Introduction

Identifying genetic differentiation between coral populations within archipelagic systems such as the Great Barrier Reef (GBR) is important for understanding the potential for adaptation on these discrete reefs. The GBR spans 2,100 km and almost 15° of latitude along the continental shelf off the north east coast of Australia. This complex reef system can be divided into three latitudinal regions: northern (∼10°S to 15°S), central (∼15°S to 20°S) and southern (∼20°S to 24°S). Inshore and offshore reefs occur in each region and those located offshore occur on both the mid- and outer-continental shelf. The proximity of inshore reefs to the coast makes them particularly vulnerable to anthropogenic stressors and as many as 28% of inshore reefs on the GBR have been affected by increased sediment runoff since the 1800s (Williams 2001; McCulloch et al. 2003). These unique ecosystems are also threatened by rising sea temperatures associated with global climate change.

Prolonged high temperature exposure results in coral bleaching, a generalised stress response that often leads to coral mortality (see reviews by Glynn 1993; Brown 1997; Hoegh-Guldberg 1999). Varying climate change scenarios have been used to model future trends in sea temperature for the GBR and these predict that rising temperatures will cause serious set backs to the ecology and appearance of reefs within a few decades (Hoegh-Guldberg 1999, 2004; Done et al. 2003). Extended periods of high seawater temperature have already caused mass coral bleaching on the GBR, with the most extensive episodes recorded in 1998 and 2002. Inshore reefs bleached more frequently and more severely than offshore reefs during both these events (Berkelmans and Oliver 1999; Berkelmans et al. 2004). Substantial drops in live coral cover were recorded on a number of inshore reefs because of these past bleaching events (Wilkinson 2000; Berkelmans 2001).

High levels of bleaching and mortality on inshore reefs of the GBR are a consequence of greater temperature extremes in shallow coastal waters during mass bleaching events (Berkelmans 2001; Berkelmans et al. 2004). However, the inshore reefs of the GBR experience typically warmer conditions than offshore reefs even during summers when no bleaching is observed. A recent principal components analysis of sea surface temperatures (SSTs) revealed intersecting latitudinal and cross-shelf temperature gradients on the GBR (Done et al. 2003). This analysis showed five zones of similar maximum-monthly sea temperature during the 1990s (reproduced in Fig. 1) and reef-building corals on the GBR show signs of acclimatisation and/or adaptation to these thermal regimes. Corals on inshore reefs, as compared to offshore reefs at similar latitudes, tolerate warmer temperatures and longer periods of high temperatures before bleaching (Berkelmans 2002). Higher temperatures and/or exposure times also occur before bleaching is observed on the comparatively warm reefs of the northern GBR compared to cooler southern reefs (Hoegh-Guldberg 1999; Berkelmans 2002).

Fig. 1
figure 1

Sampling locations and temperature zones on the Great Barrier Reef (GBR). a Reef locations, PCB Princess Charlotte Bay, MI Magnetic Island, DR Davies Reef, WS Whitsunday Islands, and KI Keppel Isles. b Location of five temperature zones on the GBR from Done et al. (2003). c Zones 1–5, based on warmest monthly-average Sea Surface Temperature (SST) from 1990 to 2000 (excluding 1998). Box colour matches colours in panel (b). Boxes show 25th and 75th percentiles, lines mark median temperature and whiskers are 90th and 10th percentiles. b and c are adapted with permission from Done et al. (2003)

Full size image

Localised adaptation of widespread coral species to different thermal regimes is likely to be driven by genetic variability, natural selection and restricted gene flow between geographically isolated populations. The present study investigates patterns of genetic differentiation between eight inshore populations of the reef-building coral Acropora millepora over ∼1,300 km on the GBR. A mid-shelf reef in the central GBR was included for comparative purposes. The nine reefs studied are distributed among all three latitudinal regions of the GBR and they fall within three of the five GBR temperature zones reported by Done et al. (2003). These same reefs were used for physiological studies of thermal tolerance and bleaching susceptibility of A. millepora (Smith 2005). Patterns of genetic differentiation are therefore interpreted in the context of geographic patterns of thermal tolerance and local thermal regimes.

