A multi-gene phylogenetic approach resulted in a much finer resolution of relationships at the tips of the tree compared with previous phylogenies. This framework allowed the estimation of dates of divergence and patterns of speciation across the family and within the recently radiated genus Fucus.
Dating inter-ocean divergence events in Fucaceae
The models used returned similar dated intervals on deeper nodes corresponding to the splitting events of ancestral Fucaceae lineages, but were less congruent in dating recent speciation events. This is likely due to the constraints of the priors used [43, 44]. Although we remain conservative by reporting the range for both models, the narrower and more recent coalescent-based intervals at the tips of the tree are more in agreement with the biological processes associated with speciation in these taxa .
The most likely origin of the Fucaceae is in the Pacific Ocean during the mid to late Miocene (19.5-7.0 Ma, estimated based on 23-7 Ma from  and 19.4-8.0 Ma from ITS; see Additional file 6), when an ancestor of the Fucaceae might have been able to colonize the North Pacific, splitting from the Australasian sister lineages Xiphophoraceae and Hormosiraceae [23, 38]. Despite support for both alternative hypotheses for the Fucaceae geographic origin, a Pacific origin involves a more direct route from the southern (Australasia) to the northern Pacific (and is supported by diversification rates and the DEC model), whereas the alternative hypothesis of an Atlantic origin requires a more complex dispersal path. A Pacific origin is also consistent with the northward drift of the Australasian landmass towards Eurasia in the Miocene and a gradual decrease in global temperatures (14-12 Ma, see Figure 2; ), which would have favored crossing of the equatorial fringe. The origin of the Fucaceae would then be due to subsequent divergence in the North Pacific.
Our data indicate that four independent Fucaceae lineages crossed the Bering Strait. The first crossing, estimated at 16.4-5.4 Ma (Figure 2), involved the splitting of the Atlantic lineage leading to Pelvetia canaliculata, and could only have taken place during the earliest openings of the Bering Strait suggested for the Late Miocene (13.0-11.0 and 7.3-6.6 Ma; ). Pacific diatoms found in Atlantic marine sediments indicate the existence of a strait at that time , supporting such early Pacific-Atlantic colonizations. The alternatives to this scenario, other than methodological bias in dating, require either accelerated lineage divergence following the trans-Arctic crossing, or the start of divergence before the trans-Arctic crossing. The latter is unlikely because Pelvetia is currently monotypic with no extant Pacific representatives. While the extreme upper intertidal distribution and stress tolerance of Atlantic P. canaliculata, makes accelerated selective ecological divergence a plausible explanation, it is unnecessary to invoke it if earlier openings of the Bering Strait occurred . A second (and probably later; 11.5-1.1 Ma, Figure 2) trans-Arctic crossing led to the Atlantic genus Ascophyllum, following a split from its Pacific sister genus Silvetia, coincident with the Bering Strait opening at 5.5-5.4 Ma . These results contradict the previous ITS phylogeny of  but agree with these data after their re-analysis with better fit models (see methods). It revealed Ascophyllum as sister to the Pacific genus Silvetia and placed the Ascophyllum-Silvetia in a basal clade to the Fucaceae, a hypothesis also raised by .
The third (possibly simultaneous) trans-Arctic crossing, and the most successful in terms of subsequent speciation, was the split between the current Hesperophycus-Pelvetiopsis in the Pacific and the lineage leading to Fucus, of which all current species are Atlantic except the circum-Arctic F. distichus complex. This divergence, estimated at 12.2-2.7 Ma, coincides both in time and reproductive mode (shifting from hermaphroditic to dioecious) with the Ascophyllum lineage split from the Silvetia clade. The timing of both lineage splitting events leading to Ascophyllum and Fucus centers around the opening of the Bering Strait 5.5-5.4 Ma when, despite moving against the predominant Atlantic-Pacific flow, the warmer climate (see Figure 2) might have facilitated stepping stone colonization and migration across the Arctic. Ancestral state reconstructions (Figure 4) place the most recent common ancestor of Fucus in the Atlantic/Arctic ocean basin, suggesting that it was here that subsequent diversification took place. The alternative hypothesis, deserving further study, is that the opening of the Bering Strait led to a vicariant split between clade 1 in the Pacific and clade 2 in the Atlantic. An additional interesting question remains as to why, following similar colonization conditions by ecologically similar lineages, Ascophyllum is currently a monotypic genus whereas Fucus underwent relatively extensive speciation.
The fourth trans-Arctic crossing involved the evolutionary history of Arctic vicariance in Fucus clade 1. The ancestor to clade 1 was estimated as Atlantic (Figure 4), and the Atlantic-Pacific dichotomy might be more accurately described as Arctic to agree with the geographical and ecological range of current representatives. The ancestral state reconstruction implies that F. serratus/F. distichus diverged in the Atlantic and/or within the Arctic basin, which represent the same side of the Bering Strait, with subsequent invasion of the Pacific by the F. distichus lineage. Although Atlantic (previously named F. evanescens) and Pacific (previously named F. gardneri) samples of F. distichus used in this phylogeny correspond to the geographical extremes of the ranges found within the F. distichus complex [24, 46], estimated Pacific-Atlantic divergence times based on coalescence are very recent (mid-Pleistocene) (Figure 2). Thus our data do not contradict the current designation of these lineages as a single species, F. distichus (see ), but do not rule out low levels of vicariant divergence (Figure 1), also in agreement with Coyer et al. .
