Population structure is not a simple function of. reproductive mode and larval type: insights from tropical corals

Transcription

1 Journal of Animal Ecology 2008, 77, doi: /j x Population structure is not a simple function of *Blackwell Publishing Ltd reproductive mode and larval type: insights from tropical corals Karen J. Miller 1 * and David J. Ayre 2 1 Institute of Antarctic and Southern Ocean Studies, University of Tasmania, Private Bag 77, Hobart 7001, Tasmania, Australia; and 2 Institute for Conservation Biology, School of Biological Sciences, University of Wollongong, Northfields Avenue, Wollongong, 2522, NSW, Australia Summary 1. For a wide range of organisms, heritable variation in life-history characteristics has been shown to be strongly subject to selection, reflecting the impact that variation in characters such as genotypic diversity, duration of larval development and adaptations for dispersal can have on the fitness of offspring and the make-up of populations. Indeed, variation in life-history characteristics, especially reproduction and larval type, have often been used to predict patterns of dispersal and resultant population structures in marine invertebrates. 2. Scleractinian corals are excellent models with which to test this relationship, as they exhibit almost every possible combination of reproductive mode and larval type. Some general patterns are emerging but, contrary to expectations, genetic data suggest that while populations of broadcast spawning species may be genotypically diverse they may be heavily reliant on localized recruitment rather than widespread dispersal of larvae. 3. Here we use microsatellites to test the importance of localized recruitment by comparing the genetic structure of populations of two broadcast spawning corals with contrasting modes of reproduction and larval development; Goniastrea favulus is self-compatible, has sticky, negatively buoyant eggs and larvae and is expected to have restricted dispersal of gametes and larvae. In contrast, Platygyra daedalea is self-incompatibile, spawns positively buoyant egg-sperm bundles and has planktonic development. 4. Surprisingly, spatial-autocorrelation revealed no fine-scale clustering of similar genotypes within sites for G. favulus, but showed a non-random distribution of genotypes in P. daedalea. Both species showed similar levels of genetic subdivision among sites separated by m (F ST = 0 03), suggesting that larval dispersal may be equivalent in both species. 5. Interestingly, as fragmentation has been considered rare in massive corals, our sample of 284 P. daedalea colonies included 28 replicated genotypes that were each unlikely (P < 0 05) to have been derived independently from sexual reproduction. 6. We conclude that the extreme life history of G. favulus does not produce unusually fine-scale genetic structure and subsequently, that reproductive mode and larval type may not be not good predictors of population structure or dispersal ability. Key-words: microsatellite DNA, population genetics, sexual and asexual reproduction, spatial autocorrelation, larval dispersal. Introduction Life history characteristics that influence the genotypic diversity of offspring and their capacity to disperse are likely to be subject to intense selection and are considered strong *Present address and correspondence: Institute of Antarctic and Southern Ocean Studies, University of Tasmania, Private Bag 77, Hobart, Australia Karen.Miller@utas.edu.au predictors of the scale of dispersal and consequent genetic structure of populations (e.g. Williams 1975; Maynard-Smith 1978; Bell 1982; Strathmann 1985; Bohonak 1999). At the extreme, asexual reproduction is expected to be favoured in the expansion and maintenance of locally adapted clones within the parental habitat (Herbert, Ward & Weider 1988; Carvalho 1994; Ayre 1995), whereas sexual reproduction is considered an adaptation to changed or uncertain conditions and, in general, involves the production of larger numbers of 2008 The Authors. Journal compilation 2008 British Ecological Society

2 714 K. J. Miller & D. J. Ayre energetically cheaper but genetically diverse and widely dispersed offspring (Williams 1975; Bengtsson & Ceplitis 2000). However, even within exclusively sexually reproducing taxa, many species display adaptations that may favour inbreeding and restricted dispersal, thus promoting the development of locally adapted demes and highly structured populations (Baker & Stebbins 1965; Shields 1982; Knowlton & Jackson 1993). Scleractinian corals provide an excellent model to test such predictions as they exhibit many combinations of reproductive mode and larval type (Harrison & Wallace 1990), the roles and consequences of which are poorly understood (Harrison & Wallace 1990; Hughes, Ayre & Connell 1992; Severance & Karl 2006) but which are vital for the conservation of coral reef ecosystems (Palumbi 2003; van Oppen & Gates 2006; Vollmer & Palumbi 2007). Investigations have been hampered by difficulties in identifying settlers and recruits (Babcock 1992) and even adults (Ayre, Veron & Dufty 1991; Stobart & Benzie 1994; Miller & Benzie 1997; Lopez et al. 1999), spatial and temporal variability in recruitment and, perhaps most tellingly, by the lack of genetic markers (Marquez et al. 2000; Ridgway & Gates 2006). When possible, genetic surveys provide the best test of realized larval dispersal and allow assessment of population structures for species with variable larval development and clonal recruitment (e.g. Hughes et al. 1992; Bohonak 1999; Hellberg et al. 2002). Previous studies have confirmed that coral species with demersal larvae may have extremely structured and closed populations (Hellberg 1994; Miller 1997, 1998) and that some branching species fragment, producing highly clonal populations (Ayre & Willis 1988; Severance & Karl 2006; although see Ayre & Dufty 1994; Smith & Hughes 1999). Clonal populations have also been described for the asexually viviparous Pocillopora damicornis (Stoddart 1984; Sherman, Ayre & Miller 2006; Whitaker 2006; although see Ayre, Hughes & Standish 1997; Ayre & Miller 2004). Moreover, these studies support the prediction that sexually viviparous species with large, free-swimming larvae must be maintained largely by localized dispersal of