Introduction

The genetic and demographic implications of connectivity play a vital role in conservation biology. Sessile marine invertebrates such as corals often have a planktonic larval stage that depends on their oceanic environment for dispersal and genetic exchange between populations (e.g. Atchison et al. 2008; van Oppen et al. 2008; Nunes et al. 2011). Larvae of broadcast spawning corals are capable of dispersal up to several 100 km (Rodriguez-Lanetty and Hoegh-Guldberg 2002; Severance and Karl 2006; Nakajima et al. 2012), although localized recruitment is more often dominant (Baums et al. 2005; Severance and Karl 2006; Underwood et al. 2009). Dispersal processes are important drivers of population size and genetic diversity, which are some of the key factors in maintaining population stability and long-term persistence. The understanding of population boundaries, source–sink dynamics, and levels of genetic diversity can, and should, be applied in the prioritization of conservation efforts and design of marine networks (Palumbi 2003), prediction of population resilience to various stresses (reviewed in van Oppen and Gates 2006), and the management of invasive species spread (Darling et al. 2008; Yasuda et al. 2009).

Localized anthropogenic impacts on the coral reefs in Singapore due to coastal development and shipping activities, which intensified during the past five decades of rapid economic growth, have resulted in the loss of approximately 60 % of its coral reefs (Chou 2006). The extent of reclamation is evidenced by the unusually straight edges along the south-western coastline of mainland Singapore, and many reefs have been lost in the merging of adjacent offshore islands (Fig. 1). The remaining reefs, which host up to 255 species of scleractinian corals (Huang et al. 2009), are subject to high levels of chronic sedimentation (Dikou and van Woesik 2006). Furthermore, recent plans (NPTD 2013) indicate more reclamation projects that may impact the remaining reefs. There is therefore a need to understand the dynamics of local coral populations. Prior to the present study, nothing was known about the genetic structure or connectivity of coral populations within the Singapore Strait, or how these factors might influence the resilience of these heavily impacted populations.

Fig. 1
figure 1

Sampling sites within the Singapore Strait are marked by stars. Bottom left panel ten sampling sites within the Southern Islands of Singapore. Bottom right panel three sampling sites along the northern coast of P. Bintan

Full size image

Platygyra sinensis, one of the most common hermatypic coral species in Singapore (Huang et al. 2009), is a key ecological component of the local reefs. It is also one of the many species of local broadcast spawning corals that participate in the mass spawning event just after the wet north-east monsoon in March/April (Guest et al. 2005). Despite the prominence of Platygyra species on tropical reefs, the population genetics of this group is understudied as compared to other broadcasting species such as the acroporids; worldwide, only four studies have thus far analysed the genetic structure of Platygyra populations: in Hong Kong (Ng and Morton 2003), East Africa (Souter and Grahn 2008), the Great Barrier Reef, Australia (Miller and Ayre 2008), and southern Taiwan (Keshavmurthy et al. 2012). Mitochondrial and nuclear gene sequences utilized in the most recently published population genetic study on a Platygyra species might have had inadequate resolving power (Keshavmurthy et al. 2012); the slow rate of mitochondrial DNA evolution in corals (Shearer et al. 2002; van Oppen and Gates 2006; Huang et al. 2008) has mostly restricted coral population geneticists to using allozymes, amplified fragment length polymorphisms (AFLPs), and microsatellites (e.g. Combosch and Vollmer 2011; Brazeau et al. 2011; Keshavmurthy et al. 2012). Furthermore, there are few available microsatellite markers due to the small size of the coral genome (Márquez et al. 2000). Hence, we developed new polymorphic microsatellite markers specific to P. sinensis in this study.

