Population genetic analysis using genome-wide SNPs identified from nine S. alveolata reef sites spanning the latitudinal and longitudinal range of the species revealed low gene flow between sites, equating to low effective dispersal, with 25 out of 36 pairwise FST values showing significant differentiation and an average pairwise FST of 0.28 ± 0.10 across the system. The STRUCTURE analysis revealed the North Atlantic site to be a separate population from all the others (FST = 0.42 ± 0.10). Genetic divergence in terms of FST was not significantly correlated with either straight-line distance or shortest ocean distance. Presence of low gene flow between reef sites is in line with a growing number of studies that have identified population structuring in marine species which, as with S. alveolata in our study, have a dispersing larval stage and no hard geographical barriers to dispersal [17, 18, 44–46]. For instance, Benestan et al. (2015)  identified strong neutral genetic structure (average pairwise FST = 0.0019) in the American lobster (Homarus americanus) along a sea surface temperature gradient using RAD genotyping despite a dispersing planktonic larval stage with a duration of 4–6 weeks. Similarly, the bat star (Patiria miniata) showed genetic population structure along a latitudinal gradient (average pairwise FST = 0.061) despite a longer larval dispersal period of 6–10 weeks , which is in line with the estimated larval duration of S. alveolata at 5–12 weeks, dependant on temperature and food availability [40, 47]. Indeed, a meta-analysis revealed that mean pelagic larval duration was not a good predictor of gene flow  and our findings support this interpretation.
Three sites were significantly differentiated from all other sites: North Atlantic (average pairwise FST = 0.42 ± 0.10 and STRUCTURE), Tyrrhenian Sea (average pairwise FST = 0.32 ± 0.12) and Bay of Biscay (average pairwise FST = 0.25 ± 0.05). The northern Irish Sea, southern Irish Sea and English Channel sites were also all significantly differentiated from each other (Table 2). The ocean circulation modelling predicted high isolation (zero to low larval density; Additional file 1) among all release sites, even in the most optimistic (dispersive) scenario. Therefore, the FST results are supported, at least in part, by patterns of ocean circulation influencing passive larval dispersal along the Atlantic and Mediterranean coastlines of Europe and North Africa.
High population sub-structuring could be partially explained by two behavioural barriers to connectivity: first, spawning in polychaetes is triggered by the action of waves, especially during spring tides, which tends to push larvae toward the coast rather than offshore  and second, S. alveolata larvae act to reduce dispersal out of the bay where theywere spawned by moving higher in the water column with an incoming- and lower with an outgoing-tide . This ‘tidal stream transport’, has been suggested as a possible mechanism for facilitating/restricting advective transport in a number of taxa (e.g. [50, 51]). Yet despite this, relatively high levels of genetic diversity were observed across the sites (He = 0.52 ± 0.04). Observed heterozygosity was very low at all sites and could be due to inbreeding , as a result of the site isolation predicted by the ocean circulation modelling and due to the reproductive strategy of gregarious colonial species, as juveniles settle together in patches and spawning of one individual triggers the spawning of those immediately adjacent . Heterozygote deficiency has also been recorded in marine invertebrates as a consequence of the Wahlund effect, due to the coexistence of genetically distinct cohorts within a sampling site . However, low heterozygosity estimations could also be a consequence of poor genome coverage and low sample size at some sites [53, 54]. Further research is needed to assess whether presence of null alleles is genome-wide, as evidence of inbreeding, or restricted to particular loci, as evidence of allelic dropout due to low coverage, in order to assess whether the species is at inbreeding risk.
Identifying the relative contribution of gene flow, genetic drift and natural selection to population structure is difficult in marine invertebrates due to their fluctuating population sizes . This is particularly difficult in S. alveolata, as in-depth local ecological knowledge, such as population size and breeding behaviour, is lacking in the majority of sites (but see [31, 55–57]). That said, the identification by STRUCTURE of only the North Atlantic site as a separate population, with all other sites forming a single population, supports the idea that low levels of gene flow are still maintaining genetic diversity within the system as a whole. Such wide-ranging genetic connectivity has also been observed in other polychaete species . This is potentially a positive sign for survival of the species in a changing climate, as the maintenance of genetic diversity is key to facilitating rapid evolution when environmental conditions change . However, S. alveolata reduce their larval duration in line with increasing temperatures  and our models show that shorter larval duration leads to reduced larval dispersal. Therefore, lower connectivity, and thus gene flow, is predicted for S. alveolata in a changing climate.
Despite some evidence of low levels of gene flow between the majority of the study sites within the system, both FST values and STRUCTURE analysis revealed isolation of the North Atlantic site and identified this reef as a separate population to all others. There are two possible processes that could, separately or in synergy, be causing the observed pattern of population structure: historical isolation and contemporary patterns of gene flow. During the last glacial maximum, geomorphology and fossil evidence suggests that southwest Ireland was partially unglaciated  and genetic data supports the presence of a glacial refugium in this area for both terrestrial [61, 62] and marine species [63–65]. In particular, Jolly et al. (2006)  found that two coastal polychaete worms (Pectinaria koreni and Owenia fusiformis) showed a private Irish Sea haplotype linking two ancestral haplotypes and they suggest this could have evolved in a small ice-free area along the southwest coast of Ireland. Therefore, one potential explanation is that this population was isolated in a different glacial refugium to the rest of the sites, but further study is needed to test this hypothesis.
