Population genetic structure between Yap and Palau for the coral Acropora hyacinthus

Study species

Acropora hyacinthus was an abundant tabular coral found on the reef slopes of Palau prior to the 1998 bleaching event. During their 2001 assessment, Bruno et al. (2001) estimated a near complete loss of this coral in the areas they sampled. By 2005, Golbuu et al. (2007) observed that the same species was dominant in the shallow reef slopes, raising the question of where the larvae originated from to allow for such a successful recovery and thereby making A. hyacinthus and ideal candidate to study coral connectivity.

A. hyacinthus is usually found between 3 and 10 m deep on barrier reefs and is readily identifiable by the rosette formation of its calices (Fig. 1). A. hyacinthus is a hermaphrodite broadcast spawning coral that produces feeding larvae (Toh, Peh & Chou, 2013) with a pelagic larval duration of approximately 90 days under laboratory conditions (Márquez et al., 2002). Little is known, however, about the pelagic duration or swimming behavior of larvae in the field, which makes realistic incorporation of biological parameters into oceanographic models difficult (Paris, Chérubin & Cowen, 2007; Woodson & McManus, 2007).

Sampling locations and methodology

In May 2012, three sites on the outer barrier reef of Palau were sampled at a shallow depth (<10 m) using SCUBA (Figs. 2A and 2B). Sites were selected to represent a range of habitats, exposures and management categories found on the barrier reef: S17 “West Palau” within a fully protected no-take area on the west side, S20 “North Palau” at the tip of the northern lagoon in a less strictly protected area and S24 “East Palau” on the east coast in a reef impacted by anthropogenic stressors (Golbuu et al., 2012). At each of these three sites, 48 colonies of A. hyacinthus were collected haphazardly in a 4 × 200 m belt transect, for a total of 144 colonies from Palau. One small branch tip (<1 cm) was cut and preserved in salt-saturated DMSO at room temperature (Gaither et al., 2010). In addition, a total of 132 samples were collected from three different sites around Yap, and another 46 colonies were sampled from a single site at Ngulu as described in Davies et al. (2015, Table 1, Figs. 2A, 2C and 2D). Importation was permitted by the Convention on International Trade in Endangered Species of Wild Fauna and Flora permit #FW 12-091.

DNA extraction and sequencing

For colonies from Palau, Genomic DNA was isolated following the DNeasy 96 Blood & Tissue Kit (Qiagen, Valencia, CA, USA) protocol. Two sites with 48 colonies each were extracted on a 96 well plate. For each colony, a 2 mm³ piece of coral from the tips of one branch was ground and incubated overnight at 55 °C in 180 µl of Qiagen Lysis buffer and 20 µl of Qiagen Proteinase K (600 mAU/ml). DNA was eluted in 200 µl of PCR grade water, with 100 µl frozen until amplified and 100 µl saved for future use at −20 °C. A sample of 50 DNAs was quantified on a SpectraMax (Molecular Devices, Sunnyvale, CA, USA) absorbance microplate reader using Accuclear Ultra High Sensitivity quantitation Kit (Biotium, Hayward, CA, USA). Quantities of DNA were very similar for all colonies, ranging from 5 to 10 ng/µl. The DNA of colonies originating from Yap and Ngulu were isolated according to Davies et al. (2015).

Eighteen microsatellite loci developed by Wang, Zhang & Matz (2009) were selected for this study (Table S1). We designed short tags according to Faircloth & Glenn (2012) to create 48 unique colony IDs on the forward primer for each microsatellite locus (Table S2). Polymerase chain reactions (PCRs) for each colony were performed in a 25 µl reaction volume containing 12.5 µl of MyTaq™ Red Mix (Bioline, Taunton, MA, USA), 1 µl of each forward and reverse primer at 5 µM, 1 µL of genomic DNA (5–10 ng/µl) and 9.5 µl of water. Thermal cycling followed a touch-down protocol with an initial denaturation at 95 °C for 3 min followed by 20 cycles of 95 °C for 15 s, 63–55 °C for 15 s (annealing temperature reduced 0.4 °C each cycle), 72 °C for 30 s followed by 20 cycles of 95 °C for 15 s, 55 °C for 15 s, 72 °C for 30 s with a final elongation at 72 °C for 3 min.

