DNA extraction, amplification and sequencing

Coral tissue samples were collected from colonies of six coral species: F. fragum, F. gravida, S. siderea, S. radians, S. stellata, and P. astreoides. Sampling took place in Panama in 2005, on the island of São Tomé in West Africa in 2006 and in three populations separated by >500 km in Brazil (Abrolhos, João Pessoa and Fortaleza) in 2002 and 2007. Genomic DNA was preserved and extracted following a protocol described previously [35]. Due to the unusually low levels of genetic variation found in the mitochondrial DNA of corals [37], [38], only nuclear loci were used. The intron and exon of β-tubulin and the intron of the Pax-C gene, which have previously been shown to contain sufficient intraspecific variation in corals for population studies [20], [39], were amplified for all species studied, with the exception of P. astreoides, for which only β-tubulin could be amplified. For P. astreoides, the mitochondrial control region (thought to be one of the most variable regions of the coral mitochondrial genome) was amplified and sequenced for 32 individuals from Panamá and Brazil, but because all individuals contained identical nucleotide sequences at this locus, this marker was not used for further analysis.

Amplification of the two nuclear loci was performed by polymerase chain reaction using published primer sequences for β-tubulin [40] and for Pax-C [39]. The thermal cycler profile had an initial denaturation step at 94°C for 2 min, followed by 38 cycles of 94°C/45 sec, 50–58°C/45 sec for β-tubulin or 50–54°C/45 sec for Pax-C, 72°C/90 sec, with a final extension step at 72°C for 5 min. An additional forward primer was designed for the amplification and sequencing of Pax-C for F. fragum (PaxFsh1: 5′- GGA GGA GCT TGC GAA TAA GA -3′). Multiple bands were amplified in P. astreoides and the Siderastrea species in β-tubulin, and for S. siderea in Pax-C. Bands at ∼750 bp and ∼600 bp were extracted and purified for products of β-tubulin and Pax-C respectively, using the Qiaquick Gel Extraction kit (Qiagen). The remaining PCR products were purified either by gel extraction or by elution through a silica-membrane column using the appropriate Qiaquick purification system (Qiagen).

Sequencing of all loci was performed directly on purified PCR products for both forward and reverse directions on an ABI 3130xl genetic analyzer with the BigDye Terminator v3.1 chemistry (Applied Biosystems). Sequence chromatographs were viewed and edited using the Sequencher v4.5 software (Gene Codes Corp). Heterozygous alleles were identified by double peaks observed in sequences from both directions. Indels were observed in the introns of β-tubulin for S. radians (site 549) and S. siderea (site 549), and in sequences of Pax-C for S. radians (site 134), F. fragum (site 205) and F. gravida (site 205). Because only a small number of individuals were heterozygous for indels within any species and indel positions were straightforward to identify, indel sites were kept and considered in the analysis.

Analysis of sequence data

Haplotypes for heterozygous individuals were reconstructed using PHASE v 2.1.1 [41], [42]. The algorithm was run three times for 100 iterations with ten thinning interval steps and 100 burn in steps. The best pairs of haplotypes for each individual resulting from the run with the highest average value for the goodness of fit were used for each nuclear locus.

Haplotype frequencies, molecular diversity indices, tests of neutrality, analysis of molecular variance (AMOVA) and estimates of population differentiation were calculated using Arlequin v.3.11 [43]. Analysis of molecular variance (AMOVA) was used to estimate levels of genetic differentiation among populations of each coral species. One AMOVA was used to estimate differentiation among all populations sampled in the Caribbean, Brazil and West Africa (Φ_{ST ALL}). In this AMOVA, populations of F. fragum (n = 1) and F. gravida (n = 3) were pooled to assess levels of genetic differentiation between these two closely related species. A second AMOVA tested for differentiation among populations within Brazil, the only region for which multiple populations were sampled (Φ_{ST BR}). Estimates of pairwise φst used distances between haplotypes. Parsimony haplotype networks were constructed with TCS [44]. Recombination in the two nuclear loci was inferred as reticulations in the haplotype network and tested by the four-gamete test in DnaSP [45]. For clarity, loops were omitted from the haplotype networks and only internal branches are shown. Due to the evidence of recombination, analyses using a coalescent framework (which assume no recombination) were not attempted.

Genetic differentiation across regions of the Atlantic

The data collected for six coral species plus an existing dataset for Montastraea cavernosa [20] suggest that significant differentiation is present between the Caribbean, Brazil and West Africa for most sampled coral species. Concordance in the pattern of regional isolation occurs at the scale of genes and species. Both nuclear markers show consistent patterns in differentiation, and regional differentiation across multiple species indicates that barriers to gene flow between biogeographic regions in the Atlantic are effective for most corals.