gametes and/or larvae, resulting in a high degree of among-site allelic differentiation and large heterozygous deficiencies (Ayre & Dufty 1994; Ayre & Hughes 2000; Maier et al. 2005; Whitaker 2006; Underwood et al. 2007). The same genetic surveys also support the prediction, based in part on biogeographical distributions, that well-provisioned larvae can also be effective long-distance colonists (e.g. Johannesson 1988; Richmond 1987), as evidenced by a lack of differentiation among regions separated by 1500 km on Australia s Great Barrier Reef (GBR) (Ayre et al. 1997; Ayre & Hughes 2000). However, the very limited set of genetic surveys for the more common broadcast spawning species have yielded less consistent and unexpected outcomes. Although such studies have shown that some broadcast spawners can form vast and apparently panmictic populations distributed over hundreds of kilometres (Ayre & Hughes 2000; Ridgway, Hoegh-Guldberg & Ayre 2001) they have also revealed that some species show significant genetic subdivision at scales of hundreds of metres to tens of kilometres, with collections often displaying large and significant heterozygous deficits consistent with the effects of limited dispersal of gametes or larvae (Ayre & Hughes 2000; Ridgway et al. 2001; Magalon, Adjeroud & Veuille 2005; Severance & Karl 2006). Although the restricted dispersal of gametes and larvae may be unexpected for broadcast spawning planktonic developers, there is increasing evidence that in many cases coralspawn slicks can remain close to their natal reefs (Black 1993) and that larvae may settle rapidly following competency (Miller & Mundy 2003). Here we use microsatellite markers to compare and contrast the fine-scale (< 100 m) and mesoscale (up to 4 km) structures of populations of two broadcast spawning massive corals with sharply contrasting early life-history characters. Platygyra daedalea Ellis & Solander 1786, like the majority of broadcast spawning corals, is self-incompatible, has postively buoyant egg/sperm bundles (Babcock et al. 1986) and has planktonic larval development. Goniastrea favulus Dana 1846 may be unique among scleractinian corals in being self-compatible, with sticky negatively buoyant eggs and larvae. We expected that the population structure and genetic diversity of these two corals would vary predictably with their contrasting reproductive modes, especially at small scales. In P. daedalea, we expect a populationgenetic structure reflecting random mating, widespread larval dispersal and with high levels of genetic and genotypic diversity. Indeed, preliminary studies (Miller 1994) found no evidence of genetic differentiation in Platygyra populations on widely separated reefs on the GBR. In G. favulus, local genetic diversity should be low due to limited larval dispersal and the effects of inbreeding, including self-fertilization. Due to their massive growth form, neither species has been thought to have the capacity for asexual recruitment other than through the fission of larger colonies to form neighbouring patches of coral tissue (Highsmith 1982). Methods STUDY SPECIES P. daedalea and G. favulus are morphologically similar massive or brain corals (Family Faviidae) that reproduce during the annual coral mass-spawning on the GBR (Babcock et al. 1986). P. daedalea spawns positively buoyant egg-sperm bundles which break apart at the water s surface, releasing gametes for fertilization. Larval development occurs in the plankton (Babcock et al. 1986), with settlement commencing from 2 5 days post-spawning (Miller & Mundy 2003), although larvae can remain competent to settle for at least 100 days (Nozawa & Harrison 2003). P. daedalea does not self-fertilize (Miller & Babcock 1997). G. favulus releases gametes freely and has sticky eggs that remain attached to the parent colony through fertilization (Kojis & Quinn 1981; Heyward & Babcock 1986). Larvae are negatively buoyant as they develop. This strategy is thought to facilitate retention of larvae in the natal habitat; however, the negatively buoyant phase may only last a couple of days, following which larvae swim actively (Kojis & Quinn 1981; Heyward & Babcock 1986). Attachment of G. favulus larvae to the substrate commences 2 days post-spawning, with metamorphosis ocurring after 5 6 days (Babcock & Heyward 1986; Miller & Mundy 2003), although the delay between attachment

3 Life history not a predictor of genetic structure 715 Fig. 1. Approximate location of the four sites on One Tree Reef, Great Barrier Reef, Australia. Photograph courtesy of the One Tree Island Research Station. and settlement may be a laboratory artefact (Miller & Mundy 2003). Interestingly, G. favulus is self-compatible (Heyward & Babcock 1986; Stoddart, Babcock & Hayward 1988), and rates of self-fertilization may be high (Stoddart et al. 1988). STUDY SITES To test if population structure in the two species varied predictably with larval type, we examined the genetic structure of P. daedalea and G. favulus populations at One Tree Reef (23 29 S, E) on the southern GBR. We established four sites ( m 2 ) on the reef flat, two each on the southern and northern sides of the reef (Fig. 1). The distance between sites on each side of the reef was m, and the northern and southern sides of the reef are 3 4 km apart. From November 2000 to December 2002 we mapped the location of every P. daedalea and G. favulus colony in each site (Fig. 2). Colony densities differed markedly between the two species; colonies per m 2 for P. daedalea and colonies per m 2 for G. favulus. GENETIC TECHNIQUES A tissue sample (approximately 2 2 cm) was collected from all P. daedalea and G. favulus colonies within each site for genetic analysis. Due to the higher density of colonies, we sampled G. favulus colonies only from an area approximately m 2 (Fig. 2) within each site to ensure a total of colonies per site. Additionally, tissue from corals < 3 cm diameter was not collected, as sampling would have destroyed the colony. Samples were transported in seawater in individual zip-lock bags and frozen subsequently in liquid nitrogen and stored at 70 C. DNA was extracted from the tissue using DNAzol (Molecular Research Center, Cincinatti) according to the manufacturer s protocol. We genotyped each colony at five microsatellite