Materials and methods

Sample collection

Acropora millepora colonies were sampled from one mid-shelf reef and eight inshore reefs spread over ∼10° of latitude and three different temperature zones on the GBR (Fig. 1). Reefs are grouped into five general locations: Princess Charlotte Bay (PCB), Magnetic Island (MI), Davies Reef (DR), Whitsunday Islands (WS) and Keppel Isles (KI). PCB and MI are found inshore in the northern and central GBR respectively and both occur in the warmest temperature zone (zone 5). DR on the central GBR mid-shelf, and WS located inshore on the border of the central and southern GBR, lie within an intermediate temperature zone (zone 3). KI is found inshore on the southern GBR in the second coolest temperature zone (zone 2). Up to three reefs were sampled at most locations and samples sizes range from 20 to 49 colonies per reef (see Table 1 for details on reef locations and sample sizes).

Table 1 Reef locations and summary genetic diversity statistics. Reefs are grouped by sampling location; ‘PCB’ Princess Charlotte Bay, ‘MI’ Magnetic Island, ‘DR’ Davies Reef , ‘WS’ Whitsunday Passage and ‘KI’ Keppel Isles
Full size table

Collections were made during November and December 2001. All collections were at depths of 2-3 m from a single area of reef flat of ∼1,500 m2. Single branches (∼5 cm long) were removed from A. millepora colonies that were separated by a minimum of 5 m. Sampling over this spatial scale should limit the inclusion of genetically identical colonies (clone mates) produced via fragmentation of coral colonies (see Smith and Hughes 1999). Branches were snap frozen in liquid nitrogen immediately after each dive and were kept frozen at −80°C for less than 6 months prior to preparation for electrophoresis.

Allozyme electrophoresis

Host coral tissue extracts were prepared by grinding coral branches on ice in 10% sucrose (w/v), 0.1% 2-mercaptoethanol. Intact symbiotic algae were removed by centrifugation at 16,000g for 30 s. Extracts were snap frozen in liquid nitrogen and kept frozen at −80°C prior to laboratory analysis. Electrophoresis for all allozyme loci was performed simultaneously for each extract within 48 h of preparation. Electrophoresis was carried out on standard 12% (w/v) horizontal starch gels (e.g., Richardson et al. 1986). Five polymorphic loci previously identified by Ayre and Hughes (2000) were analysed, two malate dehydrogenase loci (Mdh-1, Mdh-2; EC 1.1.1.37), two leucyl-tyrosine peptidase loci (Ltp-2 and Ltp-3; EC 3.4.11/13) and leucyl-proline peptidase (Lpp; EC 3.4.11/13). An additional polymorphic locus, malic enzyme (Me; EC 1.1.1.40) was also identified and scored.

Malic enzyme was run overnight at 80 V in Tris–malate pH 7.4 buffer; gel buffer 10 mM Tris, 10 mM maleic acid, 1 mM EDTA, 1 mM MgCl2, 12.5 mM NaOH, electrode buffer 100 mM Tris, 100 mM maleic acid, 10 mM EDTA, 10 mM MgCl2, 125 mM NaOH. Malate dehydrogenase was run overnight at 80 V in Tris–citrate buffer pH 8.0; gel buffer 27 mM Tris, 6.2 mM citric acid, electrode buffer 172 mM Tris, 39 mM citric acid. Tris–malate and Tris–citrate buffers were modified from (Selander et al. 1971). Peptidases, Ltp and Lpp, were run overnight at 110 V in Tris–EDTA–borate pH 8.4 buffer modified from Boyer et al. (1963); gel buffer 48 mM Tris, 1 mM EDTA, 37 mM boric acid, electrode buffer 150 mM Tris, 3 mM EDTA, 117 mM boric acid. Staining solutions were applied as agar overlays and followed standard recipes for all loci (e.g., Shaw and Prasard 1970; Richardson et al. 1986).

