The partitioned ML analysis resulted in a relatively well-resolved tree, with approximately 70 per cent of the internal branches supported by moderate to high (70–100%) bootstrap probabilities (BPs; ). Although the four families of the ‘saccopharyngiforms’ form a monophyletic group in the resultant tree, the clade is weakly supported (<50% BP) and is deeply nested within the anguilliforms as shown in the previous study (Inoue et al. 2004
). All the three currently recognized suborders (Congroidei, Anguilloidei and Muraenoidei) are recovered as polyphyletic, with two or three unnested monophyletic groups recognized for each suborder (e.g. Anguilloidei clades 1, 2, 3 in ). Of the 12 families, for which we sampled two or more taxa, nine are confidently recovered as monophyletic with 100 per cent BPs. The rest of the three congroid families (Derichthyidae, Nettastomatidae and Congridae), on the other hand, are recovered as either para- or polyphyletic within a more comprehensive clade (Congroidei-2) supported by 100 per cent BP ().
Figure 2. The best-scoring maximum likelihood (ML) tree of 58 elopomorph species based on unambiguously aligned whole mitogenome sequences (13 701 positions). Numerals beside internal branches indicate BPs of ≥50% based on 1000 replicates. Evolution of (more ...)
Apparently, the higher level classification of the anguilliforms requires substantial revision based on more extensive taxon and character sampling. Significantly, however, the present ML tree unequivocally places the freshwater eels at the top of the anguilliform phylogenies and they are nested within a more comprehensive monophyletic group (clade A) supported by 100 per cent BP. Other than the freshwater eels, clade A comprised the four saccopharyngiform families (Saccopharyngidae, Eurypharyngidae, Cyematidae and Monognathidae) and two congroid families (Nemichthyidae and Serrivomeridae), with these six families containing 47 species placed in 10 genera (Nelson 2006
). Interestingly, these 47 species are all oceanic midwater dwellers, occurring mainly at tropical and subtropical meso- and bathypelagic depths (200–3000 m) throughout their adult stages with no exception (Miller & Tsukamoto 2004
). As expected from the resultant phylogenies, the ML and Bayesian reconstruction of the ancestral growth habitats explicitly indicate that the freshwater eels evolved from an oceanic midwater ancestor (), with the character state 2 (oceanic midwater) being most likely at nodes A and B (p
How can we explain why this apparent evolutionary shift of the freshwater eel life history from the oceanic midwater to freshwater occurred? These two environments are remarkably different and require fishes to be adapted to very different ecological and physiological constraints. One possibility is that in tropical regions at the time of the divergence of anguillids, there was a productivity gradient between freshwater and marine environments as hypothesized by Gross et al. (1988)
, which made it advantageous for feeding success to invade freshwater. The importance of this relationship is supported by apparent clines in freshwater use in temperate anguillid species, with fewer eels entering freshwater at the northern margins of their ranges where productivity is much lower than in the estuarine and coastal habitats (Tsukamoto et al
). Another possibility is that because there were probably no eels in freshwater at that time, compared with the presence of multiple lineages of eels in marine environments (), including voracious predators such as moray eels (family Muraenidae), there was a vacant niche for eels in freshwater. In addition to the lack of competition with other eels, most freshwater habitats also might have had fewer large predators that could prey on eels.
Regarding the characteristics of the first freshwater eels, it should be noted that the present ancestral character reconstruction is based on the traits of the extant species and thus it does not specify the character state between nodes B and C (). If there was an ancestral eel lineage between these two nodes that went extinct long ago and lived at shallower depths than present-day mesopelagic eels, then the habitat shift into freshwater would have been more gradual. This ancestor may have eventually come to estuaries during their larval or juvenile phases and developed an adaptive behaviour of regularly inhabiting estuaries and occasionally entering freshwater in tropical regions because of higher food availability, better survival or to escape from predators (Tsukamoto et al. 2009
). Once natural selection resulted in the emergence of eels that regularly used freshwater for growth, a new catadromous life history was established, with the eels still using the open ocean as their spawning area.
The closest relatives of anguillid eels found in this study are the mesopelagic eels of the Nemichthyidae and Serrivomeridae that spawn in the open ocean, with their larvae mixing in the ocean surface layer with those of anguillids (Miller et al. 2006
). The recent capture of mature adults of the Japanese eel, Anguilla japonica
, at depths of about 220–280 m in the western North Pacific (Chow et al. 2009
), shows that anguillids have retained the apparent ancestral trait of offshore pelagic spawning. Reproductive behaviour is typically conservative and constrained by many ecological and physiological factors (Tsukamoto et al. 2002
), so the catadromous migration of freshwater eels back to their offshore habitats over an evolutionary time scale represents a remarkable relic of the reproductive behaviour of these enigmatic animals that share a common ancestry with pelagic eels of the deep ocean.