loci (Miller & Howard 2004; Table 1). For most loci the polymerase chain reaction (PCR) was in a final volume of 20 μl containing reaction buffer (10 mm Tris-HCl, 50 mm potassium chloride (KCl), 0 1% Triton X-100), 2 5 mm MgCl, 200 μm each 2 -deoxynucleosides 5 -triphosphate (dntp), 1 μg bovine serum albumin (BSA), 1 μ Taq DNA polymerase (Promega Corporation, Sydney, Australia) and 0 5 mm each primer. For locus Gf42 the PCR contained 3 5 mm magnesium chloride (MgCl). For the loci Gf12 2 and Gf69, the PCR contained 1 5 mm MgCl and 75 mm KCl, but no Triton X-100. PCR amplifications were performed on a Corbett Research thermocycler (Corbett Research, Mortlate, NSW) with cycling conditions at 94 C for 5 min; 30 cycles at 94 C for 30 s, annealing C for 30 s (Table 1), 72 C for 1 min; and 72 C for 5 min. PCR products were sized on a Beckman Coulter (Beckman Coulter, Gladesville, NSW) CEQ8000XL automated sequencer and genotypes determined using the CEQ8000 fragment analysis software. Unfortunately, two loci (Gf41 1 and Pd61) amplified unreliably and so data from these two loci were excluded. Fig. 2. Diagram of spatial plots for Platygyra daedalea (a) and Goniastrea favulus (b) at site 1 on the reef flat at One Tree Reef. Each dot represents the location of a colony within the site. The dashed square shows the relative position and scale of the G. favulus plot within the P. daedalea plot. The shaded circles represent the two subsites of P. daedalea. These were chosen to represent genetic neighbourhoods of approximately 17 m, spaced at least 17 m apart within the site, and in order to maximize the number of colonies within the subsite.

4 716 K. J. Miller & D. J. Ayre Table 1. Primers and annealing temperature used for polymerase chain reactions (PCR) for each of five microsatellite loci used to genotype colonies of Platygyra daedalea and Goniastrea favulus from One Tree Reef. Primers 5 end-labelled with WellRED dyes (D3 or D4) Locus Primer sequences Annealing temp ( C) No. of alleles Size range of alleles Goniastrea favulus Gf12 1 Forward: GACCCCATAAGCTGCATAGC Reverse: GGGTGAGACATTTTCATCTTCG Gf12 2 Forward: CCATGTCATGTTTCATTCGTCT Reverse: GGAATTGGCGAAAATTATTCAGT Gf41 1 Forward: GTGATGGTGTTAAGCTTCTGC Reverse: TTGATAAGCTAATTAGACTGGACAGC Gf42 Forward: TTGATTGGCTTCTTGTTTGC Reverse: TGTTATGGACGGGTTAATGG Gf 69 Forward: CCCTGATAGAATTGCCTTTCC Reverse: GAAGAAGTAGAAGAAGCTTAG Platygyra daedalea Pd29 2 Forward: TTGGGTGGGTGACCAAATAC Reverse: TGTCCCGCTGTTCACTTATAC Pd31 Forward: GACAAGTAATGTGTAAATCGTTGTCC Reverse: CTGTTAGAGTATCATGTCCTGAAGC Pd48 Forward: ACAAAGGAGTGGGCGTAGG Reverse: GCAACTGGATTGACTATCATGG Pd61 Forward: TGTTACTAAGCGTAAAACACTGAACC Reverse: ACCGCATTGAACATCACG Pd62 Forward: CAAACGCAAGGAATGAGAGC Reverse: CAACACAAATATTCGCCATTACC DATA ANALYSIS We chose the size of our sites in order to ensure sufficient samples for genetic analysis, rather than as a true reflection of what might represent the size of local breeding units. We therefore used spatial autocorrelation analysis to determine if the distribution of genotypes within each site was non-random, and to estimate neighbourhood size for each species which could then be used as a basis for further genetic analysis. Our spatial-autocorrelation analysis was based on the genotypes of each coral and the distance between corals (metres). We calculated the autocorrelation coefficient r (Smouse & Peakall 1999) for each site locus combination, across multiple loci for each site, and across multiple sites for each species using GenAlEx version 6 (Peakall & Smouse 2005). We used 999 permutations to test the significance of the autocorrelation, and 1000 bootstrap replications to determine the 95% confidence interval around r. For G. favulus, neighbourhood size was similar to the size of the sites. However, for P. daedalea, neighbourhoods were considerably smaller than sites (see Results). Consequently we subdivided the original sites into two subsites (~300 m 2 ; Fig. 2) that represented genetic neighbourhoods, and then analysed the data for P. daedalea as eight, rather than four populations. To characterize genetic diversity in each of the coral species, we examined genotypic and allelic diversity across all samples, and also compared observed and expected levels of heterozygosity for each population. We determined whether populations were in Hardy Weinberg equilibrium (HWE) using Exact tests and used measures of Wright s inbreeding coefficient (f) to determine the nature of departures from equilibrium. Where multiple colonies had identical genotypes, we calculated the probability that they would have been produced through random sexual reproduction as P r = n 1 C r 1 p (r 1) q (n r), where n = the number of individual colonies in the sample, r = the number of replicates of the four-locus genotype, p = the probability of occurrence of a single copy of the genotype under random mating and q = 1 p (Willis & Ayre 1985). We calculated F ST among populations to determine if there was evidence of genetic subdivision and to infer levels of gene flow as a proxy for larval dispersal. Additionally, we used an hierarchical F ST analysis to partition genetic subdivision relative to the sampling scales; between the northern and southern reef flat (~ 3 4 km), between sites on each side of the reef (~ m) and among subsites (< 20 m, for P. daedalea only). F ST was calculated according to Weir & Cockerham (1984) using Tools for Population Genetic Analyses (tfpga) version 1 3 (Miller 1997), which estimates 95% confidence intervals around F ST by bootstrapping over loci. Clonal genotypes were removed for these analyses. Because microsatellites lead invariably to low F ST estimates due to high amounts of within-population genetic variation, we standardized F ST ( F ST ) according to Hedrick (2005) and Miermans (2006). This also facilitates comparisons of F ST among species that probably have different effective population sizes. Results GENETIC NEIGHBOURHOOD SIZE Surprisingly, for P. daedalea, spatial autocorrelation analysis indicated that genotypes were not distributed randomly (Fig. 3a d). At sites 1, 2 and 4, there was a significant positive correlation (P < 0 05) among genotypes of colonies within 5 10 m, and a significant negative correlation (P < 0 05) among genotypes of colonies separated by > m. Notably, this pattern was driven by a combination of loci. For example, at site 1, a significant positive correlation at 5 m was evident at three loci: Pd31 (r = 0 09, P < 0 05), Pd48 (r = 0 16, P < 0 05) and Pd62 (r = 0 36, P < 0 05) but not Pd29 2 (r = 0 03, P > 0 05). Similarly, at site 4 strong spatial autocorrelation was apparent at 5 m for locus Pd29 2 (r = 0 11, P < 0 05), Pd48 (r = 0 12, P < 0 05) and Pd62 (r = 0 2, P < 0 05) but