Driving south: a biogeographical hypothesis for the evolution of Fucus clade 2
The earliest branching member of the clade is the dioecious lineage F. ceranoides. The contemporary cold-temperate distribution of F. ceranoides from Norway to North Portugal is similar to the present day range of F. serratus in clade 1 , which has a coincident speciation time (Figure 2). Nuclear and organelle phylogenies for F. ceranoides are congruent in the southern part of the range, while to the north of the English Channel populations harbour exclusively introgressed organellar genomes captured from F. vesiculosus that have spread by genetic surfing during postglacial range expansion . This is not the only case of organellar introgression in this clade , emphasizing that organellar sequences can be equivocal for phylogenetic inferences in taxa prone to introgression. F. vesiculosus was shown here to be polyphyletic. Two clades were well separated within F. vesiculosus according to their range distributions from: i) Iberia to the south versus, ii) the English Channel to the north. These are also differentiated at microsatellite loci [25, 29, 47], both in allelic frequencies and in the presence of private alleles, but were not recovered previously with mitochondrial markers [24, 26], possibly due to masking by extensive organellar introgression-expansion dynamics that can take place in Fucus species . Importantly, the southern F. vesiculosus share a common ancestor with the remaining members of the same lineage, all of which are hermaphroditic. The two divergent lineages in what is currently named F. vesiculosus coincide in present distribution with two marine glacial refugia (Iberia and Brittany; ). A split of southern F. vesiculosus into two clades suggested by certain analyses (Figure 1 and Additional file 4 and 5) deserves further investigation, but could result from introgressive signatures with F. guiryi, which may be found in sympatry in some regions [11, 25, 47], but not in the southernmost sites where the two species are allopatric [11, 30] (Figure 1 and Additional file 5).
Divergence of the hermaphroditic lineage in clade 2 (leading to F. virsoides, F. spiralis and F. guiryi) from their dioecious sister lineage may have been driven or at least maintained by reproductive isolation derived from a selfing reproductive mode. Once a hermaphroditic lineage arises, selfing may follow rapidly, reinforcing genetic isolation and favouring subsequent differentiation . Selfing can be advantageous in marginal and/or stressful habitats to conserve local adaptation and for reproductive assurance, both key selective pressures for intertidal broadcast spawners such as Fucus .
The earliest divergence within the hermaphroditic clade is F. virsoides, currently restricted to the northern Adriatic Sea, a possible remnant from a more extensive distribution during a cooler glacial period. More recently, the lineage split between F. guiryi and F. spiralis coincides with southern vs. northern ranges. Along the southern range, Fucus species are segregated by habitat, i.e., open coast (F. guiryi) versus estuaries and coastal lagoons (southern F. vesiculosus), whereas further north, where they co-occur, F. guiryi undergoes introgression [11, 25, 26, 47], which was hypothesized to reflect the absence of reinforcement during allopatric evolution . The phylogenetic position of the high intertidal F. spiralis reported here is incongruent with mitochondrial data , possibly another case of extensive organellar introgression in this genus.
Our data, like previous ITS and mitochondrial data [23, 24], do not resolve the relationship between the recently described F. radicans and F. vesiculosus. This is unsurprising given the suggested timescale of divergence (hundreds to at most thousands of years ), since the opening of the Baltic Sea (ca. 7 Kya), possibly facilitated by high adaptive potential of the common ancestor with F. vesiculosus [10, 50].
Mating system evolution
The evolution of reproductive mode in the Fucaceae has followed a reticulate pattern of alternating dioecious and hermaphroditic lineages that challenges current understanding of mating system evolutionary trends (Figure 3; e.g., [15, 16]). Methods to estimate the influence of species' traits on lineage diversification establish hermaphroditic lineages as ancestral in the family, evolving into dioecious lineages, folowed by switches from dioecy to hermaphroditism in the genus Fucus, contradicting earlier suggestions [24, 51]. There is considerable support for hermaphroditism (cosexuality) as the ancestral state in plants , and simple genetic mechanisms of dioecious sex determination and sex chromosome evolution have been proposed (reviewed by [52, 53]). It is intriguing that two of the three novel Atlantic lineages presumably coincided with a switch to dioecy (Ascophyllum and Fucus). The evolution of dioecy and increased evolutionary potential  may therefore have facilitated long-term establishment in the Atlantic, driven in part by the availability of extensive and novel habitats favouring large and dense populations. In contrast, hermaphroditic lineages are better colonizers of marginal habitats via increased reproductive assurance and the maintenance of locally adaptive traits.
The recent evolutionary trajectory of reproductive mode has been a switch towards hermaphroditism, and highly selfing mating systems, at least within Fucus lineage 2 [29, 30]. The transition from outcrossing to selfing is common in plants , but with little evidence for reversion, suggesting an evolutionary dead-end [16, 55]. This in turn suggests that the hermaphroditic ancestors of the dioecious lineages leading to Ascophyllum and Fucus were not highly selfing.