Reference samples were repeated on all gels and additional gels run to confirm the scoring of alleles. Samples that failed to show activity for Mdh-1, Me, Ltp-3 or Lpp were re-run. New extracts were also prepared for several such samples to check for extract degradation. Identical patterns of isozyme activity were obtained with both the original and freshly prepared extracts. With few exceptions, these re-run and/or re-extracted samples also failed to show activity for the locus in question despite displaying strong activity for all other loci. Individuals that failed to show activity on two independent gels and which showed strong activity for at least four other loci were scored as putative null (non-active) homozygotes. Loci and visible alleles were scored alphabetically in order of decreasing mobility following the conventions of Richardson et al. (1986).

Data analyses

Departures of genotype frequencies from Hardy-Weinberg Equilibrium (HWE) were calculated as F IS (Weir and Cockerham 1984), where positive and negative numbers represent deficits or excesses of heterozygotes, respectively. The significance of departures from HWE were determined by an exact probability test (Guo and Thompson 1992) employing 100,000 Markov chain iterations. Statistical independence of loci was evaluated using a Fisher exact test of linkage disequilibrium for all pairs of loci in each population using 100,000 Markov chain iterations. These tests together with F IS calculations and associated probability tests were carried out using GENEPOP Version 3.1d (Raymond and Rousset 1999).

Due to strong evidence for null alleles for Mdh-1, Me, Ltp-3 and Lpp (i.e. consistent failure of activity at just one of the six loci for some individuals) additional confirmation and assessment of null allele frequencies at each reef was undertaken using MICRO-CHECKER software version 2.2.1 (van Oosterhout et al. 2004a). This software assesses the likelihood of a null allele based on consistent deficits of heterozygote genotypes for each of the visible alleles at a given locus. For all four loci the presence of a null allele was supported by this analysis. Null allele frequency was calculated using Brookfield estimator 2 (Brookfield 1996) which takes into account both the frequency of null homozygotes and the deficits of heterozygotes. Observed allele and genotype frequencies were then adjusted for each reef according to null allele frequency and assuming Hardy-Weinberg proportions as described in van Oosterhout et al. (2004b). A table of the adjusted and unadjusted allele frequencies can be found in Appendix 1.

Analysis of molecular variance (AMOVA) was used to partition variance in allele frequencies between individuals within reefs, individuals within locations and individuals between locations (Peakall and Smouse 2001). The significance of differences in allele frequencies between all pairs of reefs was determined by exact tests for population differentiation (Raymond and Rousset 1995). Exact tests were performed across all loci and for each locus separately as implemented in TFPGA version 1.3 (Miller 1997). Patterns of genetic differentiation between all pairs of reefs was further examined by calculation of Weir and Cockerham’s (1984) unbiased estimate of F ST which takes into account sample size at each reef. Bootstrapping over loci for 5,000 replicates was performed to determine 95% confidence intervals for F ST using GDA version 1.1 (Lewis and Zaykin 2002). Significance of F ST was demonstrated when lower 95% confidence intervals did not overlap with zero. Plots of pairwise F ST versus geographic distance were examined for an isolation-by-distance effect and Mantel’s tests were conducted using GenePop Version 3.1d (Raymond and Rousset 1999). All tests for population genetic differentiation were based on observed allele frequencies for the two loci for which null alleles were not observed (Mdh-2 and Ltp-2) and adjusted allele frequencies for those affected by a null allele (Mdh-1, Me, Ltp-3 and Lpp). Sequential Bonferroni correction for α = 0.05 (Rice 1989) was used to assess the significance of all multiple tests.