5 Life history not a predictor of genetic structure 717 Fig. 3. Results from spatial autocorrelation analysis of four-locus genotypes of Platygyra daedalea from four sites on the reef flat at One Tree Reef (a d) and combined across all four sites (e). Dashed lines represent 95% confidence intervals associated with the null hypothesis of no spatial structure (i.e. r = 0). not Pd31 (r = 0 004, P > 0 05). A combined analysis across all four sites and loci was used to determine the neighbourhood size of ~17 m for P. daedalea at One Tree Reef (based on the point at which the correlogram intercepts the x-axis; Fig. 3e). In contrast, spatial autocorrelation analysis of G. favulus indicated that genotypes were distributed randomly within each site (Fig. 4), indicating that neighbourhood size in this species was likely to be equal to or larger than the smallest scale sampled (~15 m). GENETIC STRUCTURE IN P. DAEDALEA Levels of genetic and genotypic diversity were moderate in P. daedalea, which is surprising given its broadcast-spawning mode of reproduction and hypothesized widespread dispersal. The mean number of alleles detected at each microsatellite locus in P. daedalea was 12 2 [± 2 9 standard error (SE)]. In the sample of 284 P. daedalea colonies, 227 had unique genotypes but there were 28 four-locus genotypes that occurred more than once, with one of these represented by 12 colonies. Under conditions of random mating we would expect even this common genotype to occur no more than twice in our sample (P < 0 01), so repeat genotypes are likely to be the Fig. 4. Results from spatial autocorrelation analysis of four-locus genotypes of Goniastrea favulus from four sites on the reef flat at One Tree Reef (a d) and combined across all four sites (e). Dashed lines represent 95% confidence intervals associated with the null hypothesis of no spatial structure (i.e. r = 0). product of asexual reproduction. In 15 cases (54%), colonies with identical genotypes occurred within the same site (often in close proximity), so they may result from colony fission. Additionally, there were five clonal genotypes that were restricted to one side of the reef (north or south); for example, the most common genotype comprised nine colonies from site 1 and three colonies from site 2. Conceivably, ramets generated through fission or partial mortality could be transported tens of metres through wave or storm action, which may explain the identical genotypes at different sites on the same side of the reef. Over all loci and populations (sites and subsites), observed levels of heterozygosity in P. daedalea were similar to expected values (Table 2). However, at the level of genetic neighbourhood (subsites), all eight populations of P. daedalea showed significant departures from HWE (Table 3). Of the 32 singlelocus tests for departures from HWE, 13 were significant after Bonferroni correction, representing a mix of heterozygote deficits (eight cases) and excesses (five cases), as might be expected in asexually reproducing populations (Table 3). Consistent heterozygote deficits at a locus may reflect null

6 718 K. J. Miller & D. J. Ayre Table 2. Genetic diversity at four loci in Platygyra daedalea from the reef flat at One Tree Reef as four sites, and eight subsites. N = number of colonies genotyped; N A = number of alleles; H O = observed heterozygosity; H E = expected heterozygosity N N A H O H E N N A H O H E N N A H O H E Site 1 Site 1 1 Site 1 2 Pd Pd Pd Pd Overall (82) (23) (15) Site 2 Site 2 1 Site 2 2 Pd Pd Pd Pd Overall (58) (19) (12) Site 3 Site 3 1 Site 3 2 Pd Pd Pd Pd Overall (82) (21) (21) Site 4 Site 4 1 Site 4 2 Pd Pd Pd Pd Overall (62) (11) (16) Table 3. Wright s Fixation Index (f) for Platygyra daedalea in eight subsites on One Tree Reef. Significant departures from Hardy Weinberg equilibrium (where f < 0 represents heterozygote excesses and f > 0 represent heterozygote deficits) are denoted as *P < 0 05, **P < 0 01, ***P < (significance levels adjusted using a Bonferroni correction) Pd 62 Pd 48 Pd 31 Pd 29 2 Subsite * 0 614*** Subsite ** 0 859* Subsite * 0 654** Subsite Subsite * 0 697*** Subsite *** *** Subsite ** Subsite * *** Over all loci and sites (± SD) 0 19 (0 24) 95% CI alleles, and tests using Micro-Checker software (van Oosterhout et al. 2004) indicated that null alleles may exist at locus Pd48. However, such tests do not account for the effects of asexual reproduction on allele frequencies, and therefore their application to species such as coral may be problematic. To determine the potential implication of null alleles on our results, we adjusted our data for null alleles at locus Pd48 (calculated using the Oosterhout correction algorithm in Micro-Checker and using data with clonal genotypes removed to account, at least in part, for asexual reproduction) and re-ran the F ST analyses. As the adjustment made minimal difference to our F ST estimates both at locus Pd48 and overall (e.g. a difference of 0 002), we have included results only from the original data. Exact tests revealed significant differences in allele frequencies (P < 0 01) among all pairs of sites at all loci. Additionally, we detected small but significant levels of genetic subdivision among the eight P. daedalea populations on the reef flat at One Tree Reef (F ST ± SD = ± 0 036; Table 4), and higher levels of differentiation were apparent when F ST was standardized ( F ST = 0 184). Hierarchical F ST analysis indicated that most subdivision occurred at the smallest spatial scale; in fact, more than twice as much differentiation was apparent among sites separated by < 20 m than at either the m or 3 4 km scales (Table 4), and F ST was significantly different to zero (the value expected through random mating) only at the smallest scale. However, most of the subdivision among sites was attributable to a single locus Pd29 2 (Table 4).

