytoembryology, require sexual origin may resemble the mother plantthe development ofspecial plant materials and in morphology, leading to misclassificationprotocols, and the cost of associated supplies and to an overestimation of the degree ofis often beyond the means of many research apomixis. The existence of this gray area ingroups. Moreover, they may not be used with progeny plant classification was reported bymaterials that differ in origin from the Williamson (1976), after extensive progenymaterials used to identify the markers, testing in Paa sp. This makes morphologicalespecially in the case of the highly cross­ progeny tests unreliable when apomixis isspecific RAPDs (Williams et al. 1993). highly facultative, but more efficient asMorphological progeny tests are time- and apomixis expression increases. Early progenyspace-consuming because good descriptors tests using isozymic or molecular markers canare usually expressed in adult plants and a be conducted for apomixis detection on 15-25minimum of 15 to 25 offspring are needed. offspring. Only a few isozyme systems areTable 9.3 Advantages and disadvantages of important procedures for the investigation of modes ofreproduction at the plant and progeny levels. * See Ragot and Hoisington (1993) for RFLP and RAPD costs.Plant level analysesProgeny testsProcedures Cy1oembryo· Molecular markers Auxin lests Chr. counting in Adult Pionts Morphology Fingerprinlinglogy (dearing co·segregating ovaries or seeds Chr. counting'procedures) with apomixis'Information Apomixis type Depends on the Indication of Indication of Off-types of Apomixis idenlification andexpected determination nature of the apomixis apomixis 2n+n and quantification; off-types nature ifand sexual marker(s) expression; expression; n+O nature combined with chromosomepotential identified (to eslimalion of estimation of detection. counting.estimation. date linkage the degree of the degreewith apomeiosisl. parthenogenesis of apomixis.Plant 15 to 100 Already deter­ 100 flowers. 5010100 Apomixis identification:15to 25 offspring.materiak flowers, mined materials ovaries/seeds: Apomixis quantification: at least 100 offspring.required depending on in segregationlhe objectives. for markeridentification.Advantages Easy and Ana~ses can be Easy and quick Easy and Easy if flow Eosy Analyses onquick to performed to perform quick to cylometry young offspringsperform after ear~. after flowering perform after (embryo, possible.flowering. pollination. endosperm,plantlels).Constraints Expensive Preliminary work The auxin lest Expensive nme consu­ nme and space consumming.equipment to determine has been mainly equipment for ming (dassical Morphological tesls: unreliable iffor materials. Use used 10 dale in flow cytometry. methods) or opomixis is high~ focultative.microscopy. of the markers cool-season expensive ifocross accessions grasses. flow cylometryof differentis used.origins? Expensive.

Su...I.. P........ to ldeollly.od ""'.liIy Apollllxls133required to indicate apomixis and determinethe nature of the hybrids detected. RFLPs andRAPDs can also be used in the same way, butat greater expense.2. Degree of apomixis expression. Manyoffspring are needed to obtain a good estimateof the degree of apomixis. Both auxin tests andflow cytometric analyses of pollinated ovariesor seeds provide good estimates of sexualpotential, though distinguishing 2n + 0 fromn + n offspring might be difficult in certaincases. In contrast, systematic chromosomecounting within progenies is useful fordetecting 2n + nand n + 0 off-types, but it doesnot separate 2n + 0 from n + n offspring, andwithout flow cytometry it becomestremendously time consuming. Progeny testscombining cytology and marker analysesrepresent the best option for identifying thedifferent classes of offspring within apomicticprogenies. To limit cytology work (when flowcytometry is not available), markers can beapplied first to separate maternal offspringfrom (poly)haploids or hybrids. The origin ofthe latter may be determined according to thepatterns they produce (i.e., 2n + n off-typesmust carry all bands from the mother plant,plus extra bands from the pollen), and thencytologically confirmed.Choosing a ProcedureThere are four main areas of apomixis research,each with distinct constraints and objectives:(i) the search for apomixis or elements ofapomixis in new taxa, coupled with geneticstudies in wild populations, (ii) germplasmcharacterization of apomictic species,(iii) genetic and biological studies for furthermanipulation of apomixis, and (iv) breedingof apomicts and introduction ofapomixis intosexual crops.Since gametophytic apomixis is formidablylimited to perennial, polyploid, andoutcrossing species, the search for apomixis inadditional species should begin with taxapresenting these traits. The very first screeningcan be based on the expression of the alreadymentioned "indicators of apomixis," whilemore discriminative procedures may beapplied to promising specimens. Forgermplasmevaluation, a representative sampleof the collection must be chosen on the basis ofmorphological and cytological data, and traitsof agronomic value such as disease resistance.Chromosome number, repro-d uctivedevelopment, and degree of apomixis are theprimary [actors for which basic data must becollected to develop strategies for furtherresearch. Genetic studies also may beattempted to genetically dissect apomicticmechanisms (number of genes involved andtheir effects). Following this preliminary work,appropriate tools for larger-scale screeningshould be developed or chosen according tothe apomixis characteristics of the collection(e.g., callose patterns for diplospory, ES clearingfor apospory of the Panicum-type, etc.).Sexual parents involved in crosses for apomixisinheritance studies must be carefully chosenusing cytoembryology. Highly facultativeapomicts are easily misclassified as sexualsusing progeny tests. This causes distortions ofsegregation ratios for mode of reproductionamong progeny. In the same way, looking fordifferences between sexual and apomicticdevelopment at the molecular level requires theanalysis of genotypes that are wellcharacterized for mode of reproduction. Thismay allow the development of near isogeniclines, an important step in identifying thegene(s) controlling apomixis.Before apomixis can be transferred into cropsor used in breeding programs, researchers needprocedures to identify apomictic genotypes(see do Valle and Miles, Chap. 10; Savidan,Chap. 11) and to quantify apomixis ingenotypesselected for varietal release. Progeny

134 OIivie' I

lu...log Pro.."',•• 10 ldeotily Dod o.a.ti!y Apomixl. 135Grimanelli D., M. Hernandez, E. Peroni, and Y.Sovidan. 1997b. Dosage effects in theendosperm of diplosporous apomicticTripsacum (Pooceae). Sex. Plant /leprod.1 0:279-82.Grimanelli D., O. Leblanc, E. Espinosa, E. Peroni,D. Gonzalez de Lean, and Y. Sovidan.19970. Mopping diplosporous apomixis intetraploid Tripsacunr. one gene or severalgenes? Heredity 80: 33-39.Grollniklaus, U., J.P. Vie lie-Calzada, M.A.Hoeppner, and W.B. Gagliano. 1998.Maternal control of embryogenesis byMEDEA, a po~comb group gene inArabidopsis. Science 2BO: 446--50.Gupta, M.G., S. Gupta, B.V. Bhat, V. Bhat, and S1Ahmad. 199B. Estimation of frequency ofapomixis by auxin induced parthenacarpytechnique in Dichanthium annulatum(Famk.l Stapf. /lange Management andAgroforesrry 19: 149-53.Guslalsson, A. 1947. Apomixis in higher plants.Port III. Biotype and species formation.Lunds Unversitets Amkrih Nova Series 2.43: 183-370.Gustine D.L., R.T. Sherwood, and D.R. HuH.1997. Apospory-linked markers inbuHelgross. [rap Sci. 37: 947-51.Hanna, W,W" and E.C. Bashaw. 19B7. Apomixis:ils identification and use in plant breeding.[rap Sci. 27: 1036--39.Hanna, WW., K.F. Schertz, and E.C. Bashaw.1970. Apospory in Sorghum biro/or (LlMoench. Science 170: 338-39.Harlan, JR., M.H. Brooks, OS. Bargaonkar, andJ.MJ. de Wet. 1964. Nature andinheritance of apomixis in Bothriochloa andDichanthium. Bot. Gaz. 125: 41-46.Herr J.M. 1971. Anew clearing-squash techniquefor the study of ovule development inangiosperms. Amer. 1. Bot. 58: 785--90.Hignight, KW., E.C. Bashaw, and M.A. Hussey.1991. Cytological and morphologicaldiversity of nalive apomictic buHelgrass,Pennisetum ciliare (Ll Link. Bot. Gal. 152:214-18.Hillary, B.B. 1940. Uses of the Feulgen reactionin cytology. II. New techniques and specialapplications. Bot.Gal. 102: 225--35.HuH, D.R., and J.M. Bora. 1993. Determininggenetics origins of aberrant progeny fromfacultative apomictic Kentucky bluegrossusing a combination of flow cytometry andsilver-stained RAPD markers. Theor. Appl.Genet. 87: 201-08.Johansen, D.A. 1940. Plant Microtechnique. NewYork: McGraw-Hili Book Co.Jongedijk, E. 1987. Aquick enzyme squashtechnique for detailed studies on femalemeiosis in Solanum. Stain Techn. 62: 135-­41.Kindiger, B., D. Bai, and V. Sokalov.1996.Assignment of a genets) conferringapomixis in Tripsacum to a chromosomearm: cytological and molecular evidence.Genome 39: 1133-41.Kojima, A., K. Hinota, and S. Noda. 1991b. Animprovement of squash method for thecytological study of female meiosis in Alliumtuberosum, liliaceae. Chromo\{)meInformotion Service 50: 5--7.Kapma, A., and T. kowaguchi.l989. Apomicticnature of Chinese chive (Allium tuberosumRon!.) detected in unpallinated ovuleculture. lop. 1. Breed 39: 449-S6.Kojima, A., Y. Nagata, and K. Hinata. 1991 a.Degree of apomixis in Chinese Chive (Alliumtuberosum) estimated by esterase isozymeana~sis. lop. 1. Breed. 41: 73-83.Kojima, A., and Y. Nagata. 19920.Pseudagamous embryogenesis and thedegree of parthenogenesis in Alliumtuberosum. Sexual Plant /leproduction 5:79-85.--.1992b. Diplasporous embryo·sacformation and the degree of diplospary inAllium tuberosum. Sexual Plant/leprcduction 5: 72-78.Koltunow, A.M. 1993. Apomixis: Embryo socsand embryos formed lllithout meiosis offertilization in ovules. The Plant [ell 5:1425--37.Koltunow, A.M., R.A. Bicknell, and A.M.Choudhury. 1995b. Apomixis: Molecularstrategies for the generation of genetical~identical seeds without fertilization. PlantPhysiol. 108: 1345--52.Koltunow, A.M., K. Soltys, N. Nita, and S.McOure. 19950. Anther, ovule, seed, andnucellar embryo development in Gtrussinensis cv. Valenda. [on. 1. Bot. 73: 1567­82.Kowles, R.V., I. Srien

136 or..io. u.~100< awd Aod.... M.n".'.Nogler, G. 1984. Gametaphytic apamixis. In 8.M. Jahri (ed.), Embryology ofAngiosperms.Berlin: Springer,VllIlag. Pp. 475-518Nayes, R.D., and l.H. Rieseberg. 2000. TwoindependentlO

Chapter 10Breeding of Apomictic SpeciesCACILDA BORGES DO VALLE AND JOHN W. MILESIntroductionFrom a plant breeding perspective, apomixismay restrict genetic recombination, but it alsoprovides a unique mechanism for developingsuperior cultivars and preserving thesegenotypes indefinitely. Apomictic plants, likesexual plants, develop seed in the ovule of theflower, but egg and sperm nuclei do not fuseto form an embryo. Therefore, the embryo ofan apomictic plant receives all of itschromosomes from the mother plant. Unlikemost asexually propagated plants-such asbanana, potato, or horticultural crops that arepropagated from vegetative parts of the motherplant-an apomictic plant is propagatedthrough the very convenient vehicle of seed.Early investigators (e.g., Darlington 1939) wereled to believe that apomixis was anevolutionary "blind alley" due to the apparentlack of variation in natural apomicticpopulations. Indeed, obligate apomixis posesa formidable barrier to plant breeding: withoutthe new gene combinations that result fromsexual cross breeding, genetic improvementcannot occur, except by rare, random, andgenerally deleterious mutations. In truth,sexual or partially sexual plants have beenfound in native populations of most apomicticspecies, generating sufficient genetic variationto maintain the species under changingenvironments and providing germplasm forplant improvement.Aside from citrus fruits (which exhibitapomictic reproduction through seed, but aregenerally propagated vegetatively), only a fewforage and turf grasses have active apomicticbreeding programs; these include species ofEragrostis, Paspalum, Poa, Panicum, Pel1niselum,Cel1chrus, and Brachiaria. These species (oragamic complexes) have many commonattributes, which will be addressed in a generalmanner la ter in this chapter. Bashawand Funk(1987) reviewed many aspects of breedingapomictic forage grasses, and recent papershave specifically considered plant breeding inthe genus Paspah,m (Savidan 1987; Burton1992); in Cenchrus ciHaris (Bashaw and Funk1987); in Panicum maximum (Savidan et al.1989); in Pennisetum (Hanna et a1. 1992); andmost recently in Brachiaria (Miles and Valle1996). In this chapter we focus on the Brachiariabreeding programs in Brazil and Colombia toillustrate pertinent aspects of apomixis vis avis breeding programs.Prerequisites for an EffectiveForage Breeding ProgramBeef production, especially in the tropics,largely depends on pastures, either native orplanted to superior introduced species.Scientific research on forages is relativelyrecent compared to field crops. Experiencewith tropical forages is even more limited; thefew commercially available cultivars are littlemore than "side of the road" collections, whichwere often accidentally introduced, mostlyfrom Africa, evaluated in small plots, and thenmultiplied for release.

138 udda lorge. do Val. aod Joh W. MilesCameron (1983) posed a Shakespeareanquestion: "To breed or not to breed," inreference to Australian investment in breedingtropical forage plants utilizing limited geneticresources. To quote Harlan (1983): "It is fruitlessto engage in plant-breeding programs withinadequate germplasm collections...."Representative collections for most of thetropical apomictic grasses are limited,therefore, a key prerequisite for effectivetropical forage breeding projects is to acquirediverse germplasm from the centers of originof the target genus/species.Panicum maximum was extensively collected byFrench and Japanese geneticists (Combes andPemes 1970; Nakajima et aI. 1978); the resultinggermplasm collections are representative of thenatural variation (Savidan et aI. 1989). Anextensive collection of Brachiaria wasundertaken by the Centro Internacional deAgricultura Tropical (CIAT) in 1984-85 (CIAT1986). Other important apomictic tropicalforage genera (Hyparrhenia, Melinis, Urochloa,Cenchrus, and Penniselum) still need to becollected to broaden variability and to identifysexual accessions to facilitate breeding.Surveys of closely related species are relevantwhen sexual plants are not available in theapomictic species of interest or when otherdesirable traits cannot be found in the primarygene pool. In sexual crops, the search forapomixis may involve other species or generain order to find cross-compatible wild relatives,as seen in the Zea x TrypsaCllm transfer program(Savidan, Chap. 11). To accomplishhybridization, research is needed to establishphylogenetic relationships and to overcomedifferences in ploidy level, genomerelationships, and gene pools (Hanna andBashaw 1987). If the species relationship issufficiently close, and given that apomixistends to restore fertility, one should be able toproduce useful obligate apomictic, interspecifichybrids with good seed set.An extensive species relationship survey wascarried out on Paspalum, a large grass genuswith tropical and subtropical adaptations(Burson 1983; Quarin and Norrmann 1987;Burson 1989; Quarin 1992). Two species areparticularly important as forage grasses, P.nolalum and P. dilalatum, and several othershave shown promising results in agronomicand grazing trials (Grof et al. 1989b; Fernandeset aI. 1992; Pizarro and Carvalho 1992; Batistaand Godoy 2000). Approximately 400 specieshave been described taxonomically, and about80% of these are polyploids. The genus hassexual diploids and apomictic and sexualpolyploids, which range from triploids to 16x(Quarin 1992). Diploid species reproducesexually and have regular meiosis (bivalentchromosome pairing and normal distribution).Polyploidy, apomixis, and irregular meioticchromosome associations are highly correlated(Quarin and Norrmann 1987). Valuableinformation has been gathered about thisgenus, leading to more effective interspecifichybridization that may result in superiorapomictic genotypes (Quarin 1987).A second fundamental prerequisite forbreeding is adequate knowledge of biology,cytology, and reproduction of the material athand (Asker and Jerling 1992). Breeders havelong been challenged by the problems ofreproductive isolation resulting from apomixisand polyploidy. Efforts directed atdetermining the genetic basis of apomixis inseveral species have generally shown it to beunder simple genetic control (see Sherwood,Chap. 5), e.g., Bothrioc"loa-Dichanthium(Harlan et aI. 1964), Panicum (Savidan 1982),CenchYlls (Sherwood et aI. 1994), Paspalum(Burton and Forbes 1960), Brachiaria(Ndikumana 1985; Valle et aI. 1993b; Valle andSavidan 1996), TripsaCllnl (Leblanc et aI. 1995b),and possibly Eragrostis (Voigt and Burson1992). Hence it should be possible tomanipulate apomixis in a breeding programonce cross-compatible sexual or highly sexual

Breodiog .f Apomidk 5,...., 139facultative apomicts are found (Harlan et al.1964; Voigt and Bashaw 1972; Bashaw 1980;Savidan 1983; Hanna and Bashaw 1987).Differences in ploidy level are common amongsexual and apomictic species of tropicalgrasses (Burton and Forbes 1960; Carnahanand Hill 1961; Dujardin and Hanna 1983;Norrmann et al. 1989). However, in groups inwhich apomixis is found, diploid accessionsare generally obligatory sexual whilepolyploids display different degrees ofapomixis ranging from essentially sexual toobligate apomicts (de Wet and Harlan 1970;Quarin and Norrmann 1987; Valle et al. 1989;Valle 1990; Asker and Jerling 1992). In specieswith higher ploidy levels (6x or 7x), such as B.humidicola, sexuality may be found at thetetraploid level (Valle and Glienke 1991).Sexually reproducing genotypes in the tropicalforage grasses outcross and are highlyheterozygous (Bashaw and Funk 1987). Somedegree of self-incompatibility or stronginbreeding depression is common (Bashawand Funk 1987). Rates of self-fertility in sexualB. ntziziensis were not affected by chromosomedoubling and ranged from 7.2 to 8.4%,according to Lutts et a!. (1991). Whenhybridization with apomicts has beenpossible, resulting progenies are highlyvariable owing to segregation in theheterozygous parents.Since hybridization and production of fertileprogeny are more effective when progenitorsare at the same ploidy level, basic studiesleading to the determination of chromosomenumber should be undertaken early in theprogram to enhance the chances of successfulrecombination of attributes by conventionalcrossing.A third prerequisite for efficient breeding ofapomicts, as in any plant improvementprogram, is the establishment of clear,achievable objectives, and the identification ofsources of the desired attributes in the existinggermplasm. This presupposes intimateknowledge of the species of interest, in orderto identify limiting factors not readilyovercome by simple selection of superiorgenotypes from the available germplasm oramenable to improved cultural practices. Oncea constraint has been identified (e.g., diseasesusceptibility or low forage quality), thenatural germplasm needs to be screened toidentify candidates for hybridization. Ideally,the desired attribute(s) can be found inapomictic or cross-compatible sexualaccessions with a superior agronomicbackground.General Structure of aBreeding ProgramA general selection and breeding scheme forapomictic forage species is presented in Figure10.1. Note that Brachiaria serves as the examplefor the topics under discussion.Brachiaria is native to the tropical savannas ofAfrica (IBPGR 1984), encompassing about 90species with wide morphological andphenological differences (Clayton andRenvoize 1982; Ren voize et a!. 1996).Apomictic cultivars of some of these species,cylogenelicsMode of reproduction1-----+_ Genetic markers[Morphological characterizationetc.•I.. ...SEXAPO.. ...APO SEX 5 APOIIRelease of ...new cultivarsFigure 10.1 Selection and breeding scheme forapomictic forage species.

140 Ca

Breedllt of Apomktk Species 141As accessions were transferred to CIAT­Colombia and released from quarantine, largeportions of this collection were subsequentlyforwarded to Brazil (approximately 400accessions), Costa Rica (approximately 280accessions), and Peru (approximately 260accessions) for agronomic evaluation (Grof etal. 1989a; CIAT 1992). Because large numbersof accessions were involved, evaluationmethodology needed to be simple andefficient, such as that proposed by Toledo(1982), in order to discard poorly performingmaterials. Since then, more detailed andintensive evaluation has been conducted at theEmbrapa Beef Cattle Research Center inCampo Grande, MS, Brazil. Agronomicevaluation began with accessions planted insmall plots with three replications. A periodiccutting regime was imposed for three years toestimate overall and seasonal production,regrowth vigor, seed production, andresistance to spittlebug and diseases (Valle etal. 1993a). At one harvest each year, sampleswere analyzed for crude protein content andin vitro digestibility. The range of variationobserved within this collection is remarkable(Table 10.1). Nineteen selected accessions werethen evaluated in regional trials in differentecosystems and superior genotypes wereidentified (Valle et a1.1 997). The next evaluationstep involved studying the effects of livestockon the pasture. Eight apomictic accessions werecompared in paddocks to the commercialcultivar. Four of these were selected (Euclideset al. 2001) to undergo animal performancetrials, the last step prior to release as a newcultivar.Morphological characterization, applyingnumerical taxonomy and using 26 descriptors,was carried out for all 340 accessions in theBrazilian Brachiaria collection (Valle et al.1993c). The objectives were to study thediversity of the accessions, analyze thedispersion and genetic distance betweenaccessions and species, and organize thegermplasm into groups of morphologicallysimilar accessions, regardless of taxonomicclassification. This type of study helpsresearchers define closely related accessionswithin and among groups from whichindividual progenitors may be selected forfuture crosses. This analysis revealed thecontinuous polymorphism that exists amongthree species (B. deCllmbens, B. bri:mntha, andB. ruziziensis) and clearly separated typicalaccessions of B. humidicola, B. dictyoneura, andB. jubata (Figure 10.2). The selection ofaccessions for pasture trials was based on anassociation of agronomic traits withmorphological characteristics.Cytology, Reproductive Mode, andInheritance of ApomixisBasic information about mode of reproductionand cytogenetics of sexually reprod ucingaccessions was also ascertained from theTable 10.1 Agronomic evaluation of Brachiariaaccessions in BrazilN LDMY (kg/ha) OfoDSPB. brizantharange 96 2040·9420 9·27 1.9·3.8overage callee. 96 4797 19 2.6overage select. 10 7503 18 3.1B. decumbensrange 35 1348-5543 10-25 1.5·3.2overage (olle

142 C«ida Jorge. do Va" aod Jah W. Milesexperimental plots. Previous reports on somespecies of this genus established the basicchromosome number as n = 9, and the mostcommon ploidy level among commercialcultivars as 2n = 4x = 36 (Schank andSotomayor-Rios 1968; Ferguson and Crowder1974; Valle 1986). B. ruziziensis is the onlycommercially cultivated species that is diploidand obligately sexual, with normalchromosome behavior at meiosis. Otherspecies are polyploid (4x or 6x) and haveirregular meiotic configurations. Thesepolyploids are apomictic, with aposporycharacterized by a 4-nucleate embryo sac ofthe Panicum-type. One egg-cell and one(occasionally two) conspicuous polar nucleuscan be observed in cleared ovaries. The twosynergids are rarely seen. Meiotic embryo sacsof the Polygonum-type with an egg-cell, twolarge polar nuclei, and multiple antipodal cellsare found in the sexual accessions and also inthe apomicts, in differing proportions.Brachiaria is pseudogamous, therefore, pollenproduction results from normal meiosis andis abundant both in apomictic and sexualplants.PRINI10.,.----------------,·2·4.6.6·4 ·2 oPRIN2Figure 10.2 Distribution of 253 accessions ofBrachiaria (8 =B. brizantha; D=B. decumbens;R= B. ruziziensis; H= B. humidico/a; J = B. jubatc;T= B. didyoneura) in two planes (PRIN 1 andPRIN2) generated by Principal Component Analysisusing seven morphological descriptors.The diversity of the introduced collectionjustified a thorough search for sexuality. Themode of reproduction was determined byexamination of embryo sacs for 427 accessionsof 15 different species in Colombia and Brazil(Table 10.2). Flowers were fixed in FAA for 24hours and later transferred to 70% ethylalcohol. Ovaries were then extracted under astereoscope and cleared using dehydrationand methyl salycilate (Young et al. 1979).Structures were mounted on slides andexamined with interference contrastmicroscopy. Results include discovery ofobligate sexual accessions in species previouslyconsidered obligate apomicts, such as B.decumbens, B. dictyaneura, and B. brizantha, anddetermination of mode of reproduction forspecies never before studied, such as B. serra/a,B. platynota, and B. subulifolia (Valle 1990).Chromosome counts were taken onmicrosporocytes of various sexual accessionsusing traditional acetocarmin squashes. It wasdetermined that the one sexual B. brizantha andall sexual B. decumbens accessions werediplOids, whereas the majority of apomicticaccessions of these two species were tetraploid.Table 10.2 Mode of reproduction of 15 species ofBrachiaria, based on embryo-sac analysisSpedes no. accessions Range sex SEX APOB. affecta·· 3 79 -90 3 oB. bovonei 4 7·27 o 4B. brizantha 235 0- 94 1 234B. decumbens 54 0·100 22 31B. deflexa 1 91 1 oB. dictyoneura 6 0·96 1 5B. dura 1 93 1 oB. humidicola 52 0·100 2 50B. iubata 34 0·94 5 29B. miliiFormis 1 6 1 oB. nigropedata 3 5- 20 o 3B. platynoto 3 3-97 2 1B. ruziziensis 24 40-100 24 oB. serrato 2 30 ­ 100 2 oB. subuliFolia 4 7-38 o 4Total 426 65 361

....... of ApoaO

144 (..iIda 80rges cit Va" DOd Ja~. W. Mieskept in vases of water. On the day ofpollination, inflorescences were shaken overpetri dishes to collect pollen, which was usedon flowers from which stigmas had justextruded. The inflorescence from the sexualplant was prepared by removing unopenedand old flowers. After brushing the turgidstigmata with pollen from the apomicticparent, the racemes of the sexual plant wereindividually bagged and labeled. Bags werecollected when seed shattering started.Scarified seeds were individually germinated4-6 months later in styrofoam trays with asand:perlite mixture (2:1) or in petri dishes,and then transferred to plastic bags with soil,from which they were later transferred to thefield (Valle et al. 1991).Mode of reproduction was determined byembryo-sac analysis on 30--40 ovaries of 376individual first-generation hybrids fromgreenhouse crosses in Brazil. No reliablegenetic marker yet exists to determine hybridnature of the progeny, therefore attempts todiscriminate among individuals were madeusing morphological characteristics and/orelectrophoresis. Whenever parental materialsdisplay wide differences in morphology or inband patterns, hybrids showing intermediatecharacteristics can be readily identified.2n sexuali" 't4n apomictic 4n sexual 5 4n rsame orFl~ictiC1olhersf------4n apomictic 4n sexual ...F2.--L----.agronomicevaluationsFigure 10.3 Hybridization scheme for breedingBrachiaria (adapted from Gobbe et 01. 1983).Studies conducted at CIAT identified an alphabetaesterase system capable of discriminatingamong putative hybrids of carefully selectedprogenitors (Cruz et al. 1989a, 1989b; Calderonand Agudelo 1990).Second generation crosses in Brazil includedsexual x apomictic backcrosses, crossesbetween half sibs, full sibs, selfing of sexualF\s, 3-way hybrids, and facultative apomicticx apomictic crosses. Results from thisexperiment (Table 10.3) point to a singledominant gene determining apomixis, asproposed for Panicum maximum (Savidan1983), for Brachiaria (Ndikumana 1985), andCenchrus (Sherwood et al. 1994). The excessnumber of sexual plants observed in somecrosses may be due to crossing procedures;sexual maternal plants were not emasculated,and no special precautions were taken to avoidpollen circulation in the greenhouse, exceptafter pollination when flowers were bagged.Table 10.3 Segregation for mode of reproduction inBrachiaria hybridsType uoss SEX APO STER AbnF 1B.ruliliensis x B. decumbens 79 49B.ruliliensis x B. brilon/ho \25 123F 2B.ruliliensis x B.decumbens 2 0 2 0B.rulilief/Sis x B. brilon/ho 7 0 0 0BCB.ruliliensis x B.decumbens 10 9 0 0B.ruliliensis x B. brilan/ha 9 7 0 03-WB.ruliliensis x B.decumbens 24 21 9 1B.ruliliensis x B. brilon/ho 31 6 0 0FSB.ruliliensis x B.decumbens 38 27 4 0B.ruliliensis x B. brilantha 24 32 3 1HSB.ruliliensis x B.decumbens 5\ 53 6 8B.ruliliensis x B. brilon/ha 61 23 6 0BC: oockcrass, 3-W: three-way hybrids, FS: full-sibs, HS: holf·sibs.

........UpoonkIk Spede. 145The mode of reproduction of an independentset of 107 first generation Brachiaria hybridswas determined by progeny tests and byembryo-sac analysis in Colombia (Table 10.4).Embryo-sac analysis determined that 56 of theplants were sexual and 51 were apomicts(Miles and Valle 1991), a finding that agreeswith the proposed hypothesis of simpleinheritance. Although interspecific crossesmay not be ideal for studying inheritance ofapomixis, work on the agamic complexformed by B. ruziziensis, B. brizantha, and B.decumbens indicates simple genetic control forapomixis. Sexual x apomictic crosses releaseda large amount of phenotypic variation in theprogeny (plant morphology, growth habit, andflowering time).Progeny tests of the 107 open pollinated, firstgenerationinterspecific hybrids were alsoused to assess reproductive behavior (Milesand Valle 1991). Seeds harvested fromindividual plants were sown in five-plantplots, in one to four replicates. The mode ofTable 10.4 Comparison between progeny test andembryo-sac analysis for determination of mode ofreproduction for first-generation interspecificBrochiorio hybridsRate ofMode of reproduction sexuality (%)embryo-sacprogeny-test analysisembryo-sacanalysis54 hybrids sexual sexual37 hybrids apomictic apomictic 10 -734hybrids unclassified apomictic 7-832hybrids unclassified sexualFacultative apomicts first dassified as sexual541-03 sexual apomictic 10544-04 sexual apomictic 50549-02 sexual apomicfic 17554-02 sexual apomictic 63554-03 sexual apomictic 77683-0 I sexual apomictic 52687-01 sexual apomictic 70693-02 sexual apomictic 47694-07 sexual apomictic 43702-06 sexual apomictic 30reproduction of the mother plant was inferredfrom the relative uniformity or heterogeneityof the open pollinated progeny. The resultswere later compared to microscopicexamination of embryo-sac structures of thehybrid mother plants. The two methodsagreed closely, except for ten of the progenies,in which the degree of sexuality (determinedby embryo-sac analysis) ranged between 10and 77%. (Table 10.4). The degree of effectivesexuality as detected by the progeny test wasnot closely associated with the proportion ofsexual embryo-sacs observed microscopically.Whereas facultative apomictic hybrids with 10or 17% sexuality produced heterogeneousprogenies, other hybrids, in which sexualembryo-sacs were observed in up to 73% ofthe progenies, appeared to behave as obligateapomicts in the test. It is unclear what factor(s)con tribu tes to determining effectivereproductive behavior. Elucidation of themechanism of apomixis might help explain itsexpression under different circumstances.At CIAT and the Institute for Grassland andEnvironmental Research (IGER),Aberystwyth, Wales, U.K., a molecular markerfor the apomixis gene(s) is being sought, whichcould prove potentially useful for determiningreproductive mode. Pessino et al. (1997), usinga bulk segregant analysis and RFLPs andRAPDs, were able to identify molecularmarkers cosegregating with apomixis in asmall' (n = 43) Brachiaria F1 population. Twoclones (umc147 and umc72) belong to aduplicated linkage group that maps to thedistal part of maize chromosome-1long armand chromosome-5 short arm. Another,(OPC4), previously reported as a potentialmarker for apospory in Pennisetum, alsocosegregated well in Brachiaria. In later work,Pessino et al. (1998), using RFLPs and AFLPs,generated a complete map for the region inmaize chromosome 5, identifying at least twomarkers closely linked to the apospory region.

146 Cacido ....... Vole..J.b W. MIesMarkers PAM52-5 and PAM49-13 were locatedrespectively at 1.2 cM and 5.7 cM, on eitherside of the target locus. The map shows closesynteny to regions ofmaize chromosome 5 andrice chromosome 2. If proven that apomixisis, in fact, conditioned by a single dominantgene, and these markers prove to be tightlylinked, then itshould be possible to determinethe reproductive mode ofa hybrid plantbeforeflowering oreven before transplanting into thefield. This would substantially improve thegenetic efficiency ofbreedingschemes (see alsoLeblanc and Mazzucato, Chap. 9).Brachiaria breeding involves interspecifichybridization because compatible sexualplants could notbefound in the agronomicallyimportant species. Wider crosses may also berequired to transfer important traits (such ascomplete antibiosis present in one accessionof apomictic B. jubata [Lapointe et al. 1992]) tothe commercially important apomicticcultivars, using sexual plants of differentspecies as "bridges." B. humidicola, for instance,is well adapted to waterlogged soils, however,it has proven impossible to hybridize the onlysexual tetraploid B. humidicola accession withthe tetraploid apomictic varieties of B.decllmbens and B. brizantha. furtherphylogenetic studies need to be conducted todetermine possible compatible materials withwhich to attempt crosses.When conventional sexual hybridization isimpossible, direct transfer of DNA betweenspecies may be considered. Protocols for callusinduction and regeneration have beendeveloped for five Brachiaria spp. (GAT 1993)and a system for genetic transformation usingparticle bombardment has been established(Lennis 1998; Galindo 1997).Breeding PlansThe delineation of clear breeding objectives,the identification of sources of desiredattributes in apomictic Braclliaria accessionsfrom the newly enhanced germplasmcollection, and the creation of a crosscompatiblesexual material has led to thepossibility of developing large-scale, appliedplant breeding projects for this importantforage grass.The fundamental objective of any plantbreeding program for an apomictic species inwhich genetic recombination can be achievedis the identification among segregatingprogenies ofsuperior, true-breeding apomicticgenotypessuitable for cultivarstatus. Breedingplans that are being implemented for Brachiariaassume (i) simple (probably monogenic)control of apomixis and (ii) predominantlyallogamous reproduction with high levels ofself-incompatibility or strong inbreedingdepression.Information regarding inheritance of traits islimited to recent data showing a strongcorrelation between the reaction of spittlebugsto a series of parents and their top-crossprogenies (Miles et al. 1995). Eleven apomicticaccessions, chosen to represent a range ofspittlebug reactions, were each crossed to thesame susceptible sexual clone to generate 11segregating, F 1families. Spittlebug reactionwas assessed on apomictic clones and on tenrandom sibs in each of the 11 top-crossfamilies. The close parent-progenycorrelationsfound for percentage of nymphal survival [r= 0.95 (P

Iroodlog .f Apoonictk Specie. 147species rely on large-scale hybridizationbetween sexuals (facultative apomicts) andobligate apomicts to produce largepopulations, from which superior apomictichybrids are isolated (e.g., Burton and Forbes1960; Taliaferro and Bashaw 1966; Gobbe etal. 1983; Bashaw and Funk 1987; Savidan etal. 1989). Three-way or double crossesinvolving more than one apomictic male maybe required, depending on the distribu'tion ofdesired attributes among available apomicticgenotypes. Theseschemes will require carefulselection of parents and the evaluation of largepopulations to find the desired combinationof characteristics in a true breeding apomicticgenotype, Such approaches offer theopportunity to generate novel apomicticgenotypes, however, they are essentiallyconservative in the longer term because theopportunity for genetic gain is eventuallyexhausted,To continue genetic advances, a systematicscheme for recycling selected hybridgenotypes will be required (i.e., populationimprovement by recurrent selection). AnyBrachiaria breeding population must obviouslyinclude sexual genotypes to ensure geneticrecombination. An importantconsideration inthe development of populations is whether toattempt to include and maintain apomicticgenotypes in the populations. Several authorshave suggested improvement of a sexualpopulation (e.g., Pernes et al. 1975; Miles andEscandon 1997). The scheme for sexualpopulation improvement proposed by Pemeset al. involves recurrent crossing to eliteapomicts. Therefore, the superior sexualhybrid genotypes in each crossing cycle n~edto be identified to resynthesize a fully sexualpool.Miles and Escandon (1997) proposed recurrentintrapopulation improvement of aheterogeneous sexual population developedfrom sexual segregants selected from progenyof an initial series of crosses of a sexualtetraploid biotype with apomictic genotypes.In the case of Brachiaria (and otherspecies witha similar genetic control of reproductivemode), the second scheme would have theimportant advantage ofobviating the need todetermine reproductive mode in eachgeneration. As the frequency of favorablealleles is increased in the sexual population,hybridization with elite apomictic genotypeswill generate an array of improved apomicticand sexual segregants while the sexualpopulation remains fully sexual, i.e., advancedfrom purelysexualclones selected from withinthe population. Superior apomicts in thehybrid populations would be candidates forcultivar release. They could also be used in thesubsequentcycle ofsexual x apomictic crosses,although it would not be expected thatcrossing back to the parental sexual poolwould lead to maximum expression ofheterosis. A sexual population based onselected first cycle sexual x apomictic hybridsis being developed at CIAT. Thirty-two suchhybrids, involving a total of ten apomicticpaternal parents crossed to the same sexualtetraploid B. ruziziensis, were initially selected.The selection was subsequently reduced to 30,when it was determined that one clone wasvery susceptible to a virus and anotherexhibited a low, but consistent, percentage ofapomictic embryo sacs. Each clone wasvegetatively propagated ten times andgenotypes were planted in random spacing inan isolated field plot. This population will bemanaged bystandard halfsib or mass selection(Figure 10.4).A second population containing both sexualand apomictic genotypes has been formedfrom sexual x apomictic hybrids (Figure 10.5).In each generation, two types of progenies areplanted in alternating positions in a squaregrid: (i) apomictic progenies of selectedapomicts (reproductive mode determined by

148 Ca

complex) of interest. The design of breedingschemes that approach "optimum"efficiency-genetically and economicallywilldepend, as well, upon the cost andreliability of the method(s) available to assessrep rod ucti ve mode in segrega tingpopulations.An important obstacle in the recurrentselection programs is the difficulty ofachieving a full generation each year. Mostattributes of interest in perennial Brachiariaspp. are difficult or impossible to reliablyassess in a single season. Seed dormancy,which is typical in tropical wild grasses, alsodelays the breeding cycle and poses anobstacle that has not been overcome, even atthe experimental level. A detailedunderstanding of the factor(s) causingphysiological dormancy in Brachiaria seeds isessential to the rational design of dormancybreakingtreatments. In vitro techniques ofembryo rescue were developed in Brazil, andRodrigues-Otubo et al. (2000) established theage and culture medium for first generationinterspecific hybrids of Brachiaria. Embryosthat were drawn for use 9-12 days afterpollination presented the highest percentageof direct regeneration, although survival rateswere clearly genotype dependent. Embryorescue resulted in significantly highernumbers of hybrid plants being recoveredReferencesAsker, S.E., and J. Jerling 1992. Apomixis inplonts. Boca Raton, Florida: CRC Press.Boshaw, E.C 19BO. Apomixis ond ~s applicationin crop improvemenl. In W.R. Fehr ond H.H.Hadley (eds.l, Hybridization of (rap Plants.Madisan, Wisconsin: American Society ofAgronomy Press. Pp.45-63.Bashaw, E.C., and C.R. Funk. 1987. Apomicticgrosses. In W.R. Fehr led.l, Principles of(ultivar Development, Vo1.2. New York:Macmillan Publishing Co. Pp. 40-82.Batisto, LA.R, and R. Godoy 2000.Caracteriza~iio preliminar e sel~iia degermoplasmo do genero Paspolum paraprodu~iio de forragem. Rev. Bros. Zaotec.,29(1): 23-32.Bogdon, A.V. 1977. Tropical Posture and fodderPlants. New York: longmon.Burson, B.l. 19B3. Phylogenetic investigotions ofopomictic Paspolum dilatatum ond relotedsptl(ies. In Pree. XIV Int. Grossi. (angr.,lexington, Kentucky, 19B1. Boulder,(olorodo: Westview Press. Pp. 170-73.--.1989. Phylogenetks of opomicti

150 (adlda Borge. da VaNe and Joh. W. MR••Carnahan, H.L, and H.D. Hill. 1961. Cytalogyand genetics of fOfll§6'grasses. Bot. Rev.27: 1-162. ""'C1AT (Centro InternacionaLde AgriculturaTropical). 1986. Germoplosma,. InformeAnual, 1985, Pastos Tropicales. Cali,Colombia: CIAT. Pp. 5--28. Pastures.--.1992. Postures for the TropicalLowlonds: OAT's (ontribution. Cali,Colombia: C1AT.--. 1993. Biotechnology Reseorch UnitAnnuol Report 19BB-1992. Cali, Colombia:C1AT.--. 1998. Gramineas yLeguminosasTropicales: Optimizacion de la diversidadgemltica pora usos multiples (Proyecto IP­S). 1998. Informe Anual. 1997. CentroInternacional de Agricultura Tropical (C1AT),(ali, Colombia. Documento de Trabajono.174. Cali, Colombia: C1AT.aayton, W.O., and SA Renvoize. 1982.Gramineae (Part 3). In R.M. Polhill (ed.),Rora of Tropical East Africa. Kew, U.K.:Royal80tanical Gardens. Pp. 575--600.Combes, 0., and 1 Pernes. 1970. Variationsdans les nombres chromosomiques duPanicum maximum Jacq. en relation avecIe mode de reproduction. CR. Acad. Sci.Paris. 270: 782-85.Cruz, R., J.w. Miles, W. Rom, and G. de la Cruz.19890. Apomixis ysexualidad enBrachiaria. 2. Estudios citoembriologicos.Rev. (ubana de Gencia Agricola 23(3):307-12.Cruz, R., J.W. Miles, W. Rom, and H. Ramirez.1989b. Apomixis ysexualidad enBrachiaria. 1. Estudias bioquimicos. Rev.(ubana de Genria Agricola 23(3): 301­05.Darlington, CD. 1939. The Evolurion of GeneticSystems. Cambridge, U.K.: CambridgeUniversity Press.de Wet, lMJ., and J.R. Harlan. 1970.Apomixis, po~ploidy and speciation inDichanthium. Evolution 24: 270-77.Dujardin, M., and W.W. Hanno. 1983. Apomicticand sexual pearl millet x Pennisetumsquamulotumhybrids.1. Hered. 74: 277­79.Eudides, V.P.8., CB. Valle, M.CM. Macedo,andM.P.Oliveira. (2001). Evaluation ofBrachiaria brizantha ecatypes undergrazing in small plots. In Proc. XIX Int.Grassl. Congr., Piracicaba, SOo Paulo,Brazil. HALO, 2001. (in press)Ferguson, lE., and LV. Crowder. 1974. Cytologyand breeding behavior of BrachiariarUliliensis Germain et Evrard. u-op Sci. 14:893-95.Fernandes, AU., CD. Fernandes, V.P.B. Eudides,and B. Grof 1992. Avaliacoo de acessos dePaspolum spp. em consor~ia~oa comArachis pintoi, em areas umidas de baixafertilidade. In Red Internacianal deEvaluarion de Pastos Tropicales/RIEPT. 10.Reunion Sabanas. Brasilia. Documento de.trabajo Na. 117. Cali, Colombia: C1AT.Pp.555--60.Galindo-LOpez, L.F. 1997. Transformadan..: Igenetico de la graminea forrajeraBrachiaria spp., mediante la tecnica d.ebombardeo de particulas. [Genetic'transformation of the forage grassBrachiaria spp. by the technique of partidebombardment.] Undergraduate thesis.Universidad Nacional de Colombia,Palmira. 130 p. (+ Appendices).Gobbe, l, B. Longly, and B-P Louan!. 1983.Apomixie, sexualite et amelioration desgraminees tropicales. Tropicuhura 1: 5--9.Gobbe, l, A. Swenne, and B-P Louant. 1981.Diplo'ides natureb et autotetraplo"idesinduits chez Brachiario rUliliensis Germainet Evrard: criteres d'identificatian. Agron.Trop. 36: 339-46.Gro!, B., R.P. de Andrade, M.5. Fran~a-Dantas,and MA Souza. 1989a. Selection ofBrachiaria spp. for the add-soil savannas ofthe central plateau region of Brazil. Proe.XVI Int. Grassl. (ongr., Nice, France. 1989.Assoe. Fr. Prod Fourragere. v.l. Pp. 267­68.Gro!, B., R.P. de Andrade, M.A. Souza, and HM.Valls. 1989b. Selection of Paspolum spp.adapted to seasonally flooded varzea landsin central Brazil. Proe. XVI Int. Grassl.(ongr., Nice, France. 1989. Asllle. Fr. Prod.Fourragere. v.l. Pp. 291-92.Hacker, lB. 1988. Sexuality and hybridizationin signalgrass, Brachiaria decumbens. Trop.Grassl. 22: 139-44.Hallauer, A.R., and lB. Miranda. 1988.Quantitative Genetics in Maile Breeding.2nd ed. Ames, Iowa: lawa State UniversityPress. Pp. 430-34.Hanna, W.W., and E.C 8ashaw. 1987. Apomixis:~s identificotion and use in plant breeding.(rap Sri. 27: 1136-39.Hanna, W.W., M. Dujardin, P. Ozias-Akins, and L.Arthur. 1992. Transfer of apomixis inPennisetum. In lH. Bgin and LP. Mikshe(eds.), Proe. of the Apomixis Workshop.USDA-ARS. Atlanta, Georgia_ 1992.Atlanta, Georgia: USDA-ARS 104. Pp. 30­33.Harlan, lR. 1983. The scope for collection andimprovement of forage plants. In lG.Mcivor and RA Broy leds.), GeneticResources of Forage Plants. EastMelbourne, Australia: CSIRO. Pp. 3-14.Harlan, J.R., M.H. Brooks, 0.5. Borgaonkar, andJ.MJ. de Wet. 1964. Nature andinheritance of apomixis in Bothriochloaond Dichanthium. Bot. Gal. 125: 41-44.Hughes, N.R.G, C8.do Valle, M. Herrero. 1998.Estimativa da resistencia ao cisolhamento ea moagem em quatro especies deBrachiaria.. In Proe. 35 Reuniiio Anual Soe.Bras. lootee., Botucotu: SBZ, Vi~osa. Pp.43-45.Hughes, N.R.G, CB.do Valle, V. Sabotel, 1Boock, N.S. Jessop, M. Herrero. 2000.Shearing strength as an additionalselection criterion for quality in Brachiariapasture ecotypes. 1. Agrie. Sri., Cambridge,135: 123-130.Hull, F.H. 1945. Recurrent selection for specificcombining ability in corn. 1. Amer. Soc.Agran. 37: 134-4S.IBGE (Instituto Brasileiro de Geografia eEstatistica). 1995. Anuorio fstatistica doBrasilv.5S. Rio de Janeira, Brazil: IBGE.IBPGR (lnternotional Board for Plant GeneticResources). 1984. Tropical and subtropicalforages. Report of working group. Rome:FAO.Keller-Grein, G., B.LMaass, and lHanson.1996. Natural variation in Brachiaria andexisting germ plasm collections.. In JW.Miles, B.L Maass, and C8. do Valle (eds.),Brachiaria: Biology, Agronomy, andImprovement. CIAT Publicotion No.2S9.Cali, Colombia: C1AT and Campo Grande,Brazil: CNPGC/EMBRAPA. Pp. 9-15.Lapointe, S.~ G. Arango, and G. Sotelo. 1989.Amethodology for evaluation of host plantresistance in Brachiaria to spi"lebug(Homoptera:Cercopidae). Prae. XVllnt.Grassl. Congr., Nice, France. 1989.Asllle.Fr. Prod. Fourragere. v.l, pp. 731-32.Lapointe, S.L, M.S. Serrano, G.L Arango,G.Sotelo, and F. Cordaba. 1992. Antibiasisto spi"lebugs (Homoptera: Cercopidae) inaccessions of Brachiaria spp. 1 Econ.Entomology 85 (4): 1485--90.Leblanc, 0., M. Duenas, M. Hernandez, S. 8ello,V. Garcia, 1 Berthaud, and Y. Savidan.1995a. Chromosome daubling inTripsacum the production of artificial,sexual tetraploid plants. Plant Breeding114: 226-30.Leblanc 0., D. Grimanelli, O. Ganzalez de Lean,and Y. Savidan. 1995b. Detection of theapomictic mode of reproduction in maize­Tripsacum hybrids using maize RFLPmarkers. Theor. Appl. Genet. 90: 1198­1203.