Two µl of each uniquely barcoded amplicon were pooled by site for subsequent library construction and sequencing. The pooled samples were concentrated by reducing the total volume using Amicon Ultra 0.5 mL centrifugal filters (EMD Millipore, Darmstadt, Germany) and cleaned using Agentcourt Ampure XP (Beckman Coulter Inc., Brea, CA, USA) to eliminate traces of dye and unincorporated dNTPs and primers. Genomic libraries were generated using the KAPA Hyper Prep kit (Kapa Biosystems, Wilmington, MA, USA). A unique Illumina adaptor (Illumina Inc., Hayward, CA, USA) was ligated to each pool of individually barcoded amplicon samples, creating a site-specific tag (ID) and generating the following unique structure: siteID-colonyID-FWDprimer-flankingregion-tandemrepeats-flankingregion-RVSprimer. Libraries were sequenced on an Illumina MiSeq at the Hawai‘i Institute of Marine Biology (HIMB) Evolutionary Genetic core facility.

Bioinformatics (Table S3)

Splitting flanking region sequences and simple tandem repeats

We separated simple tandem repeats (STR) from their flanking regions using emboss: etandem (Rice, Longden & Bleasby, 2000) and created genotypes based on the STR length.

Analysis

Bottleneck

To test the hypothesis that A. hyacinthus on Palau has recovered from a pulse event of larval dispersal following the 1998 bleaching mortality, we looked for evidence of a recent bottleneck or founder effect using bottleneck 1.2.02 (Piry, Luikart & Cornuet, 1999). In populations that have experienced such an event, rare alleles are the first to be lost, lowering the mean number of alleles per locus. Heterozygosity is less affected, however, producing a transient excess in heterozygosity relative to that expected given the resulting number of alleles (Cornuet & Luikart, 1996). We used the graphical test from Luikart et al. (1998) based on a mode shift away from an L-shaped distribution of allelic frequencies to assess whether evidence of recent population bottlenecks could be detected, which is most appropriate for these types of data (Chaïr et al., 2011). Because we scored fewer than 20 microsatellite loci, we used the Wilcoxon signed-rank test (10,000 iterations) using both a two-phase (TPM incorporated 70% stepwise and 30% multistep mutations) and an infinite allele (IAM) mutational model.

Comparison of datasets

We detected no evidence that alternate strategies of filtering and scoring showed bias in our inference of population differentiation (Table 4). The overall F_ST and F′_ST values as well as the indices of diversity were almost identical for the 10 datasets generated. Genotypes with alleles identified by sequence length gave slightly lower F′_ST and heterozygosity because length masks some of the underlying sequence variation relative to unique IDs, reducing the mean within population heterozygosity and F′_ST (Hedrick, 2005; Jakobsson, Edge & Rosenberg, 2013). Selecting the second allele by depth biases against within population allelic diversity resulted in slightly higher F_ST and F′_ST values and lower F_IS values. As a conservative approach, we selected the first dataset generated by randomly selecting the second allele when there were more than two choices (random000_len). In each case, however, the differentiation among sites was significant (p < 0.001), and there are no cases in which inferences would differ because of this variation in magnitude. Because the different strategies for filtering reads produce qualitatively similar results in all cases we believe that the approach of selecting the second allele does not bias our interpretations.

Population differentiation

Population differentiation was significant among all sites, with a global F′_ST = 0.11. The pairwise differentiation test between islands show that F′_ST values between Yap and Ngulu (0.09), Palau and Ngulu (0.10) and Yap and Palau (0.09) were similar and all comparisons were significant (p < 0.01). The exact test between sites (Table 3) shows that both within and among islands, all comparisons are significantly different between Yap, Palau and Ngulu. For example, on Palau, site S24 shows the most differentiation with site S17 (0.12), which is of the same magnitude as comparisons among islands above. We calculated a second measure of genetic differentiation, Jost’s D which gave the same differences between sites (Table S4).

Bottleneck

The graphical test for bottlenecks, which does not require data to be in Hardy–Weinberg equilibrium, is robust to a small number of loci (fewer than 20) and detects bottlenecks that occurred within a few dozen generations (Luikart et al., 1998). All of the sites had normal L-shaped distributions and the test showed no evidence of bottlenecks for the sites around Palau.

Population structure

Microsatellite sequencing

Genotyping microsatellite loci using high throughput sequencing has become cheaper and faster than traditional fluorescent fragment length analysis. To our knowledge, however, there have been very few published papers using this technique (Roberts, Schwartz & Karl, 2004) and no standardized protocol has yet been developed to obtain the best results. The main challenge with sequenced microsatellites is to define a genotype for each colony. There were cases where regardless of the objective filtering criteria applied, more than two alleles were possible, yet corals are diploid organisms. To ensure that our results were robust to decisions about how to select among alternate possible alleles, we tested different methods of allele selection, including selecting two alleles at random from among all sequence variants within an individual. We find that the data are robust to filtering and allele selection criteria, because although the exact values differed, with consistently lower values for selection by depth, none of the indices of heterozygosity, H_O, H_E, G′_ST, or genetic structure, including F_ST, F′_ST and F_IS, fundamentally changed by using different approaches. This can be explained by previous findings that the greatest portion of the structure and diversity is driven by the most common alleles in the population (Selkoe & Toonen, 2006; Toonen & Grosberg, 2011). This is consistent with our data because neither of the selection criteria impacted the most common alleles in the populations. This robust finding indicates that scoring microsatellites using high throughput technology can give consistent and reliable results that, being based on the underlying sequence rather than length polymorphisms, can allow comparisons of the repeat motif itself and variable flanking regions and be combined reliably among labs and studies. The selection of alleles among several possibilities, however, may prove problematic for applications such as parentage analysis where the genetic pool for parentage assignment will be affected and could lead to false parentage exclusion.

