Evolutionary origins of novel conchologic growth patterns in tropical ...

644 EVOLUTION&DEVELOPMENT Vol. 10, No. 5, September^October 2008Before isotopic sampling, fossil specimens were subjected torigorous diagenetic screening to test for secondary mineralogic alteration,which could cause systematic changes in the primaryisotopic composition of shell carbonate (See Supporting Information:Diagenetic Screening). Unaltered carbonate samples weresubsequently collected from thick-sections using either point-samplingor micromilling techniques. Point-samples (50–100 mg) weretaken from the outermost shell layer using a 300-mm diameter drillbit (see Goodwin et al. 2001). Micromilled samples (20–100 mg)were collected using a computer-controlled X–Y–Z motorizedmicrodrill (Dettman and Lohmann 1995). All carbonate isotopicanalyses were performed on a Finnigan MAT 252 mass spectrometerequipped with a Kiel III automated sampling device (Departmentof Geosciences, University of Arizona). Samples were reactedwith 100% orthophosphoric acid at 701C. Repeated measurementof standard carbonates resulted in standard deviations of 0.08%.Results are presented in permil notation with respect to the VPDBcarbonate standard.Phylogenetic analysisWe conducted a species-level phylogenetic analysis, focusing onrelationships within Bothrocorbula1Hexacorbula (see Andersonand Roopnarine 2003). Also included in the ingroup were fiverepresentatives of tropical American Caryocorbula; the westernPacific Caryocorbula zelandica (placed in Anisocorbula by a numberof workers) and Notocorbula vicaria; the eastern Atlantic Corbulasulcata and Bicorbula gallica; the western Atlantic Lenticorbula?idonea (placed in Bicorbula by previous workers); and the tropicalAmerican Corbula gatunensis and C. speciosa (Table S1). In ananalysis combining conchologic and anatomic characters, thesegenera form a subclade with Bothrocorbula1Hexacorbula withinthe Corbulidae (L. C. Anderson, unpublished data). In addition, aMiocene species from Venezuela (Hexacorbula? sp.), with morphologicsimilarities to both Hexacorbula and Caryocorbula, wasincluded in the ingroup. The 20 ingroup species represent 13–19%of estimated total species diversity for the ingroup genera (TableS3), although most of this diversity (65–67%) is within Caryocorbula.InCaryocorbula, species are typically distinguished onthe basis of subtle differences in valve shape and size, providing fewdiscrete characters for coding, especially when counfounded byintraspecific and ontogenetic variation. Nevertheless, charactersused in the phylogenetic analysis are conservative across individualswithin species for the genera examined. Two corbulid species thatfall outside the selected ingroup, Juliacorbula scutata and Panamicorbulaventricosa (Anderson and Roopnarine 2003; Andersonet al. 2006), were designated as outgroup taxa.Sixty-nine multi-state conchologic characters (of which ninewere autapomorphies), describing aspects of external ornament,valve shape, hinge, pallial line and sinus, and adductor musclescars, were used in the analysis (Supplementay Material: PhylogeneticAnalysis). Most characters used had discontinuous characterstates. For those describing the degree of expression of a trait, onlycharacters with states that we could consistently distinguish becauseof morphologic gaps were retained. In addition, we coded thelargest individuals available for species (assuming these correspondto adult ontogenetic stages) to minimize ontogenetic variability incharacter states. Phylogenetic analyses were performed usingPAUP 4.0b10 (Swofford 2002). Characters were unorderedand given equal weight, and polarized using the outgroup taxa.Analyses were conducted using branch and bound searches andmaximum parsimony. Character state transformations were determinedusing both accelerated transformation (ACCTRAN) anddelayed transformation (DELTRAN). Both gave equivalent resultsfor phylogenetic patterns of growth forms, which are summarizedin the results. We calculated Bremer decay indices (Bremer 1988,1994; Ka¨llersjo¨ et al. 1992) and bootstrap values (1000 replicates)(Felsenstein 1988) with the branch and bound algorithm to characterizethe robustness of cladogram nodes.RESULTSPatterns of corbulid growthThree distinct patterns of shell development were observed,and the species Lenticorbula? idonea, Caryocorbula amethystina,andCorbula speciosa are used as exemplars in the followingsection.Lenticorbula? idonea exemplifies a mode of developmentcommon in heterodont bivalves where valve growth isdominated by commissural accretion of new material in aprimarily radial (sagittal) direction (herein referred to asGrowth Form 1, [GF1]). The result is a valve that does notsubstantially alter its direction of growth, or overall shape,during ontogeny. For example, using growth lines as aguide (Fig. 1A), the growth history of a L.? idonea valvewas arbitrarily divided into four intervals (Fig. 2A: T 1–4 ).Rotating and enlarging the cross-section at T 1 so that itcan be compared with the shapes at T 2–4 indicates that itscross-sectional shape did not change substantially as itssize increased. This developmental pattern likely accommodatesan increasing volume of internal soft-tissues. Isotopicdata for the L.? idonea specimen (Fig. 1, A and B)indicate that the individual was at least 8 years old at itstime of death: seven complete d 18 O cycles are present (Fig.1C: positive peak to positive peak) as is the latter portionof a cycle nearest to the umbo (sample nos. 1 and 2) andthe beginning of a cycle at the commissure (sample nos.35–37).The second exemplar, Caryocorbula amethystina, possessesone of two types of development apparently derived fromGF1. Examination of growth lines (Fig. 3A) suggests that thisspecies initially deposited a thin, high shell via radial accretion,followed by an interval of valve thickening. We designatethis pattern as Growth Form 2 (GF2). Sampling only theouter shell layer of C. amethystina from umbo to commissure(as in L.? idonea) would likely fail to capture all annual increments,as well as the complete range of d 18 O variability,due to time-averaging in the later portions of shell growthwhere growth lines are very closely spaced (Goodwin et al.