Breedillg .1 Apomklk Spedf' 151lennis M. 1998. Desenvolvimento de ummetodo de lramform~iio genelico deBrarhiaria spp. por bombaredoomenta deportirulos. Masters thesis. Universidade deBrasUio, DJ. Brazil. 131 p.leNeriella, G., e.B. Volle, D. Christiane, and1,1.1.0. Penteada. 1999. Otalogia emodode reprodu~iia de acessos pentaploides deB. brizantha. In Prar. 36 Reuniiia Anual doSoc Bras. Zaoter., 1999. Porto Alegre,SBl/Vidoolar, SOo Poulo, Brazil. CD· ROM.Forrogirulturo. Avalia~iio de Forrogeiras.FOR 140.LuNs, S., J. Nidikumana, B.·P Louont. 1991.Fertility of Brarhiaria ruziziensis ininterspecific crosses with Brarhiariadecumbens and Brarhiaria brizanthu.meiotic behavior, pollen Viability and seedsel. fuphytjca 57: 267-74.Miles, J.W. 1995. Applicalion of recurrentselection for specific combining ability tothe improvement of apomictic Brachiaria.In Harnessing Apomixis, An InternationalConferenre,. 25-27 September 1995.College Slation, TX, USA. Pp. 64.[Abstract). College Station, Texas: FAO/ORSTOM/USDA·ARS/RockefellerFoundation/Texas Ag.Exp.sta./Dept.SoiIond Crop Sci·TAMU/ Pioneer Hi·BredInternational, Inc.--. 1997. Abreeding scheme to exploitheterosis in apomicts. (A/mroctl. InCIMMYT. Book of Abstrarls. The Genetirsand Exploitation of Heterosis in Craps; AnInternational Symposium. Mexico,DJ:OMMYT, Pp. 182-83.Miles, J.W., and M.l. Escandon. 1997. Furtherevidence on the inheritonce of reproductivemode in Brarhioria. Can. l. Plant Sri. 77:105-07.Miles, J.W., S.l. Lopointe, M.L Es(Ondon, and G.Sotelo. 1995. Inheritance of spiNlebugresistonce in interspecific Brachiariahybri

152 Ccdda Borge. do v..... J.h W. MilesSherwood, R.T., CC Berg, ond B.A. Young.1994.lnheritonce of opospory inbuffelgross. Clop Sci. 34: 149G-94.Swenne, A., B-P louont, ond M. Dujordin. 1981.Induction por 10 cokhicine de formesoutotetro~oides chez Brochiorio ruziziensisGermain el Evrard (Grominlies). Agron.Trap. 36: 134-41.lorlOfeHo, CM., and E.C Bushaw. 1966.Inheritonce ond conlrol of ob~goleapomixis in breeding buffe~ross,Pennisetum aliarB ((enehM ciliorisl. ClopSci. 6: 473-76.Toledo, J.M. (ed.). 19B2. Manual para 10evo/uaaiin ogroniimico: Red Internocionolde Evo/uadiin de Postos Tropicoles. Coli,Colambio: CIAT.Volle, CB. do. 1986. Cytology, mode ofreproduction ond forage quolily of selectedspecies of Brochiorio Griseb. Ph.D.dissertation. University of lIfinois-Urbono.Urbano, Illinois. USA.--.1990. (oI~iio de germoplasmo dBespeeies de Brochioria no ClAT: estudosb6si!os visondo 00 melhoromento geootico.EMBRAPA·CHPGC, Documentos, 46, CompoGrande, Brazil: EMBRAPA-CHPGCVolle, CB. do, S. Calixta, and M.C Amezquha.19930. Agronomic evaluation of Brochioriogermplasm in Brozil. PrO(, XVlllnt, Grassl,Congr., HZGA, TGSA, HZSAp, ASAP·Old, ondHZIAS, Polmer>ton North, New Zealand.1993. pp. 511-12.Volle, CB, do, and CGlienke. 1991. New sexualaccessions in Brochiorio. ApomixisNBlVSietter 3: 11-13.Volle, CB. do, CGlienke, and G.O.CLeguizaman 1993b. Breeding of opomicticBroehiorio through inlerespecifichybridisotion. Prac. XVII Int. Grassl, (ongr.,NZGA, TGSA, NZSAp, ASAP-Old, ond NZIAS,Palmerllon North, New Zeolond. 1993, Pp.427-28.Volle, CB. do, G. leguizomon, and N.R. Guedes,1991.lnter>pecific hybrids of Brochior;o(Gramineoe). Apomixis Newsletter 3: 1G­Il.Volle, CB. do, B.L Moass, CB. de Almeido, andJ.CG. (0110. 1993c. Morpho/ogicolmaracterizolion of Broch;orio germplosm.Proc. XVII Int. Grassl. Congr., NZGA, TGSA,NZSAp, ASAP-Old, ond NZIAS, PalmerstonNorth, New Zealand. 1993. pp. 20~9.Volle, CB. do, Y. Sovidon, and LJonk. 19B9.Apomixis and sexuolity in Brachioriodecumbens Stapf. Proc XVI Int.Gross/.(ongr., Nice, Fronce. 1989. Assoc.Fr. Prod. Fourragilre. v.l. pp. 407~8.Volle, CB., ond Y. Sovidon. 1996. Genetics,cytogenetics and reproductive biology ofBrochiaria.. ln J.W. Miles, B.L Maoss, andCB. do VoDe (eds.), Brachiaria: Biology,Agronomy, and Improvement. (oli,Co/ombia and (ampa Grande, Brazil: ClATand (NPG(-EMBRAPA. Pp.147-163.Volle, CB., J R. Valerio, S. Colixlo, A. O.Barcellos. 1997. Characteristics of selectedgenotypes of Brachiar;a for Brazilianposlures. In Proc. XVIII Int. Grassl. (ongr.,Conadian Forage (ouncil, 1999. Winnipeg,(anada. pp.l/81-l/82. CD-ROM.v.l,section 1. 10 1358.Voigt, P.W., and E.C Bashaw. 1972. Apomixisand sexualily in fragrostis wrvu/o. (ropSci. 12: 843-47.Voigt, P.W., and B.L Burson. 1992. Apomixis infrogrostis. In J.H. 8gin Jr. and J.P. Miksme{edl.}. PrO(. Apomixis Workshop, Atlonta,Georgia, USA. February 11·12, 1992.USDA/ARS. ARS-l04. Pp. S-11.Young, B.A., R.T. Sherwood, and H. Boshaw.1979. CIeored-pistyl ond thick-sectioningtechniques for detecting aposporousapomixis in grosses. (an. 1. Bot, S7:166S-72.

Transfer of Apomixis throughWide CrossesChapter 1 1YVES SAVIDANIntroductionInterspecific hybridization has been usedextensively to transfer agronomicallyimportant genes that control resistances todiseases and insect pests. Recent advances intissue culture, especially in molecular biology,have further widened the scope of alien genetransfer-and the outlook for widehybridization in crop improvement seemsmore promising than ever. But does thatoutlook also apply to the transfer of sequencesinvolved in plant reproduction, especiallythose involved in apomixis?The interface between conventionalcytogenetical approaches and new moleculartechniques makes the "conventional" widecross approach very competitive when the traitis simply controlled and the gene(s) to betransferred is (are) available in a species thatbelongs to the secondary gene pool. Thegenetic analyses reviewed by Savidan (2000)and Sherwood (Chap. 5) suggest that apomixisa good candidate and offer support for theongoing wide cross projects. Such projects haveencountered unexpected difficulties, andseveral papers have questioned the ultimatelikelihood of transferring apomixis to any crop.Nevertheless, knowledge gathered through thequest for wide crossing apomixis into usefulcrop species, which relates to the geneticcontrol, transmission, and expression of thetrait (Grimanelli et aI., Chap. 6), has provenextremely valuable for those investigatingother approaches. Accordingly, three paths arenow being pursued in the effort to introduceapomixis into major crops: (i) the widehybridization, (ii) the identification, isolation,and manipulation of sequences from wildapomicts, and (iii) the creation of an apomicticreproduction de novo, from individualmutations (Grossniklaus, Chap. 12; Praekeltand Scott, Chap.13). In this chapter, I reviewprogress to date and the problems or questionsthat have emerged from work aimed at widecrossing of the apomictic trait.Scientists have tried for decades to use widecrosses to transfer the apomixis trait intovaluable food crops, including wheat, maize,and pearl millet. The first attempt involvedmaize and was initiated approximately fortyyears ago (Petrov et al. 1979, 1984). Crossing atetraploid maize (2n =4x =40) with a tetraploidTripsacum dacty/aides (2n =4x =72), the Russianscientists successfully produced maize­Tripsacum Fls and BC lhybrid derivatives that,according to progeny tests, reproducedapomictically. The BC lplants combined 20maize chromosomes with one complete set (18)of Tripsacum dacty/aides chromosomes. Efficienttechniques for evaluating chromosomenumbers, embryo-sac analysis, etc., were notavailable, making screening of large numbersof progenies for apomixis difficult.Consequently, little progress was made in thistransfer effort. Recently, the Russian materialswere transferred to the United States, andintrogression efforts were reinitiated. Animportant piece of information generated by

154 Yve< 5

T,..,fe, .f Apomixis ,.,... WIde Cro"e, 155cytogenetics, e.g., in Tripsacum, Galinat (Galinatet al. 1970; Galinat 1971) described four maizechromosomes that are capable of pairing withTripsamm chromosomes. Meanwhile, mappinganalyses (Grimanelli et aI., Chap. 6) suggest amore widespread colinearity between themaize and Tripsacum genomes.4. Pollen fertility. Except for the highlyfacultative apomicts, first generationhybridizations between the crop and donorspecies must use the latter as male. Severalapomictic species have been described withgreatly reduced male fertility (e.g., Elymusrectisetus, the apomictic wild relative of wheat);in such cases, a preliminary selection is needed.5. Type of apomixis. Apospory has alwaysbeen presented as an easier type of apomixisto work with, being associated with 4-nucleateembryo sacs in tropical and subtropical grassesand transmitted as a single dominant gene(Savidan 1982a; Nogler 1984; Asker and Jerling1992; Savidan 2000). Recent studies ondiplospory in Tripsacum strongly challenge thisview, bolstered by flow cytometry, which canbe used to analyze modes of reproduction(Grimanelli et al. 1997), and screens that usedifferent types of molecular markers.Nevertheless, the type of apomixis must stillbe considered, as different types ofscreens maybe applied to different types of apomixis.Whether one type of apomixis than another ismore likely to be expressed in a particular cropbackground is still largely speculative.6. Degree of apomixis (or degree of facultativeness).The degree of apomixis appears tobe a major factor related to the feasibility ofwide cross transfer of apomixis. An obligateapomixis cannot be used unless some degreeof male fertility is recovered in the F1s, whichis seldom the case in interspecific hybrids; butto produce near obligate apomictic crops,facultativeness must be low and wellcontrolled. This factor is addressed in moredetail later in this chapter.7. Agronomic characteristics. A species withpoor agronomic traits will produce hybridsand hybrid derivatives that may conserveundesirable traits for several generations,slowing the progress of the transfer.8. Previous knowledge. Previous knowledgeconcerning the interspecific or intergenerichybridization under consideration is a definiteadvantage. For example, knowing the numberof backcrosses needed to go from the maize­Tripsacum Fls to a 20-chromosome recoveredmaize (Harlan and de Wet 1977) was importantin developing the first work plan for the lRD­CIMMYT apomixis team and in maintainingits confidence about the feasibility of itsapproach.Case History: PennisetumPrnl1isetum glaucum, a cultivated pearl millet,has a basic chromosome number of x = 7. Theonly known and widespread tetraploid wildspecies with the same basic chromosomenumber is P. purpureum (211 = 4x = 28). Thoughdescribed as aposporic by Brown and Emery(1958), this species appears to be entirelysexual, as confirmed by a cyto-embryologicalsurvey made in morphologically uniform wildpopulations from West Africa (Y. Savidan,unpublished). Apomixis has been described inseveral other Pel1l1isetum species, all of whichbelong to the secondary or tertiary gene poolsand share a basic chromosome number of x =9. Dui.ardin and Hanna (1989) demonstratedthat three out of the seven apomictic speciestested were capable of producing F Ihybridswith pearl millet. The genus Pel1l1isetum,however, is one of the most complex in thegrass family. In addition, the number ofspeciesvaries greatly according to the taxonomist, themost conservative estimates beingapproximately 100 different species(Purseglove 1972), most of which are perennial,polyploid, and likely apomictic. Because of alack of available germplasm, no extensive

156 rn. s.mdaosearch for an opitimum apomictic donor forpearl millet has been conducted. Three speciesstudied by Dujardin and Hanna (1989) thatshow crossability with the crop are Pennisetumorientale, a tetraploid with 211 = 36; P. setaceum,a triploid with 211 = 27; and P. squamulatum, ahexaploid with 211 = 54. They all reproduceapomictically and their apomixis wasdescribed as obligate, which means that 100%of the observed progeny appeared to bematernal in field tests (Dujardin and Hanna1984a, b).Advantages of P. squamulatum as a donorspecies for apomixis include good pollenfertility, 4-nucleate embryo sacs, and a uniqueIssue # 1. Obligate vs. Facultative Apomixis: An Artifact?The facultativeness of apomixis has beenconsidered to be a disadvantage (Bashaw etal. 1970; Bashaw 1975) because (I) it may resultin W1controlled variation in the progeny whilefarmers require homogeneous varieties and(ii) it i apparently quantitatively inherited,i.e., under a c mplex, yet unknown geneticDiffere~ces in timing of developmentbetween meiotic and apomeioticembryosacsshould be considered in order to provide anaccurate estimate of the degree offacultativene s. This difference has beenfound in 'everal aposporous species asidefr m Pallicl/m, e.g., RanllnclIllIs al/riCOlllllscontrol. evertheless, acultativeness may be (Nogler 1984), Brachiaria spp (Ndikumananeed d in attempts to transfer apomi is tocrops through wide hybridization. Widecro ses gen rally produce highl terilehybrids that canonly be backcrossed by Llsingthem as female. If these hybrids are obligateap micts, the wide cross approach fortransferring ap mixis is a dead end. But isobligate apomiXis ever totally obligate?Asker (1979) assigned a que ti n mark toobligat apomi ·s. The developmentalproces has been described at th vule level,where meiosis. ucceed - r fails. At the plantlevel, obligate ap mixi is alreadyqu stionable. At the level of the popuJationor specie, obligate ap mi i is likely anartifact ofthe screening toot ( ee Leblan andMazzucato, Chap. 9).large number of ovules in th case of Psqllill/I//fall/III (DuJardin and Hanna 1984a)have b n examined and 8% were classifiedas abort db' d )n the absence of a normallyd el ped Jl;lbryo sac. In PlIl1icl/nt IIII1X;IIlWI/,an th r ap por us tropical forage rrass,ovul with n sac could be either abortive,in hich case they show enlarged nucellarcell' with little or no cytoplasm in an overallshrivelled ovary, or in early meioticd velopm nt stages. Sexual embryo sacs (ES)were ignificantly late as compared to thenucellar unreduced ES (Savidan 1982a).1985),P~-paltllllnotatum(MartinezetaI.1994),and dipl( p rQus Tripsacllm species (Leblancand avidan 1994), among others. Dujardin(personal camm.) confirmed that the nu eUifrom vules he classified as aborted wereperfectly normal, hence apomixis in the P.sqllall/llmlllm introduction was perhaps not asobligate as originally thought.Recent data (Hanna el al. 1993) showing highdegrees offacultativeness in later generationhybrid derivatives can possibly bereinterpreted in the light of this hypoth is.Modification of the genetic or epigenicbackground is known to affect facultativeapomixise pres ion, with an extremely highrate fs xuahty p ibly being observed. Thiswa" n in guin agrass (PauiCilm maximum),In ne natural interspecific hybrid with P.iufes/llm (see al Berthaud, Chap. 2).Though apllmi i is probably alwaysfacultative in the wild to some extent, thefacultativeness f the donor specie in atransfer attempt should be limited and/orcontrollable for apomixis to be properlymanageable in agriculture. Therefore, acompromise must be found between thefacultativeness required for male sterilehybrids to be backcrossed, and the finalobjective of relative homogeneity in thefarmers' fields.

Tra.sf.r .f Apomixis tllroogll WIde Cross.. 157potential, among the few species tested, forgiving some female and male fertility to the FIS.Disadvantages include the requirement of abridge species, P. purpureum, the different basicchromosome number (x = 9, as compared withx = 7 in pearl millet), and the hexaploid levelof ploidy. Progress made on mappingapomixis in Pennisetum and its implications forour understanding of the genetic control arepresented in Grimanelli et al. (Chap. 6).Case History: TripsacumNumerous maize x Tripsacum hybrids havebeen produced since the pioneering researchof Mangelsdorfand Reeves more than 70 yearsago (Mangelsdorfand Reeves 1931). Extensivehybridization studies have been carried outby Galinat (1971), Harlan and de Wet (1977),James (1979), and Bernard and Jewell (1985),among others. The main objective of thesestudies was to evaluate the potential role ofTripsacum in maize evolution and / or thefeasibili ty of gene transfer, though notnecessarily for apomixis. Claims ofintrogression have been made (Simone andHooker 1976; de Wet 1979; and Bergquist1981), but the Tripsacum progenitors involvedwere not tested beforehand for the target traits;consequently, the same traits couldpresumably have been present in neighboringmaize collections. However, all these studiesshowed that from a maize-Tripsacum F Ihybridit was possible, in a few generations, to recovera 20-chromosome maize with somemorphological features that were not presentin the original maize progenitor. Most of thesestudies were based on using a diploid sexualTripsacum, and most concentrated on a singlespecies, T. dilctylaides. Between 1990 and 1992,maize was successfully crossed with 66apomictic populations, representing eightdifferent species and intermediate formsbetween species (Table 11.1); 895 F Ihybridswith 2n =46 =10M + 36Tr were obtained fromthese crosses. Most of these (598, or 66.8%)involved T. dactylaides subspecies orinterspecific-like accessions involving someform of T. dactylaides. This confirmed highcrossability for T. dactylaides. The number ofPI plants per number of pollinated ears,however, showed a higher crossabilitybetween maize and T. wpilatense, which hasthe smallest area of distribution in Mexico(being found only in the Canon de Zopilote,between Mexico City and Acapulco).Advantages of using T. dilctylaides as the donorspecies include good pollen fertility and anapomixis characterized by an absence ofcallose around the megasporocyte andsubsequent cells, which is easily detected influorescence microscopy (Leblanc et al. 1995b;Leblanc and Mazzucato, Chap. 9). Diplosporyis further characterized by endosperms with aploidy level different from that ofsexual seeds,resulting from the fertilization of twoTable 11.1 Crossabilities between maize and wildTripsacum species and presumed naturalinterspecific hybridscode nb.pop ears emb. cult. Fls Fls/earZP 2 41 860 573 118 2.88DT 2 92 324 169 97 1.05iMZ 2 23 1119 140 20 0.87ilT 6 103 1143 427 83 0.81iDH 7 132 1527 452 84 0.64iPL 4 65 2169 257 33 0.51DH 30 776 10816 2892 386 0.50PL 1 4 10 1 0.25IT 5 132 779 123 32 0.24iDM 3 75 2513 444 14 0.19DM 7 121 3655 813 17 0.14LC 1 10 38 5 1 0.10BY 5 96 2091 390 7 0.07iBY 2 62 1847 352 2 0.03PR 1 20average 1732 895 0.52nb.pop.= number of populotions studied; eors= number of moize earspollinated with the TripsDcum spedes; emb.= number of countedembryos, three weeks aher pollination; cult.= number of embryoscultured; F1s= number of F 1 hybrids grown to maturity. Speciescodes: ZP= lzopilotense; DT=ldactyloides dactyloides (US types);MZ= lmaizar; IT=lintermedium; DH= ldacty\oides hirsutum; PL=lpilosum; DM= ldactyloides mexicanum; LC= llancealatum; BY;lbravum; PR= lperuvianum; i= intermediate forms (presumesnatural interspecific hybrids).

158 Yv•• Savid..unreduced polar nuclei. This trait can also beused for screening modes of reproduction insegregating populations by means of flowcytometry (Grimanelli et al. 1997). Previousstudies showing that 5-6 backcrosses areneeded to produce introgressed 20­chromosome maize plants provided anotheradvantage to using this species. Disadvantagesinclude total male sterility, which is seeminglyretained until reaching addition forms withvery few Tripsacum chromosomes, and thedifference in basic chromosome numbers (x =18 compared to x = 10 in maize).Production of Interspecific orIntergeneric F 1HybridsSeveral procedures are available to producehybrids between cultivated and distantlyrelated wild species. Special techniques,including chromosome manipulation,bridging species, hormonal treatment, embryorescue, ovary culture, and in vitro pollination,are available for overcoming the crossincompatibility and the sterility of the F1s. Thepresence of apomixis makes the cross moredifficult because it can only be performed inone direction, with the apomixis progenitorbeing used as pollinator. Therefore, the donormust exhibit good pollen fertility. Because mostapomicts require fertilization with reducedpollen to produce endosperm, pollen qualityis generally not affected by apomixis. Anexception to this rule is Elymus rectisetus, inwhich male infertility is a problem with mostaccessions O. G. Carman, personal comm.).Crossing TechniquesMost of the crossing techniques are commonto intra- and interspecific crosses. Aprerequisite is good knowledge of the selfsterilityor self-incompatibilitysystems existingwithin the crop. For most crops, however, handemasculation is preferred.Crossing species with different flower sizesand shapes may require special tricks, e.g., inthe case of maize x Tripsacum, more hybridsare produced if the silks are shortened to about2-3 cm. Most wide crosses require embryorescue techniques, using classical media suchas MS (Murashige and Skoog 1962) or N 6(Chuet al. 1975). Small embryos from maize xTripsacum F1s grew better on 50 gil sucrose ascompared with standard embryo culturemedium containing 30 gil sucrose. Severalenvironmental factors can further affect theproduction and culture of hybrid embryos. Asa result, the production of hybrids may be goodone year, but poor the next.When apomixis is not found in wild relatives,transfer may be attempted from a more distantapomictic species by using protoplast fusion.Such a transfer was started for sorghum usingapomixis from Cenchrus ciliaris (Bharathi et al.1991). However, no reports of plantregeneration have surfaced to date, apomicticor not, from such protoplast fusions. A morerecent approach, developed by Ramulu et al.(1996), explores the production ofmicroprotoplasts containing only one or twoalien chromosomes and the direct productionof monosomic addition lines after fusion withprotoplasts from the receptor species.Sterility of the FISSterility in interspecific and intergeneric F1sand supsequent backcross generations is acharacteristic of wide crosses. Restoringfertility of the F] hybrids through chromosomedoubling is the most common approach. Inboth pearl millet and maize transfer attempts,however, F1s from some wild speciesaccessions were totally sterile, while thoseobtained from other accessions showed somedegree of fertility, making the chromosomedoubling unnecessary.The transfer programs in pearl millet andwheat have produced F 1 hybrids with some

Tr...I" of Apomixis 1~,0.g11 Wide Cro.... 159degree of male fertility. However, as describedbelow for Tripsacum, this is not an absoluterequirement. Nevertheless, it obviously helps,because the F]s generally have morphologicalfeatures close to that of the wild progenitor,e.g., a limited number of fertile flowers topollinate. In maize, the Fls have less than 20flowers per inflorescence, while the recurrentmaize parent, if it could be used as female (i.e.,if the Fj hybrid had some male fertility), wouldoffer hundreds.Pennisetum setaceum (2n = 3x = 27) was the firstapomictic species crossed with pearl millet. F]hybrids had 2n = 25 chromosomes, were malesterile, but reproduced apomictically (Hanna1979). This interspecific cross was abandonedbecause of male sterility. Pennisetum orientale(2n = 4x = 36) was then crossed with pearlmillet. Fj hybrids had 2n = 25 = 18 P. orientale(Or) + 7 pearl millet (Pm) chromosomes(Hanna and Dujardin 1982). They were malesterile, but backcrossing was attempted usingpearl millet as the pollinator.Pennisetum squamulatum (2n = 6x = 54) wassuccessfully used to pollinate tetraploid pearlmillet. Crosses with diploid pearl millet failed(Dujardin and Hanna 1989). Of 20 F jhybrids,15 were facultative apomicts, based onembryo-sac analyses. One Fj was classified asan obligate apomict, although 35% of theovules were considered aborted. This maypossibly be interpreted in another way if thetiming of sexual and aposporic pathways ofdevelopment is different (see Issue # 1). Pollenfertility of this hybrid was surprisingly high(66%) and therefore it was used to pollinatetetraploid pearl millet to produce aBC]progeny. The BC] plants were totally malesterile. The breakthrough was found inmaking a tri-specific hybrid. The pearl milletx P. squamulatum male fertile F] (classified asan obligate apomict) was used to pollinate apearl millet x napier (P. purpureum) F], and1,730 hybrids were produced. A sample of 64segregated 31 apomictic (30 classified asobligate) and 30 sexual, which suggestsdominance of apomixis over sexuality.Relative crossabilities in maize x Tripsacum andpearl millet x wild species of Pennisetum areshown in Tables 11.1 and 11.2, respectively.According to J. G. Carman (personal comm.),the crossability between wheat and apomicticElymus rectisetus as measured by the same Fjs/ear ratio was less than 1%. Differences incrossability may possibly be due to relativedifferences in genetic distance between thecrop and its wild relatives or to genetic effects.Production of ApomicticProgenies throughBackcrossingFacultativeness becomes especially importantwhen interspecific or intergeneric hybrids aretotally male sterile. Dujardin and Hanna (1989)considered male sterility as an impediment tothe transfer of apomixis because theirprogenitors were apparently obligateapomicts. This was certainly reasonable basedon the available techniques and limitednumber of plants used for analysis at thebeginning of their project in the early 1980s. Inthe progenies of the maize x Tripsacum BC 3Table 11.2 Crossabilities between pearl millet andthree apomictic wild Pennisetum species(ross combination ears FIs FI slearpearl millet (2n =14) xP.arientale (2n =36) 88 20 0.23pearl millet (2n =28) xP.orientale (2n =36) 70 2 0.03peorl millet (2n =14) xP.setaceum (2n =27) 7 28 4.00pearl millet (2n = 28) xP.squamulatum (2n =54) 59 337 5.71average 224 387 1.73ears =number of pearl millet inflorescences pollinated with thePennisetum wild species; FIs= number of FI hybrids grown tomoturity.

160 Yves SoviclaoBashawet al. (1970) and Ba haw (1m)presented facultative apomixis as a difficulttrait to manipulate in breeding because ofuncontrolled variation (off-type frequency)that may result from crossing such apomictswith sexual plants. Our experience withaposporous Panicum maximum suggested thatfacultative apomixis, when the rate offacultativeness was low (1-5%), could bemaintained with the same or even lower rateofsexuality through consecutive generationsofhybridization. In suchcases, the FI and Belhybrids between sexual and apomicticguineagrass accessions had the same degreeof facultativeness as their apomicticprogenitor (Savidan 1982b). On the otherhand, crossing a highly facultative apomictwith sexual guineagrass accessions produceda largevariation for the rateoffacultativenessamong the apomictic hybrids (Savidan1982b). Whatever the complexity of thegenetic control of facultativeness, it seemedto be transmitted as a cluster along with thecontrol of apomeiosis (Savidan, 1982).Tripsacum diversity was not screened forfacultativeness, Whether sexual x apomicticTripsacum intra- or interspecific crosses mayresult in a similar conservation of the degreeof facultativeness is therefore unknown.Maize x Tripsacum hybrid derivatives couldexhibit contrasting rates of facultativeness,despite having originated from the sameapomictic F Ihybrid. Given our current stateof knowledge, this may be either acharacteristic of Tripsacllm apomixis or only aconsequence ofthe intergeneric, genetic, and/or epigenetic backgrounds.hybrid derivatives, only 0.9% of the plantsapparently resulted from fertilization of areduced egg cell, i.e., the rate of diplospory inBC 3s was 99.1%, which would probably notbe detectable if only 30 or 40 plants wereanalyzed in a progeny test.The obligate nature of apomixis may beoverestimated because of the population size,e.g., Burton et a!. (1973) classifiedapproximately 80% of their Panicum maximumaccessions as obligate apomicts based on 10­plant progeny tests. Savidan (1982b), however,found only 20% of such obligate apomictsusing a 100-ovary embryological analysis foreach accession. Therefore, the male-sterileapomictic interspecific F Ihybrid may probablyalways be used as female in the backcross,provided progenies of sufficient size can bescreened. One can expect that a few off-typeswill be produced from sexual reproduction (n+ n combinations) to help bypass the sterilitybarrier. Some may reproduce apomictically,assuming the apomixis "allele" is dominantand simplex, as observed in all sexual xapomictic hybrids produced so far in the grassfamily (see Nogler 1984 for review; Sherwood,Chap. 5). In Pennisetum, male sterile apomictichybrids could have been a good starting pointfor the transfer of apomixis if flow cytometryhad been available for screening of largeprogenies, but the technology only becameavailable to plant scientists several years afterthe project began (Galbraith et al. 1983).The BC Iplants from pearl millet x Pml1isetumorimtale hybrids had 23, 27, or 32 chromosomes.The latter were 2n + n off-types with25 + 7 Pm, as pear! millet was used aspollina!or. The 23-chromosome plants weredescribed as facultative apomicts, with a lowrate (or expression) of apomixis.From the crosses with P. setaceum, a 211 = 27BC Iplant appeared to be totally male sterile,but could be pollinated by pearl millet or P.setaceum. Pollination with pear! milletproduced no seed, while pollination with P.setaceum produced four plants, three maternaland one 2n + n. The P. orimtale pathway wasconsidered unsuitable for apomixis transferbecause of the low expression of apomixis orcomplete male sterility in the BC lderivatives.

Yraosfer of Apomilis 1~ ...gIt WIde Cro.... 161Hybrids that are totally male sterile andobligately apomictic are indeed dead ends:pollinating such hybrids with the crop pollenwill produce only maternal offspring, i.e.,perfect copies of the sterile Fl' However, ifapomixis is slightly facultative, off-types canbe produced, some of which maybe n + nandstill apomictic, representing progress towarda return to the chromosome number of thecrop. The rate of facultativeness has to be low,however, if one expects the backcrossprocedure to eventually produce an apomicticcrop germplasm with a high degree ofapomixis. Analyses made on Panicummaximum (Savidan 1982a,b)show that the rateof facultativeness, and more precisely of n + noff-types, may remain relatively conservedthrough generations of hybridization. It wastherefore suggested that a limited range ofvariation could possibly allow selection backto obligate apomixis. In the intergenericbackground of maize x Tripsacum hybridderivatives, the variation observed (Table 11.3)appeared less stable, possibly because theapomictic Tripsacum progenitor was alreadymuch more facultative than the guineagrassaccessions used by Savidan (1982). By selectingamong Tripsacum accessions for their ability toproduce hybrid derivatives in backcrossing Flswith maize, the team possibly selected one ofthe most facultative of the apomictictripsacums.Table 11.4 shows the cumulative result of theanalysis of approximately 6,000 progeniesproduced from maize x Tripsacum BetS withIssue # 3. Can apomixis be expressed at the diploid level?In the wild, apomixis is found only amongpolyploids (although a few, questionableexceptions have been cited, see Asker andJerling 1992). Population geneticists havesuggested thatsexuality wouldbeeliminatedif apomixis could be expressed at the diploidlevel (Pemes1912; Marshall and Brown 1981).Nagler (1984) claimed, with little evidence tosupport it, that apomixis is probably linkedto a lethal factor expressed at the haploid(gamete) level only. After obtaining 23­chromosome pearl millet x P. orientale Be,plants, Hanna et aI. (1993) stated thatpolyploidy is probably not needed for theexpression of genets) controlling apomixis,because these 23-chromosome plants hadonly one (simplex) set of nine P. orientalechromosomes. The genomic structure oftheseplants is likely 14 Pm + 9 Or however,suggesting that the locus involved couldpossiblybe present in triplicate. Another suchcase of apomictic expression in anonpolyploid form was previously reported(Dujardin and Hanna 1986), which related toa polyhaploid plant from a pearl millet x P.squamulatum p) hybrid which had 2n =41 =14Pm + 27Sq. This haploid had 2n = 21chromosomes. Again, as the 2n = 21­chromosome plant likely had sevenchromosomes from pearl millet and 14 fromthewild species that had a basicchromosomenumber of nine, the locus involved waspossibly in triplicate and not in duplicate.In the Tripsacum project, a few polyhaploidswere obtained in the progeny of 2n =56 =20m + 36tr BC)s (Leblanc et al. 19%). Theseplants have ne set of maize and one set ofTripsacll/ll chromosomes, as confirmed by insitu hybridization (Leblanc et al. 1996), andsome .of them could express apomixis.Whether they represent exceptional cases ofrecombination between apomixis and a lethalsystem linked to it is open to speculation (seeGrimanelli et al. 1998b). Grimanelli et al.(1998b) suggest, however, that apomixis canbeexpressed evenwhen the allele(s) involvedare in a duplex situation, a position that rej«tsthe hypotheses of dosage effect presentedearlierbyMogie (1988) and Noirot (1993), andsuggests that the transmission barrier,whatever its nature, may be overcomethrough haploidization to produce functionaldiploid apomicts.