Given the demographic dominance of P. sinensis within the Singapore Strait, their high fecundity, consistent participation during the annual mass spawning event in Singapore, positively buoyant gametes, and relatively long larval duration (almost 6 d; Babcock 1991; Tay et al. 2011), we expected a high level of larval exchange and genetic connectivity—although barriers to gene flow could have arisen due to anthropogenic impacts over the last few decades. Nevertheless, hydrodynamic simulations of broadcast coral larval transport within the Singapore Strait by Tay et al. (2012) also suggested the potential for relatively high connectivity among reefs within the Southern Islands of Singapore. Their model suggested a trend for east-to-west dispersal, likely due to the net westward current within the Singapore Strait during the annual mass spawning event, and indicated that reefs further upstream of the Singapore Strait (such as the Indonesian island Pulau Bintan; hereafter Pulau = P.) may be potential larval sources. However, some genetic structuring was expected between the populations in Singapore and P. Bintan because, for successful dispersal into or out of Singapore, larvae have to cross the >10 km width of the Strait, which is not only perpendicular to the main east-to-west direction of current, but also a major international shipping channel (Chua et al. 2000).

In the present study, we (1) document levels of genetic diversity and connectivity present in populations of P. sinensis in Singapore with comparisons to nearby Indonesian populations within the Singapore Strait, (2) investigate possible impacts on the genetic connectivity of P. sinensis within the Singapore Strait due to anthropogenic activities by comparing historical versus contemporary patterns of genetic structure, and (3) present six new microsatellite markers.

Materials and methods

Study site and sample collection, preservation and DNA extraction

A total of 428 P. sinensis samples were collected over a period of 2 years (2009–2010) from ten sites among the Southern Islands of Singapore, and three sites along the northern coast of P. Bintan, Indonesia (Fig. 1). Sample sites have been abbreviated in figures and tables as follows; Singaporean sites: Raffles Lighthouse (SGRL), P. Biola (SGBIO), P. Pawai (SGPW), P. Sudong (SGSUD), P. Hantu (SGHAN), P. Semakau (SGSEM), P. Jong (SGJO), The Sister’s Islands (SGSIL), Kusu Island (SGKI); Indonesian sites: P. Rawa (BNPR), House Reef (BNHR), P. Maoi (BNPM). Samples from two sites at P. Sudong showed non-significant genetic differentiation and were pooled to increase the sample size for the connectivity analyses.

At each sampling site, a belt transect survey at least 50 m long was performed along the reef crest (~3 m depth) to search for P. sinensis colonies. A separate transect survey was conducted when the sample size from one transect was insufficient. The locations of each transect were determined with a Garmin 76CSx global positioning system (GPS), and the x and y coordinates of each colony with respect to the transect tape were recorded. To help identify any potential cross-contamination of coral DNA extracts with those of their algal symbionts, four samples were collected during the coral bleaching event in 2010 (Guest et al. 2012) and checked under a dissecting microscope to confirm the absence of zooxanthellae. Gonads were also isolated from one gravid sample under a dissecting microscope for additional symbiont-free coral DNA. All coral tissue samples were placed in ~950 µl CHAOS solution (4 M guanidine thiocyanate, 0.1 % N-lauroyl sarcosine sodium, 10 mM Tris buffer pH 8, 0.1 M 2-mercaptoethanol; Sargent et al. 1986) for at least 5 d at room temperature for tissue digestion before DNA extraction using the phenol–chloroform method with a phenol extraction buffer (100 mM Tris buffer pH 8, 10 mM EDTA, 0.1 % SDS; Fukami et al. 2004).

Marker selection and development

Five published microsatellite markers developed for the congeneric species Platygyra daedalea (Miller and Howard 2004) were tested on 38 DNA samples of P. sinensis from Singapore. Polymerase chain reaction (PCR) amplifications were modified from Miller and Ayre (2008) and Mangubhai et al. (2007). PCRs were carried out in 10 µl reactions containing 6 ng of DNA, 1× KAPA Buffer A, 0.8 U KAPA Taq (Genome Holdings), 0.2 mM of each dNTP, 0.5–0.65 µM of each unlabelled primer, and 0.5 mM fluorescence-labelled forward primers. Cycling protocol was as follows: 95 °C for 5 min, 30 cycles at 94 °C for 30 s, 57–63 °C (locus Pd29 at 57 °C, Pd31 at 57 °C, Pd48 at 60 °C, Pd61 at 63 °C, Pd62 at 57 °C) for 30 s, and 72 °C for 1 min, ending with an extra extension period at 72 °C for 10 min. Allelic variations were visualized on an ABI 3100 capillary genetic analyser, and results were analysed using GeneMapper version 4.0 (Applied Biosystems). Of the five markers, only Pd48 gave clear, distinct peaks that were polymorphic and repeatable, despite repeats with an additional PIG-tail on the reverse primer (Brownstein et al. 1996). Hence, of these tested markers, only Pd48 was retained for use.