The most likely biogeographic barrier separating Caribbean and Brazilian coral populations are the deltas and low salinity waters of the Amazon, Orinoco and numerous rivers along the coast of northern South America. The Amazon is the largest among them, and accounts for 16% of the annual freshwater discharge into the world's oceans [47]. Its plume extends 200–500 km in width [48] and is recognizable 24–32 m below the surface [49]. The Amazon and Orinoco Rivers attained their current drainage configuration around the Late Miocene [50], indicating that low salinity and high sedimentation may have been a feature of this coastal region for ∼10 Ma. Approximately 2,300 km of coastline rich in soft sediment bottoms and low salinity coastal waters between Caribbean and Brazilian reefs may pose substantial barriers to dispersal, because the substrate near river deltas is inadequate for coral settlement and coral larvae are very sensitive to the changes in salinity [51] that these large river plumes create.

Levels of genetic differentiation between between Caribbean F. fragum and South Atlantic F. gravida were the most pronounced among all of the studied species (Table 2A and 3). This finding, combined with the observation that the species do not co-occur [9], supports the hypothesis that F. fragum and F. gravida are distinct [35]. Only a small number of mutations separate F. fragum and F. gravida (Fig. 3), but this may be due to the recent divergence of the two species coupled with the slow rates of mutation observed in corals [38].

The only species for which no significant differentiation was observed between Caribbean and Brazilian populations was P. astreoides. For β-tubulin (the only marker amplified for this species, see Methods), P. astreoides has one common haplotype that is observed both in Brazil and Panamá at similar frequencies, with all other haplotypes being unique to each respective population (see Fig. 3). This common haplotype may be an ancestral haplotype that has been maintained over time in the two populations without continued gene flow. Sequences of β-tubulin for P. astreoides have lower variation compared to the other species, perhaps insufficient to detect differences among populations. Additional support from faster-evolving markers such as microsatellites is required to determine whether continued gene flow is maintained between the Caribbean and Brazil in this species. Alternatively, P. astreoides, a brooder, may be able to raft or be more tolerant to freshwater and high sedimentation.

Significant regional differentiation was also observed with respect to West Africa (Table 3), suggesting that long distances of open ocean may be impassable for coral larvae. The easternmost point of Brazil and the island of São Tomé are separated by 4,800 km. Some mid-Atlantic islands could serve as stepping-stones for dispersal, decreasing dispersal distances by about one half. However, it appears that even these distances are too great to maintain gene flow or that the populations on these islands (such as the St. Peter and Paul rocks) are too small or ephemeral to be significant in trans-Atlantic dispersal. Interestingly, the population of M. cavernosa on the island of Bermuda is able to maintain connectivity with Caribbean populations despite being separated by at least 1,000 km [20]. Presumably the fast-moving currents of the Gulf Stream are able to maintain the influx of larvae to Bermuda. The combined effects of distance and physical oceanography are likely important isolating factors for West African coral populations.

Concordance in patterns of regional connectivity in the Atlantic is not observed in most other species of marine organisms. Four species of sea urchins sharing similar life histories and pelagic larval durations show markedly different patterns of differentiation across the Atlantic. The Amazon appears to be a significant barrier to gene flow for most species [52], [53], [54], but not all [55]. Multiple diagnostic mutations separate eastern and western Atlantic populations of some urchin species [53], [54], while others maintain continued exchange across the South Atlantic [52], [55]. The discordant patterns observed among sea urchins with similar traits in reproduction and dispersal indicate that barriers to gene flow in the Atlantic may be permeable depending on aspects other than reproduction and pelagic larval duration.

Among reef fish, similar discordance has been observed across species. Some species achieve high gene flow throughout the Atlantic (Myripristis jacobus [56]), or throughout their range (Halichoeres garnoti [57]). Others are highly differentiated [58], [59], while some display intermediate levels of gene flow between these two extremes. Interestingly, broad adult habitat preferences among surgeonfish of the genus Acanthurus correlates well with dispersal ability across the Amazon, suggesting that among reef fish, both ecology and reproduction play important roles in dispersal potential [13].

In sum, the similarities in patterns of connectivity observed among various coral species indicate that barriers such as the Amazon or stretches of open ocean are likely impassible for most coral larvae, leading to concordance of patterns. These barriers may be more permeable, however, for other organisms whose ecology and life history permit dispersal at greater distances than corals, resulting in more variable patterns of connectivity observed in other amphi-Atlantic marine organisms.