162 hOI Scrvidao211 = 56 chromosomes, i.e., 20 maize + 36 to be apomictic, while a 211 = 38 maize xTripsacum chromosomes. Note that the Tripsacum BC hybrid that produces progenyaverage rate of facultativeness at that level ranging from 211 =22 to 2n = 32 is sexual. Anwas very close to that of the Tripsacum alternative can be offered by using markersprogenitor, although variation was important. linked with apomixis, provided that apomixisis indeed controlled by one gene or smallA few dihaploids have been obtained fromsegment of DNA, and that such markers arethe progeny of 211 = 56 BCls, as 11 + 0 off-typesclosely linked.(Table 4). They grew well, flowered, andproduced a good seed set. Their progeny were A 1:1 segregation for apomixis and sexuality80% maternal and 20% 211 + 11 hybrids with was observed among maize x Tripsacum Fls,211 = 38 chromosomes. as 31 hybrids were classified as apomictic and30 as sexual, based on embryological analyses.The backcross series was continued in anThese plants were used for a bulk segregantattempt to recover apomictic maize plantsanalysis (see Grimanelli et aI., Chap. 6) aimedwith only a few Tripsacum chromosomes. Atat identifying molecular markers thateach generation, plants were screened forcosegregate with apomixis. Three RFLPapomixis and chromosome number. Embryomarkerswere first identified as linked withsac analyses, which have been usedapomixis; these belong to the same linkageextensively in several genetic analysesgroup in maize and are located on maize(Sherwood, Chap. 5), cannot be applied tochromosome-610ng arm (Leblanc et al. 1995).intergeneric hybrids or hybrid derivatives inOther markers were subsequently addedwhich inflorescences are too precious to be(Grimanelli et al. 1998a, and Chap. 6).destroyed. Modes of reproduction aretherefore estimated using progeny tests, e.g., Using both flow cytometry and markera211 = 38 maize x Tripsacum BC hybrid that assisted screening for apomixis, rare but usefulproduces mostly 211 = 38 progenies is likely apomictic plants can be selected among manyat each generation. A sourcepopulation must be grown toTable 11.3 Facultativeness of apomixis and. diplospory rate inthe Tripsacum accession used in the backcross transfer of constantly prod uce new progenyapomixis into maize and three BC progenies, showing until the next generation population1variation for this rate. D: diplospory rate.is large enough to enable progress tobe achieved in the backcrossNo. of 211+0 211+n II+nprogenies maternal off-types off-types others D% ,program. With a rate of only 3%-T.d-,-o-cty-l,-oi-:-de-s------------------ useful plants, we decided to raise the#65-1234 98 69 26 3 0 96.9 BC I population to 3,500 plants. After8(1-6-82 55 40 15 0 0 100 about 6,000 progeny had been8C I -6-52 98 73 22 1 2 99.0 analyzed, we substituted this BC I8C]"5-45 78 63 6 8 1 89.7 nursery with a BC nursery obtained3from in vitro multiplication of the 2n=Table 11.4 Chromosome numbers of BC38 apomictic off-types produced by1 (2n = 56) progenies asestimated by flow cytometrythe BC 2polyhaploids (2n = 28). Morethan 2,500 BC 3s were established inProgenies maternal off-types off-types off-typestotal no. 211+0=56 211+11=66 11+11=38 11+0 the field. The analysis of a 125,000­6259 5006 1024 218 11% 80.0 16.4 3.5 0.2

Tro..le,.1 Ap.mlxls tltr..gIo Wide Cr.sso, 163plant progeny is shown in Table 11.5, in which fertile apomictic BC 4had been confirmed asthe rate of /1 + n off-types was below 0.2%. combining 20 maize chromosomes with lessAlmost 200 hybrid derivatives have been than 16 Tripsacum chromosomes (Table 11.6).produced and classified as BC , 4with Increasing the progeny size did not change thechromosome numbers ranging from 2/1 = 20 trend, an observation suggesting that theto 2/1 = 36. Modes of reproduction could be original transfer scheme (Figure 11.2) had tobeing determined for some of them by RFLP be reconsidered, especially since its 38­markers linked with apomixis, by progeny­ chromosome plant step could not produce thetests, or by ploidy of the endosperms addition lines that were expected.evaluated through flowcytometry (Table 11.6). The Table 11.5 Maize x TripsQcum B(3 progenies, in which the B(4s areprogeny size was recently the n+ ncategoryincreased further. Progenies maternal off-types off-types off-typestotal no. 2n+0=38 2n+n=48 n+n=20·36 n+0= 10, 28 others·Screening the modes ofreprod uction through flow125916 114602 10778 158 78 300(%) 91.01 8.56 0.12 0.06 0.24cytometry is a uniqueopportunity offered byreproduction apomictic apomictic segregoting sexuol, apo apomicticdiplosporous species such , mostly 4n (reslilution nuclei)as Tripsacum dactyloides. Insexual plants, triploid Table 11.6 Maize x TripsQcum B(4 with known mode of reproductionendosperms result from Plant 2n ISH· RFLP Endo PGT plant 2n ISH RFLP Endo PGTthe fertilization, by a 1496 20 Sex 1457 27 13M+ 14Tr Sexred uced pollen, of two 1500 20 Sex 1476 27 Aporeduced polar nuclei. 1502 20 20M Sex 1460 28 20M+8Tr? Sex SexDiplosporous plants form 1503 20 Sex 1484 28 20M+8Tr? Sex Sex1516 20 Sex 1348 30 Apoendosperm as a result of1529 20 Sex 1346 31 Apothe fertiliza tion of two 1454 21 Sex Sex 1347 31 Apoumed uced polar nuclei by 1482 21 Sex Sex 1439 31 Apoa reduced pollen. The 1489 21 Sex 1453 31 Apadifference is shown in 1492 21 Sex Sex 1479 31 I7M+ 14Tr Apo1535 21 Sex Sex 1276 32 ApoFigure 11.1. Diploid sexual1275 22 Sex 1339 32 Sexplants have triploid 1338 22 Sex Sex 1426 32 Apoendosperms (peak 2 in 1345 22 Sex 1306 33 SexFigure 11.1a), while 1422 22 Sex 1349 33 18M+ 15Tr Apo Apo Apotetraploid apomictics 1499 22 20M+2Tr Sex 1493 33 Apo Apo1534 22 Sex 1532 33 Apoproduced endosperms1393 23 20M+3Tr Sex Sex 1313 34 Sex(peak 2 in Figure 11.1b), 1515 23 Sex 1394 34 16M+18TrSex? Apowith a DNA content 2.5 1229 24 Sex 1494 34 16M+ 18Tr Apa Apo Apotimes that of the embryos 1425 24 Sex Sex 1517 34 Apo Apo(Grimanelli et al. 1997). 1481 24 Sex 1521 34 Sex Apo1526 24 20M+4Tr Apo Apo Sex? 1522 34 Apo ApoPreliminary data indicated 1528 24 20M+4Tr Sex Sex 1523 34 Apoapomixis could be 1471 25 20M+5Tr Sex Sex 1544 35 Apo1501 25 20M+5Tr Sex 1308 36 20M+16TrApo Apotransmitted to the BC 4generation, although no 'ISH: in situ hybridization data; RFLP: use of morkers linked to apomixis; Endo: flow cytometryanalysis of the ploidy of the endosperms; PGT= progeny-tesl (chromosome counts).

164 r.., SaYid..Transfer of Gene(s) forApomixis from an AlienChromosome to the CropGenomePossibilities of recombination between maizeand Tripsacum chromosomes are extremelylimited before the BC 3generation. As shownin the scheme presented in Figure 11.2, the onlymeiotic event prior to this level occurs withBC Iplants. However, pairing is preferentiallymaize-maize (M-M) or Tripsacum-Tripsacum(Tr-Tr) (Engle et a1. 1974), although trivalentand tetravalent associations have beeninfrequently reported (Engle et a1. 1973).ln theBC 3s, 20 chromosomes of maize are associatedwith one haploid set of Tripsacumchromosomes, and some M-M-Tr pairing mayoccur. The same may happen in later2a. diploid sexual Tripsocumgenerations with less Tripsacum chromosomes.Associations between maize and Tripsacumchromosomes have been reported to increasewith each BC generation (Engle et at. 1973),however, they seem to involve a limitednumber of maize chromosomes.Addition lines with 211 = 21 to 24, wheneverand whatever way they are produced, areexpected to show some degree of male fertility,as observed in all previous studies. Levels offertility may vary according to the number andquality of these alien chromosomes. Most oftheir progeny, using them as male, will likelybe 211 = 20 because of chromosome eliminationand pollen competition.The next step in transferring apomixis to maizeis still to produce fertile addition lines withone to three Tripsacum chromosomes. This onits own remains a large challenge, althoughseveral indirect avenues are presently underinvestigation. Pairing and recombination3l~xljlb. tetraploid apomictic Tripsocum2c. apomictic MxT8(4I I Iindex: 1 2 3Figure 11.1 Flow-cytometric analyses on entire seeds.o. 2n =36 diploid sexuol Tripsocum; peok 1: embryo (2n =36), peak2: endosperm (2n = 54), peak 3: duplicated cells from the embryo (62stoge of cell cycle); b. 2n = 72 tetroplaid apomictic Tripsacum; peak I:embryo (2n =72), peak 2: endosperm (relative DNA content suggests2n = lOx = ISO); c. 2n =24 B(4 moize- Tripsocum hybrid; peak 1:embryo (2n =24), peak 2: duplicoted cells from the embryo (62 stageof cell cycle), peak 3: endosperm (relotive DNA cantent suggests 2n =2x +2x + x +x =6S)! Fls 2n=46=10M+3r-6Tr'---,---,2~ xLJQ!L]! BCls 2n=56=20M+36Tr31 X[JQ[J! B(2s 2n=2S=1OM+ ISTr4c::. X [JQ[J1 B(3s 2n=3S=20M+ lSTr5c:::=::III X [JQ[]! B(4s 2n=27-36=20M+7-16Tr6[::::=II X [JQ[]! S(5s 2n=21-23=20M+1-3Tr70r XIIFl xdaminonl marker slo

!,..d., of Apomi.b t~..gh Wid.

166 rves Sqyldaounanswered. However, more progress hasbeen achieved toward producing an apomicticgrain during the last ten years than ever before,mostly because of the development andapplication of new techniques. As moleculardissecting tools continue to improve, we willsee great progress in our understanding ofhow apomixis is controlled and the isolationand manipulation of its components. Anotherpromising avenue, approaches based onmutagenesis, is discussed in the following twochapters. These approaches will undoubtedlybetter our understanding of the regulation ofreproduction as a whole. In the end, apomixiscertainly cannot be manipulated without athorough understanding of how it is controlledin the wild.ReferencesAsker, S. 1979. Progress in apomixis research.Heredffcs 91: 231-40.Asker, S., ond L. Jerling. 1992. Apomixis inPlants. BolO Raton, Florida: CRC Press.Bharathi, M., U.R. Murty, K.B.R.S. Visarada, andA. Annopuma. 1991. Possibility oftransferring obligate apomixis from(enmrus cilioris L to Sorghum bicolor(LlMoench. Apomixis Newsleffer 3: 13-14.Bashaw, E.e. 1975. Problems and possibilities ofapomixis in the improvemenl of tropicalforage grosses. Tropical Forages inUvestock Production Systems. 24: 23-30.Madison, WislOnsin: American Society ofAgronomy.Boshaw, E.e., AW. Hovin, and E.e. Hoh. 1970.Apomixis, its evolutionary significance ondutilization in plant breeding. Proc. XI Int.Grassl. (ongr. Madison, Wisconsin:American Society of Agronomy. Pp. 245­48.Bergquist, R.R. 1981. Tronsfer from TripsowmdOdyloides to corn of 0 mojor gene lorusconditioning resistance to Pu«inio sorghi.Phytopathology 7: 518-20.Bernard, S., and D.C. Jewell. 1985. Crossingmaize with Sorghum, Tripsowm and millet:the products and their level ofdevelopment following pollination. Thear.Appl. Genet. 70: 474--83.Birchler, J.A. 1993. Dosage ana~sis of maizeendosperm development. Ann. /lev. Genet.27: 181-204.Brown, W.V., and H.P. Emery. 1958. Apomixis inthe Grominelle : Panicoidelle. Amer. 1. Bot.45: 253-63.Burton, GW., !.C. Millot, and W.G. Monson.1973. Breeding procedures for Paniwmmaximum suggested by plant variabilityand made of reproduction. (rap Sci. 13:717-20.Corman, 1.G. 1997. Asynchronous expression ofduplicate genes in angiosperms may lOuseapomixis, bispory, tetraspory, andpo~embryony (review). Bioi. 1. Unn.Soc.61: 51-94.Chu, e.e., e.e. Wong, e.s. Sun, e. Hsu, K.e. Yin,and LV. Chu. 1975. Establishment of onefficient medium for onther culture of ricethrough lOmparotive experiments on thenitrogen sources. Sci. Sinico 18: 659--il8.de Wet, 1.MJ.de 1979. Tripsowm introgressionand agronomic fitness in maize Ilea maysLl. Proc. (onf. Broadening Genet. Bose inCrops. Wageningen: Pudoc Publish. Pp.203-09.Dujardin, M., and WW. Han no. 19840.Microsporogenesis, reproductive behaviorand fertility in five Pennisetum species.Thear. Appl. Genet. 67: 197-201.--.1984b. Pseudogomausporthenogenesis and fertilizotion of apearlmillet xPennisetum orienlole apomicticderivotive.1. Hered. 75: 503-04.--.1986. An apomictic po~hoploidobtoined from 0 peorl millet xPenniselumsquomulatum opomictic interspecifichybrid. Thear. Appl. Genet. 72: 33-36.--.1989. Crassobility of Pellrl milletwith wild Pennisetum species. Clop Sci. 29:77-80.Engle, L.M., 1.MJ. de Wet, and J.R. Harlan.1973. Cytology of backcross offspringderived from amoize- Tripsowm hybrid.(rap Sci. 13: 690-94.--.1974. Chramosomol variationamang offspring of hybrid derivatives with20 lea and 36 Tripsowm chromosOmes.(oryolagio 27: 193-209.Galbraith, D.W., K.R. Harkins, 1.M. Maddox,N.A. Ayres, D.P. Sharma, and E.Firoozaboby. 1983. Arapid flow 1yI0metricana~is of cellcyde in intact plant tissues.Science 220: 1049-51.Galinat, W.e. 1971. The origin of maize. Ann./lev. Genel. 5: 447-78.Galina!, W.e., P. Chandravadona, and B.GSRoo. 1970. Cytological mop of Tripsowmda

Yr."" .f Apomi.i. I.rotgir Wid. Cr.ue. 167leblonc, 0., o. Grimonelli, N. Islom-Foridi, J.8erthaud, and Y. Sovidan. 1996.Repraductive behavior in moize- Tripsacumpa~haploid plants: Implications far thetransfer of apomixis into maize. 1. Hered.87: 108--11.leblonc, 0., M.D. Peel. J.G. (arman, and Y.Savidon. 19950. Megosporogenesis andmegagometogenesis in several Tripsacumspecies (Pooceoe). Amer. 1. Bot. 82: 57­63.leblanc, 0., and Y. Sovidan. 1994. Timing ofmegasporagenesis in Tripsacum species(Pooceoe) as related to the control ofapomixis and sexuolity. Polish. Bot. Stud.8: 75-8l.Mangelsdorf, P.c., and R.G. Reeves. 1931.Hybridizalion of maize, Tripsacum, andEuchlaena. 1. Hered. 22: 339--43.Marsholl, R.D., and A.H.D. Brown. 1981.Estimotion of the level of apamixis in plontpopulations. Heredity 32: 321-33.Martinez, EJ., EEspinoza, and C.L Quarin.1994. Bill progeny (2ntnl from apomicticPaspalum nolatum obtained through eor~pollination. 1. Hered. 85: 295-97.Mogie, M_ 1988 Amadel for the evolution andcontrol of generative apomixis. Bioi. 1.Unn. Soc. 35: 127-53.Murashige, 1, ond ESkoog. 1962. Arevisedmedium for rapid growth and bioassoyswith tobocco tissue cultures. Physiol. Plant.15,413-97.Ndikumona, J. 1985. Etude de rhybridationentre especes opomitiques et sexuees damIe genre Brachiaria. Ph.D. dissertation.Univ. of lauvoin, louvoin-lo-Neuve,8elgium.Nogler, G.A. 1984. Gametophytic apomixis. In8.M. Jahri (ed.), Embryology ofAngiosperms. 8erlin: Springer-Verlag.Pp.475-518.Noirot, M. 1993. Allelic rotios and sterility in lheogamic complex of the Maximae(Panicoideael: evolutionary role of theresidual sexuolity. 1. Evol. 8io. 6: 95-101.Ozias-Akim, P., E.l. lubbers, W.W. Hanna, ondJ.W. McNay. 1993. Transmission of theapomictic mode of reproduction inPennisetunr. co-inheritalKe of the trait andmoleculor markers. Theor. Appl. Genet. 85:632-38.Ozios-Akins, P., O.Rache, and w.w. Honna.1998. Tight clustering and hemizygosity ofopomixis-linked molecular morke~ inPennisetum squomulatum implies geneticcontrol of apospory by adivergent locusthot may have no allelic form in sexuolgenotypes. Proc. Nal. Acad. Sci. (USA) 95:5127-32.Pernes, J. 1972. Organisotion evalutive d'ungroupe agomique: la section des Maximaedu genre Panicum (GramineesJ. Ph.D.dissertotion, University of Poris.Petrov, OJ., N.I. 8elausava, and E.S. Fokino.1979. Inheritance of apomixis and itselements in carn-Tripsocum hybrids.Genetika 15:1827-36.Pelrov, DJ., N.I. 8elousovo, E.5. Fokina, R.M.Yalsenka, L1. laikova, and 1P. Sorokina.1984. Tramfer of some elements ofapomixis from Tripsocum to maize. In OJ.Petrov led.), Apomixis and Its Kole inEvolution and Plant Breeding. RussianTranslotion Series 22. Rotterdam: AABalkena. Pp. 9-78.Purseglove, J.W. 1972. Tropical Crops,Monocotyledons. New York: langman.Ramulu, KS., P. Dijkuis, E. Rutgers, J. 8laas, EA.Krens, JJ.M. Dons, eM. (alijn-Hoojmans,ond H.A. Verhoeven. 1996.Microprataplost-medioted transfer of singlespecific chromosomes between sexual~incampotible plants. Genome 39: 921-33.Sovidan, Y_ 19820. Nature et heredite deropomixie chez Panicum maximum Jacq.Travaux & Documents ORSTOM 153: 1­159.--.1982b. Embryologkal ana~sis offacultative apomixis in Panicum maximumJacq. [rap Sci. 22: 467-69.---.2000. Apomixis: Genetia and8reeding. Plant Breeding Keviews 18: 13­86.Simone, G.W., and A.L Hooker. 1976.Monogenic resistance in corn toHelminthosporium turcicum derived fromTripsacum floridanum. Proc. Am.PhytopathoJ. Soc. 3: 207.

Chapter 12From Sexuality to Apomixis:Molecular and Genetic ApproachesUELI GROSSNIKlAUSIntroductionSexual reproduction usually producesgenetically diverse progeny, a feature that hasbeen exploited in the selection andimprovement of agricultural crops over thecenturies. In contrast, apomixis results in theproduction of genetically uniform progeny.Apomictic embryos are derived from anunreduced cell lineage and developindependent of fertilization (Gustafsson 1947;Nogler 1984a, b; Asker and Jerling 1992;Koltunow 1993). Thus, apomictic seeds aregenotypically identical to the mother plantand usually form a genetically stable clone.The clonal nature of apomictic offspring bearstremendous potential for seed production andcrop improvement. It offers the possibility forthe immediate fixation of any desiredgenotype and its indefinite clonalpropagation. The transfer of apomixis tosexual crops will completely transformcurrent breeding strategies and seedproduction (Hanna and Bashaw 1987; Savidan1992; Savidan and Dujardin 1992; Dickinson1992; Jefferson 1993; Hanna 1995; Koltunowet al. 1996; VielJe-Calzada et al. 1996; Jeffersonand Bicknell 1996; Grossnik.laus et al. 1998a;Savidan 2000). The resulting agricultural,commercial, and social benefits for bothindustrialized and developing countrieswould be enormous (Jefferson 1994;Koltunow et al. 1995; Grossniklaus et al. 1998c;and Toenniessen, Chap. 1).Apomixis is an asexual form of reproductionthrough seeds and occurs in more than 400species (Bashaw and Hanna 1990; Asker andJerling 1992;Carman 1995, 1997). Having beendescribed in plants belonging to almost 40different families, it is thought to have evolvedindependently in several taxa from sexualancestors. Apomixis can be viewed as adevelopmental variation of the sexualreproductive pathway in which certain stepsare short-circuited (Koltunow 1993; VielleCalzada et al. 1996; Grossniklaus et al. 1998a).Thus, apomictic and sexual reproduction areclosely related to one another and share manyregulatory components. It is very likely thatthe genes controlling apomixis also play crucialregulatory roles during sexual development.Therefore, the engineering of apomixis willrequire a better understanding of the geneticbasis and molecular mechanisms that controlsexual plant reproduction. Whereasmegasporogenesis and megagametogenesishave been studied extensively at themorphological and ultrastructural levels (e.g.,Cass and Jensen 1970; Willemse and van Went1984; Russell 1985; Mogensen 1988; Huang andRussell 1992; Schneitz et al. 1995; Christensenet al. 1997), the molecular and genetic basiscontrolling these key steps in sexualreproduction are almost entirely unknown.The engineering of apomixis in sexual cropswill only be possible through aninterdisciplinary and multifaceted approach tostudying the regulation of reproduction at thegenetic and molecular levels in both sexual and

Frvm Seuallty to Apomlxb: Moleaolor ..dGeoetk App..-H. 169apomictic species. Current research focuses onfour main complementary strategies: (i)characterization of the genetic regulation ofapomixis (reviewed in Nogler 1984a; Hanna1995; Savidan, 2000; Sherwood, Chap.5;Grimanelli et a!., Chap.6); (ii) development ofapomictic model systems for molecular geneticstudies (reviewed in Koltunow et al. 1995;Jefferson and Bicknell 1996; Bicknell, Chap. 8);(iii) analyses of the genetic basis and molecularmechanisms controlling megasporogenesis,megagametogenesis, and seed development insexual species (reviewed in Drews et a!. 1998;Grossniklaus and Schneitz 1998; Yang andSundaresan 2000); and (iv) development of themolecular tools needed to introduce andcontrol the expression of candidate genes(reviewed in Gatz and Lenk 1998). In additionto identifying key regulatory genes controllingthe sexual pathway, the identification ofmutants that display certain aspects ofapomictic reproduction is one of the mostpromising of the approaches using sexualmodel systems (Chaudhury and Peacock 1993;Ohad et al. 1995; Chaudhury et a!. 1997;Ramulu et a!. 1997; Grossniklaus and Vielle­Calzada 1998; D. Page, R. Pruitt, S. Lolle andU. GrossnikJaus, unpublished data).Apomictic reproduction is under geneticcontrol (Nogler 1984a; Savidan 2000;Sherwood, Chap. 5). In studies on its geneticregulation, it was found to behave as a singledominant Mendelian trait (Nogler 1973,1975,1984b;Savidan 1980, 1982; Gadella 1987;Ozias­Akins et al. 1993, 1998; Miles et a!. 1994;Sherwood et a!. 1994; Leblanc et al. 1995;Kindiger et al. 1996; Pessino et al. 1997;Grimanelli et a!. 1998; Barcaccia et a!. 1998; vanDijk et al. 1999; Bicknell et a!. 2000; Noyes andRieseberg 2000). Ideally, the gene(s) controllingapomixis would be isolated and characterizedin an apomictic species. However, moleculargenetic analysis in apomicts is difficult becauseof their poor characterization at the molecularand genetic levels, and the obstacles posed bytheir clonal mode of reproduction. Geneticanalysis, which relies largely on recombination,can only be studied in hybrids ofsexual and apomictic genotypes betweenspecies or genera. Moreover, apomixis istightly associated with polyploidy, therebymaking genetic analysis difficult. The adventof physical mapping has had a great impacton work on apomictic species and theirrelatives. Recent advances in the establishmentof molecular marker systems and thedevelopment of apomictic model systems arediscussed elsewhere (Savidan 2000; Grimanelliet aI., Chap. 6; Bicknell, Chap. 8).In this chapter, I focus on efforts to use wellestablishedsexual model systems to elucidatethe molecular mechanisms controlling plantreproduction. A better understanding of thegenes and molecules involved in the sexualpathway and the isolation of mutants relevantto apomixis will play an important role in thetransfer of apomixis to sexual species. Sincethis chapter focuses on the use of sexualsystems, the developmental events occurringduring sexual reproduction and their geneticcontrol will be reviewed in detail. I comparesexual and apomictic reproduction and discussmutants relevant to these developmentalprocesses. Then, some of the recent advancesin the molecular and genetic characterizationof sexual reproduction in Arabidopsis tlullianawill be described. Finally, various approachesto introducing apomixis into sexual species aresurveyed and the implementationtechnologies required to engineer apomixis ina useful manner are identified.Developmental Aspects ofSexual and ApomicticReproductionThe plant life cycle alternates between adiploid sporophytic and a haploidgametophytic generation, a feature withimportant implications for the formation of the

170 lie. Gros..iklalSgametes and embryogenesis (Walbot 1996).The meiotic products ofplants undergo severaldivision cycles to form a multicellular haploidorganism. The gametes differentiate later ingametophyte development. In angiosperms,double fertilization concludes thegametophytic phase and marks the beginningof the next sporophytic generation. Inangiosperm apomixis, the plant life cycle isshort-circuited and an unreduced cell lineagegives rise to a megagametophyte (gametophyticapomixis) or directly to an embryo(sporophytic apomixis). Because of the closedevelopmental and evolutionary relationshipbetween apomictic and sexual reproduction,a better understanding of the fundamentalbiological principles governing female gametogenesisand seed development will provideinvaluable tools for the manipulation of thesexual reproductive system towards apomixis.Sexual Model SystemsTwo well-established sexual model systems,Arabidopsis tlwliana and Zea mays (maize), anda rapidly emerging system, Oryza siltiva (rice),are of particular interest for genetic andmolecular investigations. All three are wellcharacterize~at the genetic level and offer avast array of powerful genetic and moleculartechniques (Freeling and Walbot 1993;Meyerowitz and Somerville 1994; Tanksley andMcCouch 1997; McCouch et al. 1997). Versatiletransposon systems for insertionalmutagenesis and gene tagging are availableand offer the opportunity for reverse geneticapproaches (Walbot 1992; Dellaporta andMoreno 1994; Feldmann etal. 1994; Shimamotoet al. 1993; Shimamoto 1995; McKinney et al.1995; Sundaresan 1996; Parinov et al. 1999;Speulman et al. 1999; Tissier et al. 1999;Meissner et al. 1999; Krysan et al. 1999).Maize, as an agriculturally important memberof the grass family (http://www.agron.missouri.edu), has some advantages forapomixis research. It can be hybridized withits apomictic relative Tripsacum dactyloides (e.g.,Mangelsdorf and Reeves 1931; Harlan and deWet 1977; Petrov et al. 1984; Savidan, Chap.11), and the extensive synteny among thegrasses (Bennetzen and Freeling 1993; Mooreet al. 1995; Gale and Devos 1998; Keller andFeuillet 2000) allows for comparative genomicanalyses between sexual and apomictic grassspecies. The Mutator transposon system offershighly efficient methods for insertionalmutagenesis (Chomet 1994) and for sitespecifictransposon mutagenesis by reversegenetic approaches (Das and Martienssen1995; Bensen et al. 1995; Mena et al. 1996;Rabinowtiz and Grotewold 2000), originallypioneered in the fruit fly Drosophilamelanogaster (Ballinger and Benzer 1989).The small plant Arabidopsis tlwliana, a memberof the Brassicaceae, has been widely adoptedas a model system for the developmentalbiology and genetics of flowering plants(Meyerowitz 1989). The small size of the plant,its rapid life cycle, and the large number ofseeds it produces make it ideal for the isolationand study of mutants that affect biochemicaland developmental pathways. The smallgenome size (-125 Mb), high percentage ofsingle copy DNA (Pruitt and Meyerowitz1986), large number of molecular markers(http://www.arabidopsis.com). and thecomplete genome sequence (ArabidopsisGenome Initiative 2000) make Arabidopsis apowert'ul system for molecular studies. Highlyefficient T-DNA-based transformationmethods (Bechthold et al. 1993) andheterologous transposon systems for targetedgene tagging, genome wide insertionalmutagenesis, and reverse genetics areavailable (Feldmann et al. 1994; McKinney etal. 1995; Sundaresan 1996; Parinov et al. 1999;Speulman et al. 1999; Tissier et al. 1999;Meissner et al. 1999; Krysan et al. 1999). TheArabidopsis genome is the first plant genometo be completely sequenced (Arabidopsis

fro.. SullCll1l, t. Apomixis: Mole

172 Uti Gro....1aosGunning 1990). The megaspore mother cell issurrounded by extensive callose depositions,which isolate it from the sporophytic tissuesof the ovule (Rodkiewicz 1970).In Arabidopsis, the two meiotic nucleardivisions occur before cytokinesis, leading tothe formation of tetrads with a multiplanar or,more rarely, a linear arrangement (Webb andGunning 1990; Schneitz et al. 1995). Incontrast,cytokinesis accompanies meiosis in maize, firstproducing two dyad cells and finally the fourmegaspores, which form a usually linear tetrad(Figure 12.1) (Weatherwax 1919; Kiesselbach1949; Russell 1979). Little information isavailable on rice megasporogenesis, but itsdevelopment is likely to be similar to thatobserved in maize. Only the chalazal-mostmegaspore survives (functional megaspore)whereas the other three undergo programmedcell death and degenerate. Acertain variabilitywith respect to the form of the tetrads andcytoplasm allocation to the functionalmegaspore has been observed (Webb andGunning 1990; Bedinger and Russell 1994).Degenerating and surviving megaspores areinitially similar at the ultrastructural levelexcept for an enrichment of organelles in thefunctional megaspore. However, only thedegenerating megaspores are surrounded bya callose rich cell wall, whereas the functionalmegaspore remains in direct contact withnucellar tissues (Rodkiewicz 1970; Webb andGunning 1990; Russell and West 1994). It hasbeen proposed that the pattern of callosedeposition during megasporogenesis plays acrucial role for the differentiation and survivalof the chalazaI megaspore, which eventuallyforms the megagametophyte (Haig andWestoby 1986).2. Megagametogenesis. The functionalmegaspore gives rise to a mature embryo sacof the Polygonum-type by three consecutivemitotic divisions that occur in a syncytium(Randolph 1936; Kiesselbach 1949; Misra 1962;Poliakova 1964; Diboll and Larson 1966;Russell 1979; Huang and Sheridan 1994; Webband Gunning 1994; Schneitz et al. 1995;Christensen et al. 1997; Moore et al. 1997;reviewed in Grossniklaus and Schneitz 1998;Drews et al. 1998). After the first division, thenuclei migrate to opposing poles of thedeveloping megagametophyte, and aprominent large vacuole forms in its center. Asecond vacuole at the chalazal pole is foundin Arabidopsis and some genotypes of maize(Vollbrecht and Hake 1995; Christensen et al.1997). As the embryo sac enlarges, theinteguments grow to envelop the nucellus.Asymmetric growth of the integuments givesthe Arabidopsis ovule its characteristicanatropous shape (Misra 1962). The nuclei ateach pole undergo two synchronous divisionsto form the 4- and 8-nucleated embryo sac. Asingle nucleus at the chalazal pole startsmigrating toward the micropylar pole andbecomes one of the two polar nuclei in thecentral cell. Cellularization leads to theformation of seven cells: an egg cell and twosynergids at the micropylar pole, threeantipodals at the chalazal pole, and a centralcell harboring the two polar nuclei (Figure12.1). In Arabidopsis, the nucellar tissue isabsorbed as the embryo sac grows andexpands. At maturity, remnants of the nucellusare only present at the chalazal pole. Theendothdial tissues, which are in direct contactwith the megagametophyte, are ofintegumental origin. In maize and rice, ovulemorphogenesis is characterized by an initialintensive proliferation of the nucellar tissue,such that at maturity the embryo sac is stillembedded in this tissue.The two synergids and the egg cell arearranged in triangular configuration at themicropylar pole to form the egg apparatus(Mansfield et al. 1991; Webb and Gunning1988; Diboll and Larson 1966; Russell 1979;

from Sexuolity '0 Apomi.il: Molo

174 UeIGr.........embryogenesis, which is initiated by a highlyasymmetrical division of the zygote (Willemse1981; Willemse and van Went 1984; West andHarada 1993).The large central cell is highly vacuolated andcontains the two polar nuclei originating fromopposite poles. It shares a common cell wallwith both the egg apparatus and theantipodals. The polar nuclei will fuse prior tofertilization to form the homo-diploid nucleusof the central cell; in maize and rice, nuclearfusion is partial and is not completed until thearrival of the sperm cells.Three antipodal cells differentiate at thechalazal pole. In Arabidopsis, they usuallydegenerate after the fusion of the two polarnuclei (Webb and Gunning 1988, 1990; Murgiaet al. 1993; Schneitz et al. 1995; Christensen etal. 1997). In contrast, the antipodals of thePoaceae are the only cells of themegagametophyte that proliferate aftercellularization. In maize, they form a clusterof up to 40 cells that are often cytoplasmicallyconnected (coenocytic) to some degree(Kiesselbach 1949; Diboll and Larson 1966;Vollbrecht and Hake 1995). In rice, 10 to 15antipodals, which appear to have a highlyactive metabolism, are present in matureembryo sacs (Jones and Rost 1989). Theephemeral nature of the antipodals inArabidopsis is intriguing and a clear functionof these cells has not yet been established.3. Double fertilization. At fertilization, thepollen tube penetrates the receptive synergidand delivers the two sperms cells (Russell1993). They migrate to the chalazal pole of thesynergid to fuse with their targets, the egg andcentral cell. Prominent actin coronas at thepresumptive site of fusion indicate a possibleinvolvement of actin filaments for sperm cellmigration and fusion (Russell 1993).Subsequently, fertilization of both the egg andcentral cell gives rise to the diploid zygote andthe triploid primary endosperm nucleus,respectively. The site of gametic fusion ischaracterized by poorly developed cell walls,such that the cell membrane of the synergid isin direct contact with the membrane of the eggand central cell. After fertilization, thecytoplasm in the zygote undergoes extensivereorganization Oensen 1968; Olson and Cass1981; Russell 1993). The zygote elongates butdoes not divide for some time. In contrast, theprimary endosperm nucleus dividessyncytially a few times before the firstasymmetric division of the embryo occurs(Randolph 1936; Kiesselbach 1949; Jones andRost 1989; Mansfield and Briarty 1990a;Sheridan and Clark 1994; Schneitz et al. 1995;Berger 1999). The development of theendopserm is initially syncytial. The nucleithen migrate to characteristic positions at theperiphery of the embryo sac and finallycellularize in a distinct developmental pattern(McClintock 1978; Marsden and Meinke 1985;Mansfield and Briarty 1990a,b; Walbot 1994;Berger 1999; Olsen et al. 1999). Successful seeddevelopment requires the coordinatemorphogenesis of embryo, endosperm, andthe integumental cell layers that form the seedcoat (Rutishauser 1969).ApomixisDuring apomictic reproduction, the sexualpathway described above is short-circuited. Asubseq~ent developmental event is initiatedbefore the previous one is completed. Thisdevelopmental heterochronicity is a hallmarkof apomictic reproduction and various modelshave been developed to account for thedevelopmental displacement of events duringmegasporogenesis, megagametogenesis, andfertilization (e.g., Mogie 1992; Peacock 1993;Koltunow 1993; Carman 1997; Carman, Chap.7). The developmental processes leading toapomictic reproduction are diverse and havebeen described in detail elsewhere (Nogler1984a; Asker and Jerling 1992; Koltunow 1993;

From Suoolily to Ap.....i1: MoIet.lar ..dGtto.,k Appr-N' 175Naumova 1993; Crane, Chap. 3). Anultrastructural characterization of apomicticdevelopment is discussed by Naumova andVielle-Calzada (Chap. 4). Here, I will brieflydescribe the main developmental features ofapomixis in order to facilitate a comparisonwith the sexual pathway.Two fundamentally different classes ofapomictic development can be distinguished(Gustafsson 1947; Nogler 1984a; Koltunow1993) (Figure 12.2). In sporophytic apomixis,an embryo forms directly from a nucellar orintegumentary cell in the ovule (adventiveembryony). Although adventive embryos arenot derived from gametophytic cells, theirdevelopment depends on the presence of amegagametophyte, because they usually relyon sexually derived nutritive endospermtissue. Sporophytic apomixis will not beconsidered further in this chapter because anengineered switch between sexuality andgameotphytic apomixis appears moreattractive for breeding purposes. However, itshould be kept in mind that in sporophyticapomixis only embryo initiation is affectedand, thus, it may be easier to tackle sporophyticapomixis at the molecular level.In gametophytic apomicts, the embryo resultsfrom the parthenogenetic development of anegg cell produced by an unreduced embryosac. The unreduced gametophyte can originateeither directly from nucellar cells (apospory)or from a megaspore mother cell that hasundergone no or an aberrant meiosis resultingin the formation of one (mitotic diplospory) ortwo unreduced megaspores (meioticdiplospory). Aposporous embryo sacs formmitotically from nucellar cells that developduring or after megasporocyte differentiationand are similar in appearance to the megasporemother cell. Often several aposporic embryosacs are present in a single ovule in addition tothe sexually derived one. In diplosporousdevelopment, a variety ofcytologically distinctprocesses lead to a failure in meiosis; themegasporocyte switches from a meiotic to anGametophytic ApomixisLR~)-e-du-ce-d-~o-Iar~ ~"'" §)~'"'~Apomictic embryo I Endosperm I I Endosperm II~Se-xu-a-1e-m-bry-o-"( Fertiialion(I EndospermIIApomictic embryo IFigure 12.2 The main developmental features of apomixis in relationship to the sexual pathway.Unreduced cells are in rectangular boxes, reduced cel~ are in oval boxes and key events are in dark~ shaded boxes. In sexual plants, themegaspore mother cell undergoes meiosis and one of the reduced megaspores forms the embryo soc. Embryo and endosperm are formed bydouble fertilizotion. In gamelophylic apomixis, reduction is avoided and embryo soc development initiated from on unreduced diplospore oron aposporic initial cell. The embryo develops parthenogenetically from the unreduced egg while the endosperm forms eilher aUlonomous~or through fertilization of the central cell (pseudogamy). In sporophytic apomixis, the embryo forms directly from on unreduced nucellorinitioIcell. The opomictic embryo relies on sexually produced endosperm.