A new set of microsatellite markers specific to P. sinensis was therefore developed for the population genetic analyses in this study. Microsatellite enrichment and initial marker screening were performed following the protocol described in Lance et al. (2010). PCRs were carried out in 5 µl reactions containing 6 ng of DNA, 1× Qiagen PCR multiplex master mix (final concentration of 3 mM MgCl2,), 0.13–0.17 µM of each fluorescence-labelled universal primer (Plsi3.15: 0.17 µM; Plsi4.02: 0.16 µM; Plsi4.16: 0.13 µM; Plsi4.24: 0.17 µM; Plsi4.27: 0.17 µM; Plsi4.48: 0.13 µM), and 0.05–0.125 µM final concentration of each unlabelled primer (Plsi3.15: 0.125 µM; Plsi 4.02: 0.066 µM; Plsi4.16: 0.05 µM; Plsi4.24: 0.06 µM; Plsi4.27: 0.11 µM; Plsi4.48: 0.09 µM). Cycling protocol was as follows: 95 °C for 15 min, 40 cycles at 94 °C for 30 s, 57–62 °C (specific annealing temperatures in Table 1) for 30 s, and 72 °C for 1 min, ending with an extra extension period at 60 °C for 30 min. Sixty primer pair sequences were screened (2 di-, 10 tri-, 48 tetra-nucleotide repeats; Electronic Supplementary Materials, ESM, Table S1), of which only six gave reliable, consistent peaks that were suitable for scoring. Each marker was also tested to ensure good amplification of the DNA samples extracted from symbiont-free samples. Conditions and characteristics of these markers were calculated using GenAlEx 6.4 (Peakall and Smouse 2006; Table 1).

Table 1 Details for six polymorphic microsatellite loci developed for Platygyra sinensis as calculated from genotypes of 38 individuals
Full size table

Genotyping

Multiplex PCRs (Table 1) were performed in 4 µl reactions containing 3 ng of DNA, keeping the PCR cycling conditions and final concentrations of all reagents the same as stated earlier. Ambiguous alleles, or private alleles detected using GenAlEx 6.4, were re-genotyped to ensure that the alleles were real. Genotyping for difficult samples was repeated up to three times, the last of which used new extracts of DNA samples. Samples that were not successfully genotyped after four attempts were assumed to be affected by null alleles. Plsi4.24 and 4.27 had a higher frequency of missing data (4.4 and 11.6 %, respectively). Scoring was checked manually across the entire data set to ensure consistency across runs.

Statistical analyses

Since many population genetic differentiation analyses rely on the assumptions of random drift mutations and marker neutrality, Lositan (Beaumont and Nichols 1996; Antao et al. 2008) was used to detect loci that may be under positive selection, using 100,000 simulations following the stepwise mutation model. Only Plsi4.27 was found to be under positive selection and was hence excluded from analyses that make assumptions about random drift mutation.

