Marine ecosystems have historically been considered to be relatively open, with populations demographically and genetically connected over broad spatial scales. In a range of coastal taxa, however, recent molecular surveys have consistently revealed considerable phylogeographical and population genetic structure, often at seemingly small spatial scales, indicating that connectivity is frequently much lower than previously assumed. Examples include a variety of species lacking planktonic dispersive stages and/or exhibiting particularly strict ecological requirements, such as intertidal fucalean and kelp seaweeds [1–5], seagrasses [6, 7], direct-developing invertebrates and fish [8–11], high-intertidal rock-pool invertebrates , and many estuarine organisms [13–18]. In such species, shallow genetic discontinuities can be common due to intrinsic life-history and habitat constraints to dispersal. Often, however, they also display deeper genealogical splits that distinguish regional sets of populations across their ranges. Such nested patterns of phylogeographical structure can result from a number of factors and are frequently harder to interpret, especially when species are evenly distributed across transition zones.
Vicariance is usually invoked as the main driver of (neutral) genetic divergence. Extrinsic barriers to gene-flow are generally less obvious (or absolute) in marine compared to terrestrial landscapes [19, 20], but circulation patterns, coastline topography and habitat discontinuities have all been shown to potentially represent effective barriers to the exchange of individuals between adjacent marine regions [21–24]. Complex variations in habitat availability and connectivity, resulting from the Pleistocene oscillations in sea levels and surface temperatures, are also known to have produced ancient population subdivisions (and differentiation) in many coastal organisms [6, 25, 26]. Within a species, disjunct distribution of divergent genetic lineages provides strong indication for the occurrence of such vicariant processes.
Inferring the existence of a particular dispersal barrier from molecular data may not be straightforward though . In species with short dispersal range, discontinuities in individual gene trees (mostly derived from organelle markers) readily arise haphazardly within continuously distributed species simply as a consequence of idiosyncratic lineage sorting [28, 29]. Similarly, genetic drift during spatial expansions  or disruptive selection  can also result in the geographic segregation of organelle lineages across a species range even in the face of dispersal. In general, long-term isolation can only be confidently assumed when spatially concordant patterns across multiple unlinked loci are found [29, 32].
Disentangling historical from ongoing constraints to dispersal may also be problematic. Phylogeographical breaks and contemporary oceanographic barriers (or biogeographical transition zones) are often mismatched in marine restricted dispersers . Historical patterns of isolation and colonization in these organisms explain population structure better than more recent factors affecting gene-flow. Phylogeographical breaks may develop where formerly vicariant sub-populations have reassembled. The Iberian peninsula is a good example where diverse taxa such as trees , amphibians [35, 36], reptiles [37, 38] and pond-dwelling invertebrates [39, 40] are sub-divided into well defined, mostly parapatric genetic sectors that presumably formed during expansions from disjunct refugia. The temporal persistence of genetic discontinuities across marine secondary contact zones have also been demonstrated in several species . However, insight into the processes preventing steady genetic homogenization of divergent but contacting gene-pools requires finer scale genetic sampling than is common in most studies (but see [3, 6, 9, 12, 41, 42]).
Virtually all coastal organisms have some potential to disperse and colonize new habitats, as the extensive post-glacial range shifts of many demonstrate. Thus, migration would also be expected to occur between fully established populations, including between divergent populations in relatively close proximity. Incipient reproductive isolation can reduce or prevent gene flow between divergent contacting lineages [41, 43]. The persistence of fine-scale genetic differentiation in the absence of obvious reproductive and dispersal barriers seems paradoxical. In restricted dispersers, however, colonization and immigration, as sources of gene-flow, may have very different genetic effects. During expansions into vacant habitats, the original colonists can grow exponentially and contribute disproportionately to the genetic composition of the establishing population. In contrast, once the habitat patch is filled, demographic stability and increased competition can considerably reduce the impact of subsequent immigrants [44, 45]. In addition, if there is a gross disparity between the number of residents and immigrants, a common situation in low dispersal species, foreign genotypes introduced in a population will a priori be rare and have low probability of random increase due to drift alone [46–48]. In other words, established populations themselves can create a density-barrier effect buffering local changes in allele frequencies and delaying the spatial advance of genes within previously colonized areas (despite immigration). At broad geographical scales, such an effect has been invoked to explain the persistence of genetic homogeneity in recolonized areas , the asymmetrical introgression of genes from established to spatially expanding species , or the lack of gene-flow between former refugial areas that are currently connected by intermediate populations . When effective migration rates are low, patterns of non-equilibrium divergence resulting from founder and density-barrier effects can occur at much smaller spatial scales [46, 51].
In this study, we report a remarkable case of non-equilibrium divergence in the estuarine seaweed Fucus ceranoides, in which steep genetic discontinuities are preserved despite the absence of obvious barriers to dispersal. Fucus ceranoides L. (horned wrack) is a perennial, dioecious seaweed restricted to estuarine environments across much of the Northeast Atlantic. Populations of F. ceranoides from NW Iberia, at the rear edge of the species distribution, form three highly divergent genetic clusters according to both mtDNA and microsatellite markers [15, 52]. Despite their relatively close proximity (~150 km), fixed genetic differences at this scale suggest that the historical and recurrent processes contributing to their differentiation are weakly counteracted by on-going gene-flow. The poor dispersal ability of F. ceranoides certainly plays a role; fucoid algae lack planktonic dispersive stages and therefore F. ceranoides individuals typically complete their entire life-cycle within the discrete, isolated patches of the estuarine habitat they inhabit. Still, an important question remains unanswered concerning the nature and stability of genetic divergence in this system. Like in many other seaweeds, non-local (inter-estuarine) dispersal can be mediated by rafting of detached, reproductive individuals [53, 54]. Such dispersal by drifting thalli was likely responsible for the extensive post-glacial expansion of F. ceranoides into Northern Europe, including the distant colonization of Norway (across the North Sea) and Iceland [15, 52]. If F. ceranoides managed to expand its range more than 15 degrees in latitude since the Last Glacial Maximum [LGM, ~20.000 ka before present (BP)], dispersal restrictions cannot account, at least as the sole factor, for the apparent lack of population connectivity along the much narrower NW Iberian coastline.
This study aims to understand this fundamental issue in the evolutionary ecology of populations, the apparently contradicting evidence for large scale dispersal mediating vast (re)colonisations concurrently with persistent, fine scale genetic discontinuities in older refugial regions. The specific question is whether such discontinuities arise and persist due to long-lasting dispersal barriers, or simply reflect resilient non-equilibrium conditions inherited from a complex demographic past. To address this question, in this study both mtDNA sequence and microsatellite genotypic data are employed to investigate the fine-scale distribution of genetic variation in F. ceranoides from NW Iberia. This region was sampled at the finest scale of resolution achievable—a complete set of neighbouring estuaries—which was the scale over which gene-flow was more likely to be detected. We were particularly interested in the biogeographic context and the demographic processes contributing to the formation and integrity of stable genetic sectors in NW Iberian F. ceranoides.