Eurasian shad diversification
Our analysis provides strong support for the monophyly of Eurasian shads yet weak support for the internal branching order of major European lineages, as previously shown with a far more limited data set . The lack of resolution for the relationships among major lineages presumably reflects relatively rapid radiation during the Pleistocene (i.e. a hard polytomy) but whole mtDNA genome data or an extensive nuclear gene data set might still provide more resolution. Even the question of the sister species status of A. alosa and A. fallax[25, 28] cannot be unequivocally answered due to the very weak node support of the clade grouping A. fallax and BSC haplotypes (Figure 1). However, divergence estimates (Table 1) favor a considerably closer relationship of A. fallax to the BSC than to A. alosa.
For estimated divergence times among lineages we rely on a credible but rough divergence rate of 2% /Myrs for the whole mtDNA, though we provide inferences based on a range from 1-4%/Myrs (Table 1). Rates such as 1%/Myrs are very unlikely as this would result in estimates of demographic expansion in the Atlantic (Clade 1 in both A. alosa and A. fallax) ca. 25–40 thousand years ago, which is at the height of the last glacial maximum (LGM) (see Table 7). Rates of 3%/Myrs have been convincingly shown for the genus Coregonus whereby their calibration is based on a more recent split, and there is mounting evidence that there is a time-decay phenomenon for mtDNA divergence rates due to the delayed efficiency of purifying selection within a phylogeny [35, 36]. Moreover, there is little evidence for considerably higher rates of divergence for temperate freshwater fishes (see discussion on cold-tolerant fishes in [7, 37]). Thus, we limit our discussions to inferences based on a 2%/Myrs divergence rate, but emphasize that the relative as opposed to the absolute dating of events among lineages is more credible.
Using this framework, some general inferences can be drawn concerning the diversification of European Alosa lineages. First, all species or clade splits are confined to the Pleistocene implying that these cold-tolerant taxa proliferated or radiated during glacier-mediated climatic oscillations and not during pre-Pleistocene (i.e. Pliocene) climatic conditions, which were considerably warmer , or perhaps even unsuitable for Alosa in the region under study. Second, all species or clade splits occurred over time periods older than the LGM and thus this most recent glacial cycle had little to nothing to do with the origin of major lineages. The divergence between A. fallax and the Black Sea lineages started roughly 0.45 to 0.75 Myrs ago. We speculate that a radiation occurred during this period when connections between the Mediterranean and Black seas (as well as Caspian) took place , allowing the eastern expansion of Alosa into the Black and Caspian seas. These connections were subsequently interrupted several times (see  and refs. therein, ), promoting isolation and differentiation, providing a clearer biogeographic break, which has presumably played a role in promoting cladogenesis . Similar to the communication between the Mediterranean and Atlantic basins, however, there is presently no absolute barrier between the Mediterranean and Black seas. The present connection occurs over a salinity gradient along the Dardanelles channel, which connects the Aegean and Marmara seas and has bottom salinities similar to the Mediterranean. The Marmara Sea is then connected to the Black Sea via the Bosporus strait, whose total saline input into the Black Sea combined with the present freshwater inputs results in an approximately 17.5 to 19 ppt surface water salinity . This raises the question of whether the observed cladogenesis and speciation in the genus Alosa in this region is primarily promoted through allopatry stemming from phases of geographic isolation during reduced Mediterranean Sea levels, or simply the starkly differing environmental conditions (or both), as suggested for other species in the Ponto-Caspio region .
Salinity gradients or breaks have clearly played a role in the diversification of the genus Alosa. Several landlocked populations of Alosa complete their entire life cycle in freshwater and differ in terms of morphology, genetics and physiology from nearby anadromous populations [20, 27, 41, 42]. A. macedonica, endemic to freshwater Lake Volvi in Greece, is one such example. Freshwater lakes of this region were presumably colonized with Alosa immigrating from a low salinity phase of the Black Sea during one of its spills into the Aegean Sea . Alosa colonized two lakes on the Greek peninsula (Lake Volvi and Lake Vistonis), whereby the Lake Vistonis lineage has apparently gone extinct due perhaps to human-caused salt water intrusion . Alosa from Lake Volvi is considered a distinct species carrying out its entire life cycle in freshwater, whereas the Aegean Sea region, now with higher salinity harbors only A. fallax of Mediterranean origin.
Interspecific gene flow versus ancestral polymorphism
While ancestral polymorphism could in theory explain shared haplotypes between the two species, the evidence for hybridization and introgression is overwhelming, in agreement with previous published studies [20, 27, 28]. The occurrence of apparent hybrids is not geographically homogeneous, with certain rivers showing high and others low frequencies. This alone is suggestive of hybridization in contrast to ancestral polymorphism, which should show little geographic pattern. However, given the anadromy and natal site fidelity of the species, such observations are not 100% conclusive.
