The Genom of Homo sapiens.pdf

GENETIC CONTROL OF TRANSCRIPTION 113nism; a variant TF, for example, might influence the expressionof all genes within the regulon. These can be detectedin our experiments if we observe an SDP in thetrans analysis which appears to be associated with variationin the levels of mRNA in multiple genes that do notmap to this SDP.We identified two SDPs that appeared to be influencing12 different genes. The target genes of one such case includeSlc25a1 (solute carrier family 25 member 1), Xrn1(5´-3´ exoribonuclease 1), Xpnpep1 (X-prolyl aminopeptidaseP1, soluble), Ctps2 (cytidine 5´-triphosphate synthase2), BG065446 (homology with RNA polymerase1–3 16-kD subunit), C77518 (homology with PTPL-1-associatedRho-GAP), D7Ertd59e (an expressed marker), aswell as five proteins of unknown function. Our workinghypothesis is that the shared SDP defines a region thatcontains a variant transcription effector that is modulatingthe behavior of SETS of genes that comprise a regulon (aset of coregulated genes). The expectation, given that ourdata set contains 412 matches, is that any one SDP will appear412/800 times: Thus, a single SDP appearing 12times is unusual. The SDPs define a region of chromosome14 that spans 7 Mb. Using the UCSC genomebrowser, we have identified two potential transcriptionfactors within the region, Dnase1l3 and 2610511E03Rik(Rnase P-related). 2610511E03Rik contains one replacementvariation (Met132Val) in C57BL/6 compared toDBA/2J in the Celera database: The other protein containsonly silent/noncoding variants. This leads us to the workinghypothesis that variation in 2610511E03Rik is thecause of variation in the target genes. In addition, we haveidentified one case of an SDP influencing 9 genes, an SDPinfluencing 7 genes, two cases of an SDP influencing 6genes, three cases of an SDP influencing 5 genes, fourteencases of an SDP influencing 3 genes, and one hundred andtwelve cases of an SDP influencing 2 genes, and these arebeing analyzed intensively. The existence of identifiableregulons partially explains the frequency of trans influencediscussed above. We believe these preliminary dataare graphic illustrations of the potential of our analytic andexperimental analyses.CONCLUSIONWe have developed a powerful system for analyzinggenetic variation and its influence on mRNA levels. Ourapproach is readily comparable to that of Schadt et al.(2003), who used a backcross between C57BL/6 andDBA/2J to analyze mRNA level variation. Such animalsare either homozygous or heterozygous for variations, incontrast to RI strains which are homozygous, and thismay account for the greater variability seen in mRNAlevels in their analyses, where 33% of genes appeared tobe differentially expressed within the progeny. Furtheranalysis will resolve this issue.An important distinction between the use of RI orbackcross mice is that the RI lines are genetically stableand can be bred at will. We believe that ability to tailorgenotypes by selection of RI lines and other strains reinforces,once again, the great power of the mouse as a geneticmodel for human variation, because establishing therelationship of quantitative mRNA variation to ultimatephenotype is not simple. The parental C57BL/6 andDBA/2J mice differ significantly in many physical, biochemical,and behavioral respects (Festig 1998), andthese data in principle can be related to underlying geneticvariations. In practice, until we have a better idea ofthe specific effectors and genes, it is difficult to definetestable hypotheses.It is possible to extend the approach we have developedhere to humans. Unlike RI lines, humans are frequentlyheterozygous for variations, outbred, susceptible to environmentalinfluence, and not a ready source of tissue. Despitethese reservations, we believe it will be possible tocarry out preliminary experiments on tissue mRNAs isolatedfrom extended human families; specifically, thethree-generation “reference” CEPH families that havebeen very extensively typed using microsatellite markersas part of the Human Genome Project. These data, publiclyavailable at http://lpg.nci.nih.gov/CHLC/, enable usto calculate the parental origin of any genomic region,which in turn enables us to construct the equivalent of anSDP. This will be the “parent of origin distribution pattern”or PODP of the gene in the family. Analogous to ourmice experiments, concordance or discordance of expressionlevels with the PODP will indicate cis or trans influenceon expression levels.Confounding this experiment are numerous nongeneticfactors associated with the intrinsic variability of transformedlymphoblastoid cell lines, but these influences arenot expected to be identically distributed to Mendelianpatterns of inheritance, and frank genetic signal should inprinciple be isolable by the analysis we have proposed.Our observation of a significant amount of variationwithin the machinery that controls transcription, alliedwith our preliminary data and that of Schadt et al. (2003),leads us to propose a new class of project. We suggestthat identifying sequence variation in this machinery, inany organism, will provide more significant insights intothe molecular basis of phenotypic variation than conventionalcandidate gene approaches based on more limitedphysiological function.ACKNOWLEDGMENTSThis work was carried out with the support of a start-upgrant from the University of New South Wales and withthe aid of Australian Postgraduate Award scholarships toE.C. and M.K. We are indebted to the staff of the Cliveand Vera Ramaciotti Centre for Gene Function Analysisfor provision of mouse microarrays and to Matt Wand,Willam Dunsmuir, and David Nott (School of Mathematicsat UNSW) for a continuing collaboration on statisticalanalysis of our data.REFERENCESBennett S.T., Lucassen A.M., Gough S.C., Powell E.E.,Undlien D.E., Pritchard L.E., Merriman M.E., KawaguchiY., Dronsfield M.J., and Pociot F. 1995. Susceptibility to humantype 1 diabetes at IDDM2 is determined by tandem re-