FSTAT v 2.9.3 (Goudet 1995) was used to conduct tests for Hardy–Weinberg equilibrium (HWE) and linkage disequilibrium (LD) for each locus (ESM Table S2) and also to calculate allelic richness, HWE, and inbreeding coefficients (F IS) for each population (Table 2). All significance levels were adjusted with the strict Bonferroni correction for multiple tests (p < 0.05). Private alleles were detected using GenAlEx. FreeNA (Chapuis and Estoup 2007) was used to calculate null allele frequencies under the assumption of HWE. Loci with high frequencies of null alleles (>0.3) were excluded from analyses that are sensitive to null alleles (Chakraborty et al. 1994; Carlsson 2008). BOTTLENECK (Piry et al. 1999) was used to determine whether a recent genetic bottleneck had occurred in all sampled sites, and within all sites in Singapore and P. Bintan. The Wilcoxon’s test was used to test for significant deviations from population mutation-drift equilibrium, following the two-phase model (TPM) with 95 % single-step mutations and a variance among multiple steps of 12 as recommended for microsatellites.

Table 2 Population genetic indices for the 12 populations, and two locations
Full size table

Historical genetic differentiation among all sampled sites was assessed by calculating pairwise F ST values in FSTAT v 2.9.3. As there was no significant genetic differentiation detected for sites within locations (Table 3), sites within locations were pooled for an among-location pairwise F ST calculation. Retrospective power analyses were conducted in POWSIM v 4.1 (Ryman and Palm 2006) to check the adequacy of the remaining markers for resolving genetic structure among the populations, and between locations. Only individuals with complete genotype data were included in these analyses, and genetic drift to various levels of genetic divergence (0 ≤ F ST ≤ 1) for populations of 1,000 and 10,000 individuals was each simulated 1,000 times (Fig. S1).

Table 3 Pairwise F ST values calculated in FSTAT
Full size table

Contemporary differentiation patterns were assessed using the Bayesian clustering analysis implemented in STRUCTURE v 2.3.3 (Pritchard et al. 2000). Although STRUCTURE can also detect deep divergence, it is a better representation of contemporary divergence as it maximizes HWE while minimizing LD within genetic clusters. CONVERT (Glaubitz 2004) was used to format input data for use in STRUCTURE. STRUCTURE, a fully Bayesian clustering method, was used to estimate the most likely number of genetic clusters (K). Since admixture and allele frequency correlation are valid assumptions in the case of subtle population structure (e.g. F ST < 0.05; Falush et al. 2003; Hauser et al. 2006), STRUCTURE analyses were run following these assumptions with no prior population information. Each run consisted of 1,000,000 iterations with a burn-in of 100,000, for each value of K from K = 1–16. For each K, the run was replicated ten times, and the most likely value of K was determined using STRUCTURE HARVESTER (Earl and VonHoldt 2011), which implements the method of Evanno et al. (2005). Due to stochasticity between each run in STRUCTURE, the ten independent iterations at the optimal value of K were matched using the Greedy algorithm in the cluster matching and permutation program CLUMPP (Jakobsson and Rosenberg 2007) and then visualized using DISTRUCT v 1.1 (Rosenberg 2004). Since this first set of runs suggested the possibility of subtle genetic structure between Singapore and Indonesia, the procedure was repeated with these two localities set as priors, for each value of K from K = 1–5. Spatial autocorrelation analyses were also conducted to determine whether fine-scale genetic clustering was present within populations. The autocorrelation coefficient, r, which measures correlation between coral genotypes and the geographic distance between individuals, was calculated for 10-m distance classes in GenAlEx v 6.4, with 999 permutations to test the null hypothesis of no spatial structure (r = 0), followed by 10,000 bootstrap replicates to determine the 95 % CI.

Due to the lack of genetic structuring within locations (Singapore and P. Bintan), population assignments and gene flow calculations were difficult and may have been more prone to errors (e.g. Wilson and Rannala 2003; Piry et al. 2004; Beerli 2009). Hence, even though estimations of migration were tested, they are not presented here.

Results

Genetic diversity

Only one set of two identical multilocus genotypes was detected, which is in accordance with low clonality of this genus as well as with the different morphologies of adjacent colonies observed in situ. Hence, all individuals were included in the analyses. LD was not detected in any locus pair.