176 U.IGm..lln.apomictic pathway with the net result ofproducing an umed uced functionalmegaspore (see Crane, Chap. 3).Unlike in sexual development, the megasporemother cell of diplosporous spe;:ies is notsurrounded by callose. In apospoTOUS species,callose deposition around the sexualmegaspoTOcyte can be abnormal (Crane andCarman 1987; Carman et al. 1991; Leblanc etal. 1993; Naumova et al. 1993; Naumova andWillemse 1995; Peel et al. 1997). As is typicalfor the functional megaspore in sexual~,aposporic initials, which will form apomicticembryo sacs, are devoid of callose (Naumovaand Willemse 1995). The marked difference ofcallose deposition between sexual anddiplospoTOus species is intriguing, but mostlikely is not causal, instead being theconsequence of a more fundamental lesion(Crane and Carman 1987; Carman et al. 1991;Carman, Chap. 7).The unreduced megaspore of diplosporousapomicts divides mitotically to give rise to amature embryo sac. Usually, only onemegaspore of the dyad initiates embryo sacdevelopment and the other degenerates, butbisporic development, in which twounreduced nuclei are present in the same spore,also occurs (Ixeris-type). In some apomicts, thedeveloping megagametophyte undergoes onlytwo mitoses to form a 4-nucleated embryo sacwhere no antipodals form (Panicum-type)(Gustafsson 1947; Nogler 1984a, b; Crane,Chap. 3). For instance, sexual megagametophytesin Pennisetum ciliare are of the typicalseven-celled Polygonum·type, whereasaposporic embryo sacs carry only four nucleiand typically form one egg cell, two synergidsand one polar nucleus (or a variation thereof)but no antipodals (Taliaferro and Bashaw 1966;Vielle et al. 1995). The egg cell of an apomicticembryo sac initiates embryogenesisautonomously in the absence of fertilization.In Pennisetum, the aposporic egg cell iscompletely covered by a cell wall (Vielle et al.1995) whose presence may prevent the fusionof the apomictic egg with a sperm cell (Askerand Jerling 1992; Savidan 1992). Someapomictic species are truly autonomous anddo not require fertilization at all; both embryoand endosperm develop autonomously. Incontrast, seed development in most apomictsdepends on the fertilization of the central cellto produce the nutritive endosperm, which isrequired for successful seed production(pseudogamy) (Nygren 1967; Asker 1979,1980;Nogler 1984a; Richards 1986; Bashaw andHanna 1990; Asker and Jerling 1992). Unlikein sexual species, the apomictic egg cell ofteninitiates embryogenesis before the firstendosperm division occurs.Interrelationship of Sexual and ApomicticReproductionSexual and apomictic reproduction aredevelopmentally and evolutionarily related.Apomixis can be viewed as a developmentalvariation of the sexual pathway. Apomictic andsexual modes of reproduction are not mutuallyexclusive. Whereas obligate apomicts produceexclusively clonal progeny, both forms ofreproduction coexist in facultative apomicts(Asker 1980; Richards 1986; Bashaw andHanna 1990). They form both reduced egg cellsthat are fertilized to produce geneticallydiverse progeny as well as apomictic embryosacs that give rise to clonal offspring.Apomictic and sexual embryo sacs occur in thesame plant or even within the same ovule(Asker 1980; Nogler 1984a; Vielle et al. 1995).Facultative apomicts, benefiting from theadvantages of both modes of reproduction,may have an evolutionary advantage and aremore common than obligate apomicts (Nogler1984a; Richards 1986; Asker and Jerling 1992).The degree of sexuality versus apomixis infacultative apomicts is influenced by a varietyof environmental factors, the effects of whichon the reproductive system are not well

F.... s.ualn, " Aponlllis: MoIoalar .d Geeetk ApProMMl 177understood (e.g., Knox and Heslop-Harrison1963; Knox 1%7; Frost and Soost 1%8; Cox andFord 1987; Hussey et al. 1991).The developmental regulation of sexualreproduction appears to be preserved duringapomixis. Although an apomictic gametophyteor embryo has a distinct developmental origin,the sexual developmental program is largelyconserved: megagametophyte development,embryogenesis, and the development of theendosperm and seed coat are identical insexual and apomictic genotypes. At the levelof gene expression, very few differences canbe detected between obligate apomictic andsexual genotypes of Pennisetum ciliare (Vielle­Calzada et al. 1996). In apomixis, the sexualpathway is altered at the transitions betweenthe two phases of the plant life cycle, meiosisand double fertilization (Figure 12.2): (i)meiosis is aberrant or absent leading to theproduction of an unreduced cell acting as thefunctional megaspore; (ii) the egg cell initiatesembryogenesis parthenogenetically orembryos form directly from a sporophyticinitial (adventive embryony); and (iii) theendosperm develops either autonomously or,in pseudogamous species, fertilization of thecentral cell is required for endospermformation and successful seed development.In pseudogamous species, special adaptationsmay be required for successful endospermformation. Thus, apomixis can be viewed as ashort-circuited sexual pathway (Koltunow1993; Vielle-Calzada et al. 1996; Grossniklauset al. 1998a) in which part of the sexualdevelopmental program is initiated at thewrong time or in the wrong cell. Thus,apomixis is characterized by a relaxation of thespatial and temporal constraints on thereproductive developmental process. It islikely that apomictic reproduction results fromthe heterochronic or heterotopic expression ofregulatory factors that control megasporogenesis,as well as egg and cen tral cellactivation in sexual species (Mogie 1992;Peacock 1992; Koltunow 1993; Grossniklausetal. 1998a).Whereas nonreduction and parthenogeneticembryogenesiS as two of the key componentsof apomixis have been discussed extensively,endosperm formation has attracted lessattention. In pseudogamous apospecies,mechanisms preventing the fertilization of theegg cell but allowing fusion of sperm andcentral cells may rely on the formation of acomplete egg cell wall prior to sperm arrival(Savidan 1992; Vielle et al. 1995). However,specific adaptations to maintain theendosperm balance number Oohnston andHanneman 1982; Ehlenfeldt and Ortiz 1995)may be required to ensure normal seeddevelopment. In maize, endosperm formationis strictly dependent on the presence ofmatemal and patemal genomes in a ratio of2m:1 p, due to differential imprinting of theparental genomes (Lin 1984; Kermicle andAlleman 1990). This requirement is likely toexist in many plantspecies, but may be relaxedor absent in some (Haig and Westoby 1991;Messing and Grossniklaus 1999). Sinceapomictic species produce normal pollen, thefertilization of an unreduced central cell witha single reduced sperm cell would violate theendopserm balance number and lead to seedabortion. Endosperm forma tion is animportant process that must be considered forthe tra~sfer of apomixis into sexual species(Grossniklaus et al. 1998a; Spillane et al. 2000;Savidan, 2000; Grossniklaus et al. 2001;Grimanelli et al., Chap. 6).Models for Apomixis: HeterochronicInitiation of DevelopmentA developmental analysis ofapomictic eventsclearly indicates that several developmentalprocesses occur simultaneously orasynchronously. Meiosis and embryo-sacformation may occur at the same time: theapomictic initial initiates embryo-sac

178 IJeIGr.........development without entering meiosis or afterpremature meiotic abortion and nuclearrestitution (Crane, Chap. 3). Likewise,parthenogenetic embryogenesis is usuallyinitiated prior to anthesis and often beforefertilization of the central cell in pseudogamousspecies. Thus, it appears as if specificdevelopmental events are initiated prior tocompletion of the previous ones. Heterochronicdevelopment is a hallmark of apomixis, withwhich specific developmental events arereplaced asynchronously or coexist andcompete with events occurring in normalsequence.In addition to a change in the temporal orderof developmental processes there is also arelaxed constraint on cell fate decisions.Whereas in sexual species a single nucellar cel1is usually committed to the meiotic pathway,several nucellar cells in apomictic species havethe potential to form unreduced gametophytes.The regulation of individual developmentalevents appears to be conserved betweenapomictic and sexual pathways. Therefore, itis likely that key regulatory genes playingessential roles in sexual development aremisregulated in either space and/or timeleading to the developmental alterationsobserved in apomicts. Precocious initiation ofmegagametogenesis and the prematureactivation of the egg cell could be caused bymisexpressed regulatory genes that performthe same functions during sexual reproduction.Thus, the gene(s) control1ing apomixis does notnecessarily encode altered gene products, butrather could be under relaxed or aberranttemporal and/or spatial control.Several models accounting for the precociousinduction of developmental events and theinterrelationship with the sexual pathway havebeen proposed (Peacock 1993; Koltunow 1993).Developmental checkpoints similar to the onesproposed to control proper progressionthrough the cell cycle (Hartwell and Weinert1989; Murray 1992) may ensure a strictsequential order of developmental stepsduring sexual reproduction. In apomicts,developmental checkpOints and feedbackmechanisms may be ignored or altered,leading to the initiation of a developmentalevent before the completion of an earlier one(Koltunow 1993).Alternatively, rather than misexpression ofregulatory genes in the nucellus, more generalchanges in the cellular machinery could causeapomixis. For instance, an increase in theduration of the cell cycle may allow genes tobe expressed at an earlier time in developmentthan usual. Such a situation has been observedfor genes with large introns in Drosophila. Thegenes knirps (kni) and knirps-related (knrl)encode highly similar proteins, but knrlcontains a large 19 kb intron (Nauber et al.1988; Oro et al. 1988; Rothe et al. 1989). Thus,knrl is only functional at nuclear division cycle13 during cleavage, when the cell cycle hasbecome long enough to allow RNApolymeraseto transcribe the entire knrltranscription unit before the initiation of M­phase (Rothe et al. 1992). In contrast, kni isexpressed already at nuclear division cycle 9.An intron-Iess knrl gene can fully rescue thekni embryo lethal phenotype (Rothe et al.1992). Two mutants have been isolated thatallow for the functional substitution of kni byknrl. Both act by lengthening the cell cycle and,thus, al'1ow knrl to be transcribed at an earlierstage of development than usual (Ruden andJackie 1995). A similar situation may beencountered in apomictic species, withduplicated genes being activatedheterochronically. The duplicated genes maybe paralogs present in the same genome ororthologs from two genomes brought togetherthrough hybridization.The models discussed above do not take intoaccount the tight association of apomixis withpolyploidy, although they are certainly

From Su.oIily '0 Apomixis: MoJ...lar .nd Gen.t. Appr••d,•• 179compatible with it. Several models thatconsider the relationship between polyploidyand apomixis have been put forward (Nogler1982; Mogie 1992; Noirot 1993; Carman 1997;Grimanelli et al., Chap. 6). For instance, anal teration of cell cycle length as hypothesizedabove may be caused by polyploidization orwide hybridization. In many species withisolates of several ploidy levels, diploids areusually sexual and polyploids apomictic(Asker and Jerling 1992; Leblanc et al. 1995).Autoploidization of sexual diploids hasresuIted in apomictic tetraploid plants in somespecies (Burton 1992). It is attractive tospeculate that the cell cycle length is altered inresponse to changes in ploidy. However, noexperimental data is available to support thishypothesis since essentially nothing is knownabout the regulation of the cell cycle duringreproductive development in plants. Theassociation of polyploidy with apomixis may,however, be a secondary effect caused bydeleterious mutations that accumulated in thegenome of apomicts. This is supported by therecent isolation of diploid apomicts inHierncium and Allium (Bicknell 1997; Kojimaand Nagato 1997). These and earlier findingssuggest that polyploidy is not an absoluterequirement for apomixis (Savidan 1980;Nogler 1982; Hashemi et al. 1989).An attractive hypothesis, which takes bothdevelopmental and genomic peculiarities intoaccount, has recently been proposed byCarman (1997; Chap. 7) based on earliermodels put forward by Ernst (1918). In short,the duplicate-gene asynchrony hypothesisstates that duplicate sets of genes regulatingreproductive development exist in polyploids.Polyploidy originally arose throughhybridization, such that the regulatory controlof development originating from the twogenomes may not be in synchrony. Theresulting intergenomic regulatory conflict maythen lead to the developmental aberrationsobserved in apomicts and other reproductiveanomalies. An important aspect of this theoryis that apomixis results from the conflictingaction of genes that usually playa regulatoryrole during sexual development, i.e., the samewild-type genes (not mutant forms) controlboth sexuality and apomixis. This reinforcesthe need for a better understanding of themolecular and genetic basis of sexualreproduction for the engineering of apomixisin sexual crops.Another consideration that could influenceresearch strategies was raised by Jefferson andBicknell (1996). At present, there is no evidenceto indicate that apomixis is controlled by atrans-acting gene product rather than by a cisactingelement. One can envision an alterationof a cis-acting element, for instance a bindingsite for a trans-acting factor (e.g., for atranscription factor or chromatin component),with altered affinity or copy number that couldcause apomixis by changing the concentrationof the trans-acting factor in the cell. Forexample, the factor could be titrated out by anincrease in the copy number of its binding sites,which in turn would result in precocious orinappropriate development of the embryo sac.Thus, a dominant locus could readily beexplained. The recent observation that largegenomic regions that are associated with theinheritance of apomixis are not present insexual relatives (Ozias-Akins et al. 1998; Rocheet al. 1999) is consistent with such a mechanism.Genetic Control ofReproduction and CandidateGenes for the Engineering ofApomixisTo date, no fully apomictic mutants have beenrecovered in sexual species, however, severalmutants and spontaneously occurringvariations of sexual reproduction displayindividual components of apomixis. Theseinclude the production of unreduced spores

180 Ud Gr....lIoo,(Rhoades and Dempsey 1966; Franke 1975;Harlan and de Wet 1975; Jongedijk 1985; Kauland Murthy 1985), the formation ofparthenogenetic haploids (Kimber and Riley1963; Turcotte and Feaster 1963; Sarkar andCoe 1966; Chase 1969; Hagberg and Hagberg1980), and the autonomous activation ofendosperm development (Ohad et al. 1996;Chaudhury et al. 1997; Grossniklaus andVielle-Calzada 1998; D. Page, R. Pruitt, S. Lolleand U. Grossniklaus, unpublished data). Theversatility found in sexually reproducingplants suggests that sexual and apomicticmodes of reproduction share many geneticregulatory components. The engineering ofapomixis will require a better understandingof the regulatory control of femalereproductive development in sexual plants atthe molecular and genetic level. Whatdetermines the commitment of a cell to aparticular developmental pathway such asmeiosis or megagametophyte development?What are the events leading to egg cellactivation and the initiation ofembryogenesis?How are these processes connected to thecontrol of the cell cycle? Answers to these andrelated questions will constitute an importantstep for our understanding of the reproductivesystem and its manipulation.Genetic analysis of female reproduction hasmainly focused on the characterization offemale sterile mutants that disruptmorphogenesis of the sporophytic tissues ofthe ovule (Robinson-Beers et al. 1992; Lang etal. 1994; Leon-Klosterziel et al. 1994; Modrusanet al. 1994; Ray et al. 1994; Gaiser et al. 1995;Reiser et al. 1995; Klucher et al. 1996; Eliottetal. 1996; Villanueva et al. 1999; Schiefthaler etal. 1999; Yang et al. 1999). Whereas thesestudies have led to the formulation of geneticmodels for ovule development (Angenent andColombo 1996; Schneitz et al. 1997; Baker etal. 1997; Grossniklaus and Schneitz 1998;Gasser et al. 1998; Schneitz 1999), the geneticbasis and molecular mechanisms controllingmegasporogenesis, megagametogenesis, andfertilization are almost completely unknown.This section reviews some of the geneticcomponents involved in female gametogenesis,with a particular emphasis on mutantsthat are relevant to apomixis research.Megasporogenesis and NonreductionMegasporogenesis is a complex processcharacterized by the determination of themegaspore mother cell, meiosis, and theselection and differentiation of the functionalmegaspore. Gametophytic apomixis involvesthe production of an unreduced gamete andits parthenogenetic development either withor without fertilization of the associated centralcell to produce the endosperm. Thus, animportant developmental decision is whetherthe megaspore mother cell or its apomicticcounterpart undergoes a reductional division.To better understand the nature of the"decision" to undergo meiosis, it is helpful toconsider two aspects. The first is the positionalaspect of how a nuceJlar cell is selected tocommit to the meiotic pathway. The secondpertains to what steps in the meiotic pathwayare essential to its irreversible commitment forfurther development into a functionalmegaspore and eventually an embryo sac. Thefirst issue is muque to seed plants, and has littlein common with organisms with dedicatedgerm lines, or those in unicellular models, suchas yeast. The second point, which may becausally linked to the first, can be rephrasedin the terminology of the cell cycle: Are theredevelopmental checkpoints duringmegasporogenesis that can be simulated orbypassed to induce an unreduced cell toinitiate megagametogenesis?InSights into the early steps ofmegagametogenesis can be gained from ananalysis of mutants affecting megasporemother cell differentiation and meiosis.Despite the isolation of a large number offemale sterile mutants in maize and

F.... S....lity to ""'Iis: Maleu. oat! Geoetk ",....., 181Arabidopsis, relatively little is known about thegenetic control of megasporogenesis. Recently,the isolation of 270 Arabidopsis mutants withdefective spore development (megasporogenesis-defective,msd) was reported (Schneitzet al. 1997). Mutants of the msd class do notproduce a megagametophyte, however,sporophytic ovule development proceedsnormally. The developmental defects duringmegasporogenesis have not been characterizedin detail. All of these mutants also affectmicrosporogenesis and, therefore, are maleand female sterile. They may affect meiosis perse rather than female specific processes.Usually, only a single archesporiaI cell, andconsequently a single megasporocyte,differentiates in an ovule. However, theoccurrence of multiple megaspore mother cellsin some species (Eames 1961; Walters 1985;Sumner and van Caseele 1998) and of twomegasporocytes in abou t 5% of the wild typeArabidopsis ovules suggest that severalnucellar cells have the potential to enter themeiotic pathway. Once a cell is committed, itappears to inhibit neighboring cells fromdoing the same (Grossniklaus and Schneitz1998). This view is supported by a recentlyidentified mutant in maize. Plantshomozygous for multiple archegonial cellsl(mac1) contain between three and 21megasporocytes in a Single ovule (Sheridanet al. 1996). Thus, mad is only likely to beinvolved in megaspore mot~~r celldetermination. The phenotype shows certainsimilarities to apospory, in which multipleaposporic initials form around the sexualmegaspore mother cell. However, unlike inapomicts where microsporogenesis is usuallyunaffected, mad mutants also show abnormalmale sporogenesis (Sheridan et al. 1999).The genetic regulation of meiosis has beenextenSively studied in maize and the yeastSaccharomyces cerevisiae (Golubovskaya 1979;Golubovskaya et al. 1992; Mitchell 1994;Roeder 1995). In yeast, a large body ofknowledge on the molecular mechanismscontrolling meiosis has been amassed. Manyyeast mutants that regulate the entry intomeiosis and differentiate between meiotic andmitotic division have been isolated. Thesemutants share some characteristics withapospory or diplospory of the Antennaria type(Koltunow 1993), and their plant homologscould be instrumental in the engineering ofapomixis.Many genes that play roles in yeast meiosishave been characterized, generally byidentifying mutants with specific meioticdefects and studying the level of transcriptsof the corresponding genes during meiosis(Mitchell and Bowdish 1992; Mitchell 1994).These meiotic genes act at different stages ofmeiosis. Genes acting early in the pathway andregulating the decision between mitotic andmeiotic divisions are of particular interest.Among the products of early meiotic genes,the meiotic activator IMEl is a master controlgene required for the expression of the genesacting in the early phase of meiosis (Kassir etal. 1988; Smith and Mitchell 1989; Mitchell etal. 1990; Kawaguchi et al. 1992). To befunctional, IMEl has to becomephosphorylated by RIMll (Bowdish et al.1994). Upon phosphorylation, the earlymeiotic genes are activated and a starveddiploid cell undergoes meiosis to produce fourhaploid spores. In the fission yeastSchizosaccharomyces pombe, the Mei3 gene isinduced by nutrient deprivation. Mei3 inhibitsthe protein kinase Pat1, that then triggers theentry into meiosis (reviewed by Yamamoto1996). The Patl kinase, in tum, represses theMei2 protein, which is an essential positivefactor for entry into meiosis. Thus, regulatorynetworks involving phosphorylation and dephosphorylationevents responding toenvironmental signals playa crucial role.in thecommitment to the meiotic pathway.

182 Ue5Gro........Meiosis is an almost universal feature ofeukaryotic organisms, suggesting that the keyreguIatory events are conserved a t themolecular level. A search for homologs of themeiotic genes in yeast could proveparticularly useful for the engineering ofapomixis. Indeed, a putative RIMl1 homologfrom rice has recently been isolated bypolymerase chain reaction (peR) Geffersonand Nugroho 1998). The deduced amino acidsequence is 68% similar and 50% identical toyeast RIMl1, with conserved protein kinasesubdomains. Experiments to investigateexpression and function of RIMll in rice areongoing and will reveal whether this geneplays the same regulatory function in plantsand yeast (S. Nugroho and R. Jefferson,personal comm.). However, there are a verylarge number of kinases that are similar toRIMll in the Arabidopsis genome, indicatingthat the identification of the ortholog willrequire a detailed analysis and functionaltests. Similarly, several groups aimed at theisolation of Mei2 homologs (Hirayama et al.1997; 1. Siddiqi, personal comm.; B. Tinland,personal comm.), but the presence of eighthomologous genes in the Arabidopsis genomecomplicates the identification of the functionalhomolog. Despite the conservation of somemeiotic genes between yeast and plants at thesequence level, it should be remE:mbered thatthe molecular aspects of the control of meiosisare not conserved between S. cerevisiae and S.pombe and that the regulatory mechanisms inplants could be completely different.In maize, mutants that influence the"decision" between meiosis and mitosis havealso been identified, but their molecularnature is unknown. In plants homozygous forthe ameiotic 1 (aml) gene, meiosis does notoccur and is replaced by a mitotic division(Palmer 1971). The aml gene appears tocontrol the switch from the mitotic to themeiotic cell cycle and is important for theinitiation of meiotic prophase I (Golubovskayaet aI.1992). In plants homozygous for certainaml alleles, the megasporocyte does not divideat all, whereas in others meiosis is replaced byone or several mitotic division cycles(Golubovskaya et al. 1993, 1997). In absence offirst division (afd) mutants, the first meioticdivision IS replaced by a mitosis(Golubovskaya 1979). This reversion to mitosisof a cell already committed to meiosis showssimilarity to apomixis characterized by arestitution nucleus at meiosis I (TaraxacumandIxeris-type).Several mutants in plants and yeast producetwo unreduced spores reminiscent ofdiplospory. The yeast mutant spo12 (Klapholzand Esposito 1980), the elongatel (ell) mutantof maize (Rhoades 1956; Rhoades andDempsey 1966) and triploid inducer (tri) inbarley (Ahokas 1977; Finch and Bennett 1979)affect the second meiotic division. Thus, theyproduce genetically diverse progeny. Whereasell affects both sexes, tri is specific to the femaleand regular reduced pollen is producedleading to the formation of triploid embryosupon self-fertilization. In Arabidopsis, the dyadmutant produces a dyad of megaspores ratherthan a tetrad (Siddiqi et al. 2000). Based onmicroscopical analysis of chromosomesegregation and the expression of meiosisspecificmarkers, the first meiotic divisionseems to occur normally in dyad mutants butthen m~;osisarrests (Sidd iqi et al. 2000). To myknowledge, dyad is the first Arabidopsis mutantaffecting sporogenesis in a sex-specific manner.Unlike in ell and tri where unreduced viablemegagametophytes are produced, dyadmutants are fully female sterile. Although themegaspores of the dyad sometimes undergoadditional divisions, no functional embryo sacsare formed (Siddiqi et al. 2000). Based on itsgenetic mapping position, dyad may beidentical to the female-specific switchl (swil)gene whose megaspore mother cell undergoes

F_ s....trty I. Apo"'1s: Maleaolor.d Geoelk Ap,..__, 183mitotic divisions similar to the phenotypedescribed for aml in maize (Motamayor et al.2000). The difference in phenotype may be dueto allelic variation as was observed for aml, inwhich different alleles have quite distinctphenotypes, or to genetic background effects.In the yeast mutant spo13, meiosis I is omittedand a dyad of unreduced spores is formed byan equatorial division (I

184 Uel Gn...lIDo,To date, the phenotypes of about a dozenmegagametophytic Arabidopsis mutants havebeen described in the literature (reviewed inDrews et al. 1998; Grossniklaus and Schneitz1998; Yang and Sundaresan 2000). InGametophytic factor (Gf) (Redei 1965;Christensen et al. 1997), andarta (Howden etal. 1998), tistrya (Howden et al. 1998),femalegametophyte2 ifem2) and fem3, gametophytefactor4 (gfa4) and gJa5 (Christensen et al. 1998),the functional megaspore does not initiatemegagametogenesis. In prolifera (pr/) (Springeret at. 1995), cdc16 (Yang and Sundaresan 2000),and hadad (hdd) (Moore et al. 1997), thesyncytial mitotic divisions are affected andembryo sacs show an early developmentalarrest. PRL as a member of the MCM2-3-5family and a putative component of DNAReplication Licensing Factor, is an essentialgene required in all dividing cells (Springer etal. 1995). CDC16 is another gene required forthe normal operation of the cell cyclemachinery. In gfa2, gfa3, and gJa7 mutantembryo sacs, the two polar nuclei do not fuse,a phenotype also observed in some of ourmutants 0. Moore and U. Grossniklaus,unpublished results).In maize, megagametophytes carryingindeterminate gametophyte (ig) or the r-X1deficiency undergo abnormal mitotic divisionsand are transmitted through the femalegametophyte at a reduced frequency (Kermicle1971'; Lin 1978, 1981; Weber 1983; Huang andSheridan 1996). We identified an Arabidopsismutant, haumea (hma), sharing some of theseaspects with ig, namely additional divisioncycles during megagametogenesis G. Mooreand U. Grossniklaus, unpublished data).Embryo sacs mutant for lethnl ovule (/01 and102) do not produce viable seeds (Singleton andMangelsdorf 1940; Nelson and Clary 1952).The 102 embryo sacs show a defect in nucleardivision and migration, and arrestpredominantly at the 1- and 2-nucleate stages,although some 102 megagametophytesundergo all three division cycles (Vollbrecht1994; Sheridan and Huang 1997).In hdd embryo sacs, nuclear divisions at themicropylar and chalaza] pole areasynchronous, and some of the mutantmegagametophytes cellularize prematurely.Thus, nuclear division and cellularization areregulated independently. However, these cellsdo not differentiate into a particulargametophytic cell type, suggesting that cellspecification may depend on the correct spatialcontext and lor the presence of neighboringcells. Prema ture cellulariza tion andasynchronous divisions at the poles have alsobeen observed in segmental deletions in maize(Vollbrecht and Hake 1995), suggesting thatseveral loci, including hdd, are involved in thespatial and temporal coordination ofcellularization, nuclear division, andmigration. Whereas these processes arenormally tightly coordinated, they areuncoupled ofeach other in hdd mutant embryosacs, indicating that several independentdevelopmental programs control the differentprocesses during embryo-sac development.They are usually highly coordinated, possiblyby checkpoint mechanisms and regulatoryfeedback loops. Developmental aberrationsmay relax this coordinated control, and thevarious developmental programs can occurindependently of their normal context.and ParthenogenesisFertilization and egg activation have beenstudied extenSively in animal systems at thephysiological, cellular, and molecular level(reviewed in Nucitelli 1991; Whitacker andSwann 1993; Jaffe 1996). Numerous studieshave shown that the adhesion of the sperm tothe egg cell triggers a transient rise in freecalcium ions and initiates a cascade ofdownstream events after fertilization Gaffe1991,1996; Whitacker and Swann 1993; Homaet al. 1993). The release of calcium ions isEgg Acti~ation

F.... S....III., ta ApelliJlis: MoItaoIor .d Geoelle Ap,.-HS 185absolutely essential for egg activation. Case inpoint: the introduction of calcium into the eggis sufficient to induce parthenogeneticactivation of sea urchin and mammalian eggs(Steinhardt and Eppe11974; Urangaetal. 1996).Recently, in vitro injection experiments withmouse oocytes have shown that a truncated c­ki t receptor tyrosine kinase leads toparthenogenetic egg activation that requiresboth calcium and phospholipase C activity(Sette et al. 1997).In plants, our understanding of the eventsfollowing fertilization is very limited.Experimentation with angiosperm gameteshas been difficul t beca use of theirinaccessibility and the complex milieu withinthe megagametophyte where doublefertilization occurs (Russell 1993; Dumas andMogensen 1993). Specific fusion of isolatedmaize gametes has recently been achieved andoffers great promise for the study of molecularand cellular events underlying fertilization andegg activation in vitro (Faure et al. 1994; Dumasand Faure 1995; Tirlapur et al. 1995; Kranz andDresselhaus 1996; Rougier et al. 1996). One ofthe first visible changes after fertilization of anisolated maize egg cell is the formation of acell wall (Kranz et al. 1995). In the brown algaeFucus, cell wall secretion depends on anincrease in the cytosolic calcium concentration(Roberts et al. 1994; Roberts and Brownlee 1995;Belanger and Quatrano 2000). A transient riseof the cytosolic calcium concentration hasrecently been reported for the first time inangiosperms after in vitro fertilization of amaize egg cell (Digonnet et al. 1997; Antoineet al. 2000). These observations suggest thatplant and animal egg activation may be similar,however, the role played by the release ofcalcium in the fertilized plant egg and whatevents it triggers are currently unknown.Parthenogenetic egg activation is a keycomponent of apomixis and occurs in manyplant species, both spontaneously (Kimber andRiley 1963; Turcotte and Feaster 1963; Chase1969) and after induction in isolated ovariesand ovules. Certain genetic backgrounds andmutants produce a high percentage ofparthenogenetic haploids in their progeny. Abetter understanding of the genetic basis ofparthenogenetic development will proveinstrumental for the engineering of apomixis.In maize, parthenogenetic haploids areproduced spontaneously at a frequency ofabout 0.1 % (Chase 1969), and as high as 3.2%in certain genetic backgrounds (Coe 1959;Sarkar and Coe 1966). In plants homozygousfor ig, the frequency of maternal hapl.oids isincreased fourfold to 0.6% as compared toisogenic lines homozygous for the wild typeIg allele (Kermicle 1969). Interestingly, ig alsoconditions the production of patroclinous(androgenetic) offspring at a frequency of morethan 2% as compared to 0.001% (1/80,000) inwild type stocks (Chase 1963; Kermicle 1969,1994). Thus, ig not only permits extra roundsof mitosis in the embryo sac (Lin 1978, 1981;Huang and Sheridan 1996) but it also promotesthe formation of embryos in the absence ofkaryogamy. Extremely high frequencies ofhaploid production are found in barley plantshomozygous for haploid initiator (hap) and inthe Salmon system of wheat. The hap mutationconditions the formation of 30% haploidembryos when homozygous (Hagberg andHagberg 1980; Asker et al. 1983). The sperm isprevented from fertilizing the egg by anunknown mechanism whereas the central cellgets fertilized normally to produce thenutritive endosperm (Mogensen 1988).Salmon. wheat lines can produce up to 90%parthenogenetic haploids (Matzk 1995). Bothcytoplasmic and nuclear determinants areinvolved in parthenogenetic activation: thepresence of the wheat-rye translocationchromosome 1BL-1 RS in the Aegilopscytoplasm of caudata or kotschyi Salmon leadsto haploid production, whereas either thetranslocation or the Aegilops cytoplasm alone

186 Uei Gr....lIa..do not display parthenogenetic properties(Kobayashi and Tsunekami 1978; Tsunewakiand Mukai 1990; Matzk et al. 1995; Matzk1996). The nuclear factors involved in theSalmon system have been defined geneticallyand shown to involve Pig, a gene that inducesparthenogenesis, and a Spg, a supressor ofparthenogenesis. Successful parthenogenesisdepends on the presence of the Aegilopscytoplasm and Pig, and the absence of Spg. TheSalmon system offers unique opportunities tostudy parthenogenesis at the molecular geneticlevel by comparing isogenic lines carryingeither the Trilicum (sexual) or Aegilops cytoplasms(parthenogenetic); various approachesto studying the molecular basis of parthenogeneticactivation have been initiated (Matzket al. 1995; Matzk 1996; Matzk et al. 1997).Endosperm Development andGenomic ImprintingDouble fertilization involves two pairs ofgametic cells. One sperm cell fuses with theegg to form the diploid zygote whereas theother fuses with the central cell to generate theusually triploid endosperm. The endospermprovides nutrients during seed developmentand synthesizes storage reserves required forpost-germination development. Differentmodes of endosperm development have beendescribed (Vijayraghavan and Prabhakar1984), the most common of which is nuclear.The primary endosperm nucleus undergoesseveral division cycles without cytokinesis toform a large number of free nuclei, which thenmigrate to the periphery of the central cell andcellularize in a specific developmental pattern,usually from the periphery to the center. Thenuclear type of development is typical ofcereals such as maize and rice (Brink andCooper 1947) and also for the model plantAmbidopsis (Mansfield and Briarty 1990a,1990b). The interactions between embryo,endosperm, and maternal tissue are a complexand poorly understood aspect of seeddevelopment. There are many developmentalpatterns of endosperm development and therelative importance of the differentcomponents in the seed with regard to nutrientsynthesis and acquisition varies accordingly.The origin, development, and function of theendosperm has recently been reviewed indetail (Lopes and Larkins 1993; Berger 1999;Olsen et al. 1999). In this section only aspectsof endosperm development that are of directrelevance to apomixis research are discussed.1. Interrelationship of embryo andendosperm development. In apomicticspecies, normal development of theendosperm is required for the formation ofviable seeds. Although recent studies suggestthat the morphogenesis of the embryo islargely independent of endospermdevelopment (Sheridan et at. 1995), endospermforms in all apomictic species studied to date.This is not only true for gametophytic but alsofor sporophytic apomicts, in which adventiveembryos depend on a sexually producedendosperm for their pre- and/or postgerminationdevelopment (Asker and Jerling1992). Endosperm formation is achieved eitherby allowing fertilization of the central cell(pseudogamy) or by autonomous division ofthe polar nuclei to form the endosperm tissue.It is likely that the formation of a complete cellwall around the egg cell prior to anthesisprevents the egg from being fertilized (e.g.,Vielle-Calzada et al. 1995), but allows thefertilization of the central cell inpseudogamous apomicts. In addition to thisspatial block there may also be importanttemporal controls. For instance, the fertilizationof the central cell occurs after theparthenogenetic activation and autonomousdivisions of the egg cell in many apomicts(Asker and JerJing 1992). This is in contrast tosexual species, in which the endospermnucleus divides several times before the firstzygotic division. In autonomous apomicts, thecentral cell develops parthenogenetically and

From s.x...."Y 10 Apomixis: MoI."Ia, ond G...tk App.-Ile. 187the developmental program for endospermdevelopment is activated in the absence offertilization.It is possible that the same genetic control isresponsible for autonomous endosperm andegg activation. This would be in agreementwith the hypothesis that the endosperm isevolutionary and derived from a secondembryo, as first proposed by Sargant (1900).This hypothesis is supported by morphologicalanalyses of fertilization in nonflowering seedplants of the genera Ephedra and Gnetum(Friedman 1990, 1992; Carmichael andFriedman 1995). In addition, molecular andgenetic investigations have shown that thereis a large overlap in gene activity between theendosperm and embryo. For instance, amongthe 855 characterized defective kernel mutantsin maize (representing about 285 loci), the vastmajority affect both the embryo andendosperm and very few are potentiallyendosperm-specific (Neuffer and Sheridan1980). Similar results were obtained in a studyof defective kernel mutants in barley (Bosnes etal. 1987), suggesting that a very largepercentage ofseed-specific genes are expressedin both embryo and endosperm, despite theirdifferent development and physiology. Incontrast to this extensive overlap in geneexpression between embryo and endosperm,studies on several recently isolated Arabidopsismutants that allow fertilization-independentendosperm formation but not embryogenesissuggest that endosperm activation may becontrolled by different developmentalprograms (Ohad et al. 1996; Chaudhury et al.1997; Grossniklaus and Vielle-ealzada 1998;Luo et al. 1999; Kiyosue et al. 1999; D. Page, R.Pruitt, S. Lolle, and U. Grossniklaus,unpublished data).The importance of the endosperm for seeddevelopment varies among species dependingon its developmental pattern. In some speciesof the Orchidaceae the endosperm undergoesonly a few division cycles or does not divideat all. In many dicotyledonous species,including Arabidopsis, the endosperm formsbut is essentially degraded by the time seedmaturation is initiated. In contrast, theendosperm ofcereals is persistent and ofgreateconomic value. Therefore, an engineeredapomixis in grain crops will have to allow fornormal development of the endosperm. In anideal situation, endosperm formation inengineered apomictic crops could be inducedautonomously. However, successful formationof endosperm in cereals depends on thespecialized cytoplasm of the central cell andrequires contributions from both maternal andpaternal genomes. This may be because somegenes are imprinted, that is their activitydepends on their parental origin (Kermicle1970; Kermicle and Alleman 1990; Messingand Grossniklaus 1999). Thus, engineeredapomictic grain crops are likely to require thefertilization of the central cell. The vastmajority of apomictic Gramineae arepseudogamous; possible autonomousapomixis has been observed in very fewspecies, including Calamagrostis, Poa nervosaand Nardus stricta Oohri et al. 1992).2. Genomic imprinting. Imprinting in plantsis usually regarded as specific to theendosperm (Kermicle and Alleman 1990;Walbot 1996; Alleman and Doctor 2000).Forma~ion of androgenetic and gynogenetichaploids in many species (Kimber and Riley1963; Sarkar and Coe 1966; Kermicle 1969) andof asexually derived embryos in apomictssuggest that imprinting does not playa crucialrole for embryogenesis in these species,although it may exist in ·'Jhers. Thedevelopment of embryos from somatic tissue(Zimmerman 1993; Mordhorst et al. 1997) andthrough anther culture (Zaki and Dickinson1990) is taken as further evidence thatimprinting is not involved in plantembryogenesis. However, the initial

188 UtI GrOSllilaot.development of such embryos is distinctlydifferent from sexually derived embryos andit is not clear whether the same developmentalprograms control embryogenesis in thesedifferent contexts. It is possible, for example,that imprinting requirements are suppressedunder certain culture conditions. Furthermore,it is likely that the importance of imprintingfor embryo and endosperm will differ amongspecies depending on the respective roles ofthese tissues in the production and acquisitionof nutrients (Messing and Grossniklaus 1999).For instance, the Arabidopsis gametophyticmaternal effect mutation medea (mea)drastically affects cell proliferation in embryoand endosperm, resulting in seed abortion(Grossniklaus et al. 1998b). Genetic andexpression studies suggest that MEA isregulated by genomic imprinting andexpressed in both embryo and endosperm atearly stages of seed development (Vielle­Calzada et al. 1999). At later stages, the imprintat the mea locus occasionally breaks down, butreports as to which tissues are affected differ(Kinoshita et al. 1999; Luo et al. 2000). That thegenetic background has strong effects on themea phenotype suggests that these differencesmay be ecotype dependent (Vielle-Calzada etal. 1999; Grossniklaus et al. 2001). Currently,however, it is not clear which fertilizationproduct is primarily affected, but it is likelythat MEA is required in both the embryo andendosperm. Genetic interactions of mea andsimilar mutants with mutants that affect DNAmethylation and/or chromatin remodelinghave been reported (Vielle-Calzada et al. 1999;Luo et al. 2000; Vinkenoog et al. 2000;Grossniklaus et al. 2001). We are currentlyusing mea as a starting point to isolateadditional genes involved in the genomicimprinting process through second-sitemodifier screens.Genomic imprinting has not been studied inmany plant species, but it has beenunequivocally demonstrated in theendosperm of maize at the genomic,chromosomal, and individual gene levels(Kermicle and Alleman 1990; Messing andGrossniklaus 1999). In maize, properdevelopment of the endosperm is strictlydependent on the presence of maternal andpaternal genomes in a ratio of 2m:1p (Lin 1982,1984; Birchler 1993). Any deviation from thisratio leads to a failure in endosperm formationand consequently to seed abortion.Interspecific and interploidy crosses suggestthat this is likely to be true for other speciesincluding most agriculturally important graincrops (Nishiyama and Yabuno 1978; Johnstonet al. 1980; Haig and Westoby 1991). In contrast,endosperm development in Arabidopsis doesnot require a 2m:1 p ratio because interploidycrosses involving diploid and tetraploid plantsproduce v iable seeds (Redei 1964;Grossniklaus et al. 1998b), but there are distinctparent-of-origin dependent effects on seed sizein interploidy crosses (Scott et al. 1998).3. Imprinting barriers to the introduction ofapomixis into sexual species. Imprintingphenomena may be behind the high degree ofsterility observed in hybrids between sexualand apomictic genera, and so, should be givenconsideration in efforts to introduce apomixisinto sexual species. In gametophytic apomixisthe female reproductive cells are unreducedwhereas microsporogenesis is unaffected andthe male gametophytes are reduced. Thus,fertiliz~tionof the central cell generates a ratioof 4m:1p, which is expected to result in seedabortion. However, apomictic species do notshow strongly reduced fertility, suggesting (i)that the constraints for imprinting are relaxedin apospecies or (ii) that the mechanisms offertilization have been modified. Apomicts doindeed show a relaxed requirement forimprinting, which is supported by the findingthat the ploidy level of the endosperm inapomictic species can be quite variable Oohriet al. 1992). A recent study by Grimanelli et al.(1997) clearly demonstrates that endosperm

from Sexuality t. Apomixis: MoIeatIor ..d Gtto.tk Ajlpr_lIe. 189development in TripsaclIm is normal under awide range of ratios of maternal to paternalgenomes. Similarly, apomictic Paspaillm speciesare insensitive to an imbalanced genome ratioin the endosperm while the sexuals maintainthis requirem~nt (Quarin 1999).Altered modes of fertilization that are expectedto maintain the endopserm balance numberhave been reported in several cases. This canbe achieved if either both sperm cells deliveredby the pollen tube fuse with the central cell, orif only one of the two polar nuclei and a singlesperm nucleus participate in karyogamy(Rutishauser 1954; Reddy and d'Cruz 1969;Nogler 1972, 1984a). Alternatively, unreducedpollen could serve as the male parent (Chao1980). As a very successful alternative torelaxed imprinting requirements, manyapomictic grasses show apospory of thePanicum-type, where 4-nucleated embryosacsare formed, which most often contain only onepolar nucleus that fuses with a single spermnucleus (e.g.,Savidan 1980). Sexual individualsof these agamic complexes usually produce 8­nucleated Polygonum-type embryo sacs withtwo reduced polar nuclei. Thus, fertilizationof both sexually and apomictically derivedcentral cells produces endosperms withbalanced parental genomes (Reddy 1977).Apomictic species may be evolutionarilyderived from predisposed genera that hadrelaxed imprinting requirements or,alternatively, evolved specific adaptations ofthe fertilization mechanism that maintainedthe imprinting requirements (see alsoGrimanelli et aI., Chap. 6). Such predispositionsand adaptations are not thought toexist in most sexual species and imprintingmay pose a serious problem to the introductionof apomixis into sexual crop plants(Grossniklaus et al. 1998a; Spillane et al. 2000;Savidan 2000; Grossniklaus et at. 2001). Indeed,apomictic maize-Tripsawm and pearl millet­Pennisetllm hybrids show a high degree ofseedabortion (Grimanelli et al. 1995; Dujardin andHanna 1989) tha t is likely to result from agenomic unbalance in the endosperm(Grossniklaus et al. 1998a; Morgan et al. 1998).Crosses with pollen donors of higher ploidythat maintained the endosperm balancenumber could possibly restore fertility. Atpresent, our understanding of imprinting andits importance for seed development is verylimited. A sustained effort toward a betterunderstanding of the genetic and molecularbasis governing imprinting is required toovercome the potential constraints to theengineering of apomictic crops.Genetic Screens for MutantsDisplaying Apomictic Traits inSexual Model SystemsIn previous sections, I discussed a number ofgenes that control sexual development thatcould serve as powerful tools for theengineering of apomixis. As an alternative, ascreen for mutants that display apomictic traitsin a sexual species could directly lead to theidentification of key regulatory components(Peacock 1992). This approach has been takenin several laboratories using Arabidopsis as amodel system. Although no apomictic specieshave been described in this genus, the closerelative Arabis holboelli is apomictic (Asker andJerling 1992). While direct experimentationwith Arabis is difficult because of its poorgenetic characterization and long generationtime, its close relationship to Arabidopsis maybe useful for comparative and widehybridization approaches.Arabidopsis Mutants withAutonomous Seed DevelopmentScreens for Arabidopsis mutants that allow seeddevelopment in the absence of fertilizationhave been performed in several laboratories.These screens take advantage of male sterilemutants and aim to identify second sitemutations that pseudo-suppress sterility. In

190 Ud e;,....1Ia..Arabidopsis, unpollinated pistils do notelongate, so that pistil elongation is an easilyscored phenotype correlated with seeddevelopment. Different male sterile mutantshave been used, including pistillata (pi), ahomeotic flower mutant that lacks stamen(Chaudhury and Peacock 1993; Koltunow etal. 1995; Chaudhury et al. 1997); the waxbiosynthetic mutant eceriferum6 (cer6) (Dellaert1979; Preuss et al. 1993; HUiskamp et al. 1995),a conditional male sterile that is fertile underhigh humidity (Ohad et al. 1996); and thetemperature sensitive mutant TH154, isolatedin R. Pruitt's laboratory, which is male sterileat 25°C but fully fertile at 18°C (D. Page, R.Pruitt, S. Lolle, and U. Grossniklaus,unpublished results). Silique elongation underthe restrictive condition indicates an asexualmode of reproduction with full or partial seeddevelopment, or the development of a fruitwithout concomitant seed production(parthenocarpy).Using this type ofscreen with more than 15,000M1 plants, we identified three classes ofmutants (D. Page, R. Pruitt, S. Lolle, and U.Grossniklaus, unpublished results): (i) mutantssuppressing the male sterility defect, which caneasily be identified because they producefunctional pollen (some of these will be truerevertants of TH154); (ii) mutants displayingparthenocarpy; and (iii) mutants displayingapomictic traits with autonomousdevelopment of seed-like structures in theabsence of pol1en production. The latter arerather rare and, to date, mutants in only threeloci have been reported. These have beencharacterized in more detail at the genetic andmorphological level. In the fertilizationindependentendosperm (fie), mutant endospermdevelops autonomously to the free nuclearstage, the seed coat develops normally, but noembryo forms (Ohad et al. 1996). Likewise,fertilization-independent seed mutants (fisl, fis2,fis3) form autonomous endosperm, which wasshown to be diploid and can progress to thecellular stage in fisl and fis2. The seed coatdevelops properly but no embryos form(Chaudhury et al. 1997). Structures thatresemble an elongated zygote have beenobserved infisl andfis2 at a low frequency. Infieffis plants, the mutant allele is either verypoorly transmitted or not transmitted at al1through the female gametophyte and can onlybe recovered through the pol1en, becausepol1inated seeds derived from a mutantmegagametophyte abort (Ohad et al. 1996;Chaudhury et al. 1997). Thus,fie/fis mutantsdisplay a gametophytic maternal effect on seeddevelopment, which is similar to thephenotype observed for mea that also showsfertilization-independent endospermdevelopment (Grossniklaus et al. 1998b;Grossniklaus and Vielle-Calzada 1998).Indeed,jisl and two other mutants were foundto be allelic to mea (Kiyosue et al. 1999; Luo etal. 1999), as werefis3 and fie (Chaudhury et al.1997; Ohad et al. 1999).MEA and FIE encode members of thePolycomb group, proteins thought to regulategene expression by modulating higher orderchromatin structure (Grossniklaus et al. 1998b;Ohad et al. 1999). FIS2 encodes a putative DNAbinding protein with a Zn-finger domain (Luoet aI.1999). As with their animal counterparts,the MEA and FIE proteins interact directly ina protein complex (Luo et al. 2000; Spillane etal. 2000; Yadegari et al. 20(0). The function andregulation of MEA, FIE, and FIS2 have beenextensively reviewed (Goodrich 1998; Ma 1999;Preuss 1999; Mora-Garcia and Goodrich 2000;Russinova and de Vries 2000; Grossniklaus etal. 2001) and will be summarized only verybriefly. MEA was shown to be regulated bygenomic imprinting, thus explaining itsmaternal effect on seed development (Vielle­Calzada et al. 1999; Kinoshita et al. 1999). WhileFIS2 also seems to be regulated by genomicimprinting (Luo et al. 2000), FIE appears to beexpressed biparentally later during seeddevelopment (Luo et al. 2000; Yadegari et al.