Results

Genetic variation within reefs

There was a high level of genetic diversity within the reefs examined. The average heterozygosity observed within each reef was high, ranging from 0.564 ± 0.127 (mean ± SE) to 0.714 ± 0.054 after adjustment for null alleles (Table 1). Significant deficits of heterozygotes (c.f. HWE expectations) were only observed for the four loci for which there was evidence of a null allele, i.e. Mdh-1, Me, Ltp-3 and Lpp (Table 2). There was only one significant excess of heterozygotes and this occurred at North Keppel Island (KI-3) for Ltp-2 (Table 2). With the exception of Davies Reef (DR) which had significant heterozygote deficits at four of the six loci screened, all other reefs had single locus genotype frequencies that were within HWE expectations for at least four loci. Tests for linkage disequilibrium suggested that all loci were independent with no significant instances of linkage disequilibrium detected following sequential Bonferroni correction of 149 statistically valid tests.

Table 2 Single locus heterozygote deficiencies/excesses (F IS) for six allozyme loci calculated from unadjusted genotype frequencies in A. millepora from nine reefs on the GBR
Full size table

The inclusion of colonies generated asexually via fragmentation was largely avoided by sampling A. millepora colonies on the reef flat at >5 m spatial scales as fragments of A. millepora on reef flat habitats rarely migrate further than 4 m and typically have low survival and reattachment rates (Smith and Hughes 1999). There were only three instances of multi-locus genotypes being repeated within reefs. The repeated genotypes only occurred twice each in all instances. Colonies with the same genotype were assumed the product of asexual reproduction in Nelly Bay (MI) as in these two instances the repeated genotypes had been sampled immediately adjacent to each other on the reef flat. The second individual of each repeated genotype was excluded from the data set. One multi-locus genotype was repeated on the reef flat of Halfway Island (KI-1) but the two colonies of the same multi-locus genotype were well separated on the reef flat in this case. These colonies were therefore considered the product of sexual reproduction as their identical genotypes may reflect the low number of alleles observed for each locus and the limited portion of the genome examined.

Geographic patterns of genetic differentiation

Allozyme electrophoresis of six loci indicated that a small but significant amount of genetic differentiation in A. millepora is due to variation between reefs and between locations (PCB, DR, MI, WS and KI) on the GBR. Analysis of molecular variance indicated that 97% of the variance in allele frequencies was due to variance within reefs (P = 0.002). A further 1% of the variance (P = 0.002) was attributable to variance between reefs within a location and the remaining 2% was due to variance between broad scale geographic locations (P = 0.001). There were no fixed allele differences between any of the reefs examined and only a single private allele (Ltp-3 e) occurred at low frequency at Magnetic Island. There was, however, evidence for significant allele frequency differences between several of the reefs examined (Table 3). The geographic distribution of this genetic variation did not fit an isolation-by-distance model as differentiation between reefs (F ST) did not increase with increasing geographic separation (Mantel’s test P > 0.05, Fig. 2a). The allele frequencies over all six loci were significantly different for Davies Reef on the GBR mid-shelf compared to that on all eight inshore reefs including Magnetic Island ∼90 km away (Table 3). Two to three different loci contributed to these differences (Table 3). Genetic differentiation of Davies Reef from the inshore reefs is also evident from small but significant pairwise F ST estimates (ranging from 0.032 to 0.073, Table 3).

Table 3 Pairwise measures of genetic differentiation. Below diagonal is average pairwise F ST calculated over six loci
Full size table
Fig. 2
figure 2

Geographic patterns of genetic differentiation among Acropora millepora populations at nine locations on the Great Barrier Reef including one mid-shelf and eight inshore reefs. F ST over six allozyme loci versus geographic distance between 45 pairs of reefs