7 Life history not a predictor of genetic structure 719 Table 4. Results from hierarchical F ST analyses for Platygyra daedalea from eight subsites on One Tree Reef F ST Overall Among sites 3 4 km apart Among sites m apart Among sites < 20 m apart Pd Pd Pd Pd Over all loci (± SD) (0 030) (0 041) 0 03 (0 049) (0 046) 95% CI Table 5. Genetic diversity at four loci in Goniastrea favulus from four sites on One Tree Reef. N = number of colonies genotyped; N A = number of alleles; H O = observed heterozygosity; H E = expected heterozygosity Site 1 Site 2 Site 3 Site 4 N N A H O H E N N A H O H E N N A H O H E N N A H O H E Gf Gf Gf Gf Overall (93) (92) (80) (80) Table 6. Wright s Fixation Index (f) for G. favulus from four sites on One Tree Reef. Significant departures from Hardy Weinberg equilibrium (where f < 0 represents heterozygote excesses and f > 0 represent heterozygote deficits) are denoted as ***P < (significance levels adjusted using a Bonferroni correction) Site 1 Site 2 Site 3 Site 4 Gf *** 0 667*** 0 619*** 0 973*** Gf *** 0 314*** 0 19*** 0 24*** Gf *** 0 404*** 0 446*** 0 355*** Gf *** 0 361*** 0 588*** 0 576*** Over all loci and sites (± SD) (0 191) 95% CI GENETIC STRUCTURE IN G. FAVULUS In contrast to P. daedalea, levels of genetic and genotypic diversity were high in G. favulus. The mean number of alleles detected at each locus was 18 2 (± 6 5 SE) cf (± 2 9 SE) in P. daedalea. Similarly, expected heterozygosity (H E ) was slightly higher in G. favulus ( , Table 5) than in P. daedalea ( , Table 2). Of the 345 G. favulus colonies genotyped, 329 had unique genotypes and eight four-locus genotypes were repeated twice in our samples. The probability that these genotypes occurred more than once through random sexual reproduction is < 0 005, so we have assumed that they represent asexual reproduction. For five of the eight putative clonal genotypes both colonies occurred within the same site (i.e. separated by < 15 m), and in the remaining three instances the colonies were in sites on the same side of the reef (i.e. within 100 m of each other). Overall, observed levels of heterozygosity were lower than expected (Table 5) and all populations of G. favulus from the four sites showed large and significant departures from HWE. Twelve of the 16 single-locus tests for HWE showed significant heterozygote deficits at the three loci Gf12 2, Gf42 and Gf69, while locus Gf 12 1 had heterozygote excesses at all four sites (Table 6). Exact tests revealed significant differentiation (P < 0 001) among all pairs of sites at three of the four loci (Gf 69, Gf 42 and Gf 12 2) as well as across all loci. We also detected significant genetic subdivision in G. favulus among the four sites (F ST = ± 0 005; Table 7), and hierarchical analysis indicated that most of this subdivision was at the smallest scale ( m; Table 7). Standardized F ST values showed greater genetic differentiation among populations of P. daedalea ( F ST = 0 184) than G. favulus ( F ST = 0 085), a result that was counter to our expectations based on the life-history characteristics of each species. Discussion Our fine-scale analysis of the genetic structure of G. favulus and P. daedalea did not match the population structures predicted for their life-history types, nor did the genetic structure of the two species contrast sharply as we had expected. Indeed, genetic and genotypic diversity was higher in the self-fertilizing, benthic-developing G. favulus than in the

8 720 K. J. Miller & D. J. Ayre F ST overall Among sites 3 4 km apart Among sites m apart Table 7. Results from hierarchical F ST analyses for G. favulus from four sites on One Tree Reef Gf Gf Gf Gf Over all loci (± SD) (0 005) (0 005) (0 006) 95% CI planktonically developing and presumably out-crossed P. daedalea. Furthermore, we found no evidence that the benthic larval development of G. favulus leads to unusually restricted dispersal. G. favulus genotypes were distributed randomly within sites and meso-scale (among site) heterogeneity was actually less than has been reported in other allozyme and microsatellite surveys of broadcast-spawning corals, and lower than that in P. daedalea. Nevertheless, our data revealed that significant fine- to meso-scale population subdivision was present for both species and so provides general support for those recent genetic studies that have found both evidence of within-reef subdivision and genotypic diversity consistent with restricted dispersal and inbreeding (Ayre & Hughes 2000; Ridgway et al. 2001; MacKenzie et al. 2004; Miller & Ayre 2004; Baums et al. 2005; Whitaker 2006). Perhaps more surprisingly, our analyses of genotype frequency distributions revealed the presence of apparently clonal genotypes (approximately 20% of all colonies) within populations of P. daedalea, and a minor level of clonal replication for G. favulus (approximately 5% of colonies). The discovery of this level of clonality was unexpected, as these massive corals have no obvious mechanism of asexual replication other than fission as a result of within-colony damage (Babcock 1991). POPULATION SUBDIVISION Meso-scale subdivision seems readily explicable and may be a common occurrence for the broadcast-spawning corals of the GBR because spawning occurs typically under conditions that minimize water movement (Harrison et al. 1984; Babcock et al. 1986; Miller & Mundy 2005), and several studies suggest that spawn slicks can remain on or close to their natal reefs (Willis & Oliver 1990; Black, Moran & Hammond 1991). The presence of similar scales of within-reef population structuring in both coral species studied here lends support to the generalization that corals on the GBR are not panmictic and that larval dispersal, at least on ecological time scales, is localized. However, the