F.... Seu.lity to Apomixis: MaIo

192 Ueli Gro".llae,system to identify mutants that lead to theformation of an embryo sac from an unreducedcell lineage. In addition, full-sized kernels canbe scored for absence of the dominant paternalR-nj marker in the embryo, which indicatesparthenogenetic development.Because obtaining a mutation that causes bothnonreduction and parthenogenesis may beextremely difficult, a second screen aimed atthe isolation of parthenogenetic mutants hasbeen developed. It makes use of the el1mutation, which, when homozygous, producesa large fraction of unreduced embryo sacs. Inel1 homozygotes, independent assortmentduring meiosis I is not affected and theresulting gametes are not genetically identical(Roades and Dempsey 1966). Nevertheless, itprovides a reliable source for unreduced femalegametes. Lines homozygous for rl and el1 anddisplaying high Mutator activity have beenconstructed to serve as female recipients.Kernels derived from crosses with a tetraploidR-nj pollen donor can be scored for rejection ofthe R-nj marker, i.e., for the maternalphenotype, in order to identify embryos thatdeveloped without a paternal contribution.Such embryos will be diploid (through theaction of ell), which greatly facilitatessubsequent genetic characterization. Thesegenetic screens aimed at the isolation ofmutations that lead to nonreduction,parthenogenesis, or a combination of bothaspects wtll provide useful material to furtherour understanding of these processes at themolecular and genetic level.Enhancer Detection as aPowerful Tool to StudySexual Reproduction inArabidopsisThe molecular and genetic bases ofmegasporogenesis and megagametogenesisare poorly understood. To date, only onefemale-specific mutant that affectssporogenesis has been identified (Siddiqi et al.2000; Motamayor et al. 2000) and attempts toisolate genes that regulate the developmentalevents initiating female gametogenesis andembryo development are just beginning. Asan alternative to the isolation of mutations, weidentify genes expressed specifically duringmegasporogenesis and megagametogenesis.The inaccessibility of the developing embryosac and the small number of cells involvedmake this a difficult undertaking usingconventional molecular methods such asdifferential screening techniques. Therefore,we use a novel technology, enhancer detection,which allows the identification ofdevelopmentally regulated genes based ontheir pattern of expression. Enhancer detectionis one of the most powerful tools to identifytissue specifically expressed genes and theirregulatory sequences. Application of thisapproach in angiosperms will lead to theidentification of many genes that controlgametogenesis and cellular differentiation inthe female gametophyte. In addition, it willidentify many cell type- and tissue-specificregulatory regions that will be required toexpress candidate genes in a precise temporaland spatial fashion.Enhancer Detection andGene Trap SystemsEnhancer detection was first developed in thefruit fly Drosophila melanogaster and relies on amobile genetic element carrying a reportergene under the control of a weak constitutivepromoter (O'Kane and Gehring 1987). Thisminimal promoter is usually not active butideally it can be activated in all tissues and atall developmental stages. Ifit comes under thecontrol of genomic cis-regulatory elementssuch as enhancers, the reporter gene isexpressed in a specific temporal and spatialpattern (Figure 12.4, p. 197). This patternreflects the expression of a nearby genecontrolled by the same regulatory elementsand, thus, allows the identification of genes

F"'II\ Sexoallty ,. Apolllixl" MoIeaolar.d G...tk Approadoe. 193based on their pattern ofexpression rather thanon a mutant phenotype (Bellenet al. 1989; Bieret al. 1989; Grossniklaus et al. 1989; Wilson etal. 1989). Enhancer detector screens have beenextremely successful in Drosophiladevelopmental genetics and similarapproaches were rapidly adapted to othermodel systems. Because of the largeintergenomic regions in mice, gene traps weredeveloped that depend on a modification ofthis approach involving the generation oftranscriptional fusions to the reporter gene(Gossleretal.1990;Skarnes 1990; Friedrich andSoriano 1991). In Arabidopsis, similar systemsbased on T-DNA insertional mutagenesis(Topping et al. 1991; Fobert et al. 1991;Kertbundit et al. 1991) or the AciDstransposable element system from maize(Sundaresan et al. 1995; Springer et al. 1995;Smith and Fedoroff 1995) have beendeveloped.Enhancer detection and gene trap systemsoffer an added benefit: they allow theidentification of genes that are not readilyamenable to classical genetic analyses (Bellenet al. 1989; Bier et al. 1989; Grossniklaus et al.1989; Wilson et al. 1989). They have beenespecially useful in studying developmentalprocesses occurring late in development, i.e.,after the effective lethal phase of acorresponding mutation. For example, a genethat is required for essential steps during bothembryo and ovule development would beidentified as an embryo lethal mutation andits function during ovule development wouldbe masked. Enhancer detection allowedidentification of the first embryo lethal genesin Drosophila that are also required foroogenesis or eye development (Grossniklauset al. 1989; Mlodzik et al. 1990). The dissectionof processes characterized by functionalredundancy and high complexity is alsogreatly facilitated by enhancer detection(Wilson et al. 1990; Bellen et al. 1990).Enhancer detection has some importantadvantages over classical genetic screens forthe identification of genes required in thegametophytic phase of the life cycle: (i) manygenes that encode components of the basiccellular machinery display a gametophytelethal phenotype if disrupted. Essential genesare expected to show widespread although notnecessarily ubiquitous expression, whereasexpression in particular cell types of themegagametophyte suggests a function in cellspecification and differentiation; (ii) genesrequired for both micro- and megagametogenesiscan be isolated, because a largepercentage of enhancer detector insertions donot disrupt gene function. Mutants affectingboth male and female gametophytes canusually only be recovered as rare partiallypenetrant mutations or in genomic regions thatcan be covered by a duplication (Vollbrecht1997); (iii) enhancer detection is the onlyefficient technique that allows theidentification of genes expressed in very fewor even single cells. By focusing on the cellsand tissues where a gene is expressed, subtlephenotypes can be identified that may noteasily be recognized in phenotypic screens;and (iv) most importantly, enhancer detectorand gene trap transposons greatly facilitate themolecular cloning of genomic sequencesflanking the insertion site. In addition, theyallow a detailed genetic analysis of the detectedgene through remobilization and the recoveryof additional alleles, regional chromosomalrearrangements, and revertant sectors (e.g.,Grossniklaus et al. 1992; Springer et al. 1995;Tsugeki et al. 1996; Grossniklaus et al. 1998b).Generation of Transposants andOngoing ScreensTo identify genes involved in femalegametogenesis, we have generated nearly4,300 lines carrying randomly insertedenhancer detector or gene trap transposons (U.Grossniklaus, J. Moore, W. Gagliano, J-P. VielleCalzada, unpublished data). We are using the

194 UeI Gm..ikIonsystem developed by Sundaresan et al. (1995),which is based on the Ac/ Os transposon ofmaize and allows the recovery of unlinkedtransposition events throughout theArabidopsis genome. In brief, an enhancerdetector or gene trap transposon is mobilizedby crossing a starter line homozygous for Osto a line carrying a stable Ac transposasesource. Self-pollination of the F] plants, whichcontain both the Os starter locus and the Actransposase construct, results in some F 2progeny carrying a transposed Os element(transposants). By pOSitively selecting for thepresence of Os but negatively against thedonor Os locus and Ac, unlinked stabletransposition events can be recovered(Sundaresan et al. 1995). Negative selectionagainst the donor locus ensures the recoveryof unlinked or loosely linked transpositionevents, a prerequisite for genome-widerandom insertional mutagenesis. Similarstrategies are currently being developed forrice (Chin et al. 1999; R. Jefferson, personalcomm.).Using six independent starter lines, wegenerated approximately 45,000 Fjs, of whichmore than 35,000 were grown to maturity toharvest their F 2seeds. About 23,000 of the F 2families have been put through the positive/negative selection process to recovertransposants. Between 20% and 25% of the F2families yielded an unlinked transpositionevent. The transposant library of about 4,300lines that we generated serves as the basis forfour large-scale screens aimed at identifyinggenes involved in female reproduction.Two of the screens we are performing aredesigned to identify genes expressed duringovule and megagametophyte development.The first one targets early ovule development,which encompasses the key events ofmegasporogenesis (Vielle-Calzada et al. 1998).More than 1,000 transposants have beenscreened and about 30 lines have beenidentified with a restricted expression patternin young ovule primordia O-P. Vielle Calzadaand U. Grossniklaus, unpublished data).Many of the expression patterns reflect thehighly polar organization of the ovule andmay be involved in establishing or interpretingpositional information. Other patterns areassociated with the meiotic cell lineage andare of particular interest to the engineering ofapomixis. The second screen is aimed at theidentification of genes expressed in individualcell types of the embryo sac and concentrateson late stages of ovule ontogeny, includingpost-fertilization stages (R. Baskar, J. Moore,w. Gagliano, U. Grossniklaus, unpublisheddata). Some of these patterns are specific tothe individual cell types of the embryo sacincluding the egg cell and will serve asimportant tools to direct expression ofcandidate genes to the megagametophyte. Theresults of this screen are discussed in moredetail in the next section.Independently, our transposant library isbeing used for classical genetic screens toisolate insertional mutants that affect fertility.The first phenotypiC screen identifies mutantsthat disrupt the development or function ofthe female gametophyte. These mutants areidentified in a two-step screen for reducedfertility and a non-Mendelian segregationratio, both indicating a gametophytic defect(Moore et al. 1997; Feldmann et al. 1997;Howd~n et al. 1998; Christensen et al. 1998;Bonhomme et al. 1998; Grini et al. 1999). In asecond screen we identify families segregatingsterile plants that show a sporophyticrequirement. Of 3,200 families that werescreened, nearly 40 sterile mutants wereidentified (Vielle-Calzada et al. 1998).Reciprocal outcrosses to wild type showed thatmost of these mutants are male sterile or affectboth sexes, but six were found to be femalespecific.These recessive female sterile mutantsaffect ovule morphogenesis or megasporo­

f..... s....Ii,y 10 Apalllhls: MoIeaoIar .d GHtlk Appr-MS 195genesis and are currently being analyzed inmore detail at the molecular and genetic level(J-P. Vielle-Calzada and U. Grossniklaus,unpublished data).For the rapid isolation and sequencing ofgenomic regions flanking the Os insertion weadapted a PCR-based procedure, TAIL-PCR(Liu et al. 1995), to be used in conjunction withOs elements (see also Tsugeki et al. 1996).Using a set of Os-specific primers(GrossnikJaus et al. 1998a) in combination witharbitrary primers, we isolated at least oneflanking fragment for more than 150 lines thatwe identified in various screens. Using twosets of arbitrary primers, the success rate wasgreater than 95% (GrossnikJaus et al. 1998a).We have identified insertions into a multitudeof genes that encode essential proteinsinvolved in basic metabolic and cellularprocesses, putative regulatory proteins, andseveral novel sequences of unknown function.Based on sequence information, the majorityof the detected genes appear to be involved ingene regulation and signaling processes.Identification of DevelopmentallyRegulated Genes and Their PromotersVery few genes expressed in themegagametophyte have been described(Nadeau et a1.1996; Belostotsky and Meagher1996), and genes expressed in individual cellsof the embryo sac have not previously beenidentified and characterized. Recently, cDNAlibraries obtained from isolated egg cell andin vitro fertilized zygotes of maize have beengenerated, leading to the identification ofgenes expressed in the embryo sac and embryo(Dresselhaus et al. 1994, 1996, 1999a,b).Although no cell type-specific genes have beenisolated yet, this approach holds great promisefor the identification of embryo sac-specificgenes. We use enhancer detector and gene traptransposons carrying the uidA reporter geneencoding b-g1ucuronidase (GUS) Oefferson etal. 1986; Jefferson 1987). The expression of GUScan be visualized by histochemical stainingOefferson et al. 1987; Kavanagh et al. 1988). Toidentify genes expressed during ovuledevelopment and female gametogenesis, weanalyzed GUS expression in maturing ovulesof about 2,300 transposants (R. Baskar, J.Moore, W. Gagliano, U. Grossniklaus,unpublished data). Between 9% and 10% ofthe enhancer trap lines and 2% to 3% of thegene trap lines show spatially restricted GUSexpression in mature ovules. Approximatelyhalf of these enhancer detector lines showexpression restricted to sporophyte andgametophyte, respectively, whereas very fewshow regional expression in bothgametophytic and sporophytic tissues.Although many Arabidopsis promoters havebeen found to be highly compact (e.g., Dwyeret al. 1994; Thoma et al. 1994; Xia et al. 1996),enhancers that drive reporter gene expressioncould be at a considerable distance from thesite of insertion. This makes the isolation ofthe detected gene and its promoter morelaborious, requiring that the expressionpattern of an isolated gene be confirmed. Todate, we have analyzed the expression of threegenes expressed in the megagametophyte; insitu hybridization indicates that they areindeed expressed as expected (Vielle-Calzadaet al. 2000; R. Baskar, J-P. Vielle-Calzada andU. GrossnikJaus, unpublished data). It wouldbe preferable to identify gene trap insertionswithcell type-specific expression because theyhave to be inserted within the transcriptionunit in order to function. However, becausegene traps must integrate within the gene inthe correct orientation (Sundaresan et al. 1995)and GUS activity is often weak, the frequencyat which highly specific expression patternsare recovered is very low. The screeningprocess for cell-type specific expression in theovule and megagametophyte is extremelylaborious, requiring preparations for highresolutionlight microscopy. Therefore, we

196 Ud Gro...ild...concentrate our current screens on enhancertraps because the recovery of highly specificpatterns using gene traps requires screening amany more transposants.About half of the enhancer transposants withGUS activity in the ovule are expressed in themegagametophyte. Some are expressed in allcells of the embryo sac (Figure 12.5, p. 197),whereas in others, GUS expression is sharedby only a subset of cells, for example, the threecells of the egg apparatus. Most importantly,we also identified transposants withexpression in individual cells of themegagametophyte, such as the synergids, theegg cell, and the antipodals. More lines wereexpressed in the synergids than in any othercell type of the embryo sac. This finding isconsistent with earlier reports suggesting thatthe synergid is, metabolically, the most activecell of the megagametophyte Oensen 1974;Russell 1993). The synergid serves importantfunctions in pollen tube guidance, spermdischarge and transport, and fertilizationOensen et al. 1985; Dumas and Mogensen 1993;Russell 1993). Egg cell-specific expression andexpression in the central cell are very rare.Some of the genes expressed in individual celltypes of the female gametophyte may serveimportant regulatory functions during sexualreproduction and could be involved in cellspecification and differentiation processes. Ifthe corresponding genes control importantdevelopmental decisions during megagametogenesis,as suggested by theirexpression pattern, mutant phenotype and/or sequence, they may be useful for theengineering of certain aspects of apomicticreproduction. Most importantly, theregulatory regions of these genes will proveto be invaluable tools for the misexpression ofcandidate genes in particular cell types.Regulatory regions that direct embryo sac-,egg apparatus-, and egg cell-specificexpression have been identified (R. Baskar andU. GrossnikJaus, unpublished data; W. Yang,R. Jefferson, and U. Grossniklaus, unpublisheddata). We intend to use these to probe thepotential of the egg cell for autonomousactivation through misexpression ofcandidategenes such as cell cycle control genes andgrowth regulators. Enhancer detection is themost promising technique for providingnumerous temporally and spatially regulatedpromoters for use in modifying thereproductive system.Introduction of Apomixis intoSexual SpeciesThe introduction of apomixis into sexual cropplants can be achieved through two distinctroutes: (i) the study of the genetic control ofnaturally occurring apomicts could provide uswith the molecular tools necessary tointroduce apomixis into sexual species, or (ii)apomixis could be engineered by synthesizingindividual traits such as nonreduction,parthenogenesis, and normal endospermdevelopment through the introduction ofmutations and/or transgenes controlling theseprocesses. The versatility of the reproductivesystem and the interrelationship betweensexuality and apomixis suggest that this canbe achieved even if naturally occurringapomixis is controlled by an entirely differentand possibly complex mechanism(s).However, knowledge of the genetic andmolec\llar basis controlling apomixis andsexuality does not directly lead to the abilityto manipulate or reconstruct the trait in asexually reproducing plant. The technologicalconstraints to implementation are substantialand must not be underestimated indevelopment of a suitable research strategy.Introgression and Genetic SynthesisThe introduction of apomixis to sexual speciesthrough classical genetic means is particularlyattractive because it does not require priorknowledge of the molecular nature of the geneand can be achieved even if the genomic

From Sexuality la Apomixis: Molerolar and Geneli, Approa'hes 197Figure 12.3 The R-Novojo (R-n,) dominantmaker system for embryo and endosperm.(a) Bath embryo and endosperm carry the R-njmarker and show anthocyanin pigmentation. [b}Only the endosperm carries the R-nj marker while theembryo does not, i.e., is nof pigmented. This kernelwas obtained from a mixed pollination throughheterofertilizatian, but a parthenogenetically formedembryo will also lack the paternal R-nj marker.~ nP "GUSmin"~ Enhancer DetectionArabidopsis gene~ .. .J ~ c:O:---fuOs elementstigmo-specificenhancerGene Trapping((::::;OPSiS genestigma-specificenhancer~Ds elementFigure 12.4 The principle of enhancer detection and gene trapping.la) an inflorescence of an enhancer detector transpasant with reporter geneactivity specifically in stigmatic papillar cells of mature flowers. (b) enhancerdetection relies an a Ds element carrying a reporter gene (GUS) under the (ontrolof a minimal promoter (Pmin)' If the Pmin is influenced by genomic cis-regulatorysequences, e.g., a stigma-specific enhancer, the GUS gene isexpressed in a lissue- ond time-specific manner in the sameway as the gene that is usually controlled by the detectedenhancer. (e) gene traps are a modification of enhancerdetectors relying on splice acceptor (SA) sites in all threereading frames. If the gene trap inserts into a transuiplionsunit frans-splicing can produce a GUS fusion protein Ihatreflects the expression of the detected gene.MiFigure 12.5 Enhancer detector transposant with GUSexpression restricted to the megagametophyte.Abbreviations: Ch, chalaza; Mi, micropyle; ES, embryo sac; 01,outer integument; II, inner integument.

198 lleIG

Fro.. Sexoalny to Aponixls: MoIeatIo, .d G...tl< A,p,....s 199have been conducted in angiosperms and themutants described so far have been identifiedfortuitously. Well-defined sexual modelsystems are best suited for such screens. Ratherthan performing large-scale genetic screens forreproductive mutants in many different cropplants, it may be easier to isolate the relevantgenes from Arabidapsis (or maize) and mimicthe mutant phenotype in crops through geneticengineering.De Novo Engineering ofApomixis through BiotechnologyAs outlined above, the introduction ofapomixis into a wide variety of sexual cropswill be most efficiently achieved throughgenetic engineering. To maximize itsusefulness and versatility in a bioengineeredform apomixis will have to be dominant;otherwise the fixation of hybrid genotypes willbe very slow. The engineering of apomixisthrough biotechnology will require a concertedeffort in three main areas: (i) the identificationand characterization of candidate genes thatcan be used to manipulate the reproductivesystem; (ii) the isolation of promoters that allowa precise control of gene expression at thespatial and temporal levels; and (iii) thedevelopment of efficient technologies tointroduce and control transgenes and/or toperform targeted mutagenesis of endogenousgenes.The identification of regulatory genes that canbe used to control apomixis is being pursuedby a variety of approaches using bothapomictic and sexual systems. Insertionalmutagenesis in Hieracium and Tripsacum(Bicknell, Chap. 8; Grimanelli et aI., Chap. 6)or positional cloning based on mappingapproaches and comparative genomics willlead ~9 .the identification of the componentscontrolling apomixis in natural apomicts. Insexual model systems, the characterization andmolecular isolation of existing reproductivemutants that show individual components ofapomixis is underway and will provide us withnovel tools to manipulate sexuality towardsapomIXIS. These efforts are beingcomplemented by new screens that speCificallytarget relevant aspects of reproduction insexual species. Several laboratories are alsotrying to isolate plant homologs of yeastmutants that could play crucial roles indetermining the meiotic lineage andnonreduction. It must be emphasized that ourunderstanding of the molecular mechanismsthat control plant reproduction are stillextremely limited.In addition to regulators of the sexual pathway,many other genes may be useful for theengineering of apomixis. Such genes includeglucanases that degrade callose, the absenceof which serves as a consistent indicator ofcellsinitiating megagametogenesis. Genespromoting cell wall formation could also beuseful to prevent fertilization of the egg celland promote parthenogenesis. Anotherimportant class is made up ofgenes that controlthe cell cycle. Heterochronic or heterotopicexpression ofsuch regulators could potentiallybe used to trigger cell division and initiatedevelopmental events such asmegagametogenesis and embryogenesis.Recent studies in Arabidopsis have shown thatmisexpression of eye/inlAt in root cells cantrigger extra rounds of cell division (Doerneret a!. 1996) and that eye/inD controls the growthrate irftobacco (Cockcroft et a!. 2000). It will beinteresting to see whether similar experimentscan induce cell proliferation in the egg cell.Other growth regulators, such as planthormones, have not been studied in detailduring sexual reproduction, but may playcrucial roles in initiating developmentalprograms relevant to apomixis.Targeted misexpression ofcandidate genes willrequire well-defined regulatory sequences thatcan be used to drive transgene expression. Todate, only a few promoters have been

200 UtlG09".ila,ndescribed that are active in the ovule, andpromoters specific to the gametophyte andits constituent~ells have just been isolated (R.Baskar and U. Grossniklaus, unpublisheddata). Enhancer detection bears greatpotential for the identification of genes thatplay crucial roles during sexual reproduction,and also because it serves as an entry point toisolate a multitude of highly specificpromoters that are restricted to specific celltypes and regions of the ovule and femalegametophyte.Although the introduction of transgenes isnow readily achieved in many crop species,present technology only allows for theintroduction of dominant traits. Transgeneactivity is based either on overexpression oron homology dependent gene silencingphenomena, such as antisense suppression orsense cosuppression (e.g., Matzke and Matzke1995; Jorgensen 1995; Jorgensen et al. 1996).By virtue of the epigenetic nature of thesephenomena, it may prove difficult toeffiCiently and stably introduce the traits thatcontrol apomixis. Currently, it is not routinelypossible to disruptendogenous genes, and thelack of efficient homologous recombinationtechniques, which would allow us to mutateor otherwise modify endogenous genes,remains a serious obstacle to geneticengineering in plants. However, recentreports on successful homologousrecombination in Arabidopsis (Miao and Lam1995; Kempin et al. 1997) may soon lead tothe development of more efficient protocolsfor targeted gene disruption. In Arabidopsisand maize, gene disruption by site-specifictransposon mutagenesis has become anefficient way to inactivate specific genes.Although this approach has been adapted torice (Izawa et al. 1997), it is unlikely that itwill be implemented in a wide variety of othercrop species.Field-level Regulation of Apomictic TraitsTo maximize benefits, it will be necessary tocontrol the expression of apomixis such that achoice can be made between the sexual andapomictic modes of reproduction at any stageof a breeding program. Apomixis as aconstitutive trait could potentially pose a threatto genetic diversity, which would preclude theuse of apomixis for crop improvement in aversatile and creative way. For instance,apomixis would constitute the defaultcondition, wherein application of anexogenous condition or compound wouldsuppress apomixis to allow for sexual breedingto introgress new germplasm and createsegregating populations. Alternatively,sexuality could be the default condition,addressing the concern that apomixis couldpose a threat to biodiversity (van Dijk and vanDamme 2000), and apomixis would only beinduced at specific steps of the breedingprogram and for seed production.Although several inducible systems have beendescribed and shown to work efficiently underlaboratory conditions, none of these systemswould allow the induction or suppression ofa trait under field conditions. These systemsallow either repression or induction of a geneupon addition of a compound. All of theexisting inducible (suppressible) systems,including the tetracycline repressor (Gatz et al.1992; Weimann et al. 1994), the copperinducible system of the yeast metallothioningene (Mett et al. 1993; Mett et al. 1996), andthe glucocorticoid receptor (Schena et al. 1991;Aoyama and Chua 1997) have seriousdrawbacks for field use, although they arepotent systems for use in the laboratory andgreenhouse. These systems often have a highbackground of noninduced expression and useexpensive and/or environmentallyunacceptable inducers that often have poormobility in the plant.

from S...a1ily 10 Apomb

202 UeIGfo...ll...Two additional points should be stressed. First,the technologies needed to engineer andcontrol apomixis under field conditions are notyet available and must first be developed, apoint that cannot be overstated as theidentification of potential regulatory genesprogresses rapidly. Second, this review of plantreproduction has focused on a geneticperspective with Mendelian traits controllingsexual and apomictic development. However,gene regulation by genomic imprinting maynot be the only epigenetic mechanism thatshould be considered. Current knowledge ofthe regulation of apomixis is fully compatiblewith an epigenetic view of this trait and weshould keep our minds open to alternativeexplanations that are epigenetic in nature. Ifapomixis is, for instance, due to epigenetic.in teractions between genomes, theengineering of apomixis in sexual species willbe much more complex.Finally, the accessibility of apomiXIStechnology to a broad community of plantbreeders in the public and private sectorworldwide must be ensured. If universal andequitable access to apomixis technologycannot be achieved through innovativepatenting and licensing, then the e.xcitingscience discussed in this volume will have littlepositive impact (http://billie.btny.purdue.edu/apomixis/).AcknowledgmentsMy special thanks go to Richard Jefferson forhis valuable contributions and comments, andto Gian Nogler for proofreading themanuscript. I would also like to thank Jean­Philippe Vielle-Calzada, Robert Pruitt, andRoss Bicknell for their stimulating discussions,and Alison Coluccio for help with thebibliography. SpeCial thanks go to RobertPruitt, Susan Lolle, Richard Jefferson, WeiYang, Satya Nugroho, lmran Sidiqqi, BrunoTinland, and the members of my laboratoryfor allowing me to cite data prior topublication. I am grateful for the supportprovided by the European Molecular BiologyOrganization, the Human Frontiers of ScienceProgram, the Janggen-Poehn-Foundation, andthe Demerec-Kaufmann-Hollaender­Fellowship in Developmental Genetics.Research in my lahoratory was funded by theCold Spring Harbor Laboratory President'sCouncil, Pioneer Hi-Bred International, grantMCB-9723948 of the National ScienceFoundation, and grant # 9801207 of theNational Research Initiative CompetitiveGrants Program of the U.5.D.A.ReferencesAhokos, H. 1977. Amutant of barley: triploidinducer. Barley Genet. Newsl. 7: 4-6.Allemon, M., ond, J. Dooor. 2000. Genomicimprinting in planls: observations andevolutionary implications. Plant Mal. Bial43: 147-61.Angenent, G.c., and LColombo. 1996.Molecular control of ovule development.Trends Plant Sci 1: 228-32.Antoine, A.F., J·E. Faure, S. Cordeiro, C. Dumas,M. Rougier, and J.A. Feija. 2000. Acalciuminflux is triggered and propagates in thezygote as awavefront during in vitrofertilization of flowering ~ants. Proc. Natl.Acad. Sci. USA 97: 10643-48.Aoyama, 1, and N·H. Chua. 1997. Aglucocorticaid·mediated lranscriptionalinduction system in transgenic plants. PlantJ.ll:60S-12.Arabidopsis Genome Initiative 2000. Analysis ofthe genome sequence of the floweringplant Arabidapsis thaliana. Nature 408:796-8lS.Asker, S.E. 1979. Progress in opomixis research.Herooitas91: 231-40.--.1980. Gametophytic apomixis:Elements ond genetic regulation. Hereditas93: 277-93.Asker, S.E., A. Hagberg, and G. Hagberg. 1983.Apomixis in barley? Sver. UtsiidesFiiren.Tidskr. 93: 7s-76.Asker, S.E., and LJerling. 1992. Apomixis inPlants. Boca Raton, Florida: CRC Press.Baker, S.c., K. Robinson·Beers, J.M. Villanueva,J.c. Gaiser, and C.S. Gasser. 1997.Interactions among genes regulating ovuledevelopment in Arabidopsis tholiono.Genetics 14S: 1109-24.Ballinger, D.G., and S. Benzer. 1989. Targetedgene mutations in Drosophila. Proc. Natl.Acad. Sci. (USA) 86: 9402-ll6.Barcacoa, G., A. Manucato, E. Albertini, J.lethaf, A. Gerats, M. Penotti, M. Falcinelli.1998. Inheritance of parthenogenesis inPoa pratensis L: auxin test and AFLPlinkage ana~ses support monogeniccontrol. Theoretical and Applied Genetics97: 74-82.8ashaw, E.C., and W.W. Hanna. 1990. Apomicticreproduction.. In G.P. Chapman led.},Reproductive Versatility in the Grosses.Cambridge, U.K.: Cambridge UniversityPress. Pp. 10~30.8echtold, N., J. Ellis, and G. Pelletier. 1993. Inplanta Agrobaeterium gene transfer byinfiltration 01 adult Arabidapsis thalianaplants. CR. Acad. Sci. Paris. 316: 1194-99.

F",," s.....Uy I. Apomixis: MoIeuIor ..dGHelk Ap,.-M, 203Bedinger, P., and S.D. Russell. 1994.Gametogenesis in maize. In M. Freeling andV. Walbot (eds.), The Maize Handbook. NewYork: Springer-Verlag. Pp. 48-61.Belanger K.D., and R.s. Quatrano. 2000.Polarity: the role of localized secretion. Curr.Gp. Plant Bioi. 3: 67-72.Bellen, HJ., CJ. O'Kane, C. Wibon, U.Grossniklaus, R_K_ Pearson, and WJ.Gehring. 1989. P-element-mediatedenhancer deteelion: aversatile method 10study development in Drosophila. Genes andDev. 3: 1288-1300.Bellen, HJ., C. Wilson, and WJ. Gehring. 1990.Dissecting the complexity of the nervouss~tem by enhancer detection. Bioessays 10:199-204.Belostotsky, D.A., and R.B. Meagher. 1996. Apollen-, ovule-, and ear~ embryo-specificpo~ (A) binding pratein from Arabidopsiscomplements essential functions in yeast.Plant Ce1/8: 1261-75.Bennelzen, lL, and M. Freeling. 1993. Grassesas asingle genetic system: genomecomposition, colinearity ond compatibility.Trends in Genetics 9: 259-61.Bensen, RJ., G.S. Jahal, V.c. Crane, J.T. Tossberg,P.S. Schnable, R.B. Meeley, and S.P. Briggs.1995. Ooning and characterization of themaize AnI gene. Plant Ce1/7: 75-84.Berger, F. 1999. Endosperm development. Curr.Op. Plant Bioi. 2: 28-32.Bicknell, R.A. 1997. Isolation of adiploid,apomictic plant of Hierodum aurontiacum.Sex. Plant Reproo. 10: 168-72.Bicknell, R.A., N.K. Barsl, and A.M. Kahunow.2000. Monogenic inheritalKe of opomixis inIwo Hieradum species wilh dislinctdevelopmental me

204 Uel G.....ilcmDiboll, A.G., and DA larson. 1966. An electronmicrographic study of the maturemegagametophyte in lea mays. Amef. 1.Bot. 53: 391-402.Dickinson, H.G. 1992. Plant reproduction: Pasr,present and future. In Y. Dottee, e. Dumos,and A. Gallais leds.), Reproductive Biologyand Plant Breeding. Berlin: Springer-Verlag.Digonnet, e., D. A~on, N. leduc, e. Dumas, andM. Rougier. 1997. First evidence of acolcium transient in flowering plonls atfertilization. Development 124: 2B67-74.Doerner, P., J·E. Jergensen, R. You, l Sreppuhn,and e. lomb. 1996. Conlrol of rool growthand development by cydin expression.Nature 3BO: 520--21Doutriaux, M.P., F. Couleau, e. Bergounioux,and e. White. 1998. Isolation andcharocterisolion of the RADS1 and DMCIhomologs from Arabidapsis Ihaliana. Mol.Gen. Genet. 257: 283-291.Dresselhous, 1, Cordts, S., Heuer, S., Souler M.,H. LOrz, and E. Kranz. 1999b. Novelribosomal genes from maize aredifferentiol~ expressed in the zygotic andsomatic cell cydes. Mal. Gen. Genel. 261:416-27.Dresselhous, l, Cordts, and H. lorz 19990. Atranscript encoding translation initiolionfactor elF-SA is stored in the unfertilisedegg cell of maize. Plant Mol. Bioi. 39:1063-71.Dresselhous, l, e. Hagel., H. LOrz, and E. Kranz.1996. Isolation of a full· length cDNAencoding calretirulin from a PCR·librory ofin vitro zygotes of maize. Plant Mal Bioi31: 23-34.Dresselhous, l, H.lolZ, and E. Kranz. 1994.Representative cDNA libraires from fewplant cells. Plant J5: 605-10.Drews G.N., D. lee, and e.A. Christensen. 1998.Genetic ona~is 01 femole gametophytedevelopment and function. Plant (ell 10:5-18.Dwyer, K.G., M.K. Kandasamy, 0.1. Mahosky, lAniai, B.I. Kudish, J.E. Mille, M. Nasrallah,and J.B. NOllallah. 1994. Asuperfomi~ ofS·lom related sequences in Arabidapsis:Diverse structures and expression patterns.Plant (eI16: 1829-41DUjardin, M., and W.W. Hanna. 19B9.Developing apomictic pearl millet:characterization of a B~ plant. 1. Genet.Breed. 43: 145-51.Dumos, e., and J-E. Faure 1995. Use of in vitrofertilization and zygote (U~ure in cropimprovement. (urf. Opinion Biotech. 6:183-88.Dumos, C, and H.L Mogensen. 1993. Gametesand fertilization: maize as a model systemfor experimental embryogenesis inflowering plants. Planl (ell 5: 1337--48.Eames, AJ. 1961. Morphology ofAngiosperms.New York: McGraw· Hill.Ehlenfeldl, M.K., and R. Ortiz. 1995. Evidence onrhe nature and origin of endosperm dosagerequirements in Solanum and otherangiosperm genera. Sex. P1anl Reprod. 8:189-96.Biotl, R.c.. AS Betzner, E. Huttner., M.P. Oakes,W.QJ. Tucker, D. Gerentes, P. Perez, andD.R. Smyth. 1996. AINTEGUMENTA, anAPElALA2-like gene of Arabidopsis withpleiotropic roles in ovule development andfloral organ growth. Plant (ell. 8: 155-68.Ernst, A. 1918. Die Bastordierung als Ursache derApogamie im PIIonzenreiche. Jena,Germany: Fischer.Faure, lE., e. Digonnet, and e. Dumas. 1992. Anin vitra system for adhesion and Mion ofmaize gametes. Science 263: 1598--1600.Feldmann, KA, D. Coury, and M.l. Christianson.1997. Exceptional segregation of aselectable marker (Kan!) identifies genesimportant for gomelophytic growth anddevelopment. Genetics 147: 1411-22.Feldmonn, K.A., R.L Malmberg, and CDean.1994. Mutagenesis in Arabidopsis.. In E.M.Meyerowitz and CR. Somerville (eds.),Arobidopsis. Cold Spring Harbor, USA: ColdSpring Harbor laboratory Press. Pp. 137­72.Finch, RA, and M.D. Bennett. 1979. Action oftriploid inducer(tn) on meiosis in barley(Hordeum vulgare l.l. Heredity. 43: 87-93.Fobert, P.R., B.L Miki, and V.N.lyer. 1991.Detection of gene regulatory signals inplanrs revealed by T-DNA·medioted fusions.Plant Mal. Bioi 17: 837-51.Franke, R. 1975. Uber das Aulreten vonunreduzierlen Gameten bei Angiospermen.Archiv fUr liichtungsforschung 5: 201-j)8.Freeling, M., and V. Walbol. 1994. The MoizeHandbook. New York: Springer Verlag.Friedman, W.E. 1990. Double ferlilization inEphedra, a non-flowering seed plant: itsbearing on Ihe origin of angiosperms.Science 247: 951-54.--. 1992. Evidence of a pre·ongiospermorigin of endosperm: implication for theevolulion of Rowering plants. Science 255:336-39.Friedrich, G., and P. Soriano. 1991. Promotertraps in embryonic stem cells: a geneticscreen to identify and mutatedevelopmental genes in mice. Genes andDev 5: 1513-23.Frost, H.B., and R.K. Soost. 1968. Seedreproduction development of gametes andembryos. In WReuther, lD. Botchelor, andHJ. Webber (eds.), The Citrus Industry, VolII. Berkeley, California: University ofCalifornia Press. Pp. 290--324Gadella, lWJ. 1987. Sexual tetraploid andapomictic pentaploid populations ofHieradum pilosella(Compositoe). PlantSyst. Evo1157: 219-45.Gaiser, le., K.Robinson-Beers, and e.s. Gosser.1995. The Arabidapsis SUPERMAN genemediates asymmetric growth of the outerintegument of ovules. Plant (ell 7: 333­45.Gale. M.D., and K.M. Devos. 1998. Camparotivegenetics in the grasses. PrO(. Not!. Acod. SciUSA 95: 1971-74.Gasser, CS., l Broodhvest, and B.A. Hauser.1998. Genetic ona~is of ovuledevelopment. Ann. Rev. Plant Physiol. PlontMal. Bioi. 49: 1-24.Gotz, C1998. An intoxicating swilth for planttransgene expression. Nature Biotech. 16:140.Gotz. C, e. Frohberg, and R. Wendenburg.1992. Stringent repression andhomogeneous de·repression by tetra·cydine of a modified CaMV 35S promoterin intact transgenic tobocco plants. Plant J.2: 397-404.Gotz, C, and I. lenk. 1998. Promoters thotrespond to chemicol inducers. Trends PlantSci 3: 352-58.Goodrich l 1998. Plant development: Medea'smalernol instinct. (Uff. BioI. 8: 480--84.Golubovskaya, LN. 1979. Genetic contlol ofmeiosis. Int. Rev. (yfal. 58: 247-90.Golubovskaya, I.N., N.A. Avalkin, and W.F.Sheridan. 1992. EHect of several meioticmutants on femole meiosis in moize. De vI.Genet. 13:411-24.--.1997. New insights into the role ofthe moize ameiafic 1locus. Genetics 147:1339-SO.Golubovskayo, LN., Z.K. Grebennikovo, N.A.Avolkina, and W.F. Sheridan. 1993. Therole of the ameiotic 1gene in the initiationof meiosis and in subsequent meioticevents in maize. Genetics 135: 1151--66.Gossler, A., A.L Joyner, J. Rossant, and we.Skornes. 1989. Mouse embryonic stemcells and reporter conslructslo detectdevelopmentally regulated genes. Science244: 463-65.Grimonelli, D., M. Hernlindez, E. Peratti, and Y.Savidon. 1997. Dosage effects in theendosperm of diplosporous apomicticTripsacum (Pooceae). Sex. Plant Reprod.10: 279-82.Grimonelli, D., O.leblanc, E. Espinosa, E.Perotti, D. Gonzalez de lean, and Y.Sovidan. 1998. Mapping diplosporousapomixis in tetraploid Tripsacum: one gene01 several genes? Heredity 80: 40-47.

fro.. Smarrty to Apoml.ds: Molecolor ood Gn.tk A,,.-Ms 205Grimanelli, D., O. Leblanc, D. Gonzales de leon,and Y. Savidan. 1995. Apomixis expressionin maize- Tripsacum hybrid derivatives andthe implications regarding its control andpotential for manipulation. ApomixisNewsletter 8: 35-37.Grini, P. E., A. S