Full size image

There was also evidence for genetic differentiation of inshore reefs along the length of the GBR. The three inshore reefs in the Keppel Isles (southern GBR) were not significantly differentiated from one another but each had a significantly different allele frequency from the two northern-most reefs within Princess Charlotte Bay (Table 3). The difference in allele frequency was significant over all loci and for between one to three individual loci (Table 3). The Keppel Isles reefs also had significantly different allele frequency from Magnetic Island in the central GBR involving between one and three different loci (Table 3). Pairwise F ST estimates support the differentiation between reefs in the Keppel Isles and the inshore reefs of the northern and central GBR with small but significant F ST values (ranging from 0.023 to 0.077, Table 3). UPGMA clustering supports the relative isolation of the reefs in the Keppel Isles (data not shown). In contrast, the six allozyme loci examined gave no indication of genetic differentiation between the two inshore reefs in Princess Charlotte Bay and between these reefs and the reef in Nelly Bay, Magnetic Island (Table 3).

There was no consistent pattern of genetic similarity between fringing reefs of the Whitsunday Islands, located at the southern end of the central GBR, and inshore reefs located to the north and south. Two reefs within the Whitsunday Island group, Daydream Island (WS-1) and Mid Molle Island (WS-2) were sampled within ∼2 km of one another. The A. millepora populations on these two reefs had significantly different allele frequencies from one another, particularly at the Mdh-1 locus (Table 3). The A. millepora population at WS-1 appears less differentiated from reefs to the north (F ST from 0.009 to 0.017) compared to those reefs to the south (F ST from 0.033 to 0.053). Significant allele frequency differences between WS-1 and the southern reefs in the Keppel Isles were detected for up to three loci (Table 3). None of the six loci showed significant allele frequency differences between WS-1 and the inshore reefs to the north. In contrast, there was a tendency for A. millepora at WS-2 to be more differentiated from reefs to the north (non-significant pairwise F ST from 0.013 to 0.032) compared to inshore reefs to the south (non-significant pairwise F ST from 0.001 to 0.012). There were significant allele frequency differences for at least one locus (Mdh-1) between WS-2 and the three inshore reefs to the north. Fewer instances of significant allele frequency differences could be detected in comparisons between WS-2 and reefs of the Keppel Isles to the south (Table 3). An apparent inconsistency between the lack of significance of pairwise F ST and the significance of exact tests for allele frequency differences may be the result of intra-locus sampling error associated with small sample sizes and the low values of F ST obtained in this case (Waples 1998).

The frequency of null alleles made an important contribution to the patterns of genetic differentiation observed between A. millepora populations on the GBR. Differences in the frequency of null alleles at a number of loci may be largely responsible for the differentiation observed between Davies Reef on the GBR mid-shelf and the inshore reefs studied. The apparent absence of null alleles for Me at Davies Reef (Fig. 3b) and the relatively high frequency of null alleles for Lpp at the same reef (Fig. 3c) are likely contributors to the pattern of differentiation described. The latitudinal pattern in genetic differentiation may also be attributed to differences in the frequency of null alleles at a number of loci. A relatively high frequency of Mdh-1 null alleles indicated for the three Keppel Isles reefs (Fig. 3a) combined with a low frequency of Me null alleles for the same reefs (Fig. 3b) contributes to the pattern of relative genetic isolation of these inshore reefs on the southern GBR.

Fig. 3
figure 3

Spatial distributions of alleles for a Mdh-1; b Me and c Lpp enzyme encoding loci among nine Acropora millepora populations on the Great Barrier Reef (GBR). Pie charts indicate allele frequency for reefs as in Table 1 and latitudinal regions of the GBR are indicated on the far left. A vertical bar separates the inshore and mid-shelf reef sampled in the central region