Coalescent-based estimates under a model of isolation-with-migration support gene flow between A. alosa and A. fallax. Strong population substructure or complex demographic scenarios may provide violations of model assumptions. However, several lines of argument strongly support that introgression is occurring between the two species. First, it has been shown that migration estimates based on isolation-with-migration models are robust to population structure and departures from simple demographic scenarios . Second, the occurrence of morphological hybrids between these species has been widely documented . And finally, previous studies showed a correlation between genotypes based on nuclear markers and gill raker counts, with the hybrids presenting intermediate morphological and genetic compositions, strongly supporting the occurrence of hybridization .
Interestingly, the levels of introgression are particularly high in A. fallax populations of Usk and Tywi (Wales, UK), although A. alosa basically disappeared from this region [27, 32]. These results are concordant with the higher mtDNA introgression levels previously found in these same rivers . The interpretation of these results under a wider geographic perspective suggests that hybridization between A. alosa and A. fallax varies among river basins from 0 to 63%, reaching a maximum in Tywi (Wales). The causes of different levels of introgression among river basins are unknown but might be related to either differing degrees of anthropogenic disturbance, mainly the construction of migration obstacles such dams and weirs [20, 21], and/or innate differences relating to behavior or genetic composition. Whether natural or anthropogenically induced, changing environmental conditions have long been thought to play a major role in increasing rates of hybridization in fishes . Several recent freshwater fish studies [46, 47] underscore the importance of broader ecological degradation leading to increased hybridization rates, a situation which easily applies to the UK and Portuguese rivers where high levels of hybridization are documented. Nevertheless, shads represent a very unique case of introgressive hybridization among European fishes, something that needs to be further addressed using also nuclear markers.
Geographic distribution of intraspecific genetic variability
Phylogeographic structure within A. fallax revealed by AMOVA and SAMOVA clearly relates to the major basins (Atlantic and Mediterranean), but also to smaller-scale patterns evidenced by regionally specific haplotypes. Therefore, it appears that differentiation was mainly shaped by historical processes of fragmentation between the Mediterranean and the Atlantic as described for many other species (e.g. species of the Sparidae family- marine [48, 49]; brown trout- anadromous ), but also by regional barriers within these oceanographic regions. The divergence time estimates between the main clades marking the divide between the Mediterranean and the Atlantic A. fallax populations (0.20 to 0.30 Myrs between clade 2 and clade 1; Table 1) suggest that separation occurred after the Mindel glaciations, when a hypothetical ancestral population expanded its range through the three regions, allowing subsequent divergence. However, we do not know with certainty if fragmentation was caused by sea level drop during glacial maximums, shifting marine currents, or adaptation to different environmental conditions existing between the two basins. Thus, at least for A. fallax, the Strait of Gibraltar or a nearby region (e.g. the Almeria-Oran front) act in restricting gene flow between the Mediterranean and Atlantic, as described for many other species . This barrier may not be absolute for A. fallax, considering the existence of shared haplotypes (Af3 and Af10) between the two basins and the coalescent-based migration estimates, with significant gene flow from the Mediterranean into the Atlantic (but not vice versa; Table 2). A simulation study testing the violations to isolation-with-migration models showed that gene flow with a third population, can inflate migration and effective population size estimates . However, our estimates remained significant even when we corrected for this possibility (Additional file 1: Figures S1D, 1E and 1F), namely migration between Morocco and European Atlantic coast, suggesting that A. fallax migration from the Mediterranean into the Atlantic inferred using mtDNA is indeed real. Although migration over long distances through the Straits of Gibraltar is possible, we cannot exclude alternative possibilities associated with past connection between headwater captures of Mediterranean and Atlantic draining rivers, migration through artificial canals, or non-documented human-mediated transport related with stocking. Nonetheless, as no A. fallax haplotypes typical for the Atlantic region have been thus far found in the Mediterranean and A. alosa is currently limited to the Atlantic basin, it would appear as if the Atlantic-Mediterranean corridor is at least a contemporary isolating mechanism for these species.
The geographic origin of clade 3 in A. fallax is not clear. The fact that it reaches higher frequencies in the southern Atlantic populations from Morocco and Guadiana suggests an African or Southern Iberian origin. However, the lack of variability in the southern most population (Sebou) is puzzling. Huge population declines and possible local extinction has been suggested for A. fallax in Morocco , which could have resulted in a depletion of variability on that region. However, further studies using nuclear markers and samples from other African populations (if available) are needed to evaluate this hypothesis and to elucidate about the putative origin of clade 3.
At a smaller geographic scale, significant population structure (Table 6) supported by the existence of fixed or regionally restricted haplotypes is observed in A. fallax (Table 4). This pattern is especially evident in the Mediterranean, where the existence of relatively frequent haplotypes restricted to some regions (Af9 in Greece and Turkey; and Af19 in Corsica and Sardinia; Table 4) explain the higher differentiation found in this oceanographic region compared with the Atlantic (Table 6). This can either suggest a higher fidelity in terms of homing behavior of the A. fallax populations inhabiting the Mediterranean or a longer history of isolation accompanied by a more demographic stability of these populations. However, this is difficult to evaluate with the present data set, as the system of rivers draining to each of these oceanographic regions is not easily comparable.