206 UeRG'....ild...--.1974. Reproduction in floweringplants. In: A.W. Robards (ed.), DynamicAspect5 of Plant U#rostrudure. New York:McGrow-Hill. Pp. 481-503.Jensen, W.A., M.E. Ashton, and c.A. ~easley.1985. Pollen tube-embryo soc interaction inca"an. In D.L Mulcahy and E.O"aviana(eds.), Pollen: Biology and Implicotions forPlant Breeding. New York: EkevierBiomedical. Pp. 67-72.Jahnslan, SA, T.P.M. den Ni~, SJ. Peloquin, andR.E. Hanneman Jr. 1980. lhe significance ofgenic balance to endosperm development ininte~pecific crosses. Thear. Appl. Genet. 57:>--9.Johnston, SA, and R.E. Hannemon. 1982.Manipulutions of endosperm balancenumber overcome crossing barriers betweendiploid Solanum species. Science 247: 446­48.Johri, B.M., K.B. Ambegaakar, and P.S.Srinivasta. 1992. (omparotive Embryologyof Angiosperms. Vois. 1and 2. New York:Springer-Verlag.Jones, TJ., and 1.L Rosl. 1989. Histochemistryand ultrastructure of rice (Oryza sativa)zygotic embryogenesis. Amer. J. Bot. 76:504-20.Jongedijk, E. 1985. The po«ern ofmegasporogenesis and megagamelogenesisin diploid Solanum spedes hybrids: ilsrelevance to the origin of 2n-eggs and theinduction of apomixis. Euphytica 34: 599­611.Jorgensen, tA. 1995. Co-suppression, flowercolor potlerns, and metastable geneexpression states. Science 268: 686-91.Jorgensen, RA., P.D. auster, J. English, Q. Que,and CA. Nopoli. 1996. Chalcone synthasecosuppression phenotypes in peluniaflowers: comparison of sense vs. antisenseconstructs and single-copy VI. cocnplex l­DNA sequences. Plant Mol. Bioi. 31: 957­73.Kossir, Y., D. Granol, and G. Simchen. 1988.IMEl, a positive regulator gene of meiosisin 5. cerevisioe. (ell 52: 853--62.Kaul, M.LH., and 1.G.K. Murthy. 1985. Mutanlgenes affecting higher plant meiosis. Thear.Appl. Genet. 70: 449--{)6.Kovanagh, TA., R.A. Jeffe~n, and M.W. Bevan.1988. Targeting a foreign protein tochloroplasts using fusions 10 the Iransitpeptide of a chlorophyll alb protein_ Mo/.Gen. Genef. 215: 38-45.Kowaguchi, H., M. Yoshida, and I. Yornoshija.1992. Nutrilional regulation of meiosisspecificgene expression in Saccharomycescerevisiae. Biotech. Biochem. 56: 289-97.Keller, B., and C. Feuillet. 2000. Calineority andgene density in gross genornes. Trends PlantSci. 5: 246-51.Kempin, SA., SJ. Liljegren, LM. Block, S.D.Raunsley, MJ Yanofsky, and E. Lom. 1997.Targeted disruption in Arabidopsis. Nature.389: 802-113.Kerbundit, 5., H. DeGreve, F. Deboeck, M. vonMantagu, and J.-P. Hernalsteens 1991. Invivo random b-glucuronidase gene fusion inArobidopsis thol;ana. Proe. Natl. Acad. Sci.(USA) 88: 5212-16.Kermicle, J. 1969. Androgenesis conditioned bya mutation in maize. Scienre 116: 1422­24.--. 1970. Dependence of Ihe R-motlledaleurone phenotype in maize on the madeof sexual transmission. Genetics 66: 69­85.--.1971. Pleiotropic effects on seeddevelopment of the indeterminategametophyte gene in maize. Amer. 1. Bot.58: 1-7.--. '994. Indeterminate gametophyte(ig): biology and use. In M. Freeling and V.Walbol teds.), The Maize Handbook. NewYork: Springer-Verlag. Pp.388-93.Kermicle, J., and M. Alleman 1990. Gameticimprinting in maize in relation to theangiosperm life cycle. DevelopmentSuppl.1:9-14.Kiesselbach, TA 1949. The structure andreproduction of corn. Univ. Nebraska (011.Agric. Exp. Station Res. Bull. 161. 50 thanniversiary edition. Cold Spring Horbar(USA): Cold Spring Harbor laboratoryPress.Kilian, A., DA. Kudra, A. Kleinhofs, M. Yano, N.Kurato, B. Steffenson, and 1. Sasaki. 1995.Rice-barley synteny and ils application tosaturation mapping of the barley Rpg1region. Nucleic Acids Res. 23: 2729-33.Kimber, G., and R. Riley. 1963. Haploidangiosperms. Bot. Rev. 29: 480-531.Kindiger, B.K., B. Dapeng, and V_ Sakalav. 1996.Cytological and malecular studies ofapomictic maize- Tripsorum hybrids.Agronomy Abstroct5. pp. 132.Kinoshita, 1., R. Yadegari, JJ. Harada, R.B.Goldberg, and R.l. fIScher. 1999.Imprinting of the MEOEA Polycomb gene inthe Arabidopsis endosperm. Plant (ell 11 :1945-52.Kiyosue, 1., N. Ohad, R. Yadegari, M. Hannon, J.Dinneny, D. Wells, A. Katz, LMargossian,JJ. Harada, R.B. Goldberg, and R.L Fischer.1999. Control of fertilization-independentendosperm development by the MEOEAPo!ycomb group gene in Arobidopsis. Proc.Natl. Acad. 56. USA 96: 4185-91.Klapholz, S., and R.E. Esposito. 1980. Isolation ofspo 12-1 and spo' 3-1 from a naturalvariant of yeast that undergoes asinglemeiotic division. Genetics 96: 567...,ll8.Klimyuk, V.I., and J.D.G. Jones. 1997. AIDMC1,the Arabidopsis homologue of the yeastDMCl gene: characterization, transposoninducedallelic variation and meiosisassociated expression. Plant Journal. 11: 1­14.Klucher, K.M., H. Chow, l. Reiser, and R.L fIScher.1996. The AINTEGUMENTA gene ofArobidapsis required for ovule and femalegametophyte development is related to thefloral homeotic gene APfTAlA2. Plant (ell. 8:137-53.Knox, R.B. 1967. Apomix~: seosonal andpopulation differences in agrass. 5cience157: 32>--26.Knox, R.8., and J. Heslop-Harrison. 1963.Experimental control of oposporousapomixis in 0 gross of the Andropogoneae.Bot. Not. 116: 127-41.Kobayoshi, 1, Y. Ho"a, and S. Tabato. 1993.Isolation and characterization of ayeastgene that is homologous with a meiosisspecificcDNA from a plan!. Mol. Gen. Genet.237: 22>--32.Kobayashi, 1., E. Kobayoshi, S. Sato, Y. Ho"a, N.Miyajima, A. Tanaka, and S. Tabato. 1994.Characterization of cDNAs induced in meioticprophose in lily microsparocyfes. DNA Res.1: 15-26.Kobayoshi, 1, ond K. Tsunewoki. 1978_ Hoploidinduction and its genetic mechonism inalloplasmic common wheat. 1. Hered. 71: 9­14.Kojima, A., and Y. Nagoto. 1997. Discovery ofhighly apomictic ond highly omphimicticdihaploids in Allium tuberosum Sex. PlontReprod. 10: 8---12.Koltunow, A.M. 1993. Apomixis: embryo sacs andembryos formed without meiosis orfertilization in ovules. Plant (ell 5: 1425­37.Koltunow, A.M_, R.A. Bicknell, and A.M.Choudhury. 1995. Apomixis: Molecularstrotegies for the generotian of geneticallyidentical seeds without fertilization. PlantPhysiol. 108: 134>--52.Koltunow, A.M., 1 Hidoka, and S.P. Robinson.1996. Polyembryony in citrus. Accumulationof seed storage proteins in seeds and inembryos cuhured in vitro. Plant Physio/.11 0: 599-609.Kranz, E., and 1. Dresselhaus. 1996. In vilrofertilization with isolated higher plontgametes. Trends Plant Sci. 1: 82-89.Kranz, E., P. von Wiegen, and H. lOrz. 1995.Early cytological events aher induction of celldivision in egg cells and zygote developmentfollowing in vitro fertilization withangiosperm gometes. Plant 1. 8: 9-23.Krysan, PJ., J. C. Young, and M. R. Sussman.1999. T-DNA os on insertional mutagen inArabidopsis. Plant (ell 11: 2283--90.

from S...oIily '0 Apomixis: MoIeaolar lIIId G...li< Appr_hes 207Lang, lD., S. Roy, and A. Roy. 1994. sinl, amulatian affecting lemale lertility inArobidopsis, interacts with modI, itsrecessive modifier. Genetics. 137: 11Dl-1D.leblanc D., M.D. Peel, lG. Corman, and Y.Sovidan. 1993. Megasporogenesis in sexualand apomictic Tripsocum species usinginterlerence contrast and floorescence.Apomixis Newsletter 6: 14--17.--.1995. Megasporogenesis andmegagametogenesis in several Tripsocumspecies (Poaceae). Amer. 1. Botony. 82: S7­63.leon·Klasterziel, K.M., C. l Keijzer, and M.Koornneel. 1994. Aseed shope mulant inArobidopsis that is affected in integumentdevelopment. Plont Cell. 6: 385-92,lin, 8.·Y. 1978. Structural modilications olthelemale gametophyle asso

208 Uel GtO,,"ikIa..MIodzik, M., Y. HirOlri, U. Weber, C.S. Goodman,and G.M. Rubin. \990. The Drosophilaseven-up gene, amember of the steroidreceptor superfamily conlrab photoreceptorcell fates. Ce1/60: 1237-49.Modruson z., L. Reiser, K.A. Feldmann, R.LF!sdler, ond G.W. Haughn. 1994. Hameotictransformation of ovules into carpel.likestructures in Arahidapsis. Plant (ell. 6:333-49.Mogensen, H.L 1988. Exclusion of malemitochondria and plastids during syngamyin barley as a basis for maternalinheritance. Proe. Natl. kad. Sa. (USA) 85:2594-97.Mogie, M. 1992. The evolution of asexualreproduction in plants. london: Chapmanand Hall.Montamoyor, J.c., D. Vezon, C. 8ajon, A.5ouvanet, O. Grandieon, M. Marchand, N.Bechtold, G. Pelletier, and C. Horlow. 2000.Switch I (swill, on Arahit/opsismutantaffected in the female meiotic switch. Sex.Planl Reprod. 12: 209-18.Moore, G., K.M. Devos, Z. Wong, and M.D. Gale.1995. Grosses: line·up and form acircle.(urrent Bioi. 5: 73S-42.Moore, J., J.p' Yl8l1e Calzoda, W. Gagliana, andU. Grossniklaus. 1997. Geneticcharacterization of hadad, a mutantdisrupting female gametogenesis inArahit/opsis Iholiana. Cold Spring HarborSymp. Quanl. Bioi. 62: 35-47.Mora·Garcia, S., and J. Goodrkh. 2000. Seeds ofconflict. (urr Bioi. 10: 71-74.Mordhorst, A.P., MAJ. Toonen, and S.c. de Vries.1997. Plant embryogenesis. (ril. Rev. PI.Sci. 16: 535--76.Morgan, R.N., P. Ozios·Akins, and W.W. Hanna.1998. Seed set in on apomictic B~ perlmillet. Inl. J. Planl Sci. 159: 89~7.Murroy, A.W. 1992. Creative blocks: Cell cyclecheckpoints and feedback cantrob. Nature359: 599-604.Murgia, M., 8.·a. Huang, S.c. Tucker, and M.E.Musgrove. 1993. Embryo soc lockinganlipodal celb in Arahidopsis thalianaIBrassicaceoe). Amer. 1. Bal. BO: 824-38.Nadeau,J.A., X.S. Zhong, 1 Li, and S.D. O'Neill.1996. Ovule develollffient: Identification ofstage specific and tissue specific cDNAs.Plont Cell 8: 213-39.Nouber, U., MJ. Pankratz, A. Kienlin, E. Seifert,U. K1envn, and H. Jiickle. 19B8. Abdominalsegmentation of the Drosophila embryorequires a hormane receptor·like proteinencoded by the gop gene knirps. Nature336: 489-92.Naumavo, IN. 1993. Apomixis in Angiosperms:Nucel/ar and Inlegumentary Emhryony.Boca Raton, Florida: CRC Press.Naumava, ltt, A.P.M. Den NilS, and M.IM.Willemse. 1993. Quantitative analysis ofaposporous parthenogenesis in Poapralensis genotypes. Ada Bol. Neerl. 42:299-12.Naumava, IN., and M.T.M. Willemse. 1995.Ultrastructural characterization of aposporyin Panicum maximum. Sexual Plant Reprod.8: 197-204.Nelson, O.E., and G.B. Gory. 1952. Genic controlof semi·sterility in maize. 1. Hered. 43:205--10Neuffer, M.G., and W.F. Sheridan. 1980.Defective kernel mUlants of maize. I.Genelic and letholity studies. Genetics 95:929-44Nishiyama, I., and I Yabuno. 1978. Causalrelationships between the polar nuclei indouble fertilization and inte~pecific cross·incompatibility in Aveno. (ylologia (Tokyo)43: 4S3-66.Nogler, G.A. 1972. Genetik der Aposporie beiRanunculus ouricomus. II.Endospermzyfologie. Ber. Schweiz. Bot. Ges.82: 54-03.--. 1973. Genetik der Aposporie beiRanunculus aurkamus.1I1. FfRlkkkreuzungsbaslarde. Ber. Schweiz. Bol.Ges. 83: 295--305.--. 1975. Genetics of apospory inRanunculus auricomus. IV. Embryology of F3and F 4backcross offspring.Phylomorphology 25: 485--90.--.1982. How to obtain diploid apomicticIlanunculus ouricomus plonts not found inthe wild state. Bot. Helv. 92: 13-22.--. 19840. Gametaphytic apomixis. InB.M. Johri led.), &nhryologyofAngiosperms. Berlin: Springer·Veriag. Pp.475·518.--.1984b. Genetics of apospory inapomictic Ilanunculus auricomus. V.Conclusion. Bol. Helv. 94: 411-22.Noirot, M. 1993. Allelic ratios and sterility in theagamic complex of the Maximeo{(Panicoideae): evolutionary role of theresidual sexuality. 1. Evol. BioI. 6: 95--101.Noyes, R.D., and L. H. Rieseberg. 2000. Twoindependent loci control agamospermylapomixis) in the triploid flowering plantErigeron annuus. Genelics 155: 379-90.Nucc~el~, R. 1991. How do sperm activate eggs?(urr. Topics Devl. Bioi. 25:1-16.Nygren, A. 1967. Apomixis in the angiosperms.Handh. der PFlanzenphys. 18: 551-96.Ohad, N., LMargossian, Y.·c. Hsyu, e. Williams,P. Repehi, and R.L Fischer. 1996. Amutation that allows endospermdeveloprnentwithaut fertilization. Proe.Natl. Acad. xi. (USA) 93: 5319-24.Ohod, N., R. Yadegari, LMorgossian, M. Hannon,D. Michoeli, JJ. Haroda, R.B. Goldberg, andR.L Fischer. 1999. Mutations in FIE, aWDpolycomb group gene, allow endospermdevelopment withoUl fertilization. Planl (elf11: 407-16.O'Kane, c., and WJ. Gehring. 1987. Detection insitu 01 genomic regulatory elements inDrosophila. Proe. Nafl. Acad. xi. (USA) 84:9123-27.Obon, A.R., and D.O. Coss. 1981. Changes inmegagametophyte structure in Popovernudieoule following in vitro placentalpollinotion. Amer. J. Bol. 68: 1338-41.Oben, O·A., C. Linneslad, and S.E. Nichols. 1999.Developmental biology of the cerealendosperm. Trends Planf Sci. 4: 253-57.Oro, A.E., E.S. Ong, J.5. Margolis, J.W. Posokony,M. McKeown, and R.M. Evans. 1988. TheDrosophila gene knirps·reloted is amemberof Ihe steroid·receptor gene supenamily.Nature 336: 493-96.Ozias·Akins, P., E.L. lubbe~, W.w. Hanna, andlW. McNay. 1993. Tronsmission of theapomictic mode of reproduction inPennisefum: Co-dominance of the trait andmoleculor markers. Thear. Appl. Genet. 85:632-38.Ozias·Akins, P., D. Roche, and W.W. Hanna. 1998.Tight clustering and hemizygosity ofapomixis·linked molecular markers inPennisefum squamulatum implies geneticcontrol of apospory by a divergentlDC1Jsthot moy hove no allelic form in sexualgenotypes. Proe. Nafl. kad. Sci. (USA) 95:5127-32.Palmer, R.G. 1971. Cytological studies of ameioticand normal maize with reference topremeiotic pairing. Chromasoma. 35: 233­46.Parinov, S., S. Moyolagu, D. Ye, W·e. Yang, M.Kumaron, and V. Sundaresan. 1999.Ana~is of flanking sequences fromDissociation insertion lines: Adatabase forreverse genetics in Arahidopsis. Planl (elf11: 2263-70.Paterson, A.H., 1,·H. lon, IlP. Reischmann, e.Chong, Y.·R. Lin, S.,c. Liu, M.D. 8urow, S.P.Kowalski, e.S. Katsor, T.A. DelMonte, K.A.Feldmann, K.F. Schertz, and J.F. Wendel.1996. Toward a unified genetic map ofhigher plants, transcending the monocot·dicot divergence. Nature Gene/ics 14: 380­82.Paherson, E.B. 1978. Properties and uses ofduplicate deficienl chromosome complimentsin maize. In D. Waldon led.!, MoizeBreeding and Genefics. New York: JohnWiley and Sons. Pp. 693-71 O.Peacock, l.W. 1992. Genetic engineering andmutagenesis for apomixis in rice. ApomixisNewsletter 4: 3-7.

From Sexotality 10 Apomixis: MoIe

21 0 Ud Gromild.1S--.1982. Nature et hered~e deI'apomixie chez Panicum maximum Jacq.Travaux et Documents ORSTOM, Paris1S3:1-159.---. 1992. Pragre5s in re5e0rch onapomixis and its transler to major groincrops. In Y. Donee, CDumos, and A. Gallais(eds.), Reprodudive Biology and PlantBreeding. Berlin: Springer-Verlag.---. 2000. Apomixis: Genetics andbreeding. Plant Breeding Reviews 18: 13­86.Sovidan, Y., and J Berthoud 1994. Maize xTripsacum hybridization ond the potentiallor apomixis tronsler lor maizeimprovement. In Y.P.S. Baiai led.),Biotechnology in Agriculture and Forestry,Vol. 25 Maize_ Berlin: Springer·Veriag. Pp_69~3.Sovidan, Y., and M. Duiardin. 1992. Apamixie: 10prochoine revolution verte? to Recherche241: 326-34.Schena, M., A.M. Uoyd, and RW. Davis. 1991. Asteroid-inducible gene expre5sion systemlor plant celk. Proc. Natl. Acad. Sci. (USA)88: 10421-25.Schiehhaler, U., S. Balosubromonion, PSieber,D. Chevalier, E. Wismon, and K. Schneitz.1999. Molecular analysis 01 NOZZLE, agene involved in panern formation andearly sporogene5is during sex organdevelopment in Arabidapsis thaliana. Proc.Natl. Acad. Sci.IUSA) 96: 11664-69.Schneitz, K. 1999. The molewlor and geneticcontrol 01 ovule development. (urr. Op.Plant. BioI. 2: 13-17.Schneitz, K., M. Hukkamp, and R.E. Pruin. 1995.Wild-type ovule development in Arabidopsisthaliana: a light microscope study 01 clearedwhole-mount tissue. Plant 1. 7: 731-49.---. 1997. Dissection 01 sexual organsontogenesis: a genetic analysis 01 ovuledevelopment in Arabidopsis thaliana.Development 124: 1367-76.Scon, RJ., M. Spielman, J Bailey, and H.G.Dickinson. 199B. Parent·af·origin el/ects inseed development in Arabidopsis thaliana.Development 125: 3329-41.Sene, C, A. Bevilacqua, A. Bianchini, F. Mangia,R. Geremia, and PRossi. 1997.Parthenogenetic activation of mouse eggsby microiniection of atruncated c-kittyrosine kinase pre5ent in spermatozoa.Development 124: 2267-74.Sheridan, W.F., N.A. Avalkina, I.! Shamrov, lB.Batygina, and I.N. Golubovskaya. 1996.The mac 1gene: controlling the commitmentto the meiotic pathway in maize. Genetirs142: 1009-20.Sheridan, W.F., and JK. {lark. 1994. Fertilizationand embryogeny in moize. In M. Freelingand V. Walbot leds.}, The Maize Handbook.New York: Springer-Verlag. Pp.3-IO.Sheridan, W.F., lK. Gark, and M.G. Neuffer.1995. Embryo-endosperm interaction inearly seed development. Proc. 1sf Int. (onl.on Apomixis: "Harnessing Apomixis. ANewFrontier in Plant Science:' College Station,Texas,USA.Sheridan, W.F., EA Golubeva, 1.I. Abrhamova,and I.N. Golubovskaya. 1999. The marlmutation oilers Ihe developmental late ofthe hypodermal cells and their cellularprogeny in the maize anther. Genetics 153:933-41Sheridan, W.F., and B.-O. Huang. 1997. Nudearbehavior is delective in the maize IZeomays l) lethal ovulel lemalegametophy1e. Plant J. 11: 1029.Sherwood, R.T., CC Berg, and B.A. Young. 1994.Inheritance 01 apospary in bullelgrass.(rap. Sci. 34: 1490-94.Shimamato, K. 1995. The molecular biology 01rice. Science 270: 1772-73.Shimamata, K., Miyazaki, C, Hashimoto, H.,Izowa, 1, Itoh, K., Terada, R., and S. lido.1993. Trans-activation and stableintegration of the maize transposableelement Os cotranslected with the Actransposase gene in transgenic rice plants.Mal. Gen. Genet. 239: 354-60.Siddiqi, I., G. Gane5h, U. Grossniklaus, and V.Subbiah. 2000. The dyad gene is requiredlor progre5sion through lemale meiosis inArabidopsis. Development 127:197-207.Singleton, W.R., and PC Mangelsdorf. 1940.Gametic letha Is on the laurth chromosome01 maize. Genetics 25: 366-90.Skarne5, w.c. 1990. Entrapment vectors: anewfool lor mommolion genetics.BioTechnology. 8: B27-31.Smith, D.L, ond N.V. Fedorol/. 1995. lRP1, agene expressed in lateral and adventitiousroot primardio 01 Arabidopsis. Plant Cell 7:735-45.Smith, H.E., and A.P Mitchell. 1989. Atranscriptional cascade governs entry intomeiosis in Saccharamyces cerevisiae. Mol.(el/Biol. 9:2142-52.Speulmon, E., PU. Metz, G. von Arkel, B. Ielintel Hekkert, WJ. Stiekema, and A.Pereira. 1999. Afwo-componenl enhancer·inhibitor transposon mutagenesis system lorfunctionol analysis 01 the Arobidopsisgenome. Plant (ell 11 :1B53-66.Spillane, c., J-P Vielle-Calzada, and U.Grossniklaus. 2000. Parent-ol-llrigin effectsand seed development: genetics andepigenetics. In Y.H. HUi, G.GKhachatourians, A. McHughen, and W.K.Nip, R. Scorza leds.!, Handbook ofTransgenic Food. New York: Mocel DekkerInc. (in press)Springer, PS., W.R. McCombie, V. Sundareson,and R.A. Martienssen. 1995. Gene trapfogging of PROLIFERA, on e5sentiaIM(Ml­J-5·like gene in Arabidopsis. Science 26B:877-BO.Steinhardt, RA, and D. Eppel. 1974. Activationof sea urchin eggs by a calcium ianaphare.Proc. Natl. Acad. Sci. IUSA) 71: 191 >--19.Sundaresan, V. 1996. Horizontal spread oftransposan mutagenesis: new USel for oldelements. Trends Plant Sci. 1: 184-90.Sundaresan, V., PS. Springer, 1 Volpe, S.Howard, JD.G. Jone5, C. Dean, H. Ma, ondR.A. Martienssen. 1995. Panerns of geneaction in plant development revealed byenhoncer trap and gene trap transposableelements. Genes and Dev. 9: 1797-1 Bl O.Sumner, M.J, and Lvon Coseele. 19BB. Ovuledevelopment in Brassica campesfris. Alightmicroscope study. (an. 1. Bot. 66: 2459­2569.Talialerra, C.M., and E.C. Bashaw. 1966.Inheritance and control of obligateapomixis in breeding buffelgrass,Pennisetum ciliare. (rop Sci. 6: 473-76.Tanksley, S.D, and S.R. McCauch. 1997 Seedbonks and molecular mops: unlockinggenetic potential from the wild. Science227: 1063-66.Thoma, S., U. Hechl, A. Kippers, J Botello, S.c.de Vries, ond C. Somerville. 1994. Tissuespecilicexpression of the corrol gene EPilipid tronsler protein gene. Plant (elf 3:907-21.Tirlapur, U.K., E. Kranz, and M. Cresti. 1995.Characterization 01 isoloted egg cells, invitro fusion products and zygotes of Zeamays l. using the technique of imogeana~sis and confocal loser scanningmicroscopy. Zygote 3: 57-64.Tissier, A.F., S. Marillonnel, V. Klimyuk, K. Patel,M. Angel Torres, G. Murphy, and J. D.G.Jones. 1999. Muhiple independentdefective supprelsor-mutalor transposaninserlions in Arobidopsir. a 1001 forfunctional genomics. Plant (elf 11: 184,­52.Topping, JJ, WWei, and K. Undsey. 1991.Functionollogging of regulotory elem.:nlsin Ihe plont genome. Development 112:1009-19.

from Selllalily to Apomixl!: MoIe"Io...dGe.elk App.oocile. 211Tsugeki, R., E.Z. Kochievo, ond N.V. Fedoroff.1996. AIransposon insertion in theArabidopsis SSII16 gene causes on embryodefectivelethal mutation. Plant 1. 10:479-89.Tsunewoki, K., and Y. Mukai. 1990. Whealhaploids through the Salmon method. InY.P.S. Baiai (ed.!, Biotechnology inAgriculture ond Forestry, voln Wheat.Heidelberg: Springer-Verlag. Pp. 460-78.Turcame, E.L, and e.v. Feoster. 1963. Haploids:high-frequency production from singleembryoseeds in 0 line of Pima cotton.Science 140: 1407-08.Urongo JA, R.A. Pedersen, ond J. Arechogo.1996. Porthenogenetic octivotion of mouseoocytes using calcium ionophores ondprotein kinose Cstimulotors. Inl. 1. Dev.BioI. 40: 515--19.von Diik, PJ., I.G.O. Tos, M. Folque, ond T. Bakx­Scholmon. 1999. Crosses between sexuoland apomictic dandelions (Toraxacuml.1IThe breokdown of opomixis. Heredity 83:715--21von Diik, P., ond J. von Domme. 2000. Apomixistechnology and the paradox of sex. TrendsPlant Sci 5: 81-84.Vielle, J-P., 8.L. Burson, E.e. Boshow, ond M.A.Hussey. 1995. Eorly fertilizotion events inIhe sexual ond oposporous egg apporatusof Pennisetum ciliare ILl link. Plant J. 8:309-16.Vielle·Calzada, J-P., R. Baskar, and U.Grossniklaus. 2000. Delayed activation ofthe paternal genome during seeddevelopment. Nature 404: 91-94.Vielle·Calzada, J-P., CF. Crone, and D.M. Stel~.1996. Apomixis. The asexual revolution.Science 274: 1322-23.Vielle-Colzada, J-P., J.M. Moore, W.B. Gagliano,and U. Grossniklaus. 1998. Altering sexualdevelopment in Arabidopsis. J. Plant Bioi.41: 7Hl.Vielle-Calzada, J-P., J. Thomas, e. Spillane, A.Coluccio, M.A. Hoeppner, and U.Grossniklaus. 1999. Maintenance ofgenomic imprinting ot the Arabidopsismedea locus requires zygotic DDM1activity. Genes Dev. 13: 2971-82.Villanueva, J.M., J. Broodhvest, B.A. Houser, RJ.Meister, K. Schneilz, ond e.S. Gosser. 1999.INNEII NO OUTEII regulates abaxial/adaxial pottering in Arabidopsis ovules.Genes Dev.l3: 3160-69.Vinkenoog, R., M. Spielma, S. Adams, R.LRscher, H.G. Dickinson, and RJ. Scott.2000. Hypamehtylatian promatesoutonomous endosperm development andre\{ues post-fertilizalion lethality in fiemutants. Plant Cell 12: 2271-82.Vijayraghavan, M.R., and K. Prabhakar. 1984.The endosperm. In B.M. Jahri led.!,Embryology ofAngiosperms. Berlin:Springer Verlag. Pp. 319-76.Vizir, I.Y., M.L Anderson, Z.A. Wilson, and BJ.Mulligan. 1994.lsolalion 01 deficiencies inthe Arabidopsis genome by g·irradiation 01pollen. Genetics 137: 1111-19.Vollbrecht, E. 1994.lethol ovule2 causesaberrant embryo soc development. MaizeNewslel/er 68: 2-3.Vollbrecht, E. ond S. Hoke. 1995. Deliciencyono~is of female gometogenesis in moize.Devl. Genet. 16: 44-63.Wokana, A., and S. Uemoto. 1987. Adventiveembryogenesis in Citrus. I. The occurrenceof adventive embryos without pollinafionor lertilization. Arner. 1. Bol. 74: 517-30.Walbot, V. 1992. Strategies for mutagenesis ondgene cloning using transposan togging andT·DNA mutagenesis. Ann. lIev. PlantPhysiol. Plant Mol. BioI. 43: 49-82.--.1994. Overview af the key steps inaleurone development.. In M. Freeling andV. Walbot leds.), The Maize Handhaak.New York: Springer Verlag. Pp. 78-80--.1996. Sources and consequences ofphenotypic and genotypic plasficity inflowering plants. Trends Plant Sci. 1: 27­32.Wolters, M.S. 1985. Meiasis readiness in lilium.Can. 1. Genet. Cyrol. 27: 33-38.Weatherwax, P. 1919. Gametogenesis andlecundation in lea mays as the basis 01xenia and heredity in the endosperm. Bull.Torrey Bat. Ouh 46: 73-90.Webb, M.e., and B.E.S. Gunning 1988. Themicrotubular cytoskeleton during embryosocdevelopment in Arahidapsis thaliana. InR.B. Knox, M.B. Singh, and LF. Troianileds.!, Pollination 'BB. Melbourne,Australia: School of Botany, University ofMelbourne.--.1990. Embryo soc developl1).ent inArohidopsis thaliana. I. Megasporogenesis,including the microtubular cytoskeleton.Sex. Plant lIeprod. 3: 244-56.--.1991. The microtubular cytoskeletonduring the development of the zygote,proembryo and Iree-nuclear endosperm inArahidopsis thaliana III Heynh. Planla184: 187-95.--.1994. Embryo soc development inArabidapsis thaliana. II. The cytoskeletonduring megagamelogenesis. Sex. PlantlIeprad. 7: 153-63.Weber, OJ. 1983. Monosomic ana~sis in diploidcrop plonts. In P.K. Gupla and U. Sinha(eds.), Cytogenetics of Crop Planfs. NewDelhi, Indio: Macmillan Indio lid. Pp. 352­78.Weinmann, P., M. Gossen, W. Hillen, H. Bujard,and e. Galz. 1994. Achimerictransactivator allows lelra-cyclineresponsivegene expression in wholeplants. Plant 1. 5: 559--69.West, MAL, and JJ. Harada. 1993.Embryogenesis in higher plants: Anoverview. Plant Cell 5: 1361-69.Whitacker, M., and K. Swann. 1993. lighling Ihefuse at lerlilizalion. Development 117: 1­12.Willemse, MIM. 1981. Polarity duringmegasporogenesis andmegagamelogenesis. Phytamorphalogy31: 124-34.Willemse, MIM., and J.L von Went. 1984. Thefemale gamelaphyte. In B.M. Johri led.),Emhryalogy of Angiosperms. New York:Springer-Verlag. Pp. I 59-96..Wilson, e., HJ. Bellen, and WJ. Gehring. 1990.Posilion effects on eukaryatic geneexpression. Annu. lIev. Cell. Bial. 6: 679­714.Wi~on, C, R.K. Pearson, HJ. BeIIen, CJ.O'Kane, U. Grossniklaus, and W.J. Gehring.1989. P-elemenl-mediated enhancerdetection: on efficient melhod for isolatingand characterizing developmentallyregula led genes in Drosophila. Genes andDev. 3: 1301-13.Xia, Y., BJ. Nikolau, and P.S. Schnable. 1996.daning and characterizalion of CEII2, onArahidopsis gene Ihat affects cuticular waxaccumulolion. Plant Cell 8: 1291-1304.Yodegari, R., T. Kinoshila, O. Lotan, G. Cohen, A.Katz, Y. Choi, A. Katz, K. Nakashima, JJHarada, R.B. Goldberg, R.L R\{her, and N.OOOd. 2000. Mulalions in the FIE and MEAgenes Ihal encode interacting Po!ycamhproteins couse parenl-of-origin effects onseed development by dislinct mechanisms.Plant Cell 12: 2367-82.Yomamalo, M. 1996. Regulalion 01 meiosis inf1ssianYellst. CellStrud. Fund. 21:431­36.Yang, W-e., and V. Sundaresan. 2000. Genetks01 gamelophyte biogenesis in Arahidapsis.CUff. Op. Plant Sci. 3: 53-57.Yang, W-e., D. Ye, J. Xu, and V. Sundaresan.1999. The SPOIIOCYTELfSS gene ofArohidapsis is required lor initiation 01sporogenesis and encodes a novel nuclellrprole in. Genes Dev. 13: 2108--17.Zaki, M.A.M., and H.G. Dickinson. 1990.Structural chonges during the first divisionsof embryos resuhing from anther and freemicrospore culture in Brassica nopus.Protoplasmo 156: 149-62.Zimmerman, lL 1993. Somatk embryogenesis:a model lor ear~ development in higherplants. Plant Cell 5: 1411-23.

Chapter 13Induction of Apomixis inSexual Plants by MutagenesisUTA PRAEKELT AND ROD SCOTTIntroductionThe ability to manipulate reproduction in cropplants from the sexual to the apomictic mode,and vice versa, is highly desirable (Hanna1995; Jefferson and Bicknell 1996). Severalchapters in this volume deal with the variousapproaches that have been taken toward thisgoal. The most promising sh'ategy so far hasbeen the transfer of natural apomixis genesfrom wild species into related sexual cropplants by introgression (Hanna et al. 1992;Leblanc et al. 1995; Savidan, Chap. 11).Unfortunately, this is likely to remain limitedto those crops that have apomictic relatives,and therefore will not be widely applicable.In light of this situation, efforts are being madeto identify the gene(s) that confer apomixis,both to gain a better understanding of thegenetic regulation of the trait and to eventuallyfacilitate transfer to a wider range of speciesby genetic engineering methods. As describedby Grimanelli et a!. (Chap. 6), one key steptoward gene isolation is the genetic mappingof apomixis genes, and considerable progresshas been made with some species (Kindigeret al. 1996). However, because of the intrinsicdifficulties in mapping apomicts, and becauserecombination around the apomixis locusappears reduced (Grimanelli et a!., Chap. 6),map-based cloning of the apomixis gene(s) islikely to proceed slowly.Recently, an alternative approach, usingmutagenesis, has been considered by severalgroups both for the identification of naturalapomixis genes and for the de novo inductionof apomixis in sexual plants (Koltunow et al.1995). Mutagenesis has been widely andsuccessfully applied to the study of manyaspects of plant growth and development.Thanks to rapidly advancing methods in allilTeas of DNA technology, improvements inplant transformation methods, and theaccumulation of mapping and sequencingdata, the isolation of genes via their mutantalleles has become a feasible approach in manyareas of plant research. Bicknell (Chap. 8)describes the development of Hieracium as amodel apomict (see also Bicknell 1994a,b,c;Bicknell and Borst 1994). The aim is to inducereversion tosexual reproduction by insertionalmutation of the responsible gene(s) and toisolate them via the inserted sequence.In this chapter we describe the reciprocalapproach, the mutagenesis of a model sexualplant in an attempt to induce apomixisde novo.Mutated alleles conferring apomixis orindivid ua1 components of apomicticdevelopment can be identified with relativeease in a model plant, and the cloned geneswould then be available for transfer to otherspecies. The possibility of a mutagenicapproach resulting in the isolation of thedesired mutants greatly depends on themethodology employed. Therefore, in order tomaximize the opportunities for the inductionand detection of these mutants, severalconsiderations must be taken into account.Details of the mechanisms of apomixis have

lod"lIo. of ApomIxis 10 SelIai PIoo,. ~y ..'......Is 213been described elsewhere in this volume,therefore only a brief summary is given hereof the individual components of apomicticversus sexual development that could beseparately affected by mutations. Wesummarize some of the earlier work withmutagenesis and describe some of the mostinteresting mutants with elements of apomixisthat have been isolated in various plants. Thefact that none of these potentially usefulmutations has been characterized at the DNAlevel underlines the importance of using wellcharacterizedmodel plants for this work. Thefeasibility of obtaining mutants of Arabidopsiswith apomictic characteristics has beenconfirmed recently by the identification ofseveral mutants with partial seed development.We describe the various approachescurrently underway in several laboratories,which are aimed at the induction of apomicticcharacteristics in model sexual plants.ConsiderationsComponents of ApomixisThe two major types of gametophyticapomixis that can be distinguished, diplosporyand apospory, differ from sexual developmentin more than one aspect (for reviews see Asker1980; Asker and Jerling 1992; Nogler 1984; andCrane, Chap. 3). Since mutants with individualcomponents of apomixis have been identifiedpreviously, and can be expected in futuremutagenesis experiments, the main differencesare summarized briefly. Also, as pointed outbelow, apomictic tendencies are occasionallyobserved in sexual plants and can beinfluenced by environmental stimuli.1. Avoidance of meiosis. In diplosporousapomicts, meiosis is avoided by twoprincipally different mechanisms: in the first,meiosis of the archesporial cell is directlyreplaced by a mitotic division; in the second,failure of chromosome association or synapsisis associated with nuclear restitution, andfollowed by equational division of the entirechromosome complement in the secondmeiotic division. Subsequent developmentleads to the formation of a functionally normalembryo sac with an unreduced egg and centralcell. Sexual plants occasionally produceunreduced gametes, and through fertilizationthese give rise to the polyploid series that canbe observed in many plant species (Asker1980). The production of unreduced embryosacs in some plants, including Brassica species,is revealed after distant or "prickle" pollination(which stimulates parthenogenesis; see below)by the appearance of matromorphous progenyat considerable frequencies (Eenink 1974a).2. Fonnation of aposporous embryo sacs. Inaposporous apomicts, unreduced embryo sacsarise directly by mitosis from nucellar cells,usually in addition to a sexual, reducedembryo sac with normal meiosis. In facultativeapomicts, sexual and apomictic embryo sacsdevelop side-by-side or in separate ovules andgive rise to a mixture of sexual and maternalprogeny. Although female meiosis is mostlynormal, in many cases the aposporous embryosacs, which are not delayed by meiosis,develop at a faster rate than the products ofmeiosis, which frequently degenerate at themegaspore stage.3. Parthenogenesis. The unreduced egg cellin both diplosporous and aposporous embryosacs d~velops directly into an embryo, withoutfertilization by a sperm nucleus. It is not clearwhether fertilization is prevented actively oras a consequence of precocious parthenogenesis.Again, parthenogenesis occasionallyoccurs in sexual plants, and can be induced inmany plants by certain stimuli such as pricklepollination. Of particular interest to the presentinvestigation is the frequent occurrence ofmatromorphy, or diploid parthenogenesis, inBrassica oleracea, a species that belongs to thesame family as Arabidopsis, the Brassicaceae.