All loci were highly polymorphic, mostly with 13–16 alleles (Plsi4.27 had 33 alleles; ESM Table S2). Expected heterozygosities were high, ranging from 0.663 to 0.894. Additionally, allelic richness was high, ranging from 8.59 to 10.19, with a mean allelic richness of 9.33 ± SE 0.17. No significant heterozygote excesses were found, while significant heterozygote deficits were detected in five of the seven loci, and across most populations for Plsi4.24 and Plsi4.27. High inbreeding coefficients (>0.4) were found in Pd48, Plsi4.24, and Plsi4.27. The significant heterozygote deficits for Plsi4.24 and Plsi4.27 were likely due to the high frequencies of null alleles (up to 21.5 ± 0.011 and 25.0 ± 0.014 %, respectively; ESM Table S2). However, null allele frequencies exceeding 0.3 only occurred in Plsi4.24 and Plsi4.27 at P. Jong. Private alleles were detected in six loci; comparing among the 12 sites, low frequencies of seven private alleles were found in the sites in Singapore (0.012–0.045), and three from the sites in P. Bintan (0.014). F IS values were calculated across all loci, and excluding loci with high null allele frequencies (Plsi4.24 and Plsi4.27). As expected, both the F IS values and their significance levels were considerably reduced, although highly significant deviations from HWE persisted in a few sites in Singapore (Table 2). BOTTLENECK analyses found no signs of recent genetic bottlenecks in any of the 12 populations, as indicated by the non-significant probabilities of heterozygote excess (p ≥ 0.71), and the L-shaped mode-shift graphs.

High levels of historical and contemporary genetic connectivity

Low and non-significant historical genetic structure within the nine Singapore sites (overall F ST = 0.005) and the three Indonesian sites (overall F ST = 0) was detected (pairwise F ST values between sites in Table 3). Across the Singapore Strait, very subtle historical structure was detected between Singaporean and Indonesian populations (pairwise F ST = 0.003, p < 0.01), which could be due to the two instances of significant genetic differentiation between P. Hantu in Singapore and two sites off of P. Bintan, although the absolute F ST values were still low in these two comparisons (F ST = 0.006 and 0.011, respectively; Table 3). Retrospective power analyses indicated sufficient power of the markers in resolving the low genetic structure among the individual sampling sites, with >99 % probability of detecting a true F ST = 0.006 between individual sites, but a >91 % probability of detecting a true F ST = 0.003 across the Singapore Strait (ESM Fig. S1).

With just three genetic clusters detected as the optimal number (K = 3), contemporary genetic structure analyses also indicated low genetic structure within the two locations (Fig. 2). All three clusters were detected in every population. However, as with historical differentiation, a subtle difference in genetic composition was found between the Indonesian and Singaporean populations. This structuring was more prominent when location was set as a priori in the STRUCTURE runs. Given the high frequency of genetic admixture at the individual level, however, it is also possible that there is no structure at all (where K = 1). Nevertheless, we found greater support for the subtle genetic divergence (when K = 3) due to the following reasons: (1) the posterior probability (LnPD) values were low at K = 1, and highest (least negative) at K = 3, and (2) the observed mixed population assignments in each individual were not symmetrical as they would be if there were truly no population structure (Pritchard et al. 2000).

Fig. 2
figure 2

Bayesian model-based cluster analysis as implemented by STRUCTURE when K = 3. Each vertical bar in the bar plots represents one sample, with the probability of membership of each individual to a genetic cluster represented by a different shade. Sampling localities are given on the x axis and separated by a black line; the nine sites on the left are from Singapore (SG-) and the three on the right from P. Bintan in Indonesia (BN-)

Full size image

Within sites (≤110 m transect), autocorrelation coefficients were not significantly different from random, including at the first distance class (0–10 m), indicating genetic homogeneity and a lack of fine-scale structuring (ESM Fig. S2).