Full size image

Discussion

Population genetic structure is the product of opposing forces that lead to either genetic homogenization (i.e. gene flow) or divergence (i.e. genetic drift, mutation and natural selection). Neutral genetic markers, like the allozyme markers examined in the present study, are useful indicators for the extent of genetic divergence resulting from gene flow and genetic drift (Slatkin 1987). A latitudinal component of genetic differentiation was observed among the eight inshore populations of A. millepora examined. There was a consistent pattern of small but significant differentiation of the three inshore reefs at the southern end of the GBR (KI-1, 2 & 3) from the three inshore reefs in the northern and central GBR (PCB-1, 2 & MI). The extent of this differentiation (F ST = 0.048 ± 0.005; mean ± SE) was similar to that observed between Davies Reef on the mid-shelf and the inshore reefs (F ST  = 0.051 ± 0.005). Putative null alleles contribute strongly to the patterns of genetic divergence described, however, emerging molecular data support the relative separation of warm-northern versus cool-southern populations of A. millepora. A recent examination of four intron sequences, using single strand conformational polymorphism (SSCP) analysis and a sub-set of the present samples, supports the genetic differentiation of PCB and MI from WS and KI (Ridgeway and Smith, unpublished data).

Gene flow is typically moderate to high, and the extent of population differentiation correspondingly low, for marine organisms with pelagic larval stages (Waples 1998; Luttikhuizen et al. 2003). In the present study, moderate levels of gene flow (F ST = 0.023 and maximum F ST = 0.077), are inferred between the northern A. millepora populations (Magnetic Island and Princess Charlotte Bay) and the cool southern GBR populations (Keppel Isles). In contrast, inferred levels of gene flow between the two northern reefs within Princess Charlotte Bay, and between both of these reefs and the centrally located Magnetic Island are very high (F ST not significantly different from zero), as is gene flow among the most southern reefs at the Keppel Isles. Levels of gene flow in excess of one migrant per generation are sufficient to prevent extensive population divergence (i.e. fixation of alleles) at neutral loci via genetic drift (reviewed in Slatkin 1987). The low level of population differentiation observed between the northern/central and southern regions of the GBR in the present study are consistent with weak genetic drift in the face of moderate levels of gene flow. The lack of complete panmixis does, however, suggest that past and present day barriers to gene flow have influenced genetic structure of the coral populations examined.

During past low sea level stands the GBR fauna from the northern/central and southern GBR retracted to two separated plateaus, namely the Queensland and Marion Plateaus, respectively (Davies 1994; Doherty et al. 1995; Worheide et al. 2002). Past separation and relatively slow expansion of southern GBR fauna against prevailing southward currents is a likely contributing factor to present day genetic differentiation of northern versus southern populations of several reef fish species (Doherty et al. 1995), and between northern/central and southern GBR populations of the calcareous sponge Leucetta chagosensis (Worheide et al. 2002). In the present day, a persistent frontal convergence zone to the south of the Whitsunday Islands (Burrage et al. 1996) may also limit gene flow between northern and central versus southern reefs on the GBR. A study of five reef building corals on the GBR mid-shelf showed greater differentiation between a central and southern reef (mean F ST = 0.114) than between a northern and central reef (mean F ST = 0.035) (Ayre and Hughes 2004). Genetic differentiation of southern populations of two species of clams, Tridacna maxima and Tridacna derasa, on the outer GBR shelf has also been reported (Benzie 1994). Although levels of differentiation between northern/central and southern GBR reefs are generally low, several coral reef invertebrates across the GBR shelf show evidence of genetic differentiation between these regions. If this differentiation is due to present day restrictions to gene flow then the southern inshore reefs of the GBR may be less resilient to rising sea temperatures than the more highly connected northern and central GBR reefs. This is especially important as global climate change may see the warm temperatures currently experienced in the northern GBR extending to the southern GBR by ∼2050 to 2100 (Done et al. 2003).