214 Uta Praebll aod Rod ScottCrosses with pollen from different speciesresult in various proportions of matromorphsarising from the parthenogenetic developmentof unreduced egg cells (Eenink 1974b). Thusin Brassica, two of the most importantingredients ofapomixis are revealed by distantpollination: (i) the presence of unreducedembryo sacs, and (ii) the inherent capacity forparthenogenesis. Haploid parthenogenesishas recently been induced both in Arabidopsisand in Brassica juncea by the application ofbrassinolide, a steroid hormone first isolatedfrom Brassica pollen (Kitani 1994).4. Endospenn development. In autonomousapomicts, endosperm development occursspontaneously, but in the case ofpseudogamous apomicts, it depends onfertilization of the central cell nucleus by asperm nucleus. In some cases, pollinationwithout fertilization has been suspected oftriggering endosperm development.Endosperm plays a crucial role in theformation of viable seed, and requires specialconsideration in the design of a mutagenesisscreen. The problem of the endosperm isdiscussed later in more detail.Genetic Control of ApomixisWith perhaps one exception (Carman 1997;Carman, Chap. 7), it is now generally acceptedthat apomixis has evolved from sexualancestors by mutation rather than being aconsequence of polyploidization andheterozygoci ty. This is an importantassumption for mutagenic approaches to thestudy of apomixis. Much discussion focuseson whether apomixis is regulated by a singlegene and could therefore be induced by asingle mutation in a sexual plant. Given thedifferent components of apomixis, it seemsmore likely that a number of mutations wereneeded in the evolution of a viable apomictfrom a sexual ancestor. The apparent singlelocus inheritance that has been reported inseveral cases could be explained by the tightlinkage of several genes, a possibility that isconsistent with the lack of recombinationobserved between molecular markersassociated with apomixis (Grimanelli et a!.,Chap. 6). Even if a Single mutation, perhapsresulting in the inhibition of meiosis, wasresponsible for the evolution of apomixis, it islikely to have occurred in a background thatpermitted its expression (Mogie 1988) and hastherefore evolved only in a subsection of generaand families. For these reasons, we do notexpect that a single mutation in a sexual plantcould result in the production of viable seed inthe absence of fertilization. However, given thevariety of apomictic forms that can bedistinguished, such as in diplospory andapospory, apomixis has probably arisenindependently in different species and perhapsinvolves different genes in each case.Consequently, there may be ampleopportunities for the induction of someelement of apomixis in a sexual plant bymutations in a number of different genes.An important aspect to consider is the apparentdominance, in many cases, of apomixis oversexuality (Nogler 1984; Mogie 1988; Leblancet al. 1995). Mutations that completely abolishgene function, such as deletions, can bedominant only if expression of that gene issubject to gene dosage. One of the hypothesesthat have been put forward on the control ofapomixis is that the responsible gene(s) encoderegulatory functions that initiate or represscertain developmental programs (Koltunow etal. 1995). Ifapomictic development is repressedin sexual plants by a negative regulatoryprotein whose concentration is crucial, then areduction in the level of this protein could besufficient to induce a developmental pathwaythat is suppressed only when two copies of thegene are present. The fact that many apomictsare facultative, and that the proportion ofapomictic progeny can be influenced byenvironmental factors, supports the hypothesis

IId.cllo••f ApOllllxl. 10 Sex..1Pia.,. by MII......i. 215of a finely tuned regula tory function forapomixis genes.An alternative route to a dominant allele is amutation leading to a change in function of theprotein, e.g., an altered specificity. Mutationsof this type could be very rare indeed.However, many regulatory proteins are activeonly after forming a complex with otherproteins, and changes in many amino acidresidues, particularly those exposed on thesurface, could alter their binding affinity,resulting in reduced activity of the entirecomplex. These possibilities have to be takeninto account in the design of any mutagenesisexperiments, as they may influence the choiceof mutagen (see below).How Important is Polyploidy?The majority of apomicts are polyploid. Ifpolyploidy is a prerequisite for apomixis, ashas been suggested by some researchers,screening for apomictic mutants in a diploidspecies is not a promising proposition.Fortunately, there are a few examples of diploidapomicts, the most relevant to this work beingArabis, a genus comprising species with a rangeof base numbers and ploidy levels (Bacher1951; Carman 1997). Some of the diploids aresexual, while others are pseudogamousdiplosporous apomicts. The observed variationin ploidy of pollen nuclei of sexual andapomictic species, and the occasionaloccurrence of unreduced embryo sacs in thesexual species, form the basis for the evolutionof this agamouscomplex, and support the viewthat polyploidy is a result, rather than aprerequisite, of apomixis. However, accordingto Carman (personal comm.), care should beexercised in assigning a particular ploidy tosexual or apomictic plants because anapparently diploid species may represent adiploidized polyploid (paleopolyploid).Notwithstanding this cautionary note, one ofthe most encouraging examples of apomixisat the diploid or near diploid level is the recentdevelopment of a polyhaploid hybrid betweendiploid (sexual) maize and tetraploid(diplosporous) Tripsacum, containing onecomplete set of chromosomes from each of theparent species. Although male-sterile, thisautoallotriploid hybrid (three genomes)reproduces apomictically after pollinationwith maize, demonstrating that a genetic locusconferring apomixis normally in a tetraploidcan also function in a diploid(-like)background (Leblanc et al. 1996).In addition to being a consequence ofapomixis, polyploidy has an importantfunction in apomicts. As discussed by Nogler,the apospory "factor" or allele in Ranunculusauricomus can only be transmitted by diploidor polyploid gametes, and acts as a recessivelethal factor in haploid gametes. Apparently,the normal (sexual) allele is required for someaspect of gamete formation or function. Thissuggests that, while not essential in anindividual plant that produces unreducedgametes heterozygous for the apomixis allele,polyploidy may be important for themaintenance and spread of apomicticpopulations.The Problem of the EndospermThere is overwhelming evidence for theimportance of a balanced ratio between thematernally and paternally inherited genomesin the endosperm of many species. Endospermis usuplly triploid, resulting from the fusionof two maternal polar nuclei with one pollennucleus, producing a ratio of two maternal toone paternal genome (2m:1 p). Deviation fromthis ratio often leads to embryo abortion or tosubfertile, abnormally shaped seed. The mostplausible explanation for these observationsis the parental imprinting of genes, which,unlike in animals, does not affect the embryobut does affect the endosperm of thedeveloping seed. Various strategies seem tohave evolved during the evolution ofapomixisto ensure normal endosperm development. In

216 Uta Praektlt.. Rod Scana minority of apomicts, endosperm productionis autonomous, i.e., it does not requirefertilization, and here the endosperm is diploidwith no paternal contribution. However, themajority of apomicts are pseudogamous andrequire fertilization of the central cell. Becausemost apomicts produce normal reduced malegametes, the expected m:p ratio in theseendosperms is 4m:1p, given that the polarnuclei are diplOid. This ratio, however, has notbeen observed in many of the examined cases,and it appears that the correct (ancestral) ratiois obtained in a variety of ways. In Diclumthiumannulatum, the unreduced polar nuclei remainunfused at fertilization and either (i) a reducedsperm fuses with only one polar nucleus andthe other polar nucleus degenerates, or (ii) eachpolar nucleus is fertilized with a reducedsperm (Reddy and d'Cruz 1969). Similarly, inRanuneulus auricomus, two fused unreducedpolar nuclei are fertilized by two reducedsperm (Rutishauser 1954). Acommon strategyamong apomictic grasses is to produceunreduced embryo sacs with four rather thaneight nuclei; the Single polar nucleus isfertilized by a reduced sperm (Warmke 1954).These observations are probably the mostconvincing evidence that the evolution ofbothautonomous and pseudogamous apomictsmust have occurred in a genetic environmentcontaining either pre-adaptations oradditional mutations. Autonomous apomictsmay have evolved in a background where thestringent requirement for a balancedendosperm had been relaxed.It was recently shown that in Arabidopsis thepresence in endosperm of maternal orpaternalexcess, resulting from reciprocal crossesbetween diploid and tetraploid parents,greatly affects seed size (Scottet aI.1998). Seedswith maternal excess (4m:1p) are smaller, butseeds with paternal excess (2m:2p) are largerthan normal. However, these seeds havenormal viability. As indicated earlier, haploidseeds can be induced in Arabidopsis andBrassiea juneea by the application ofbrassinolide to the stigmas of emasculatedflowers. Endosperm formation in these seedsoccurs in the absence of any paternalcontribution, and as expected, the seeds in bothcases are smaller than normal diploid seeds;however, they give rise to stable haplOid plants.These are important observations with respectto the screens for apomixis in Arabidopsis,because they suggest that, should a mutationlead to parthenogenetic embryo development,an imbalance in the m:p ratio of the endospermshould not hinder viable seed production,provided of course that the endospermproliferates and supports seed production. Theeffect on seed size by an imbalancedendosperm could, furthermore, be exploitedfor the screening of mutants. However, amutation resulting in parthenogenetic embryodevelopment does not necessarily induceendosperm proliferation. This may require anadditional stimulus, such as that normallyprovided by pollination or a second mutation.The problem of the endosperm is one of themost important aspects to consider in amutagenesis program, and because of it, wedo not expect to be able to induce viable formsof apomixis by a single round of mutagenesisof sexual plants. The aim of severalmutagenesis and screening programscurrently in progress is to facilitate theidentificlltion of mutants that have only somecharacteristics of apomixis, such asparthenogenesis or autonomous endospermdevelopment, and which do not necessarilyproduce fertile seed in the absence ofpollination.Which Mutagen?The choice of mutagen is an importantconsideration because it determines the typesof mutations obtained; here we need onlydistinguish between the two categories of"change of function" and "loss of function"

mutations. Change of function ·mutations arethe result of amino acid substitutions or ofchanges in gene expression level. Loss offunction mutations abolish the gene productaltogether. As discussed earlier, apomixis maybe induced by a dominant allele, and this ismore likely to be the result of a change offunction mutation. There are numerous reportsof dominant mutations resulting from singleamino acid substitutions that have beenobtained by EMS or by in vitro mutagenesis(e.g., Hemerly et al. 1995; Kim et al. 1996;Wilkinson et al. 1997). For this reason, it couldbe important to use a mutagenic agent that caninduce subtle mutations. Ethylmethanesulfonate (EMS) has been shown to be the mostversatile mutagen, as it causes point mutationsat a high frequency, and whilst most of theseshould result in amino acid substitutions, theymay also create stop codons, and thus lead toloss of gene function (for a summary andreferences, see Feldmann et al. 1994). However,as already described, some useful mutantshave been obtained by radiation treatment,which frequently causes large deletions, andtherefore the value of mutagens that can causegene disruptions or deletions should not bedismissed.A widely used method of mutagenesis isinsertional inactivation, or gene tagging, withT-DNA or transposons (Feldmann 1991;Lindsey et al. 1993; Bouchez et al. 1993; Aartset al. 1995). As with deletions, this type ofmutagenesis most likely will create loss offunction mutations due to insertion into thecoding region of a gene, although it isconceivable that changes in gene expressionlevels could result from insertion into a genepromoter. However, a major advantage ofgenetagging is the ease with which the mutatedgene can subsequently be identified. Itsapplication to the study of apomixis inHieracium is examined by Bicknell (Chap. 8).We briefly summarize below the work byRamulu (1997) on transposon mutagenesis inArabidopsis and Petunia. For the purpose ofinducing apomixis in Arabidopsis, it may beimportant to choose a mutagen that is capableof producing change of function mutations.Arabidopsis is the best-characterized plantspecies, and map-based cloning of nontaggedmutant alleles is no longer a major obstacle.Some Early Work with MutantsInduction of Sexuality in ApomictsImproving cultivars of apomictic crop speciesby breeding is difficult, and therefore manyattempts have been made to increase thefrequency of sexual reproduction in apomictsby mutagenesis (Bashaw and Hoff 1962;Hanson and Juska 1962; Gustafsson and Gadd1965; Asker 1966). Complete reversion tosexuality was not observed in any of thesestudies. In the few cases where sexuality wasobserved, it was transient and plants revertedto apomixis in subsequent generations. Anumber of factors probably contributed to thislack of success. First, the induction of mutantsin polyploid species is inherently difficultbecause mutations can be masked byadditional copies of wild type genes. Second,in many experiments, the number of seedsmutagenized and screened (200 per treatmentin one case) was probably inadequate fordetecting potentially rare mutations. Whetherthe choice of mutagens, in most casesirradiaJion, also contributed to the lack ofsuccess depends on whether a change offunction rather than a loss of function mutationwould be required for a reversal to sexuality.We do not believe that the lack of success todate indicates that efforts to induce sexualityin apomictic species must necessarily fail.Researchers will need to work with a muchlarger number of mutants, to choose anapomict that has a low ploidy level, andperhaps to use a more subtle mutagenictreatment that induces both change and lossof function mutations.

21 8 Uta P...ke~ .d Rod S

1'"""10••f ApomIxl. 10 S....I Plaot. by M.I......1s 219The above mutants of maize and barleyillustrate that mutations at a number of loci caninduce one element of apomixis-theformation of unreduced gametes. Furtherdevelopment is dependent on fertilization, andnormal endosperm development depends ona balanced maternal to paternal ratio.The preferential detection of unreduced eggsby crosses with tetraploids is a useful reminderof the importance of choosing a suitable pollenparent for mutagenesis studies involvingpollination. On one hand, mutants withunbalanced endosperm may be difficult toisolate initially because of low viability; on theother hand, the shrunken endospermphenotype could serve as a useful criterion inthe selection for meiotic mutants.2. Parthenogenetic mutants. The EMS-inducedhap mutant of barley was initially isolated inthe form of a chlorophyll deficient mutantcontaining at least three linked mutations, tig,let (pollen lethality), and hap (Nielsen 1974).Presence of the hap allele results in a lowfrequency of haploid progeny. After separationof hap from the other two mutations, Hagbergand Hagberg (1980) showed that hap isincompletely dominant over the wild typeallele: heterozygous (hap / +) plants produce3-6% haploid progeny, whereas homozygous(hap/hap) plants produce up to 40% haploidprogeny. Perhaps not surprisingly, crossesbetween a homozygous mutant (hap/hap )andwild type (+ / +) plants produce different resultsin the F I, depending on whether the mutant isthe male or the female parent: a hap/hap femaleplant pollinated by wild type (+/+) pollenresults in a high frequency of haploid F Iprogeny, whereas no haploids are producedwhen a wild type (+ / +) female plant is crossedby a hap/hap male. This indicates that the haplocus acts only through the maternal tissue,either to prevent fertilization of the egg cell orto stimulate the egg cell nucleus to divideprematurely. In this mutant, parthenogenesisis the only element of apomixis that has beeninduced. The formation of a perfectly welldevelopedendosperm, which supports theproduction of a viable seed, is presumablyfacilitated by normal fertilization events in thecentral cell that result in a genomicallybalanced endosperm.3. Aposporous mutants. In pearl millet, twomutants that produce aposporous embryo sacshave been reported. The first,Jemale sterile (js),is a recessive mutation induced by radiationtreatment (Hanna and Powell 1974; Arthur etal. 1993). Homozygous mutants are femalesterilebut produce normal viable pollen.Mutant ovules are small and immaturecompared with the wild type, and only abouthalf of them contain embryo sacs. Of these, themajority are multiple embryo sacs that appearto be aposporous, although sexual embryo sacsare observed in some ovules. Only a very smallproportion of ovaries display any endospermor pro-embryo development, and all ovulesdegenerate five days after pollination. Pollentube growth in the mutant is abnormal, andthe inhibition of fertilization has been proposedto explain the absence of seed set.The second, stubby head, was discovered inprogeny of seed treated with both thermalneutrons and diethyl sulfate (Hanna andPowell 1973). This recessive mutation causes apleiotropic phenotype, including twin ovules,shortened internodes, flattened stems, and astubby inflorescence. It produces both normalsexual embryo sacs in some ovules andmultiple embryo sacs, which arise fromnucellar cells, in others. Test crosses confirmedthat stubby head is a facultative apomict,producing maternal progeny at frequenciesranging between 23% and 77%.Mircrosporogenesis in this mutant is normal,however, seed set is low; this has beenattributed partly to nonfertilization because ofcompetition between the multiple embryo sacs.

220 Ura P"",k.h .d Rod ScottStubby head could indeed be a very usefulmutant, as it produces viable seed by apospory.Although the nature of the mutation is notknown, the mutation affects several elementsof apomixis in a single step. Whilst this couldbe due to a single gene mutation, thepleiotropic nature of the mutant phenotypecould well indicate that it is caused by adeletion encompassing a number of genes.4. Conclusions. The above examples ofm u tants with apomictic characteristicsillustrate a number of points relevant tomutagenic approaches in model plants.The most important conclusion is that all theelements of apomixis can be induced bymutation in sexual plants. In one case, stubbyhead, several elements were inducedsimultaneously to produce a viable form offacultative apospory. It would be veryinteresting to analyze these mutants at theDNA level, but the isolation of genes fromthese crop species would not be a Simple task.Unlike Arabidopsis, the cloning of genes viatheir mutant alleles is not routine.None of the mutants described herein wereoriginally isolated in screens for apomicticcharacteristics. The majority were isolated asreduced fertility mutants and some aspleiotropic mutants. Therefore, it seems worthconsidering which kinds of mutantphenotypes, other than apomixis itself, couldbe screened for in Arabidopsis. First, it mightbe worthwhile to screen for reduced fertilitymutants, a simple screen that would involvetesting for seed set in the M 2. A large numberof male-sterile mutants have been isolatedfrom Arabidopsis. Some are present in thecollection of T-DNA tagged lines (availablefrom the Arabidopsis Stock Centers atNottingham, U.K. and in Ohio, U.s.A.); thesecould all be screened for maternal progenyafter pollination with a dominantly markedpaternal line. Second, the tri and el mutantsboth give rise to shrunken seeds, and since ithas been shown that an unbalanced ratioaffects seed size in Arabidopsis too, thisphenomenon could be used as a screen for theisolation of Arabidopsis mutants that produceunreduced egg cells.Current Approaches to theIsolation of Apomictic Mutantsin Model Sexual PlantsThe most important precondition for the largescalescreening of apomictic mutants is theavailability of male sterile lines that do not setseed in the absence of pollination. Arabidopsisis ideally suited for such an undertakingbecause it offers several ways of establishingmale-sterile lines. The most useful of thesecauses conditional male sterility, which allowsthe propagation of homozygous seed byselfing under permissive conditions. Inaddition, Arabidopsis provides easily scoreddominant and recessive markers forsubsequent screening of mutagenizedprogeny.The most obvious screen for apomicticmutants of a male-sterile sexual plant is forseed set in the absence of pollination. However,there have been no reports of mutants (in anymodel plant) that produce viable seed in theabsence of pollination. This may well bebecause the number of progeny screened inthis way was not large enough to detect suchmutants, as these could indeed be very rare.For the reasons already outlined, however, italso seems likely that more than one genemutation is required to induce a viable formof apomixis.Current efforts in several laboratories aredirected toward the induction of partialapomictic development. Basically two maintypes of screens are being conducted. The firstidentifies mutants that show partialdevelopment of fruits in the absence of

h••"..., Apooixl. it So.... Ploor. ~,MIItapIoe

222 Uta Praekelt Old Rod S,ollelongation of the silique. As discussed by Ohadet a!., these two processes both involvematernal developmental programs that mustbe induced by signals from the developinggametophyte. The discovery of embryolessseeds with normal seed coat and elongatingsiliques suggests that the signal originates inthe endosperm rather than in the embryo.It is not clear what the relationship is betweenfislfie and the genes controlling apomixis. Thefact that the mutations affect a latedevelopmental stage after embryo-sacdifferentiation make it unlikely that they arealleles of the apomixis genes per se. Theycould, however, be the targets of apomicticregulatory genes. It is noteworthy that fislfieare female lethal mutations, as this is consistentwith the observation that apomixis appears tobe controlled by a dominant gene, the recessive(sexual) allele of which may be required forsome function in development. It would beinteresting to test the effect of these mutationsin a tetraploid background to see if diploidembryo sacs heterozygous for fis/fie haveautonomous endospenn development and areable to transmit the mutant allele.The three mutants described above wereidentified in two independent large-scalescreens of M] and M 2progeny, respectively;each mutant was isolated in the fonn ofseveralallelic mutations, suggesting that the level ofmutagenesis must have been near saturation.The fact that all mutants detected in these twoscreens show autonomous endosperm, andthat many of the seed-like structures containedno embryos, could mean that a mutationresulting in autonomous embryo development,but one that does not activateendosperm, would not be detected in a screenfor elongated siliques.Finally, a screening program for elongatedsiliques in Arabidopsis has been described byRamulu et al. (1997). Again, the conditionalmale-sterile seed parents are waxless eceriferummutants, cerl and cer6-2 (popl), which are selffertileat high humidity and male sterile at lowhumidity. M] plants have been screened forsectors with elongated siliques, and M 2families derived from selfed individual M]plants have been screened for segregation ofplants containing elongated siliques.Screening for Dominant Mutations in theM, after PollinationWe are currently conducting a screen fordominant mutations that result in theproduction of autonomous embryos but thatmay require fertilization of the central nucleusfor the development of endosperm. Althoughmost natural pseudogamous apomicts havefertile pollen, induced mutations that give riseto autonomous embryos as a result of meioticdisturbances could also have defective pollen.Such mutants would not be recovered in theM 2generation after selfing. For this reason, andto take advantage of the smaller number ofplants that need to be screened to obtain amutant, we decided to screen in the M 1.In one version of the screen, mutants aredetected in the form of maternal progeny afterpollination by a pollen parent containing adominant marker. This screen depends on theformation of fertile seed and would not detectmutant apomictic seeds that fail to germinate.Therefore, we devised a second version thatallows thee detection of autonomous embryosat an immature seed stage, and which can beemployed should the first version fail.The seed stock used for mutagenesis is aconditional male sterile line (OTA 03) in theArabidopsis C24 background. This line ishomozygous for a transgene encoding atemperature sensitive diptheria toxin underthe control of the tapetum-specific promoterA9. Growth at 18°C results in male sterility dueto the absence of functional pollen. At 26°C,the plants are fertile, allowing the propagationof homozygous seed by selfing.

IndUdton of Apomixis in Sexual Plants by MlIItagenesis 223Seeds were m utagenized wi th EMS for 6 hours,washed, and dried for storage on filter paper.The seeds were sown in batches of 100-200 atregular intervals, and seedlings were grownto flowering in individual pots. After growthat 18°C for at least one week, the plants werecut back to two mature flowers on a singleinflorescence. These M jflowers werepollinated by a transgenic line in the WSbackground homozygous for a Basta herbicideresistance gene. Progeny resulting from crossfertilizationare resistant to Basta, whereasmaternal progeny are Basta-sensitive (Figure13.1). However, since Basta-sensitive plantscannot be rescued, this selection presents aproblem in cases in which only a small numberof seeds are available. We are thereforecurrently using a simplified selective criterion,spoecifically, the dominant leaf morphology of(rosses(mutagenized seed parentxpollen parent)Normal fertilizationA. (24 OTA x 063 Basta RB. (24 OTA xWSthe pollen parent (Figure 13.1). The WSecotype produces a highly characteristic leafrosette consisting of tightly-spaced roundleaves with abundant trichomes that aredistinguishable from the smooth elongatedleaves of C24 at an early stage of growth.Heterozygous seedlings are indistinguishablefrom the homozygous pollen parent WS.As the screen is conducted in the M 1, mutantsthat give rise to maternal progeny are mostlikely to be heterozygous for the mutation.Maternal progeny could potentially arise in anumber of different ways: (i) the parthenogeneticdevelopment of a reduced egg incombination with pseudogamous endospermresulting in a haploid seedling; (ii) the parthenogeneticdevelopment of an unreduced eggcell with pseudogamous endosperm andhaploidBastal/R 1:1ParthenogenesisdiploidBasta l~OO ~C. (24 OTA x EM2

224 Uta P...kek .d Rad Scaliresulting in a diploid seedling; and (iii) selfingof the conditional male-sterile plant byinfrequently fertile pollen. As illustrated inFigure 13.1, haploid parthenogenesis arising inthe M 1should produce equal frequencies ofmaternal and hybrid progeny. In the case ofdiploid parthenogenesis resulting from adominant mutation, all progeny could bematernal. However, in both cases the expectedfrequency would be less if the mutation resultsin facultative parthenogenesis or if the mutationhad incomplete penetrance.Considering pseudogamous development ofthe endosperm, both haploid parthenogenesisand selfing would result in a balanced maternal:paternal ratio, and therefore the size and shapesof seeds are expected to be normal. However,in the case of diploid parthenogenesis, theendosperm may have either a balanced orunbalanced m:p ratio, depending on whetherone or both sperm nuclei fertilize the centralcell. Therefore the resulting seed could be eitherof normal size or smaller (maternal excessphenotype). Small seeds might bedistinguishable at the time of sowing, andappropriate steps could be taken should theynot readily germinate on normal soil.The scale of any screening program involvingpollination is limited by its labor intensity andby the space required for growing the progeny.For future screens, we have thereforeincorporated a recessive marker into the seedparent that is detectable at a very early stage ofseedling growth. We crossed the conditionalmale sterile line DTA Q3 with the thiamineauxotrophic tz mutant, and selected progenyhomozygous for both conditions (DTA tz). Afterpollination with wild type pollen, heterozygousprogeny are thiamine prototrophic, butmaternal progeny are auxotrophic (Figure 13.1).Homozygous tz mutants emerge with greencotyledons, but the first true leaves are white.These seedlings can be rescued by sprayingwith thiamine. The advantage of this system isthe early detection of maternal progeny thatallows screening at a much higher density andprovides results soon after sowing. If thisscreen for maternal progeny mutants is notsuccessful, the second version is employedthat allows the identification of autonomousembryos (Figure 13.1). For this purpose,mutants are pollinated by a pollen parenttransgenic for an embryo-specific GUSreporter gene (Topping et al. 1994). GUSexpression is easily detectable in thedeveloping seeds from eight days afterpollination, well in advance of seed ripening.GUS-negative embryos again could arise fromhaploid or diploid parthenogenetic egg cellsor from selfing of the plant.In principle, the viable and nonviable seedscreens could be carried out Simultaneouslyby combining the use of the thiamineauxotrophicmale-sterile seed parent with apollen donor containing the embryo-specificGUS reporter. One of the siliques resultingfrom the cross could be stained for GUSactivity at an immature stage and the otherleft on the plant until seed maturation.Putative apomictic candidates can be furthertested to confirm whether endospermdevelopment is autonomous orpseudogamous (Figure 13.1). For this purpose,a pollen donor line has been established in theecotype WS that is transgenic for anendosperm-specific marker gene, the GUSgene ~nder the control of a high molecularweightglutenin wheat gene (Colot et al. 1987).This is expressed only in the cells of thedeveloping endosperm. After crossing withthis marker line, GUS expression is tested insiliques at an immature stage, beforeendosperm absorption. This system could alsobe used to screen for autonomous endospermmutants independent of embryo formation.If the cited screens for viable and nonviableapomictic seed prove unsuccessful, a screeninvolving pollination will be conducted in an

lod"1101 of ApomIxi. II S....IPIlII,. ~ MlfatHe.1s 225M 2. This allows the detection of recessive aswell as dominant mutants. Also, it is wellknown that mutations occur in sectors of M]plants, and by screening only a singleinflorescence per plant, many mutants may belost. This loss can be avoided by screening intheM 2Transposon Mutagenesis for theIsolation of Apomictic Mutants ofArabidopsis and PetuniaTransposon mutagenesis, like T-DNA tagging,creates mutations by insertion of thetransposon into a gene. Its advantage is thattagged genes can be isolated by using theinserted sequence as a molecular probe. Also,a large number of mutants can be producedsimply by repeated selfing of the plants. Thisapproach is currently being applied toArabidopsis and Petunia, as described in detailby Ramulu et al. (1997). We shall brieflysummarize the main points.For Arabidopsis, a two-element system, derivedfrom the maize transposable element En- [,was used (Aarts et al. 1995). The maizetransposon has a 13 bp inverted repeat at eachterminus and encodes a transposase requiredfor transposition. The two-element systemconsists of a nonautonomous "wings-elipped"En-transposase under the control of the CaMV355 promoter and a nonautonomous mobileI-element with flanking inverted repeats thathas been inserted into a kanamycin resistance(nptII) gene. Both elements are containedwithin a T-DNA that also carries a hygromycinresistance marker for selecti on oftransformants. Several lines that containedabout 20 I-elements and the En-transposasewere crossed with homozygous cerl and cer6-2mutants, and homozygous male sterile lineswere selected from the segregating F 2population. Propagation of these lines forseveral generations under permissiveconditions is expected to create a large numberof new mutations, which can be screenedunder nonpermissive conditions for apomicticmutants.For transposon mutagenesis in Petunia,Ramulu et al. (1997) are using a two-elementtransposon system found in Petunin, whichconsists of a nonautonomous element, dTphl,and an autonomous element carrying thetransposase, ACTl (Doodeman et al. 1984;Gerats et al. 1990). A line containing more than200 copies of dTphl, which produces a highfrequency of unstable mutations in selfedprogeny, was used to establish a number oftransposon genotypes, which were eachcrossed with a conditional male-sterile plant.Male sterility in this line results from theabsence of flavonols, which is caused by achalcone synthase antisense gene (Ylstra et al.1994). The application of flavonols, which arerequired for pollen tube growth, restoresfertili ty and allows selfing. Plants homozygousfor the malesterilityphenotype will beselectedin F 2populations, and screening for apomicticmutants will be conducted on a large numberofF 3and F 4plants in the absence of flavonols.Branching Out in the BrassicasA benefit of mapping data from severalimportant crop plants and from Arabidopsis hasbeen the discovery that groups of genes withinlarge segments of the chromosomes arearranged in the same linear order betweenrelated species regardless of differences ingenome size (Flavell and Moore 1996). Inmany cases, molecular markers identified foronespecies are found to map to correspondinglocations in a related species. This high levelof synteny can be exploited for the isolationof genes from species with large genomes. Thehomologous gene can first be isolated from arelated species with a small genome, such asArabidopsis, where fine mapping andchromosome walking are feasible; it can thenbe used as a probe for direct isolation of thegene from the species of interest.

226 Uta P,..kelt ..dRod S

I.d,dio. of Apomilis i. I..ual PIa.,s by MI'age••sls 227pseudogamy can be concluded from theinduction of haploid parthenogenesis at highfrequency by the application of brassinolide,the steroid hormone present in pollen that maybe the trigger for endosperm developmentfollowing normal pollination (Kitani 1994).The fact that to date mutagenesis has notresulted in fertile maternal seed confirms thehypothesis that viable apomixis can only beobtained in species that have acquired thenecessary preadaptations. A long-termconsideration in our pursuit of apomicticmutants in Arabidopsis may well be to combinemutations obtained from different screeningprocedures. For example, it is possible that amutation that produces unreduced embryosacs, in combination with a ftslfie mutation(perhaps in the heterozygous state), wouldresult in a viable form of apomixis. Althoughsuch experiments could be carried out withoutfurther knowledge of the genes involved, animportant step toward a controllable systemof apomixis will be the isolation andcharacterization of the mutations and theirwild type alleles. The mutant genes could betransferred to normal Arabidopsis via T-DNA,to determine what effect they have in abackground where the wild type allele is alsopresent. It would be interesting to determineif the ftslfie mutations are transmissible byfemale gametes and, if dominant, give rise toseeds with autonomous endosperm andembryos.Efforts to identify apomixis genes from Arabis,as well as other apomicts, should eventuallycome to fruition. An intriguing area of inquiry,when that comes to pass, will be the nature ofthe relationship between these naturalapomixis loci and any mutations conferringapomictic characteristics that have beenidentified in the related sexual species. Thecombined strategies of mutagenesis in sexualplants to induce apomixis, and in apomicticplants to identify natural apomixis genes,should eventually enable us to understand theregulation of these traits and to manipulatethem in the best interests of agriculture.ReferencesAarts, M.G.M., P. Carzaan, WJ. Stiekema, and A. Pereira. 1995. Atwaelement Enhancer· Inhibitor transposon system in Arabidopsisthaliana. Mol. Gen. Genet. 247: 555-64.Ahokos, H. 1977. Amutont of barley: Triploid indu

228 Uta P..utt aod Rod 5

Chapter 14Genetic Engineering of Apomixis inSexual Crops: A Critical Assessment ofthe Apomixis TechnologyTHOMAS DRESSELHAUS, JOHN G. CARMAN, AND YVES SAVIDANIntroductionAccording to projections, world populationwill increase from six billion people today toeight billion in 2020, stabilizing at 9-11 billionpeople around the middle of the 21'1 century(Lutz et al. 1997; Evans 1998; Toenniessen,Chap. 1). Profuse quantities of high quality andsafe food products will be required to feed thisgrowing population. At the same time, strongpressures are at work demanding that this foodbe produced in an environmentally friendlymanner, e.g., using less agrochemicals. InEurope, agricultural production has steadilyincreased while population has begun todecrease, resulting in an overproduction offood products. By contrast, the developingworld will need to produce two or three timesas much food as it does today (Toenniessen,Chap. 1). By 2020, cereal production, forexample, will need to increase by 41 %, and rootand tuber production by 40% (Spillane 1999).To meet this dramatically increasing demand,new plant varieties are needed that are bothhigher yielding and better adapted to specificclimatic conditions. Essentially, this challengemust be met without a significant expansionof agricultural area.Although less agricultural production will beneeded in the developed world, new products,so-called 'novel foods,' 'functional foods,''designer foods,' as well as renewable rawmaterials will soon gain more agriculturalmarket share. It is expected that most of thesenew products will be produced throughbiotechnology. Therefore, it is not surprisingthat the global market for agriculturalbiotechnology products is expected to increasefrom US$500 million in 1996 to US$20 billionwithin the next 15 years Games 1997).One biological process in particularapomixis---couldrevolutionize 2pt centuryagriculture in both developed and developingcountries. The harnessing of apomixis isexpected to launch a new era for plant breedingand seed production. Mastering apomixiswould allow (i) immediate fixation of anydesired genetic combination (genotypes, F lsincluded); (ii) propagation of crops throughseed that are currently propagatedvegetatively (seed is easier to transport and tosow); (iii) faster and less expensive plantbreeding and seed production (e.g., hybridseeds could be easily produced); (iv) a largerpool of germplasm to be used to create morelocally adapted varieties (once apomixis isintegrated into breeding schemes); and (v) acarryover of beneficial phytosanitary sideeffects through seed propagation, because veryfew pathogens are transferred through seeds(Grossniklaus et al. 1998a; Bicknell and Bicknell1999). Furthermore, exploiting apomixiswould allow breeding with obligate apomicticspecies (e.g., Pennisetum spec.), whereintrogression of new traits is currently verylimited (do Valle and Miles, Chap. 10), and theuse of male sterile plants for seed production.In tum, this would prevent the migration oftransgenes from crop plants to wild relatives.

230 1\omas Dresselhaus, Joh. G. (IJrmall, aid YVeJ SavidaftAll these advantages taken togetherundoubtedly would lead to large increases inagricultural production and prompted VielIe­Calzadaet al. (1996a) to coin the term"AsexualRevolution" to describe the potential impactof the technology.The possible economic benefits of thetechnology are also considerable. In rice,added productivity would total more thanUS$2.5 billion per year (McMeniman andLubulwa 1997). It is prOjected that the heterosiseffect alone would result in yield increases ofmore than 30% (Yuan 1993; Toennissen,Chap. 1). Of today's US$15 billion globalmarket in commercial seed, hybrid seedaccounts for 40% of sales (Rabobank 1994), afurther indication of the enormous economicpotential of apomixis for agriculturalenterprises.Unfortunately, scientific and economicpotential shed little light on the actualintricacies of how the genes involved inapomictic reproduction work. Many haveconcluded that the genes that control apomixisare also crucial for sexual development,indicating that apomixis is a short-circuitedsexual pathway (Koltunow et al. 1995;Grossniklaus, Chap. 12). The geneticengineering of apomixis, therefore, requires abetter understanding of both apomictic andsexual pathways of reproduction.In general, apomixis is thought to occur inpolyploid species (Asker and Jerling 1992),especially in the Rosaceae, Asteraceae, and inthe Poaceae (for review see Berthaud, Chap. 2).For most species in which apomixis has beendescribed, diplOids reproduce sexually, whilepolyploids of the same species are apomictic.Most natural apomicts reproduce throughfacultative apomixis (Asker and Jerling 1992;Berthaud, Chap. 2). The degree of apomicticreproduction is influenced by the geneticbackground, ploidy level, modifier genes, andthe environment. There is also a great diversityof apomictic behavior: nine types ofgametophytic apomixis have been describedin addition to sporophytic apomixis(adventitious embryony) (Crane, Chap. 3).Unfortunately, apomixis is not found in themost important cultivated crops, which couldbe a resultofcrop domestication, selection, andsegregation analysis (Grossniklaus, Chap. 12).There are three main options for theengineering of apomixis into sexual crops:(i) transfer the trait into crops from wild,naturally apomictic relatives throughnumerous backcrossings, (ii) screen sexualcrops for apomictic mutants, and (iii) de novosynthesize the apomictic trait directly intocrops. These approaches will be discussed inthe following pages.Transfer of the Apomixis Traitto Sexual CropsBreeding and Introgressionfrom Wild RelativesGenerally, breeding apomictic species is verydifficult, consequently, there have been only afew breeding programs, and these focused ona very limited number of tropical grass species.The basic structure of such breeding programsis described in this book, using Brachiaria asan example, an important forage grass in SouthAmerica, (do Valle and Miles, Chap. 10).Obligate apomicts cannot serve as maternalplants and breeding of such species is thereforeimpossible. The polyploid and highlyheterozygous nature of most apomictic plantsfurther complicates genetic analysis. Inaddition, controlled pollination is needed toanalyze reproductive behavior (methods aredescribed by Sherwood, Chap. 5). Additionaltechniques are needed to monitor reproductionbehavior in progeny plants of new varieties.Such techniques are described in this book byBerthaud (Chap. 2), Crane (Chap. 3), andLeblanc and Mazzucato (Chap. 9). Thetechniques described include chromosomecounting, flow cytometry, clearing and

G...Ii, Elgil..';"!!o/ Apomlxl. i. 5....1Cropo: ACr~iuII A.....me.t of the Apanixl. w ..1ogy 231squashing techniques, sectioning, molecularmarkers, and the "auxin test." Ultrastructuralstudies using electron microscopy (Naumovaand Vielle-Calzada, Chap. 4) reveal even moreinformation, but are very laborious, timeconsuming,and poorly suited to large-scaleprogeny analysis. Flow cytometry analysis ofseeds is a fast and easy tool and thus probablythe method of choice for first progeny testings.This is because large numbers of progenypopulations have to be produced andinvestigated at each generation in order toanalyze reproductive behavior (Matzk et al.2000; Savidan, Chap. 11).Several sexual crop plants are closely relatedto wild apomicts, and introgression of theapomixis trait through wide crosses hassuccessfully been performed with wheat,maize, and pearl millet (reviewed by Bicknell,Chap. 8; Savidan, Chap. 11). Nevertheless,there are some limitations: total male sterilitywas observed frequently in F jhybrids of widecrosses, representing a dead end once theapomixis trait is obligate. In wide crossesbetween Tripsacum and maize, fertile apomicticBC 4with less than 11 Tripsacum chromosomescould not be identified (Savidan, Chap. 11),resulting in maize lines devoid of agronomicvalue. Another disadvantage of this approachis that transfer of natural apomixis genes fromwild species into related sexual crops byintrogression is likely to remain limited to thosecrops that have apomictic relatives and so willnot be applicable to other species.Mutagenesis ApproachesMutagenesis approaches have been describedin great detail earlier in this book byGrossniklaus (Chap. 12) and Praekelt and Scott(Chap. 13). Therefore, we will discuss only themain conclusions here.The basis for all mutagenesis approaches is theassumption that apomictic reproductionpathways are developmental variations of thesexual pathway, thus a short-circuited sexualpathway. Mutant screens have therefore beendesigned to induce sexuality in apomicts andapomictic mutants in sexual plants by theinactivation of genes. Many mutants wereidentified as being defective in meiosis,megasporogenesis, and gametogenesis (forreview, see Yang and Sundaresan 2000;Grossnikiaus, Chap. 12). Mutant analysis ofmegagametogenesis, for example, suggeststhat a large number of loci are essential forembryo-sac development. Other mutants aredescribed as displaying autonomous embryoand/ or endosperm development. Thecorresponding genes have been recentlycloned. Mealfisl (medea/jertilization independentseed 1) is a gametophyte maternal effect geneprobably involved in regulating cellproliferation in the endosperm and alsopartially in the embryo (Grossniklaus et al.1998b; Luo et al. 1999). Fis2 shows a similarmutant phenotype and encodes a putativezinc-finger transcription factor (Luo et al. 1999).Autonomous endosperm development wasobserved in the fie (fertilization independentendospermlfis3) mutant. Mealfisl and fie/fis3display homology to Polycomb proteins(Grossnikiaus et al. 1998b; Ohad et al. 1999),which are involved in long-term repression ofhomeotic genes in Drosophila and mammalianembryo development (pirrotta 1998).The most important conclusion derived fromthe des~ription of these mutants is that all theelements of apomixis can indeed be inducedby mutations in sexual plants. In addition, it isobvious that more than one mutation will benecessary to obtain vital apomictic seeds insexual crops. Nevertheless, a combination ofsuch isolated genes could be used for knowngene approaches, but additional genes will beneeded to obtain fully developed seeds. Untilnow, most mutagenesis screens haveconcentrated on the partial or completeinactivation of the genes that are needed for

progression or inhibition of development.Future screens will also include activationtagging in order to induce genes under aspatial, temporal, or developmental regimethat differs from that in the sexual wild typeplants.Known Gene ApproachesKnown genes used for genetically engineeringthe apomixis trait should lead to the followingbiological processes:(1) avoidance and bypassing of meiosis(apomeiosis);(2) formation, ideally, of one functionalunreduced embryo sac within each ovule;(3) autonomous development of theunreduced egg cell by parthenogenesis;(4) development of a functionalendosperm-this could be autonomousor pseudogamous after fertilization of thecentral cell; and(5) an inducible/repressible system that isnecessary to switch between apomicticand sexual reproduction pathways,because sexuality and recombination willbe required for the introduction of newtraits into crops, which will result in newand improved plant varieties.Based on analyses of mutants in apomictic andsexual plant species, it is unlikely that theapomixis trait can be engineered using a singlegene. Tms is supported by the fact that in mostcases apomixis is facultative and that theproportion of apomictic progeny can beinfluenced by different factors, e.g., byenvironmental factors. Variability witmn thedifferent apomictic reproduction pathwaysfurther indicates that asexual seeddevelopment cannot be explained on the basisof a single gene.One possibility for engineering apomixis isbased on isolating the apomixis gene(s) fromnatural apomicts and inserting them intosexual crops. Molecular mapping of apomixisgenes and gene isolation by map-based cloningor transposon tagging (described byGrimanelliet aI., Chap. 6) are performed in variouslaboratories, but until now no apomixis genescould be isolated and markers still lie withincM distance. One major problem with severalapomicts is suppression of recombinationaround the apomixis loci (e.g., Pennisetum andTripsacum; Grimanelli et aI., Chap. 6). Inaddition, apomictic species do not belong tothe classical model plant species, and thereforepositional cloning is difficult because of therelatively low number of available markers,which are needed to "walk" to the apomixisgene(s). Transposon tagging is not possible formost apomicts (TripsQwm is an exceptionbecause it can easily be crossed with maizelines carrying active transposon elements), andfor the near future, T-DNA tagging will remainrestricted to dicotyledonous apomicts such asHieracium, which are accessible toAgrobacterium tumefaciens transformation(Bicknell, Chap. 8). Moreover, it is also possiblethat because of the polyploid nature of naturalapomicts, no such phenotype exists.Known genes/promoters from sexual speciesthat could be used for genetic engineeringinclude those involved with (i) ovule development,(ii) initiation of meiosis, (iii) femalegametophyte development, (iv) parthenogenesis,and thus autonomous embryodevelopment, and (v) initiation of endospermdevelopment. Grossniklaus (Chap. 12)specul~tes that the genes controlling apomixisare under relaxed or aberrant temporal and/or spatial control, thus developmentalcheckpoints and feedback mechanisms may beignored or altered, leading to precociousdevelopment of the megaspore mother celland/or the unreduced egg cell.Ovule- and nucellus-specific genes/promotersare now available as tools (see Tables 14.1 and14.2). The molecular control of meiosis is wellcharacterized in yeast (Vershon and Pierce2000) and some animal systems, e.g.,

G...tk bgltoeerlog.' Apomixis 10 Seu.1 Cr.ps: ACrilkal A.........I., tile Apomixis Tedo..logy 233Caenorhabditis elegans (Zetka and Rose 1995),and many genes have been isolated andcharacterized during the last few years. Muchless is known about the genes involved in plantmeiosis. However; the first homologs to yeastmeiosis genes were recently isolated (reviewedby Grossniklaus, Chap. 12), and many meiosismutants remain available for furthercharacterization (e.g., in maize and Arabidopsis;Neuffer et al. 1997; Yang and Sundaresan 2000).Genes that are expressed during the inductionof meiosis have been identified in lily(Kobayashi et al. 1994). Most work on meiosisin plants has been accomplished throughinvestigating male meiosis, but for geneticengineering, female meiosis genes will be ofparticular interest. Some genes involved withfemale gametophyte development have beenidentified, of which some are specificallyexpressed in different cells of the femaleTable 14.1 Examples of isolated genes and their promoters that might be useful as tools for de novosynthesis of the apomixis trait in sexual cropsProcess to be monipulatedGene (expression/function)'Apomixis genes'not isolated yet (?)Ovule and nucellus-specific target gene expressionFBPI promoter (ovule-specific)DEFH9 promoter (ovule-specific)WM403 promoter (nucellus-specific)Nucellin eDNA (nucellus-specilic)Prevention of meiosis/ apomeiosisdiverse cDNAs (early meiosis-specific)pAWlL3 eDNA (early meiosis-specific)DMO gene (MMC* -specific)5YNl gene (chrom. condensation/pairing)Parthenogenesis (autonomous embryo development)5ERK gene (competence to form embryos)LfCl gene (competence 10 form embryos)BBM1gene (competence to form embryos)ImE51-4 promoter (embryo sac-specific)(Autonomous) endosperm developmentMWF/51 gene (suppressor)FI52 gene (suppressor)FJEIF/53 gene (suppressor)ImE51-4 promoter (embryo sac-specific)ImprintingMfTl als (hypomethylationJIndudble/ repressoble systemsSteroid-inducible promoterCopper-inducible promoterTetracycline-inducible!-inoctivalable promolerEthanol-inducibele promoter-MMC: Mega- and Microlpore molher cellI.(Origin)(Petunia)(Anthirrhinum)(waler-melon)(barley)(lily)(wheat)(Arabidopsis)(Arabidopsis)(carrot, Arabidopsis)(Arabidopsis)(Brossico, Arobidopsis)(moize)(Arobidopsis)(Arabidopsis)(Arabidopsis)(maize)(Arobidopsis)(mammals)(yeast)(bacterium)(fungus)ReferenceColombo el 01., 1997Rotino el 01., 1997Shen et 01., unpublishedChen and Foolad, 1997Kobayashi et 01., 1994Ji and langridge, 199~Klimyuk and Jones, 1997Bai et 01., 1999Schmidt et 01., 1997lotan el 01., 199BBoutilier et 01., unpublishedAmien and Dresselhaus, unpublishedGrossniklaus et 01., 1998bluo et 01., 1999luo et 01., 1999Ohad et 01., 1999Amien and Dresselhaus, unpublishedAdams el 01., 2000Vinkenoog et 01., 2000Schena et 01., 1991Melt et 01., 1993Weinmann et 01., 1994Caddick el 01., 1998