Discussion

This paper presents the first population genetic analysis of P. sinensis based on highly polymorphic microsatellite loci, of which six were developed for the present study. Consistent with most other broadcast spawning coral species, the populations within the Singapore Strait propagate mainly by sexual reproduction that disperses gametes away from the parental colonies and are genetically diverse. Analyses of contemporary as well as historical genetic structure indicate virtual panmixia within locations (<45 km), and subtle genetic differentiation between Singapore and a nearby Indonesian island, P. Bintan. This increase in genetic divergence suggests that the Singapore Strait is becoming a mild barrier to gene flow.

Population genetic structure

The overall low genetic structure among P. sinensis populations within the 70-km stretch of the Singapore Strait (F ST values ≤0.018) is similar to the only other study of this species (i.e. Ng and Morton 2003: mean F ST of 0.004 across all allozyme loci) conducted in Hong Kong, a site that is also heavily impacted by anthropogenic activities. Even though different markers were used, the similarity of these results suggests analogous population dynamics at both locations. The low genetic structure detected in our study, however, could also be due to the simultaneous formation of local coral reefs during the rapid flooding of the Sunda Shelf (Hanebuth et al. 2000; Bird et al. 2006) and maintained by high gene flow among populations. In support of the low overall F ST, P. sinensis genotypes in this study were randomly distributed from the smallest (0–10 m) to the largest (0–110 m) scale examined, indicating that routine gamete dispersal is >100 m from parental colonies. Interestingly, a previous microsatellite study of P. daedalea on the GBR detected genotype clustering within 5–10 m of each other (Miller and Ayre 2008). Hence, even for closely related species, routine gamete dispersal distances may vary significantly depending on location. Routine transport of larvae >100 m from the parental colonies in Singapore and Indonesia is important from a management perspective, as it indicates that sites are more likely to be well-connected and that panmixia occurs over a larger scale than ‘site’.

Significant heterozygote deficiencies, such as those found in the present study, are common among marine invertebrates, and especially in corals (e.g. Miller and Ayre 2008; Combosch and Vollmer 2011). Ng and Morton (2003) found only a few instances of heterozygote deficits in P. sinensis populations in Hong Kong, but this could be due to the lower allelic richness of allozymes. Heterozygote deficits in Singapore’s P. sinensis populations are likely due to null alleles, as significant reductions in F IS values were observed with the removal of the two loci with the highest null allele frequencies. Inbreeding effects are also possible as significant F IS values persisted in some populations after removal of loci with high null allele frequencies—although this is unlikely due to the high connectivity within and among populations.

Land use and connectivity

Both historical and contemporary analyses suggest that P. sinensis in Singapore is an effectively panmictic population. This high genetic connectivity is reassuring, but surprising given the anthropogenic impacts over the past few decades. P. sinensis has a generation time of ~10 years (Carpenter et al. 2008), within the time frame of the rapid economic and coastal development in Singapore. Land reclamation activities in the 1980s, for instance, caused an annual average of ~0.8 % increase in Singapore’s land area with a spike of ~3.5 % in the early 1980s (Tan et al. 2010). Given that environmental selection can bias the successful establishment of dispersed coral planulae (Hodgson 1990; Marshall et al. 2010), the massive reductions in coral reef populations (Chou 2006) and/or the fragmentation and degradation of the local reef environments (Dikou and van Woesik 2006) were expected to have caused changes in local larval dispersal processes (e.g. reduction in larval output or prevention of successful dispersal and recruitment of larvae), resulting in increased genetic divergence among sites over the past few decades. However, this result was not detected using available analyses and can likely be attributed to the high reproductive capacity and dispersal ability of broadcast spawning corals.

Furthermore, no significant difference in genetic diversity (assessed by paired t test of allelic richness, p = 0.28) was detected between populations in Singapore and the less-disturbed reefs on P. Bintan that were expected to have higher levels of genetic diversity. BOTTLENECK analyses also showed no signs of a genetic bottleneck, indicating that perhaps the anthropogenic activities and loss of coral populations thus far have not been sufficient to cause significant impacts to the local populations of P. sinensis.