scale of retention of surface slicks of gametes and larvae on reefs is too coarse to explain clustering of multilocus genotypes at the scale of metres seen in P. daedalea. Asexual reproduction through fission may result in a clustering of identical genotypes resulting in a fine-scale spatial autocorrelation. To test this, we re-ran our spatial analyses for P. daedalea at all four sites with the clonal genotypes removed, but this made no difference to our results. This finding implies that fine-scale population structures are not linked to asexual reproduction but are either being driven by an unknown element of larval behaviour or development, or that there is intense but general site-specific selection occurring in this species. Both these alternatives seem unlikely. Certainly, everything that is known about larval development in P. daedalea suggests that developing corals must spend a minimum of 2 3 days in the water column before settlement (Miller & Mundy 2003). Moreover, strong site-specific selection seems unlikely in a broadcast-spawning species, especially within an environment as spatially and temporally complex as a tropical coral reef. Tellingly, we have found little evidence of localized settlement even for the co-occurring but asexually viviparous P. damicornis (Ayre et al. 1997; Ayre & Miller 2004). Nevertheless, we have confidence that this is a robust result, as clustering was apparent at all four loci and because our data match the generally overlooked analyses of Stoddart (1988), who found similar genetic structuring in the planktonically dispersing Acropora digitifera. We did not find small-scale clustering in G. favulus, the species considered a strong candidate for localized settlement, because of its negatively buoyant eggs and larvae. Indeed, levels of genetic differentiation in G. favulus (F ST = 0 024, Table 7) are in fact lower than those found in brooders with rapid larval settlement (e.g. Seriatopora hystrix F ST = 0 089, Maier et al. 2005; F ST = 0 095, Underwood et al. 2007) but are also lower than broadcasting coral species with planktonic larval development (e.g. P. daedalea F ST = (Table 4), A. palmata F ST = 0 036, Baums, Miller & Hellberg 2005; P. meandrina F ST = , Magalon et al. 2005). The unique reproductive mode of G. favulus may have less to do with retaining larvae in natal habitats (Kojis & Quinn 1981) than other factors, such as maximizing fertilization success (Miller & Mundy 2005). However, it is conceivable that benthic larval development in G. favulus may merely be a consequence of lipid retention from eggs that favour sinking or, alternatively, may be a strategy which enables the larvae to optimize the timing of their planktonic period. In laboratory trials, most benthic larvae of G. favulus became planktonic in a narrow window between 12 p.m. and 6 a.m., 6 days post-spawning (Miller & Mundy 2003), which coincided with the incoming tide. While the switch from a benthic to a planktonic phase has been attributed to laboratory artefacts (Kojis & Quinn 1981; Miller & Mundy 2003), it may be that larvae choose to become planktonic to coincide with water movement that will carry them away from parents (to minimize inbreeding), but which will limit their being swept off-reef where settlement opportunities may be few. We acknowledge that this seems to be an unsatisfactory explanation for the co-occurrence of the

9 Life history not a predictor of genetic structure 721 unusual traits of self-compatibility and negative larval buoyancy and cannot suggest why this one species should require unusually fine control over its planktonic phase. However, these traits clearly imply some singularity and such a dispersal strategy may well account for the meso-scale genetic subdivision recorded here. If true, then it will be critical to understand more clearly the intricacies of behaviour to enable any simple prediction of dispersal from larval type. Ironically, G. favulus is one of the only corals that selffertilizes (Stoddart et al. 1988), although neither the levels of allelic or genotypic diversity nor the presence of heterozygote excesses at some loci indicate that self-fertilization is important within the One Tree Reef populations. Mounting evidence suggests that self-fertilization is likely to be a strategy to ensure fertilization success in the absence of non-self sperm (Heyward & Babcock 1986; Miller & Mundy 2005). This is unlikely to be a problem on the reef flat at One Tree Reef, where populations are well established and colony densities approach two colonies per m 2, but would certainly facilitate establishment of new populations. Indeed, G. favulus is a widespread species, although rarely common throughout its range (Veron 2000). CLONAL RECRUITMENT Fragmentation seems an unlikely method of reproduction in massive corals simply because they are hard to break. None the less, our data suggest that it may occur at least occasionally in both species and more commonly in P. daedalea, although the levels of clonality reported here for the two massives are considerably less than the extremely clonal populations that have been reported for coral species with a branching morphology (Ayre & Willis 1988; Ayre & Hughes 2000; Baums et al. 2006). Interestingly, Babcock (1991) reported asexual reproduction through fission was almost twice as common in P. sinensis than either G. aspera or G. favulus although, notably, this was based on tissue damage resulting in multiple ramets, rather than skeletal fragmentation. We specifically avoided sampling from colonies that were clearly the product of fission and where skeletal connections were still obvious, but despite this still found evidence of clonality in both species. At least some of the adjacent clones we