Evolution: In search of the whales' sisters
Nature, Vol. 404, No. 6775, p. 235-239 (1998) © Macmillan Publishers Ltd.
Which mammals are the closest relatives to cetaceans (dolphins, porpoises and whales)? That is, which mammalian group is their 'sister taxon'? There is a wide gulf between the morphological and molecular evolutionary studies on the question, for they give conflicting answers. In a paper in Systematic Biology, O'Leary and Geisler[1] highlight the importance of the early divergent lineages, or extinct fossil taxonomic groups, in resolving this intriguing problem.
The descent of whales from land-dwelling mammals is a compelling example of evolution. It is documented by a rich fossil record of intermediate forms spanning the land–water transition; and by morphological and molecular lines of evidence, both of which testify that cetaceans have a close affinity to the artiodactyl mammals, or ungulates with even-toed feet and double-pulley ankle, such as today's cow, camel and hippopotamus. But two central issues have yet to be settled. One is the identity of the cetacean sister taxon. The other is whether the cetaceans are 'nested' within the group of artiodactyls as a branch on the family tree of the living artiodactyls — in other words, whether cetaceans represent some ancient artiodactyls that became adapted to aquatic life.
What O'Leary and Geisler have done is to evaluate all the available morphological and molecular evidence[1, 2. They show that both issues hinge on the long-extinct fossil groups that were split early from extant cetaceans. The topology, or branching sequence, of the family tree of living artiodactyls and cetaceans depends on inclusion of these divergent lineages of cetaceans and their putative extinct relatives.
Among those possible fossil relatives are the mesonychids that lived from 60 to 30 million years ago. Their feet were even-toed and adapted for running, like those of modern artiodactyls, and their fish-eating or carrion teeth closely resemble those of the earliest whales. Since the 1960s3, several morphological studies1, 2, 4, 5 (with one exception6) have considered that extinct mesonychids and cetaceans are sister taxa (Fig. 1a). The combined mesonychid–cetacean group is, in turn, related to living artiodactyls. In this scheme, living artiodactyls are recognized as a clade1, 2, 5, 7, a genealogical group that includes all descendants of a common ancestor.
Figure 1 Vectorial representation (not to scale) of thrust generation and aeroelastic response at low running speeds (bold symbols represent vectors).
In contrast, molecular studies8-12 have generally concluded that hippopotamuses and cetaceans are more closely related to each other than either group is to other living artiodactyls (Fig. 1b). Consequently, the artiodactyls would be a grade that has similar evolutionary adaptations, but not a genealogical clade. Moreover, some molecular studies12 suggest that mesonychids are not closely related to cetaceans; this is because the mesonychid–cetacean relationship implies a larger gap in the fossil record9, and is therefore less plausible than the hippo–whale connection12. So, for all their congruence on the broad picture of ungulate–cetacean evolution, there is a big disagreement between morphological and molecular studies over these two phylogenetic issues.
What does that mean in terms of the origin of cetacean adaptations to aquatic life? If hippos and whales are sister taxa to the exclusion of mesonychids (as molecular studies suggest), then cetacean adaptations such as underwater nursing of offspring and nearly hairless skin could have originated in the most recent common ancestor of both groups (Fig. 1b); this implies that certain aquatic features had evolved before the origin of cetaceans9. The similar ear regions4, 5 and the fish-eating and carrion teeth must have evolved independently in mesonychids and in the earliest cetaceans because these derived features are absent in living hippos and their artiodactyl allies.
Alternatively, if mesonychids and cetaceans are sister taxa (as the best morphological evidence has it), then the aquatic adaptations of hippos and living cetaceans must be convergences that occurred well after the split of their respective lineages. This is because certain primitive cetaceans (pakicetids) have many ear4 and ankle6 structures typical of a land mammal, and mesonychids were fully terrestrial and adapted to running (Fig. 1a).
What about the relative strengths of the two lines of evidence? Morphological studies of ungulate–cetacean phylogeny take in a much wider range of taxa than the molecular studies. This is because the living artiodactyls and cetaceans available for molecular analyses represent only a few twigs of their bushy family trees that have survived pruning by extinction. As O'Leary and Geisler1 point out, 90% of ungulate genera and more than 86% of the cetacean genera are extinct. Taking the mesonychid and cetacean fossil taxa into account produces a markedly different evolutionary history (Fig. 1a). From this, O'Leary and Geisler conclude that artiodactyls are a clade to the exclusion of cetaceans. Here we have a good example of the principle that including data from early divergent fossil lineages can shake, and reshape, the trees of living taxa based solely on molecular evidence.
However, for living taxa, use of sequences of genes and proteins is in some ways more powerful than use of morphological characters. In recent years, more genes in larger sequence samples have been added to the arsenal for estimating ungulate and cetacean phylogeny9-12. And the poor taxonomic sampling of some of the earlier molecular work has at least in part been redressed10. All in all, because molecular characters can vastly outnumber morphological features, they often prevail in simultaneous analyses of conflicting data sets. That is, in such analyses they effectively swamp the morphological estimates of ungulate–cetacean phylogeny.
Both morphological and molecular data are vulnerable to the problem of homoplasies — reversals to ancestral conditions or parallel changes in different lineages that can camouflage the true phylogeny. In this sense, neither approach is better than the other. For instance, the ear region of the skull, traditionally considered to be a good source of highly stable characters, shows some glaring homoplasies among the ungulates and cetaceans4, 5. Moreover, the fossil record of many early divergent fossil taxa is incomplete, resulting in ambiguities in morphological estimates.
On the molecular side, DNA and protein sequences have parallel and back mutations. Even the newest studies using retroposons, which are the RNA-mediated insertion sequences interspersed in the genome, have their limitations. Retroposons show a low level of homoplasy12, 13, but mutational decay of the flanking region of retroposons may make them difficult to detect in older lineages. This means that retroposon-based estimates may not be effective for resolving lineages that go back more than 50 million years14. Cetaceans were already diversified by 53.5 million years ago, and their divergence from extant artiodactyls goes back much further15, 16 than that.
A way to untie this Gordian knot may be to seek out compatible aspects of the molecular and morphological data sets. Measurement of the hidden support and conflict between them10 can help extract additional information. Also, morphological features should be better analysed for some living taxa for which extensive molecular data are available. Such studies may be just as helpful as discoveries of new fossils and genes in resolving details of the cetaceans' phylogenetic tree — most particularly, the question of which group is their sister taxon.
Zhexi Luo
Department Vertebrate Paleontology,
Carnegie Museum of Natural History,
Pittsburgh, Pennsylvania 19213, USA
e-mail:luoz@carnegiemuseums.org