Connectivity between Palau, Yap and Ngulu

Golbuu et al. (2012) show oceanographic connectivity between Palau and Yap, and argue that the reefs of Palau recovered surprisingly quickly from the mass mortality following the 1998 bleaching event due to a pulse of coral larvae from Yap (including the atoll of Ngulu) in 1999. Bruno et al. (2001) describe populations of A. hyacinthus in Palau as having suffered virtually 100% mortality in areas that were surveyed. Given the generation time of A. hyacinthus (Wallace, 1985), we would expect roughly three generations in the 14 years between this mortality and our sampling in 2012. If Palau had recovered from a pulse recruitment from Yap after such a widespread and dramatic loss, we would expect evidence of a bottleneck in the recovering population, and should be able to observe a significant founder effect such that Palau would contain a strict subset of the total diversity of Yap and Ngulu from which it was recolonized. Instead, we did not observe signs of a bottleneck, contradicting the hypothesis that A. hyacinthus on Palau recovered from a pulse recruitment of larvae from a single source. These data, however, do not exclude the possibility that Palau received a larval pulse from Yap and/or Ngulu, or from more distant populations such as Phoenix Island as demonstrated by Davies et al. (2015), which then mixed with larvae from surviving local populations.

The pairwise comparisons between islands and among sites show similar F′_ST values, highlighting the fact that there is as much differentiation among sites within a single island as among sites on different islands. Although all pairwise F′_ST values are significant (p < 0.01) and there is clear population structure among all sites sampled in this study, it is important to note that F′_ST values are small enough to reflect some degree of gene flow through time. This result is more consistent with the prediction of high local retention of larvae in the simulations of Golbuu et al. (2012) than the hypothesized mass recruitment from Yap. Overall, these data support the idea that recovery of A. hyacinthus in Palau did not come exclusively from a single source. Instead, the lack of any evidence of a genetic bottleneck and the unique genetic diversity seen in Palau indicate that either mortality was less than 100% and that recovery came from a combination of areas, including sites within Palau. Further, these data indicate that the catastrophic mortality of A. hyacinthus reported by Bruno et al. (2001) was likely not as widespread as thought and that there were enough surviving colonies to reseed the barrier reef of Palau as well as maintain genetic diversity.

Structure of sites within Palau

Sites within Palau separated by as little as 5 km show as much population genetic differentiation as sites between islands separated by as much as 452 km. Both the pairwise and PCA analyses show that Palau’s site S24 (West Palau) is as close to Yap sites S27, S29 and S30 or Ngulu site S28 as it is to Palau site S17 (East Palau). It is interesting to also note that site S17 shows the most differences with other Palau sites S20 and S24 and even greater differences with Yap and Ngulu sites. The eastern lagoon and reef of Palau has historically been impacted by development such as the construction of the airports and roads (Maragos & Cook, 1995) and has been described as being impacted from sediment run-off (Golbuu, 2011; Golbuu et al., 2012). These anthropogenic impacts may have triggered similar population changes to those observed in terrestrial habitats when there is habitat degradation and fragmentation in terrestrial habitats which lead to an erosion of genetic variation and increased genetic divergence between populations due to increased random genetic drift, elevated inbreeding, and reduced gene flow (Young, Boyle & Brown, 1996).

Implications for conservation

We show that populations of A. hyacinthus on Palau did not recover from a single pulse recruitment of larvae, and that mortality was likely less widespread than originally thought, with at least some pockets of surviving colonies within the Palauan archipelago that preserved unique genetic diversity there. Furthermore, there are significant differences among sites around Palau indicating that if there is exchange, it is insufficient to homogenize the populations, supporting the larval dispersal simulations of Golbuu et al. (2012) that there is a high level of self-recruitment among sites. In terms of conservation, these data support increasing the area of conservation by either increasing the number of MPAs or increasing the size of existing MPAs around Palau to protect a wide array of the genetic diversity.