identified may have arisen through fission, where past skeletal connections were no longer evident. Alternative modes of asexual reproduction in massives, e.g. polyp expulsion (Kramarsky-Winter, Fine & Loya 1997) or physical damage from storms, may explain the more distant distribution of some clones (i.e. tens of metres). Asexual reproduction as a strategy in massive corals is becoming apparent through genetic studies (Severance & Karl 2006; Foster, Baums & Mumby 2007). Given the longevity of massive species (Cameron & Endean 1985; Babcock 1991), long-term studies will be needed to gain sensible estimates of the frequency and success of asexual reproduction through fragmentation or other means. However, much could be learnt from tracking the performance of individual genotypes through multiple storm events. POPULATION STRUCTURE We would expect to see a mix of heterozygote excesses and deficits in asexually reproducing populations and this was apparent in P. daedalea, the species in which we observed the higher level of clonality. For G. favulus, heterozygote excesses were apparent only at a single locus (Gf12 1), while the remaining three loci showed large heterozygote deficits. Given the low expected rate of asexual reproduction in massive corals (e.g. mean annual probability of fission = 0 029; Babcock 1991), we suspect that clonality associated with fragmentation is unlikely to be a major force influencing local population genetic structure of massive corals although notably the longevity of ramets may confound this. It is possible that the patterns observed at Gf12 1 may reflect balancing selection (which favours heterozygotes and results in uniformity of alleles even when gene flow is limited, e.g. Karl & Avise 1992). Equally, for P. daedalea, unusually large F ST values at Pd29 2 may also be linked to selection. For corals, many microsatellite loci may actually be under selection or may mirror selection acting on linked gene regions, e.g. through genetic hitchhiking (Slatkin 1995). Certainly, selection could explain the fine-scale population structuring we have observed in these two coral species, which are expected to have at least moderate levels of dispersal. The application of additional markers might clarify the role of selection in structuring coral populations at fine scales and strengthen our conclusions with regard to the scale of larval dispersal, although even this could be problematic. We attempted to amplify a fifth microsatellite locus in both species, but the data proved unreliable and were discarded. Interestingly, one of these loci (Pd61) has been used to compare populations of P. daedalea at scales > 10 km in East Africa (Souter & Grahn 2007), so it may be that the usefulness of microsatellite loci will vary spatially or even temporally. Generally, the application of microsatellite markers in corals has proved problematic, associated with an apparently low abundance of microsatellites linked possibly to the small size of the coral genome (Marquez et al. 2000; Miller & Howard 2004) and the overall conserved nature of the coral genome, resulting in low levels of intraspecific variation (e.g. Romano & Palumbi 1997; van Oppen et al. 1999; Calderón, Garrabou & Aurelle 2006). Recently published studies that have applied microsatellites to population studies of hermatypic corals include data from only three (Maier et al. 2005), four (Magalon et al. 2005), five (Baums et al. 2005; Souter & Grahn 2007) and eight loci (Underwood et al. 2007) and average ~15 alleles per locus. None the less, despite the limitations of few moderately variable loci, the application of microsatellites to reef corals still has the potential to provide a greater understanding of coral biology and ecology than we have currently. Conclusions Differences in life-history traits of marine invertebrates are thought to correlate with larval dispersal ability and to reflect

10 722 K. J. Miller & D. J. Ayre population structure; empirical studies that support this paradigm abound (e.g. see reviews by Palumbi 1995; Bohonak 1999). Equally, however, the high incidence of studies that, like ours, are exceptions to this rule (whereby marine species with apparently limited dispersal show little genetic differentiation (e.g. Hedgecock 1986; France, Hessler & Vrijenhoek 1992; Imron et al. 2007) or where subdivision is apparent in wide-ranging species (e.g. Burton & Feldman 1982; Porter, Ryland & Carvalho 2002) emphasizes the role that factors such as larval behaviour, environment, selection and historical events are likely to have on present-day population genetic structure (Palumbi 1995). In this study we have shown that despite the strongly contrasting spawning mode and larval development in the corals P. daedalea and G. favulus, genetic neighbourhood size and population structure are surprisingly similar in both species. This finding is notable as one of the few studies of marine invertebrates that demonstrate these effects at the scale of metres, and among individuals as well as populations. Withinas well as between-population genetic patterns appear not to be dispersal-driven, based on expectations of planktonic vs. benthic development of larvae. Additionally, locus-specific selection acting on fine scales, linked possibly with high levels of environmental heterogeneity (Ayre & Miller 2004), may well shape the evolutionary potential of reef corals in ways that will not be readily predictable either from life-history traits or large-scale genetic surveys. Our results emphasize that larval development and oceanography alone are not perfect descriptors of dispersal potential or population structure in corals. However, the meso-scale genetic subdivision we have observed in both species highlights the importance of