234 T1lo.... Dr......... Jo~. G. (..-.. ood Yv•• Sovida.Table 14.2 Examples of patents linked with the engineering of the apomixis trait in sexual crops.Sources: Intellectual Property Network (hflp://www.delphion.com). European Patent Office (http://ep.dips.org/dipsl, and Bicknelland Bicknell (1999).Apomixis technologyPatent number'(Publication date)Breeding strategiesW08900810(Feb. 9. 1989)CN1124564(June 19. 1996)US5710367(Jon. 20.1998)W0971 0704(Sep. 22, 1998)W09833374(Aug. 6, 1998)W0007434(Feb. 17,2000)Stimulation of apomictic reproductionEP0127313(Dec. 5, 1984)SU 1323048(Ju~ 15, 1987)US4818693(April 4. 1989)US5840567(Nov. 24, 1998)Title (and content)De noyo synthesis of apomixis (genes and promoters)W09743427(Nov. 11, 1997)W09808961(March 5, 1998)W09828431(July 2, 1998)US5792929(Aug. 11. 1998)W09836090(Aug. 20, 1998)W09837184(Aug. 27, 1998)US5907082(May 25. 1999)W09935258(July 15, 1999)W09953083(Oct. 21, 1999)W0024914(Moy 4, 2000)Asexual induclion Df heritoble mole sterility and apomixis in plants(use of mole sterility factors).Hybrid vigor fixing breeding process for rice opomixis(breeding ond selection strategy).Apomictic moize (intragression of apomixisfrom Tripsawmlo moize).Apomixis for producing true-breeding plont progenies (introgressionof apomixis from Pennisetum squamulatum to cullivars).Methods for producing apomicitic plants(breeding program).Novel genetic material for transmission intomaize (introgression of apomixis from Tripsawm).The production of haploid seed, of doubled haploids ond ofhomozygous plant lines therefrom (causing apomixis by applyingon apomixic agent).Stimulator of floral apomixis(no file available).Methods and materials for enhanced somalicembryo regeneration in the presence of auxin.Simplified hybrid seed production by latent diploid parthenogenesis andparthenote cleavage (induced by controlled environmental conditions).Production of apomictic seed (using a SERK gene forembryogenic potential).Endosperm and nucellus specific genes, promoters anduses thereof.Transcriptional regulation in planls(using a meiosis specific promoter).Plants with modified flowers (modifying flower cells aftertransformation with foreign DNA).Meonsfor identifying nucleotide sequencesinvolved in apomixis (isolation and modification of sexual genesfor the expression of apomixis in Gramineae).Leafy cotyledon 1genes and their use (using embryo specific genesand their promoters).Ovule-specific gene expression(using ovule·specific genes).Nucleic acid markers for apospory-specificgenomic region (from the genus Paspalum).Seed specific polycomb group gene andmethods of use for same (using repressors of embryo andendosperm development).Apomixis conferred by expression of SERKinteracting proteins (see above W097434271.• WOo US. EP. CN and SU refer 10 World polenls, US-. European. Chinese ond former Sowjet Union polents.Applicant{s)Maxell Hybrids INCChen J.USDAUSDAUniversity of Utah StaleEubanks M.w.Rohm & HaasPollav Selskokhaz IGNikitskijPGSUniversity of CaliforniaNovartis and inventorsDaan. D.N.P., Olsen,O.-A. and Unnestod, CJohn Innes CentreInnov. LTD and inventorsPGSIRD and CIMMYT-ABCUniversity of CaliforniaUniversity of CaliforniaUniversity of GeorgiaResearch Found. INCCold Spring Harbor Lob.Novar1is

Go..'k 1091....i.9 .1 Aporo" io S....ICr.ps: A(,ilkal A..........'.I ... Apomlxi. Tech.logy 235gametophyte (Grossniklaus, Chap. 12; Cordtsand Dresselhaus, unpublished results).Through the use of mutant approaches(Vollbrecht and Hake 1995; Drews et a!. 1998;Yang and Sundaresan 2000; Grossniklaus,Chap. 12; Praekelt and Scott, Chap. 13), we cananticipate that many more genes involved infemale gametophyte development will soon beisolated. Gene trap screens such as T-DNAinsertional mutagenesis, transposonmutagenesis, and enhancer detection(Grossniklaus, Chap. 12) are very powerfulmolecular tools for isolating the correspondinggenes and!or their promoters from sexualmodel plants like maize and Arabidopsis.Further tissue!cell-specific genes and theirpromoters will be isolated by transcriptprofiling methods (e.g., Liang and Pardee 1992;Welford et al. 1998; Matsumura et a!. 1999) andfrom tissue!cell-specific cDNA libraries (e.g.,Dresselhaus et a!. 1994; Diatchenkoetal. 1996).Initial attempts have been made to comparegene expression profiles between sexual andapomictic lines within the same species. A fewgenes that are specifically expressed in theovules of either sexual or apomictic lines wereisolated (Vielle-Calzada et al. 1996b). Thesegenes may eventually be useful tools forinducing apomictic development in sexuallines or sexual development in apomictic lines.Parthenogenetic embryogenesis fromunreduced eggs is the next required step forsuccessfully engineering the apomixis trait.Whether this will occur spontaneously oncethe egg is diploid has yet to be shown. Quarinand Hanna (1980) found that doubling a sexualdiploid Paspalum line generated a tetraploidthat was facultative aposporous, thusunreduced egg cells developed parthenogeneticallyinto embryos. Spontaneousparthenogenetic development was observed ata low frequency in maize (Chase 1969; Bantinand Dresselhaus, unpublished results). Wheatlines have been described that produced up to90% parthenogenetic haploids (Matzk et al.1995). Very little molecular data concerningparthenogenesis are available for higherplants. One protein (a-tubulin) was identifiedwhose expression is associated with theinitiation of parthenogenesis in wheat (Matzket a!. 1997). And auxin (2,4 D) treated sexualeggs from maize can be triggered to initiateembryo development at a low frequency'r.(Kranz et al. 1995), however, the molecularmechanism is not understood. Three geneswere used to successfully initiate the formationof embryo-like structures on vegetative tissue(lec1: leafy cotyledon 1, Lotan et a!. 1998; andbbm1: baby boom1, Boutilier et a!., unpublishedresults) or to enhance the rate of somaticembryos in culture (SERK1: somaticembryogenesis receptor-like kinase 1, Hecht et a!.,unpublished results), respectively. It remainsto be demonstrated whether these genes arealso useful for inducing embryo developmentin reproductive cells.Parthenogenesis may also arise as a functionof timing, taking into account thatparthenogenetic embryogenesis is usuallyinitiated before anthesis. In contrast to sexualeggs, parthenogenetic eggs (e.g., Pennisetumciliare and wheat) contain ample amounts ofribosomes and polysomes and a large numberof cristae in mitochondria, thus suggesting ahighly active metabolic status prior topollination (Naumova and Vielle-Calzada,Chap"4; Naumova and Matzk 1998). Incontrast to sexual eggs, degeneration ofsynergids in aposporous Pennisetum ciliarefemale gametophyte was precocious andrapid. In addition, a complete cell wall aroundthe eggs was already generated before thearrival of the pollen tube (Vielle et al. 1995). Inmaize, zygotic gene activation (ZGA), theswitch from maternal to embryonic control ofdevelopment, occurs soon after fertilization(Sauter et al. 1998; Dresselhaus et al. 1999;Bantin and Dresselhaus, unpublished).Precocious expression of zygotic genes beforepollination!fertilization could thus eventually

236 TIoo.... D........., J.~. G. (annal, God rYe' SaYId.be used as a tool to induce parthenogeneticdevelopment of sexual eggs, and perhapsthose samegenes might be useful for inducingendosperm development. Although theexistence of repressor molecules that preventunfertilized eggs from initiating embryodevelopment has not been proven, it isreasonable to postulate their reality. Onceisolated, they might be a useful tool forengineering parthenogenetic embryodevelopment as a component of apomixis.Induction of endosperm development willprobably be the biggest obstacle to the utilizingapomixis in sexual crop species (discussedfurther under "Main Limitations").Nevertheless, an in vitro system forendosperm development in maize wasreported recently (Kranz et al. 1998), providingimpetus to molecular investigations aboutgene expression and regulation during theearliest steps of endosperm development.Transformation andInducible Promoter SystemsTremendous progress has been made in plantgenetic engineering since the first reports ofsuccessful plant transformation appeared inthe early 1980s, and many commerciallyrelevant genes have been transferred to cropplants (Christou 1996). Agrobacteril/mmediatedtransformation has been the methodof choice for introducing exogenous DNAintodicotyledonous plants. Agrobacteril/mtransformation has proven difficult withcereals, and consequently, alternative methodssuch as particle bombardment have beenemployed. Nevertheless, because Agrobacterillm-mediatedgene delivery offers manyadvantages (easy protocols, often low- or evensingle-copy integrations, mostly full-lengthintegration of transgenes, short or no tissueculture period), considerable effort has beendedicated to establishing this method forcereals (Komari et al. 1998). Agrobacteril/mtransformation of rice is now routine, whilesuccessful transformation of maize and wheathas also been reported (Ishida et al. 1996;Cheng et al. 1997). Even so, particlebombardment of wheat and maize immaturescutellum tissue remains the most widely usedmethod in most public laboratories. Relativelyefficient transformation systems are nowavailable for all major crops as well as someforage grasses (Spangenberg et al. 1998).Development of transformation systems forapomictic species is in progress, andtransformation protocols for pearl millet willbe established once interesting apomixis genesbecome available (P. Ozias-Akins, personalcomm.). Transformation of Brachiaria andTripsacum are foci ofapomixis programs at theInternational Center for Tropical Agriculture(CIAT) and the International Maize and WheatImprovement Center (CIMMYT), respectively.A major problem related to transgene activityis the instability of expression Oorgensen 1995;Matzke and Matzke 1995). Often inactivationof transgene expression is accompanied by anincrease in DNA methylation (Meyer 1995). Inaddition, transgenes may be integrated inhypermethylated chromosomal regionsdisplaying a spatial and temporal change ofmethylation during plant growth anddevelopment (position effect). Transgenes withhomologous sequences to endogenous genesmay be silenced through the cosuppressioneffect Oorgensen 1995; Matzke and Matzke1995). All the same, plants stably expressingthe transgenes can be selected overgenerations, although this is time-consumingand expensive. Suggestions have been madeas to how vectors used for genetictransformation can be optimized in order tominimize the cosuppression effect (Meyer1995). Single-eopy integration of transgeneswill be enabled by the deployment ofAgrobacterillm-mediated gene delivery. This intum will increase the rate of plants that stablyexpress the transgenes. Gene targeting byhomologous recombination, i.e., the

G...1i< bgl...n.g.f A,oooixis io S.loaI Crops: ACritkal A.....'_t.f tH Aponll. TeO..1ogy 237generation of null mutants, is probably theideal way to stably silence genes. Thedeployment of this approach, however, is stillrelatively limited for higher plants (Puchta1998). An alternative is homology-dependentgene silencing (HOGS; for review, see Kooteret a!. 1999), especially through the use ofdouble-stranded RNA (RNAi: RNAinterference technology) as a template for genesilencing (Bass 2000). Gene silencing at ratesup to 100% was reported with transgenicplants using the latter approach.Inducible/repressible systems are necessary toengineer the apomixis trait, because geneticrecombination through sexual crossing willalways be required for the introduction of newtraits into crops. In a panel discussion withindustrial representatives during the ThirdEuropean Apomixis Workshop (April 21-24,1999, Gargnano, Italy), it became very clear thatinducible systems for engineering theapomictic trait are highly desired (http:/ /www.apomixis.de;seeworkshops).mainlybecause they serve as a natural means ofprotecting intellectual property rights (see"Intellectual Property Rights," this chapter).The question is whether such systems arepractically possible, given the problemsencountered with the application ofgametocides. Various chemical induciblesystems have been reported, e.g., thetetracycline inducible/inactivatable promotersystem, and steroid-, copper- and ethanolinducible promoter systems (for review, seeGatz and Lenk 1998). Whether these systemsare applicable and acceptable for use underfield conditions is doubtful; sprayingantibiotics, steroids, and heavy metals isenVironmentally unacceptable. Ethanolsystems might offer an alternative. Most ofthese systems, however, are leaky and havesome background activity, or they may be toosensitive. In addition, there is the question ofhow homogeneously the induction works indifferent organs, especially in embedded cellslike megaspore mother cells and the cells ofthe embryo sac, which are the main target cellsfor the genetic engineering of differentapomixis components. Seed producersanticipate efficiency rates as high as 99% forsuch systems (http://www.apomixis.de; seepanel discussion during the Third EuropeanApomixis Workshop). EXisting systems,therefore, must be optimized, or preferably,new systems using natural, easilybiodegradable, and harmless chemicals asinducers must be developed to satisfy seedproducer demands and environmentalnecessities.Main LimitationsPerhaps the biggest obstacle to geneticallyengineering apomictic grain crops is thatfertilization of the central cell is likely to berequired because of dosage effects (Birchler1993; Savidan, Chap. 11) and becauseautonomous endosperm development occursat low frequenCies in cereals. A balancedmaternal:paternal genome ratio (2m:1p) is anabsolute requirement for endospermdevelopment in cereals-(Birchler 1993). In mostcases, deviation from this ratio leads to embryoabortion or seeds with diminished fertility(Birchler 1993; Praekelt and Scott, Chap. 13).In contrast to cereals, Scott et al. (1998) haveshown that in Arabidopsis, 2m:2p, 4m:1p and4m:2p ratios are allowed. Also observed inmost pseudogamous apomicts are ratios of4m:1p and 4m:2p. In apomictic lines of themaize relative Tripsacum, Grimanelli et a1.(1997) identified 2m:2p, 4m:1p, and 8m:1pratios. Imprinting of gametic nuclei is thegenetic reason behind this phenomenon: oneset of alleles is silenced on the chromosomescontributed by the mother, while another setis silenced on the paternal chromosomes. Eachgenome thus contributes a different set ofactive alleles (Vinkenoog et al. 2000; Allemanand Doctor 2000). A few imprinted loci have

238 T1tonoas Dr......... Joino G. 'arJnlIlI, aod Yv.. S.vidaobeen investigated in plants (e.g., Kinoshita etal. 1999; Vielle-Calzada et aI. 2000; Allemanand Doctor 2000; Crane, Chap. 3), but we arejust beginning to understand the molecularmechanisms underlying these processes.Nevertheless, the combination of maternalhypomethylation in combination with a lossof fie function was recently shown to enablethe formation of differentiated endospermwithout fertilization in Arabidopsis (Vinkenooget aI. 2000). It remains to be demonstratedwhether this approach is also feasible forcrops, especially cereals, but it represents apromising step in assembling the manycomponents needed to engineer apomixis intosexual crops.Another obstacle that needs to be overcome isthe relatively high number of genes/promoters that are required; in addition toinducible/repressible systems, it is likely thatthe precise and controlled interaction of manygenes will have to be engineered. In naturalapomicts, genes from different chromosomesare required for the expression of apomicticreproduction pathways. Blakey et aI. (1997)have shown that in apomictic Tripsacum, genesrequired for seed set are located on at least fiveTripsacum linkage groups, which are syntenicto four maize chromosome arms. Sherwood(Chap. 5) observes that the expression ofapospory requires the dominant allele of amajor gene or linkat and that the degree ofapomixis may be further influenced by manyother genes (e.g., modifiers). Fewer data areavailable for diplospory, but in this case aswell, a single master gene or a number ofgenesthat behave as a single locus may be requiredfor the expression of apomixis. The technicaldifficulties of introducing multipl~ geneswithin a single transformation event weresuccessfully resolved recently usingAgrobacterium-transformation with rice (Ye etaI. 20(0). Four genes were integrated on oneconstruct; by crossing transgenic lines carryingother transgenes, a whole biosyntheticpathway was engineered into rice endosperm(Ye et aI. 2(00).To sum up, our understanding of themolecular regulation of apomictic andamphimictic reproduction pathways in crops,especially cereals, is still in its infancy, and thus,due to the complexity of these biologicalprocesses, modifying or controlling thepathways will probably not be achievedwithin the next five years.Intellectual Property RightsIntellectual property rights (IPR) are a meansof promoting commercially relevantinnovation and for sharing resources. The IPRowner obtains the right to use the intellectualproperty (IP) exclusively, license it, or not useit at all for a limited period (e.g., 20 years). Inagricultural biotechnology and plant breeding,both scientific knowledge and its commercialapplications are increasingly being claimed bycompanies, but also by public institutions suchas universities and research centers (Spillane1999). With hundreds of millions of dollarsinvested every year in plant biotechnology andbreeding research, companies need effectiveIP protection to provide an incentive forniaking large research investments. Theseresearch results offer enormous benefits foragrochemical and seed companies, farmers,and the society as a whole. In the United States,cIPR include (i) geperal utility patents, (ii) PlantVariety Protection (UPGV), and (iii) plantpatents for asexually repro.duced plantsUondle 1999).Given this context, it is not surprising that IPRfor methods and .genes/promoters that areuseful for the genetic engineering of apomixishave been claimed (Table 14.2). Most of thepatents were filed during the last five years,probably because of improvements in plaJiltgene technology and in recognition of theenormous economic potential of utilizing

Ge••1i< Eo,i.....iog of A,....I. io s...aI Cr.,.: ACrlliuol As........' of Iloo Aponolxl. WoolotY 239apomixis for crop improvement. Theseapomixis patents raised concerns about the useof apomixis technology. The RuralAdvancement Foundation International(RAFI), a nongovernmental organization,recently expressed the concern that apomixisIPR could wind up in the hands of only a fewdominate global agrobusiness players, and thatfarmers in both developed and developingcountries might become totally dependent ontheir seed products. Other concerns are thatgenetic diversity could significantly declineand that developing countries will not haveaccess to this technology because they will beunable to afford the required rights andlicenses (RAFI 1998). The latter concern isshared by leading apomixis researchers andwas formalized in 1998 in the BellagioApomixis Declaration (for full text, see http:/ /billie.harvard.edu/apomixis). Signatories tothe declaration were interested in how todevelop novel approaches for generating theenabling technology, and how to patent andlicense it. Currently, patents related to apomixisenabling technology are dispersed amongmany parties (Table 14.2). Furthermore, it isexpected that the number of patents willgreatly swell as numerous public and privateresearch institutions continue investigatingdifferent aspects of apomictic and amphimicticreproduction pathways using different speciesand approaches (see e.g., Bicknell and Bicknell1999).Another negative impact stemming fromapomixis patents is that communication ofresearch results to the scientific community iseither delayed until patents have been filed orthey are Simply not communicated at all. AWidespread phenomenon in today'sbiomedical research is that while IPR isgrowing rapidly, scarce resources are poorlyutilized because too many patent owners areblocking one another. Paradoxically, more IPRmay lead to fewer useful products for theimprovement of human health (Heller andEisenberg 1998). In regards to apomixis, it isunlikely that the situation will change in thenear future because it is still pOSSible to file verybroad apomixis patents.The question of whether farmers in developingcountries will get access to disclosed apomixistechnology remains unanswered. One canhope that many of the relevant patents will besecured by public organizations such as theConsultative Group on InternationalAgricultural Research (CGIAR) and otherpublic institutions (see Hoisington et al. 1999),thus giving interested parties in developingcountries the possibility of acquiring freeaccess to this powerful technology. Certainly,the public image of the big agrobusinessplayers would benefit from freely licensing thetechnology to CGIAR institutions or directlyhelping farmers in developing countries usethis technology. The bulk of profits, after all,will be earned in the more developedcountries. Introducing the apomixis trait intolocal varieties would give farmers indeveloping countries access to powerful andproductive hybrid technology (Hoisington etal. 1999). To some extent, these farmers shouldhave the right to save seed for subsequentreplanting, thus allowing them to significantlyincrease their crop yield and personal income.Risk Assessment StudiesRisk assessment research and studies relate tothe use and or release of genetically modifiedorganisms (GMOs) into the environment. Sincethe first release of genetically m·odified plants(GMPs) some twelve years ago, many shorttermstudies have been conducted (de Vries1998). Short- and long-term risk assessmentstudies are also needed to evaluate theenvironmental implications ofnovel apomicticcrops. One key issue for investigation iswhether the apomixis trait can move to thelandraces and wild ancestors of food crop

240 no- P,ossdlaos, Jo. G.'-.... Y.os ScrrioI..plants, and if so, what would be the impact.This issue is especially important in the centersof origin for the crop plants. Furthermore, theissue of how apomixis might affect geneticdiversity, and whether it would increase ordecrease monoculture farming needs to beexplored. Based on field studies on herbicideand/or insecticide resistant plants, we canprobably expect engineered apomixis genes tomove through vertical gene transfer (transferof a gene from plant to plant via sexualreproduction/pollen) (Lutman 1999). The rateof horizontal gene transfer (asexual gene flowbetween organisms) is relatively low and therisk negligible, however, microbiological riskassessment studies in this area could be useful(Syvanen 1994). Given our current knowledge,it appears unlikely that microorganisms couldgain some advantage over wild relatives afteruptake of apomixis genes.Ifapomixis is controlled by multiple genes, theprobability of diffusing this trait to wildrelatives is extremely low. The transfer ofseveral genes to a wild plant should lower itsfitness to a level unacceptable for survival inthe wild (Berthaud, Chap. 2). If apomixis iscontrolled by a single gene, which would resultin obligate apomictic wild races, these raceswould lose their potential to evolve. Ifdominant, an apomixis gene could rapidlybecome fixed in an outcrossing sexualpopulation. Therefore, in theory, apomixistransgenes could possess advantages thatmight result in the uncontrollable spread ofthe transgenes (van Dijk and van Damme2000). Inducible apomictic systems and malesterility might circumvent these problems.Nevertheless, the described possibilitiesindicate that risk assessment studies andresearch to investigate the ecologicalimplications of novel apomictic crops (onceavailable) to the environment are an absolutenecessity. In addition, socioeconomic studieson the positive and negative implications ofthis technology for breeders, seed companies,and farmers-in both developing and developedcountries (see also IPR) will be required, andthe research results should be communicatedto all potential users.SummaryThe extensive introduction of apomixis intosexual crops will undoubtedly rely on geneticengineering, as we anticipate that morecandidate genes (especially regulatory genesand tissue/cell-specific promoters) andenabling techniques will be identified anddeveloped in the near future. Transformationtechnology for all major crops is now availableand inducible systems are currently beingdeveloped and optimized, allowing the controlof transgene expression and activity evenunder field conditions. Adventious apomixisusing already described or novel genes underthe control of ovule-, nucellus- or archesporespecificpromoters is probably the easiest wayto engineer the apomixis trait. Plant breedersand seed producers would like to generateinducible obligate mitotic diplospory incombination with autonomous endospermdevelopment. The latter is probably the mostdifficult aspect of engineering apomixis,espeCially for cereals such as wheat, rice, andmaize, because of dosage and imprintingeffects.Although apomixis is a hot topic in plantresearch, our current understanding of bothapomictic and amphimictic reproductionpathways in higher plants is still extremelylimited. The economic potential of apomixismight provide the impetus to bring apomicticcrops to the marketplace, and in the process itmay well contribute significantly to our futureunderstanding of the molecular regulation ofthe many different sexual and apomictic plantreproduction pathways.International and interdisciplinary approachesand efforts are now needed to study andmanipulate seed reproduction. It will be

Ge..tk Eagloeeri.g .1 Ap.mlxi. io Se..al Ct.,.: A(ritkal A.........' .1 the Apamlli. Tedooology 241necessary (i) to characterize the genetic reproduction through seeds in apomicticregulation of apomixis and isolate the systems and sexual crops," coordinated by T.responsible genes, (ii) to analyze the genetic Dresselhaus). In 1999, a transatlanticand molecular bases of sexual reproduction consortium was initiated between two publicand to isolate the corresponding genes, and institutions (CIMMYT and IRD) and three(iii) to produce the tissue/cell-specific and private companies (pioneer Hi-Bred, Novartis,inducible/repressible promoters that will be and Group Limagrain). This is just a beginningneeded to control the expression of the target and more concerted projects are needed ingenes. Concerted international research efforts order to reach the ambitious aim ofhave been made in Europe aimed at manipulating the apomixis trait in crops.understanding apomictic and sexualApomixis technology will offer many excitingreproduction pathways in order to developopportunities for the agriculture of the 21 sftools for the manipulation of the apomicticcentury, and indeed many patents alreadytrait (e.g., an E.U. Research Technology andhave been filed with many more yet to come.Development (RTD) project entitled "TheIt is critically important that these patents bemanipulation of apomixis for theheld and used for the good of all. Publicimprovement of tropical forages," coordinatedinstitutions in particular must safeguard theby M. D. Hayward; a RTD project entitledaccess of developing countries to these"Apomixis in agriculture: a molecularenabling technologies. In all likelihood,approach," coordinated by M. van Lookerenconstraints to the broad and generous use ofCampagne; and a Concerted Action Projectapomixis technology will be political andentitled "Introducing and controlling asexualeconomic rather than technical in the future.ReferencesBlakey, CA., CL Dewald, and S.L Goldman. Colombo, L, 1. Franken, A.R. Von der Krol, PE.1997. Co-segregation of DNA markers with Wittich, HJ. Dons, and G.C Angenent.Adams, S., R. Vinkenoog, M. Spielman, H.G.Tripsacum fertility. Moydica 42: 363-69. 1997. Dawnregulotion of ovule-specificDickinson, and RJ. Scott. 2000. Porent-of­Caddick, M.X., AJ. Greenland, I. Jepson, K.P.MADS box genes from petunia results inorigin effects on seed development inKrouse, N. Qu, K.V. Riddell, M.G. Solter, W. maternally controlled defects in seedArabidapsis thaliana require DNASchuch, U. Sonnewold, and A.8. Tomsett. development. Plant Cell 9: 703-15.methylation. Development 127: 2493­1998. An ethanol inducible gene switch for Diatchenka, l., Y.·F. Chris Lou, A.P Campbell, A.2502. plonts used to manipulate carbon Chenchik, EMoqodom, B. Huang, S.Alleman, M., and J. Dodar. 2000. Genomicmetabolism. Not. Biotechnal. 16: 177-80. Lukyanav, K. Lukyanaw, N. Gurskaya, E.D.imprinting in plants: observations andChose, S. 1969 Monoploids and monoploid. Sverdloc, ond PD. Siebert. J996.evolutionary implications. Plant Mol. BioI.derivatives of maize (leo mays U.Suppression subtractive hybridizotian: A43: 147-61.Botanical Review 35: 117-fJ7.method for generating differentiallyAsker, $.E., and LJerling. 1992. Apomixis inChen, E, and M.R. Faolod. 1997. Molecularregulated or tissue-specific eDNA probesPlants. Boca Raton, Florida: CRC Press.organization of agene in barley whichand libraries. Proe. Notl. Acad. Sri. (USA)Boi, X., B.N. Peirson, EDong, CXue, ond CA.encodes aprotein similar to aspartic 93: 6025-30.Mokoroff. 1999.lsolotion ondpralease and its specific expression in Dresselhaus, 1, S. Cordts, S. Heuer, M. Souter, H.characterization of SYNI, a RAD21·likenucellar cells during degenero1;on. Plant liirz, and E. Kronz. J999. Novel ribosomalgene essential for meios~ in Arabidopsis.Mol. BioI. 35: 821-31.genes from maize are differentiallyPlant Cell 11: 417-30.Cheng, M., J.E. Fry,S. Pong, H. Zhau, CM.expressed in the zygotic and somatic cellBoss, B.L 2000. Double-stranded RNA as a Hironaka, D.R. Duncon, 1W. Conner, ond Y. cycles. Mol. Gen. Genet. 261 : 416--27.template for gene silencing. CellI 01:Won. 1997. Genetic transformation of Dresselhaus, 1, H. Uirz, and E. Kranz. 1994.235-38.wheat medioted by AgrobocteriumRepresenlolive cDNA libraries from fewBicknell, R.A., and K.B. Bicknell. 1999. Who willtumefariens. Plant Physial. 115: 971-80. plant cells. Plant 1. 5: 605-10.benefit from apomixis? Biotechnology andChristou, P1996. Transformation technology. Drews, G.N., D. Lee, and CA. Christensen. 1998.Development Monitor 37: 17-21.Trends Plant Sci. 1: 423-31.Genetic analysis of female gametophyteBirchler, JA 1993. Dosoge analysis of moizede Vries, G.E. 1998. Post, Present And Future development and function. Plant Cell] 0:endosperm development. Annu. Rev.Considerations In Risk Assessment When 5-17.Genet. 27: IBI-204.Using GMOs. Billhaven, The Netherlands: Evans, L.1 199B. Feeding the Ten Billion: PlantsCommission Genetic Modificotion.and Population Growth. New York:Cambridge University Press.

242 1110""" 0,........, M. G. c,.._ IIIld rYeS SavldooGalz, C, and I. Lenk. 1998. Pramalersthalrespond to chemical inducers Trends PlantSci. 3: 352-58.Grimanelli, D., M. Hernandez, E. Pero"i, and Y.Savidan. 1997. Dosage eHects in theendosperm of diplosparaus opamicticTripsacum (Pooceael. Sex. Plant Reprod.10: 279-92.Grossniklaus, U., A. Kallunow, and M. vonLaakeren Campagne. 19980. Abrightfuture for apomixis. Trends Plant Sci. 3:415-16.Grollniklous, U., J -Po Vielle-Colzada, M.A.Hoeppner, and W.B. Gagliano. 1998b.Malernal conlrol of embryogenesis byMEDEA, a Polycomh group gene inArobidopsis. Science 280: 446--50.Heller, M.A., ond RS. Eisenberg. 1998. Canpalents deter innovalion? The anticommonsin biomedical research. Science 280: 698­701.Hoisington, D., M. Khoirallah, 1 Reeves, J-M.Ribaut, B. Skovmand, S. Tabo, and M.Warburton. 1999. Planl genefic resources:What can they contribule toward inueasedcrop productivity. Proe. Natl. Acad. Sci.(USA) 96: 5937-41Ishida, Y., H. Saito, S. Ohta, Y. Hiei, 1 Komori,and 1 Kumashira. 1996. High eHicienttransformation of maize (lea mays IImediated by Agrobaderium tumefadens.Nature Biotechnology 14: 745-50.James, C1997. Global StatUI of TtransgenicCraps in /997. ISMA Briek No.5. Ithoco,New York: ISMA.Ji, L-H., ond P.langridge. 1994 An ear~ meiosiscDNA done from wheol. Mol. Gen. Genet.243: 17-21Jondle, RJ. 1999. P1anl breeders' rights andplanl biotechnology. In A. Altman, M. Ziv,and S. Izhor (edl.!, Plant Biotechnology andIn Vitro Biology 01 the 2/st (entury.Dordrecht, The Netherlands: KluwerAcodemic Publishers. Pp. 747-49Jorgensen, RA 1995. Co-supprellion, Rowercolor patterns, and metostable geneexpression Ifates. Science 26B: 6B6--91.Kinoshita,l, R. Yadegori, JJ. Harada, R.B.Goldberg, and R.L Fischer. 1999.Imprinting of the MEDEA Polycomb gene inthe Arabidopsis endosperm. Plant (ell 11 :1945-52.Klimyuk, V.I., and JD. Jones. 1997. AtDMG, theArabidopsis homologue of the yeasl DMGgene: characterization, transpason·inducedallelic varialion and meiosis-ossociatedexpression. Plant 1. 11: 1-14.Kobayashi, 1, E. Koboyashi, S. Sato, Y. Ha"a, N.Miyajima, A. Tanaka, and S. Tabata. 1994.Characterization of cONAs induced inmeiotic prophase in Ii~ miuospores. DNARes. 1: 15-26.Kallunow, A.M., R.A. Bicknell, and A.M.Choudhury. 1995. Apomixis: Molecularstrategies for the generation of genetical~identical seeds without fertilization. PlantPhysial. 108: 1345-52.Komori, 1, Y. Hieri, Y. Ishida, 1 Kumashiro, and1 Kubo. 1998. Advances in cereal genetransfer. (urr. Opin. Plant BioI. 1: 161-65.Kooter, J.M., MA Matzke, and P. Meyer. 1999.Ulfening 10 the silent genes: tronsgenesilencing, gene regulation and pathogencontrol. Trends Plant Sci. 4: 340-47.Kranz, E., P. von Wiegen, and H.LOrz. 1995.Early cytological events aher induction ofcell division in egg cells and zygotedevelopment following in vitro fertilizationwilh angiosperm gametes. Plant 1. 8: 9-21Kranz, E., P. von Wiegen, H. Quader, and H.Larz. 1998. Endosperm development aherfusion of isolated, single moize sperm andcentral cells in vitro. Plant (ell 10: 511-24.Uang, P., and A.B. Pardee. 1992. DiHerentialdisplay of eukaryatic messenger RNA bymeons of the po~merase chain reaction.Science 257: 967-71.Lotan, 1, M. Ohto, K.M. Vee, M.A West, R. Lo,RW. Kwong, K. Yomogishi, R.L. Fischer, R.B.Goldberg, and JJ. Harada. 1998.Arobidopsis LEAFY COTYl£DONl is sufficientto induce embryo development invegetative celk. (eI193: 1195-1205.lua, M., P. Bilodeau, A. Kollunaw, E.S. Dennis,WJ. Peacock, and A.M. Choudhury 1999.Genes controlling fertilizalion·independentseed development in Arobidopsis thaliana.Proe. Natl. Acad. Sa. (USA) 96: 296--301.Lutmon, PJW. 1999. Gene now and agricullure:Relevance for transgenic uops. British (rapProtedion (ounci/Symposium ProceedingsNo. 72.lulz, W., W. Sanderson, and S. Scherbov. '1997.Doubling of world population unlike~.Nature 387: B03-05.Matsumura, H., S. Nirasawo, and R. Terauchi.1999. Transuipt profiling in rice (Oryzasativa l) seedlings using serial analysis ofgene expression (SAGE). Plant J. 20: 719­26.Matzke, MA, and AJ.M. Matzke 1995. Howand why do plants inactivate homologous(Irans)genes? Plant Physiol. 107: 679-85.Matzk, E, A. Meister, and I. Schubert. 2000. AneHicient screen for reproductive pathwaysusing mature seeds of manacols and dicots.Plant J. 21: 97-108.Matzk, E, H.-M. Meyer, H. Baumlein, H.-JBalzer, and I. Schubert. 1995. Anovelapproach to the ana~sis of the initiation ofembryo development in Gramineae. Sex.Plant Reprod. 8: 266--72.Matzk, F., H.-M. Meyer, CHorstmann, H.-J.Balzer, H. Baumlein, and I. Schubert. 1997.Aspecific alpha·tubulin is associated withthe initiation of parthenogenesis in'Salmon' wheat lines. Hereditas 126: 219­24.McMeniman, S., and G. Lubulwa. 1997. ProieclDevelopment Assessment: An EconomicEvaluation of the Potential Benefits ofIntegrating Apomixis in Hybrid Rice.Canberra: Australian Centre forInternational Agricultural Research.Me", V.L, L.P. Lochhead, and P.H. Reynolds.1993. Capper·controllable gene expressionsystem lor whole plants. Proe. Natl. Acad.Sci. (USA) 90: 4567-71Meyer, P. 1995. Understanding and controllingtransgene expression. TlBTE(H 13: 332­37.Naumova, IN., and F. Malzk 199B. Differencesin lhe initiation of the zygotic andparthenogenetic pathway in the Salmonlines of wheal: ullrastructuralsludies. SexPlant Reprod. 11: 121-30.NeuHer, M.G., E.H. Coe, and S.R. Wessler. 1997.Mutants of Maize. Cold Spring Harbor, NewYork: Cold Spring Harbor Laboratory Press.Ohad, N., R. Yadegari, L MargDlsian, M.Hannan, D. Michaeli, JJ. Harada, R.B.Goldberg, and R.L Fischer. 1999. Mutationsin FIE, aWD Polycamb group gene, allowendosperm development withoutfertilization. Plant (ell 11 :407-15.Pirra"a, V. 1998. Palycombing the genome: PcG,trxG, and chrommin silencing. (eIl93:333-36.Puchta, H. 1998. Towordstargetedtransformation in planls. Trends Plant Sa.3: 77-78.Quarin, C.L and w.w. Hanna. 1980. Effect ofthree ploidy levek on meiosis and mode ofreproduction in Pospolum hexosfachyum.Crop Sci. 20: 69-75.Rabobonk International. 1994. The World SeedMarket: Developments and Strategy.Ulrechl, The Netherlands: Rabobank.RAFI (Rural Advancement FoundalianInternational). 1998. Terminator Trends:The 'Silent Spring' of 'Formers Rights': SeedSaving, The Public Sedor and TerminatorTransnationals. Occasional Paper Vol. 5,No.1. Winnipeg, Canada: RAFI Publications.Rotina, G.t., E. Perri, M. Za"ini, H. Sommer, andA. Speno. 1997. Genetic engineering ofparthenocarpic planls. Nat. Bio/echnol. 15:1398-1401.

Ge••H, l.gi•••,I1'9 0' Apomixis i. 1....1Crops: AC,itkal A.....me.t.' the Apomilis redo••logy 243Sauter, M., P. van Wiegen, H. lorz, and E. Kranz.1998. Cell cycle regulatory genes frommaize are differential~ controlled duringferlilizalion and firsl embryonic celldivision. Sex. Plant Reprad. 11: 41-48.Sovidon, Y.(2000). Apomixis: genetiC5 andbreeding. Plant Breeding Reviews 18: 13­86.Schena, M., A.M. Uoyd, and RW. Davis. 1991. Asteroid-indicible gene expression systemfor plant cells. Proc. Natl. Acad. SQ. (USA)88: 10421-425.Schmidt, E.D., F. Guzzo, M.A. Toonen, S.C deVri~. 1997. Aleucine-rich repeatcontaining receptor-like kinase markssomalic plant cells competent 10 formembryos. Development 124: 2049-62.Scon, RJ., M. Spielman, J. Boiley, ond H.G.Dickinson. 1998. Parent-of·origin effectson seed development in Arabidopsisthaliana. Development 125: 3329-41.Spangenberg, G., Z.Y. Wang, and I. Potrykus.1998. Biotechnology in lorage and lurfgrass improvement. Monographs anThearetical and Applied Genetics. Vol. 23.Heidelberg: Springer Verlag.Spillane, C1999. Recent Developments inBiotechnology os They Relate to PlantGenetic Resources for Food and Agriculture.FAD Bockground Sludy Paper NO.9. Rome:FAD.Syvanen, M. '994. Horizontal gene transfer:Evidence and possible consequences. Annu.Rev. Genet. 28: 237-61.van Dijk, P., and J. van Damme. 2000.Apomixis technology and the paradox ofsex. Trends Plant Sci. 5: 81-84.Vershon, A.K., and M. Pierce. 2000.Transcriptional regulation of meiosis inyeast. (Uff. Dpin. (el/Bial. 12: 334-39.Vie lie-Calzada, l-P., R. Baskar, and U.Grossniklaus. 2000. Delayed activalion ofthe paternal genome during seeddevelopment. Nature 404: 91-94.Vielle, l-P., B.L Burson, H. Bashaw, and M.A.Hussey. 1995. Early fertilization evenls inIhe sexual and aposporous egg apporalusof Pennisetum riliare (l) link. Plant 1. 8:309-16.Vie lie-Calzada, J.-P., c.F. Crane, and D.M. Slel~.19960. Apomixis: The asexual revalulion.Science 274: 1322-23.Vielle-Calzada, l-P., M.L Nuccio, M.A. Budiman,IL Thomos, B.L 8urson, M.A. Hussey, andR.A. Wing. 1996b. Comparative geneexpression in sexual and apomictic ovari~of Pennisetum ciliare (l) link. Plant Mol.Bioi. 32: 108>-92.Vinkenoog, R., M. Spielmon, S. Adoms, R.LFischer, H.G. Dickinson, and RJ. Scon.2000. Hypomelhylolion promol~autonomous endosperm development andres(U~ postfertilization letholity in fiemutants. Plant (eI/12: 2271-82.Vollbrechl, L, and S. Hake. 1995. Deficiencyanalysis of femole gameloge~is inmaize. Dev. Genetics 16: 44-63.Weinmann, P., M. Gossen, W. Hillen, H. BUiard,and C. Gatz. 1994. Achimerictransoctivator alloWltetracyciineresponsivegene expr~sion in wholeplants. Plant 1. 5: 559--{)9.Welford, S.M., 1 Gregg, E. Chen, D. Garrison,P.H. Sorensen, CI Denny, and S.F. Nelson.1998. Defection of dilferenlial~ expmsedgenes in primary tumor tissu~ usingrepr~entational dilferenc~ ana~siscoupled to microarroy hybridization. Nue/.Acids Res. 26: 3059-65.Yong, W.-L, and V. Sundor~n. 2000. GeneliC5and gamelophyte biogen~is inArabidop~s. (urf. Dpin. Plant Bioi. 3: 53­57.Ye, 1, S. AI-Babili, A. Kloti, 1 Zhang, P. lucca, P.Beyer, and I. Potrykus. 2000. Engineeringthe provitamine A(b-carotene) biosynthelicpothway inlo (corotenoid-free) riceendosperm. Science 287: 303-05.Yuan, LP. 1993. Progr~s of fWo-line system inhybrid rice breeding. In K. Muralidhoranand E.A. Siddig (eds.), New Frontiers inRice Research. Hyderabad, India:Directorate of Rice R~earch. Pp. 86-93.Zetka, M., and A. Rose. 1995. The genetia ofmeiosis in (aenarhabditis elegans. TrendsGenet. 11: 27-31.