The second hypothesis, that contemporary gene flow is very low between the North Atlantic and other S. alveolata sites, is supported by predictions of larval dispersal as seen in the ocean circulation modelling. In even the most optimistic scenario for larval dispersal, the North Atlantic site was not predicted to have interchange of individuals with any of the other sampled sites and had by far the highest predicted proportion of larval retention (79% of released individuals were retained compared to an average of 20 ± 13% for all other sites). This can be explained by the hydrodynamic modelling in terms of current patterns within Galway Bay limiting dispersal. The presence of the North Atlantic Current, which moves eastwardly towards Ireland and then continues northwards, is also an isolating factor for the North Atlantic reef. Any larvae that do move beyond their spawning reef at the North Atlantic site are drawn northwards, beyond the current northern range limit for the species (the Solway Firth, Scotland ), and there are no reef sites along the north coast of Northern Ireland to act as stepping stone populations (Firth, unpublished). This oceanic barrier is also a likely cause of the observed genetic divergence between the northern/southern Irish Sea sites and the English Channel, and the reason that the dispersing larvae from more southerly sites do not reach the North Atlantic site. Although larval dispersal of many other marine species has been found to be mediated by the North Atlantic Current (e.g. [66, 67]), isolation due to the North Atlantic Current has not previously been reported, likely because it depends upon species-specific occurrence range. Therefore, reduced contemporary gene flow is a likely cause of the observed population isolation of the North Atlantic S. alveolata reef. These findings highlight the North Atlantic reef site as at risk if population size reductions occur as recruitment and genetic augmentation is unlikely from elsewhere within the species’ range. Therefore, conservation management is needed to ensure that population size does not decrease at this site.
Despite the importance of ocean circulation to larval dispersal in general [18, 19, 68] and to S. alveolata in particular, as identified by our data in regards to the isolated North Atlantic site, our ocean circulation models of larval dispersal did not show a significant correlation with observed genetic distance between sites. This is surprising as both FST values and larval dispersal models suggest low connectivity between sites. This lack of correlation between our larval dispersal modelling and genetic distance values is in contrast to studies on a number of species with a dispersing larval stage including the Mediterranean shore crab (Carcinus aestuarii) , the bat star (P. miniata)  and the American lobster (H. americanus) ; these studies found genetic structure was directly related to ocean currents or to estimates of potential larval connectivity obtained with coupled physical-biological models. However, these studies were conducted over a smaller geographical area than our study. As per our study, Jorde et al. (2015)  found that when looking at large-scale differentiation patterns in the north Atlantic, geographic distance and larval drift alone explained only a minor portion (2.5–4.7%) of genetic isolation in the northern shrimp (Pandalus borealis).. Galindo et al. (2010)  modified a biophysical model for Monterey Bay in California to simulate dispersal of the acorn barnacle (Balanus glandula) but it also did not match an observed genetic cline in the species. They discovered that their model fit was improved by including natural selection, larval retention, and input values from an additional source population .
There are several factors likely to reduce the similarity between modelled larval dispersal and observed gene flow. Gene flow is representative of multigenerational mixing, while the model used in this study is representative of the dispersal patterns of only one generation. It is possible to simulate multigenerational gene flow based on dispersal models [45, 70, 71] but this would be confounded by the lack of known sites that could be used as “stepping stone” sites, which may facilitate dispersive spread over multiple generations. Such sites are likely to exist but are currently undocumented. These undiscovered sites represent missing source populations within our model and may well be a cause of the observed mismatch between predicted larval dispersal and observed gene flow. The distribution and occurrence of S. alveolata reefs are not well documented, particularly in the southern range of the species (Firth, unpublished), and the prediction of our model that genetic differentiation between our sample sites would be high could be due to the absence of intermediate reef sites within the model that would create higher levels of admixture within the system as a whole [9, 72]. This is further complicated by the fact that reefs can vary temporally in their presence at a site . Therefore, increased knowledge of the location of S. alveolata reefs and ongoing monitoring of reef sites is required to generate a more clear picture of the role of connectivity in meta-population maintenance in this species and to further inform the ocean circulation model. Our study highlights the importance of comparing bio-physical models with observed population structure of a species in order to create accurate dispersal predictions.
Loci under selection
Twenty-seven SNPs were identified as outlier loci, all of which were putatively under balancing selection. Identification of a higher number of loci potentially under balancing rather than divergent selection is common in genome scan studies, including those on marine invertebrates [12, 46, 73–75]. Our finding that 5.6% of SNPs were potentially under balancing selection is reflective of a growing body of evidence that suggests that balancing selection, which acts to preserve polymorphism, is more important in the genome than previously considered [12, 46, 76]. In their study of the sea anemone, Nematostella vectensis, covering a broad geographical range, Reitzel et al. (2013)  identified 37 polymorphic sites inferred to be under balancing selection, but none under divergent selection, as with our study. However, further evidence in the form of elevated polymorphism, reduced differentiation, and shifts towards intermediate allele frequencies are needed to confirm that these loci are indeed under balancing selection .
The lack of identification of loci under divergent selection using either outlier analyses or environmental association tests is surprising given that local adaptation to temperature is known to be present in this system . We previously reported that S. alveolata individualsfrom sites along the latitudinal range of this species (all of which are included in this study) showed different responses to thermal regime changes in terms of their membrane lipid composition dependant on their site of origin . This is likely a reflection of poor genome coverage leading to a low number of genotyped loci (as per ). Further studies are needed to search for adaptive loci and their functions in this system.