At a larger spatial scale across the Singapore Strait, however, the observed patterns in historical and contemporary genetic structuring showed some evidence of increased genetic divergence: a distinct clustering of Singaporean versus Indonesian sites was detected in STRUCTURE analyses, whereas the F ST value, despite being statistically significant, was extremely low at 0.003 (the probability of detecting a true F ST of ≤0.003 with the current available loci is only ≤91.2 %; hence, the actual F ST could potentially be lower; ESM Fig. S1), and the presence of low genetic structure between only one site in Singapore and two sites in P. Bintan suggests weaker historical genetic structure across the Strait. In addition, private alleles found in only either location (21 in Singapore, 5 in P. Bintan; Table 2) are signs of recent genetic divergence and relatively large population sizes. Here, we suggest that the Singapore Strait has been a mild barrier to gene flow due to the (1) strong currents within the Strait, and/or (2) intensive shipping activities in one of the busiest and most important shipping waterways in the world (Chou 2006; Qu and Meng 2011): copper leached from ships can inhibit coral larval motility (Kwok and Ang 2013) and hence dispersal, while navigation-induced wave splash can interfere with the currents and cause larval displacement (Lechner et al. 2013).

Management implications

The high genetic connectivity of P. sinensis populations within the Singapore Strait has management implications at both local and cross-strait levels. For a reef system that potentially consists of multiple sources and sinks such as in Singapore waters, conservation efforts should focus on the prevention of multiple populations from being impacted simultaneously, such as by staggering land reclamation or dredging activities. Prioritization of conservation, if necessary, could target areas with higher genetic diversity such as P. Sudong or Kusu Island. The Sister’s Islands and nearby reefs have recently been designated as the first Marine Park in Singapore. This will help to maintain the genetic connectivity of the local reefs, because this site has been predicted to be a good source site of coral larvae within the Southern Islands (Tay et al. 2012).

Resilience of P. sinensis populations in Singapore waters is likely high, due to its relatively widespread distribution and abundance (large effective population size), high genetic diversity, and indications of continued sexual reproduction. Its high connectivity within Singapore means these populations can likely recover quickly from extremely localized perturbations. This potential for recovery can also be inferred from the recent natural recruitment of corals on the artificial seawalls at P. Semakau within 15 years after their construction (Ng et al. 2012).

Given the high genetic relatedness across the Singapore Strait, the broadcast spawning coral populations in Singapore and northern P. Bintan may be managed as a metapopulation instead of individual reefs. While designating the entire Singapore Strait as an MPA is not realistic both politically and economically, transboundary management efforts between Singapore and Indonesia could include assistance in species reestablishment: species that are low in abundance or have become locally extinct in Singapore, for instance, could be re-established via transplantation of coral colonies from P. Bintan and vice versa, without much disruption to the genetic composition of these populations.

The adaptive abilities and long-term viability of Singapore’s coral populations are nevertheless a concern given a wider context, i.e. the effects of climate change (e.g. two catastrophic thermal-bleaching events in 1998 and 2010; Guest et al. 2012), constant direct human pressures such as high sedimentation rates (Dikou and van Woesik 2006), and lack of good quality larval settlement sites. Even though their high genetic diversity can promote thermal-tolerance either directly or via internal symbionts (i.e. genetic resilience; Császár et al. 2010; LaJeunesse et al. 2010; Baums et al. 2013), as a whole, the reefs within the Singapore Strait may be vulnerable to large-scale disturbances. Recent evidence shows that it is the host, not the endosymbiont, that determines the coral’s upper thermal-tolerance threshold (Wooldridge 2014), and as they share the same gene pool, they are probably susceptible to diseases or other impacts, as suggested by the mass bleaching of massive Porites coral populations in both Singapore (Guest et al. 2012) and P. Bintan (Y.C. Tay pers obs.) during the 2010 high sea surface temperature event. Furthermore, if the Singapore Strait becomes a stronger barrier to gene flow and Singapore’s reefs become reliant on self-seeding, a large-scale disturbance could devastate the local population. Management efforts could therefore focus also on understanding the causes, and mitigating the effects, of the genetic barrier across the Singapore Strait.