Genetic differentiation within regions

Although patterns in inter-regional differentiation are concordant among most coral species, different estimates of gene flow are observed within regions across coral species, and the extent of gene flow is correlated with reproductive traits. In this study, multiple populations within a region were sampled only in Brazil. For each species, two or three populations each separated by >500 km across a total of 2,000 km along the coast of Brazil were sampled for all species except F. fragum (present only in the Caribbean) and P. astreoides, for which specimens were only taken in João Pessoa (see Fig. 2). Strong differentiation observed at both loci between populations of the brooder F. gravida in Abrolhos and João Pessoa (Table 3) indicates that little exchange is occurring between populations of this species along the coast of Brazil. Intermediate levels of differentiation were observed for the brooder S. radians, where AMOVA indicates that differentiation is significant between populations within regions for both loci (Table 2), but not all pairwise population comparisons showed significant differentiation (Table 3). In contrast, gene flow appears to be maintained between three populations spanning 2,000 km of Brazilian coast of the two broadcasting corals studied, S. siderea and M. cavernosa (Table 2).

The two broadcasting species that are able to maintain gene flow along the coast of Brazil are both gonochoric with one reproductive cycle per year, while the two brooding species with fragmented populations are hermaphroditic and spawn multiple times per year (Fig. 1). Egg sizes for broadcasting M. cavernosa and S. siderea are larger than for the brooder F. gravida, but only marginally larger than brooding S. radians. However, the dispersing propagules for brooders, the planula larvae, are larger than broadcast eggs and bear zooxanthellae (Fig. 1). The findings of this study indicate that neither the frequency of spawning nor the size of the dispersing propagules (and associated energy reserves) appear to provide an advantage for long distance dispersal. Brooded larvae may allocate much of their energy reserves towards settlement and growth rather than dispersal. Because broadcast gametes must spend time in the water column before being competent for settlement, their chances of being entrained in currents that take them away from their natal reef are greater, resulting in greater dispersal distances.

The species with the greatest population differentiation, F. gravida, is a hermaphroditic brooder, possibly capable of self-fertilization like its sibling species F. fragum [60]. Self-fertilization may ensure reproductive success when population densities are low and sperm limitation reduces chances of fertilization, but with the disadvantage that populations may become inbred and dominated by clonemates. Self-fertilization will also result in apparent reduced gene flow, so at least part of the high differentiation among populations of F. gravida could reflect inbreeding.

Some Caribbean coral species are able to maintain gene flow throughout the Caribbean and Bermuda, such as M. cavernosa [20] and M. faveolata [61], but dispersal within the Caribbean is more restricted for other species. Among the seven Caribbean corals studied to date, gene flow is restricted to <500 km for most species [62]. Acropora palmata [63] and A. cervicornis [62] are able to disperse widely, but their populations are subdivided between the eastern and western Caribbean. Gene flow appears to be restricted to even shorter distances for species such as Agaricia agaricites [64] and some members of the M. annularis species complex [61], [65]. All of the aforementioned Caribbean corals are broadcasters with the exception of A. agaricites. In agreement with findings for Brazilian brooders, A. agaricites in the Caribbean has a more fragmented population than broadcasting species, although studies on additional Caribbean brooders are required to confirm the generality of this trend.

Among Caribbean broadcasters, there are no clear explanations for differences in dispersal ability. Dispersal ability in A. palmata and A. cervicornis appears to be more restricted than for M. cavernosa and M. faveolata, even though Caribbean Acropora eggs are at least 1.4 times greater in size [23]. Egg size may not be a good predictor of dispersal ability, as nutritional reserves may be allocated for settlement rather than survival in the water column. Most striking is the difference in dispersal ability among closely related species with similar reproductive traits, such as short-ranging M. annularis and Caribbean-wide M. faveolata [61].

In the Indo-Pacific, trends in dispersal ability show even less association with reproductive mode. A survey of nine co-distributed species along 1200 km of the Great Barrier Reef showed no correlation between connectivity and reproductive mode [18]. Likewise, reproductive mode was not a good predictor of dispersal ability among several coral populations along the GBR and the peripheral populations on Lord Howe Island [19]. Paradoxically, species with highly different larval longevity estimates and life history characteristics have been observed to have similar levels of fine to meso-scale population subdivision [66].