localized recruitment in structuring populations within reefs, and more widely (e.g. Butler & Keough 1990; Strathmann et al. 2002). Quite possibly, the role of benthic development for local recruitment in G. favulus may have been misinterpreted, and its dispersal ability may be more similar to typical broadcast-spawners than had been thought; although notably the likelihood of long-distance dispersal, even in typical broadcast-spawners, is probably low. With an increasing emphasis on connectivity as a basis for management of marine protected areas (e.g. Palumbi 2003), our results caution the application of models that assume both a simplistic relationship between larval type and dispersal in marine invertebrates, and connectivity mediated through widespread dispersal of pelagic larvae. Acknowledgements This research was funded by the Australian Research Council through a postdoctoral fellowship to K. J. M. We thank Craig Mundy for assistance with mapping coral locations and collecting tissue samples for genetic analysis. References Ayre, D.J. (1995) Localised adaptation of sea anemone clones: evidence from transplantation over two spatial scales. Journal of Animal Ecology, 64, Ayre, D.J. & Dufty, S. (1994) Evidence of restricted gene flow in the viviparous coral Seriatopora hystrix on Australia s Great Barrier Reef. Evolution, 48, Ayre, D.J. & Hughes, T.P. (2000) Genotypic diversity and gene flow in brooding and spawning corals along the Great Barrier Reef, Australia. Evolution, 54, Ayre, D.J., Hughes, T.P. & Standish, R.J. (1997) Genetic differentiation, reproductive mode, and gene flow in the brooding coral Pocillopora damicornis along the Great Barrier Reef, Australia. Marine Ecology Progress Series, 159, Ayre, D.J. & Miller, K.J. (2004) Where do clonal coral larvae go? Adult genotypic diversity conflicts with reproductive effort in the brooding coral Pocillopora damicornis. Marine Ecology Progress Series, 277, Ayre, D.J., Veron, J.E.N. & Dufty, S.L. (1991) The corals Acropora palifera and Acropora cuneata are genetically and ecologically distinct. Coral Reefs, 10, Ayre, D.J. & Willis, B.L. (1988) Population structure in the coral Pavona cactus: clonal genotypes show little phenotypic plasticity. Marine Biology, 99, Babcock, R.C. (1991) Comparative demography of three species of scleractinian corals using age- and size-dependent classifications. Ecological Monographs, 6, Babcock, R.C. (1992) Measuring Coral Recruitment. Report no. 7. Bolinao Marine Laboratory, Marine Science Institute, University of the Phillippines, Bolineo, Phillippines. Babcock, R.C., Bull, G.D., Harrison, P.L., Heyward, A.J., Oliver, J.K., Wallace, C.C. & Willis, B.L. (1986) Synchronous spawnings of 105 scleractinian coral species on the Great Barrier Reef. Marine Biology, 90, Babcock, R.C. & Heyward, A.J. (1986) Larval development of certain gamete spawning scleractinian corals. Coral Reefs, 5, Baker, H.D. & Stebbins, G.L. (1965) The Genetics of Colonizing Species. Academic Press, New York. Baums, I.B., Miller, M.W. & Hellberg, M.E. (2005) Regionally isolated populations of an imperilled Caribbean coral, Acropora palmata. Molecular Ecology, 14, Baums, I.B., Miller, M.W. & Hellberg, M.E. (2006) Geographic variation in clonal structure in a reef building Caribbean coral, Acropora palmata. Ecological monographs, 76, Bell, G. (1982) The Masterpiece of Nature: The Evolution and Genetics of Sexuality. University of California Press, Berkeley, CA. Bengtsson, B.O. & Ceplitis, A. (2000) The balance between sexual and asexual reproduction in plants living in variable environments. Journal of Evolutionary Biology, 13, Black, K.P. (1993) The relative importance of local retention and inter-reef dispersal of neutrally buoyant material on coral reefs. Coral Reefs, 12, Black, K.P., Moran, P.J. & Hammond, L.S. (1991) Numerical models show coral reefs can be self seeding. Marine Ecology Progress Series, 74, Bohonak, A.J. (1999) Dispersal, gene flow, and population structure. Quarterly Review of Biology, 74, Butler, A.J. & Keough, M.J. (1990) A comment on short-supply lines. Trends in Ecology and Evolution, 5, 97. Burton, R.S. & Feldman, M. (1982) Population genetics of coastal and estuarine invertebrates: Does larval behavior influence population structure? Estuarine Comparisons (ed V.S. Kennedy), pp Academic Press, New York. Calderón, I., Garrabou, J. & Aurelle, D. (2006) Evaluation of the utility of COI and ITS markers as tools for population genetic studies of temperate gorgonian. Journal of Experimental Marine Biology and Ecology, 336, Cameron, A.M. & Endean, R. (1985) Do long lived species structure coral reef ecosystems? Proceedings of the Fifth International Coral Reef Symposium, 6, Carvalho, G.R. (1994) Evolutionary genetics of aquatic clonal invertebrates: concepts, problems and prospects. Genetics and Evolution of Aquatic Organisms (ed. A.R. Beaumont), pp Chapman & Hall, London. Foster, N.L., Baums, I.B. & Mumby, P.J. (2007) Sexual vs. asexual reproduction in an ecosystem engineer: the massive coral Montastraea annularis. Journal of Animal Ecology, 76, France, S.C., Hessler, R.R. & Vrijenhoek, R.C. (1992) Genetic differentiation between spatially disjunct populations of the deep-sea, hydrothermal vent-endemic amphipod Ventiella sulfuris. Marine Biology, 114, Harrison, P.L., Babcock, R.C., Bull, G.D., Oliver, J.K., Wallace, C.C. & Willis, B.L. (1984) Mass spawning in tropical reef corals. Science, 223, Harrison, P.L. & Wallace, C.C. (1990) Reproduction, dispersal and recruitment of scleractinian corals. Coral Reef Ecosystems (ed. Z. Dubinsky), pp Elsevier, Amsterdam. Hedgecock, D. (1986) Is gene flow from pelagic larval dispersal important in the adaptation and evolution of marine invertebrates. Bulletin of Marine Science, 39,