Genetic variation of Alosa in the Atlantic basin shows both concordance between the two species, with respect to the haplotype networks, as well as strong differences in overall diversity and phylogeographic structure. In both species, three similarly divergent clades are present, most likely reflecting evolution in the same refugia during glacial maxima. Despite the fact that A. alosa displays little phylogeographic structure compared to A. fallax (Table 6), clades as a whole are broadly spread throughout the Atlantic basin in both species, where they co-exist (Figure 3). This pattern implies a concordant history of both species presumably due to the same climatic events and the same long-term refugia, yet differing in post-glacial demography and/or ecological responses to climatic amelioration (see below). A. alosa has a considerably more limited distribution and spawns much higher in the river systems than A. fallax, an obviously more demanding ecological niche, through periods of natural hydrological instability and glacial advance as well as in more contemporary times due to numerous anthropogenically caused interruptions in river corridors. Consequently, the task of inferring the geographic origin of the three main clades observed for A. alosa is comparatively much more challenging, as the species disappeared from some of the putative candidate regions (Mediterranean and Morocco). Future studies should thus make use of existing material (ancient DNA) and of A. alosa haplotypes “available” in populations of A. fallax through introgression, prior the local extinction of A. alosa, to identify the geographic origin of the clades identified here.
Demographic responses to Pleistocene climate
Interestingly, although the two species share parallel intraspecific clade structures supporting parallel refugia, mtDNA haplotype diversity of A. alosa is much lower than for A. fallax. The rapidly declining range of A. alosa, including its disappearance from the Mediterranean basin and Northern Africa can explain the differences found between the two species [20, 21], and have perhaps permanently clouded any trace of its evolutionary origins in geographic terms.
The demographic responses detected in our analysis were largely lineage specific, most likely reflecting relatively large-scale environmental changes. Assuming a neutral or nearly-neutral evolution of the mtDNA molecule, strong statistical support for post-glacial growth was seen for both A. alosa and A. fallax lineages in the Atlantic basin, with roughly similar age estimates, presumably reflecting expansion after the LGM (Table 6). Reproductive migrations of A. fallax are thought to be inhibited by water temperatures below 11°C  implying that this species must have been purged from, or at least drastically reduced in the North Atlantic during the LGM. However, not only present, but also historical effective population sizes of A. fallax may have been larger than for A. alosa. This may suggest that even before anthropogenically caused environmental changes A. alosa had a more limited capacity to survive climatic oscillations. However, we cannot exclude that a population of A. alosa may have survived in the North Atlantic, where haplotypes of clade 2 are more frequent (Figure 3B), a hypothesis that needs to be tested in the future with nuclear data. Regardless of the refugia location, it appears that the refugial populations of A. alosa have only carried limited mtDNA haplotype diversity at our level of resolution compared to A. fallax, which seems to have experienced more stable conditions along the Iberian coast compared to the North Atlantic.
Demographic growth in A. fallax (as suggested by Fu’s F
s) seems to have started earlier in the Mediterranean than in the Atlantic (Table 7; Additional file 2: Figure S2), suggesting either a faster amelioration or the maintenance of adequate environmental conditions in the Mediterranean during the last glacial period (Würm), allowing the permanence of stable populations of A. fallax in the region. The actual rarity of Alosa on the southern shores of the Mediterranean probably reflects that the present day sub-tropical temperatures are far from optimal for this species. In general, for both A. alosa and A. fallax, the contemporary southern limit of their distributions appears to reflect marginal habitat conditions, due to warm temperatures. During the last century, the occurrence of A. alosa along the Mediterranean was mainly reported for the coast of the Iberian Peninsula and France, where A. fallax is presently abundant. The fact that this region exhibits the coolest water temperatures along the entire Mediterranean basin  further supports our hypothesis.
The distribution of the genetic diversity of European Alosa species throughout the Atlantic, in particular of A. fallax (clade 1), suggests that a southern refuge existed along the Iberian Peninsula (or further south), similar to that reported for many other species (e.g. marine- the common goby, Pomatoschistus microps; freshwater- saramugo, Anaecypris hispanica; anadromous- brown trout, Salmo trutta; but see  for a review). However, contrary to what has been suggested for other anadromous fishes with Atlantic distribution, namely the cold-tolerant species Atlantic salmon S. salar and brown trout S. trutta (see also  for marine species) for which glacial refugia in Northern Europe have been suggested, Alosa northern populations seem to have been directly affected by the advance of ice sheets, as suggested by the low nucleotide diversity observed in these populations, when only the haplotypes of the clade with a putative Atlantic origin (clade 1) are considered. Thus, Alosa’s comparative cold-intolerance may limit the potential range of glacial refugia in the Atlantic, which, together with the fact that contemporary climatic conditions appear to be causing problems for the species at the southern limit of their distribution, underscores, especially for A. alosa, the sensitivity of these species to climatic change. In contrast to the Atlantic, the Mediterranean seems to provide more long-term stable habitat for lineages of Alosa during colder periods, but the opposite may also be true during interglacials, as suggested by the recent disappearance of A. alosa and the distribution of A. fallax throughout the region, with lower abundance in areas with higher seawater temperature.