Determining which life history traits impart greater success in long-distance dispersal is not straightforward. In Atlantic corals, reproductive mode appears to provide an advantage for long-distance dispersal, but this trend is not observed for corals in the Indo-Pacific. These results indicate that single aspects of coral life history, such as reproductive mode, propagule size or larval longevity, cannot alone predict dispersal ability, and that complex interactions of multiple factors likely play a role in determining connectivity. In other organisms, aspects of reproduction and life history, when taken alone, have at times also not been good predictors of dispersal ability and connectivity. In reef fish of the Atlantic, for example, pelagic larval duration is not a good predictor of population connectivity [56].

While dispersal ability is an important factor for maintaining connectivity, environmental factors also play an important role, as migrants must also be able to adapt to the environment at a distant reef in order to become established and contribute to gene flow. Interactions with other organisms, such as competition for space, can also affect successful recruitment. In sum, patterns in connectivity are difficult to predict based on reproductive traits alone and are likely influenced by multiple biotic and abiotic factors.

Patterns of genetic diversity

Mean values of molecular diversity indices overlap among regions for most coral species, although for some comparisons, diversity appears to be lower in the peripheral populations, such as gene diversity in S. siderea of West Africa for both loci and F. gravida in West Africa for Pax-C (Table 1). On average, however, there does not appear to be a strong trend in decreasing genetic diversity towards the edges of the species range for most species.

This finding is at odds with patterns observed in populations of the coral M. cavernosa, where genetic diversity was found to be lower in Brazil and West Africa compared to the Caribbean [20]. Similarly, reduced allozyme allelic diversity and heterozygosity was also found for several coral species on the isolated Lord Howe Island relative to populations on the Great Barrier Reef [19] as would be predicted by small population size and founder effects. Peripheral populations that are isolated from the center of species range or center of diversity may have smaller effective population size (N_e) and can suffer loss of diversity via genetic drift. In addition, peripheral populations colonized by a small group of founders may only contain a subset of the full diversity of alleles observed in the main population. However, loss of genetic diversity in Brazil and West Africa does not appear to be a consistent trend among the coral species studied here, or if differences in genetic diversity exist among regions, they have not been captured with the markers used here.

Overall, levels of polymorphism and the number of haplotypes for the brooding species F. fragum, F. gravida, P. astreoides and S. radians were much lower than for the broadcasters S. siderea and M. cavernosa. For example, >40 haplotypes were observed for M. cavernosa and S. radians, while the number of haplotypes observed for the four brooding species ranged from 6–12. High levels of polymorphism in M. cavernosa may have made the patterns of reduced variation in the peripheral populations more pronounced and readily detected. In S. siderea, decreasing trends in gene diversity (h) are observed, but there is overlap in the mean values of other diversity indices among the three biogeographic regions (Table 1).

Regional differences in genetic diversity may have gone undetected, but another reasonable alternative is that loss of variation has occurred in peripheral populations of M. cavernosa, but not in other species. Although regional isolation is observed for nearly all species, the mechanisms that lead to reduced variation may be less effective in some species relative to others. Loss of variation resulting from genetic drift is more pronounced in populations with small effective size (N_e). Differences in N_e between the species may result in different rates in the loss of genetic diversity. Founder effects may also be less pronounced if peripheral populations have received multiple founding events. Some of the brooding species studied have multiple mating cycles per year and reproductive strategies that may allow them to establish larger populations over shorter time periods, thereby increasing N_e and buffering the loss of genetic diversity. Another possible scenario is that historically some of the other species were able to maintain gene flow up until more recently, while gene flow for M. cavernosa ended a longer time in the past – subjecting the species to a longer period of time for drift to act upon.

Conclusions

Patterns of connectivity across broad regions of the Atlantic are congruent across multiple species of corals, suggesting that barriers to dispersal such as the Amazon freshwater plume and the long distances that separate the east and west South Atlantic are effective for most coral species. At shorter, regional scales, it appears that some aspects of reproduction such as mode of larval development can influence gene flow, although this pattern cannot be generalized to all scenarios. Broadcasters have greater dispersal ability in the Atlantic, but several exceptions exist in the Indo-Pacific, suggesting that complex interactions between biotic and abiotic factors can limit connectivity rather than single traits of a species. Loss of genetic diversity in peripheral populations of corals is also not a general trend observed among all amphi-Atlantic corals, likely due to differences in population size and timing of isolation across species. Regional isolation may mean that the persistence of peripheral populations relies primarily on local to regional recruitment since migration from other regions must occur rarely. These populations may be more vulnerable to disturbance on ecological time scales as a result of their isolation. On the other hand, isolated peripheral populations with small effective population size may become the sites for local adaptation and allopatric speciation. In both scenarios, these regions require special attention for conservation as a result of their potential vulnerability to environmental change, but also their importance in generating diversity.