The geographic patterns of genetic differentiation reported for A. millepora on the GBR are consistent with a recent physiological study. Smith (2005) examined the bleaching susceptibility and thermal tolerance of A. millepora from four of the same locations (PCB, MI, WS and KI) used in the present genetic study. A thermal stress experiment conducted after several weeks of laboratory acclimation under identical temperature conditions identified the relative thermal tolerance of A. millepora from these locations as Princess Charlotte Bay = Magnetic Island > the Whitsunday Islands > the Keppel Isles (Smith 2005). This ranking of thermal tolerance is in keeping with local thermal regimes (as shown in Fig. 1) and the pattern of genetic differentiation reported here for the host coral populations. In particular, the high and equal thermal tolerance ranking of A. millepora from Princess Charlotte Bay and Magnetic Island is consistent with the lack of genetic differentiation between these populations and their co-occurrence in the warmest temperature zone. The genetic differentiation of the Keppel Isles from both of these populations is also consistent with its low thermal tolerance (Smith 2005) and its location within the second coolest temperature zone on the GBR.

The present allozyme study has examined patterns of genetic differentiation for just one component of the coral–algal holobiont, namely the coral host. Specific taxonomic differences in symbiotic algal strain are also observed for A. millepora in the locations examined (Smith 2005). In particular, the Magnetic Island population of A. millepora hosts Symbiodinium Clade D algal symbionts (as determined by nuclear ribosomal DNA ITS1 type). Possession of this particular clade of symbionts may contribute to relatively high thermal tolerance in A. millepora (Berkelmans and van Oppen 2006) and other scleractinian corals (Glynn et al. 2001; Baker et al. 2004; Fabricius et al. 2004; Rowan 2004). However, Symbiodinium D is not the predominant algal strain in thermally tolerant populations of A. millepora in Princess Charlotte Bay. The warm-adapted population of A. millepora in Princess Charlotte Bay in fact shares Symbiodinium C2 symbionts with the less thermally tolerant Whitsunday Islands population (Smith 2005). Dissimilarity between the geographic distribution of algal symbiont strains and thermal tolerance of A. millepora populations suggests factors other than algal genotype may determine thermal tolerance in most of the populations examined.

Strong natural selection will overcome the homogenizing effects of gene flow (Wright 1931; Slatkin 1987) and population structure of quantitative traits (i.e. Q st) often exceeds the extent of genetic differentiation observed with allozyme and other neutral markers (Merila and Crnokrak 2001). Natural selection in the face of moderate to high gene flow is known to have shaped quantitative traits such as salinity tolerance in toads (Gomez-Mestre and Tejedo 2004) and morphological variation in mollusks (Wilding et al. 2001; Luttikhuizen et al. 2003). Thermal tolerance is a quantitative trait for which an underlying genetic component has been identified in other organisms (e.g. Ottaviano et al. 1991; Perry et al. 2001; Hoffmann et al. 2002). In the case of Drosophila melanogaster populations along the east coast of Australia, divergent selection from environmental temperature extremes has been implicated as the driving force behind a latitudinal cline in thermal tolerance (Hoffmann et al. 2002) and a corresponding cline in the frequency of alleles encoding classic heat shock response proteins (Bettencourt et al. 2002; Frydenber et al. 2003). Given the concordant patterns of thermal tolerance, slight genetic differentiation at neutral allozyme markers and divergent summer temperature regimes identified in the present study it is therefore possible that natural selection plays a role in shaping the thermal tolerance of A. millepora on the GBR.

The present study of A. millepora raises the interesting possibility that thermal tolerance in this reef building coral species could have an underlying genetic component on which natural selection may operate. Determination of the extent to which natural selection shapes thermal tolerance in this species will require additional studies on a greater number of populations throughout the range of thermal regimes on the GBR. Such studies should examine phenotypic variance in thermal tolerance between populations, correlations between this quantitative trait and both upper and lower temperature extremes, as well as patterns of genetic differentiation at candidate thermo-resistance genes or associated quantitative trait loci. Controlled experiments estimating the heritability of thermal tolerance in a range of populations would also be of substantial value given recent debate about the impact of global warming and rising sea temperatures (including extreme temperature events) on coral reefs. In light of the present data it is arguably time to move towards a more quantitative genetics approach aimed at providing a deeper understanding of the heritability and molecular-genetic basis of variance in thermal tolerance within coral species like A. millepora.