Evolutionary biology

Even-toed fingerprints on whale ancestry

Article metrics

Both morphological1 and molecular2 studies indicate that cetaceans (whales, dolphins and porpoises) and artiodactyls (even-toed ungulates, which include pigs, hippos, camels and ruminants) form a clade or monophyletic group — that is, they have a common ancestor that is not shared by any other group of mammals. This is counter-intuitive, because it implies that a cow is more closely related to a dolphin or a whale than to a horse, yet it is one of the best examples of congruence between morphological and molecular estimates of mammalian phylogeny.

The molecular analyses of Shimamura et al.3, reported on page 666 of this issue, further disrupt phylogenetic dogma. Indeed, not only do the authors confirm the close relationship between artiodactyls and cetaceans, but they propose that cetaceans are deeply nested within the phylogenetic tree of the artiodactyls. These results strikingly contradict the common interpretation of the available morphological data (supporting artiodactyl monophyly) and, if correct, would make a cow or a hippopotamus more closely related to a dolphin or a whale than to a pig or a camel (Fig. 1).

Figure 1: Twisted or untied?

Shimamura et al.3 propose to attach the lineages of hippos and cetaceans (curved arrows, red) to the ruminant branch on this phylogenetic tree of artiodactyls — a marked diversion from the traditional view (white branching pattern). The two smaller boxes summarize some of the morphological evidence that disagrees with the new data.

Although it is compatible with earlier molecular analyses (for example, refs 4, 5), the idea that cetaceans are highly derived artiodactyls was first suggested in 1994 on the basis of mitochondrial and nuclear DNA and amino-acid sequences6. The idea was corroborated by other phylogenetic analyses of DNA sequences7. But the issue is still controversial, because the exact means by which molecular sequence data should be analysed remains debated — although analytical settings that are particularly meaningful with respect to phylogenetic inferences can probably be identified in specific instances8. But, basically, many morphologists consider that molecular data are necessarily more noisy than morphological data.

The analyses by Shimamura and colleagues now provide a remarkable example of molecular markers, which should lead morphologists to re-examine what might have misled them for more than a century. The authors report phylogenetic interpretations of nine retropositional events that led to the insertion of so-called ‘short interspersed elements’ at particular loci in the nuclear genome of various artiodactyl and cetacean ancestors. Three of these events unambiguously support the grouping of cetaceans, hippos and ruminants in a clade, and the other six provide a partial resolution of the relationships within that clade. Because the likelihood of these elements being independently inserted at the same locus in different lineages (or precisely excised) seems virtually nil, these markers can reasonably be considered to be essentially noise-free.

So, these molecular results may prompt a serious revision of how we view morphological transformations in whales and artiodactyls. Three salient features of artiodactyls are: first, the axis of symmetry of hand and foot runs between the third and fourth digit (paraxony); second, the heel has developed an extremely mobile joint at a place where most mammals have a barely mobile joint9; and third, the last lower milk molar consists of three rows of cusps (three-lobed deciduous lower premolar 4, DP/4). Although paraxony is a striking feature, it also occurs in primitive whales10, so it is uninformative for the issue at hand. But the remodelled heel is a different matter — because it is present in all artiodactyls and in no other mammal, this character is classically interpreted as derived and, therefore, as supporting artiodactyl monophyly.

The presence of the mobile joint in artiodactyls can be inferred from bones: there is a pulley (or trochlea) on the distal part of the astragalus (one of the heel bones) — a so-called trochleated astragalar head. This led to efficient and fast locomotion in the earliest artiodactyls. Although some features of the artiodactyl heel are present in other mammals such as rabbits, the astragalar head is never trochleated. In modern cetaceans, the hindlimb is so greatly reduced that the heel cannot be recognized, and no complete functional astragalus is known for a fossil whale. However, the mesonychians — a group of land mammals that is considered to be the closest extinct relative of cetaceans — also lack a trochleated astragalus. So if Shimamura and colleagues’ hypothesis is correct, then either mesonychians are not closely related to cetaceans (and many dental characters are convergent), or the specialized heel morphology is not the exclusive character that many morphologists take it to be. It may have evolved several times independently in artiodactyls, or have been lost in the mesonychian/cetacean clade. The complete astragalus of an early cetacean would probably shed light on this issue.

The hypothesis put forward by Shimamura et al. also clashes with the DP/4 character — the tooth has three lobes in all artiodactyls, but not in early cetaceans or mesonychians. If the new molecular data are correct, the morphology of this tooth has, like the trochleated astragalus, a complicated phylogenetic history that includes reversals, convergences or both. Dental differences could reflect dietary differences, because both early cetaceans and mesonychians were probably carrion feeders or carnivores, whereas all early artiodactyls were omnivores or herbivores. But many morphological systematists are reluctant to let functional arguments influence their evaluation of characters.

What characters, besides the molecular ones described by Shimamura et al.and by Gatesy7, do hippos, ruminants and cetaceans have in common that makes them different from pigs, peccaries and tylopods (camels and llamas)? Morphological studies have usually upheld close genealogical ties among pigs, peccaries and hippos (and the larger group that includes hippos: anthracotheroids), but the similarities are usually uninformative primitive characters that are also present in the ancestral artiodactyl, or features that are subject to rampant convergences (such as certain dental characters, see Fig. 1). Shimamura et al. also suggest that artiodactyls and cetaceans diverged in the Cretaceous — about 15 million years before they are found in the fossil record. However, if hippos are closely related to cetaceans, and if mesonychians are not, then the discrepancy in the times of origin is considerably less: it amounts to the difference between the origin of cetaceans (approximately 50 million years ago) and that of anthracotheroids (around 49 million years ago). But because of the rapid and unique specialization of cetacean morphology, few characters can be recruited to bolster the grouping of ruminants, hippos and cetaceans into a clade. Recovery and study of the earliest cetaceans would probably help to resolve this problem.

Few molecular studies rule out previously accepted morphological trees as convincingly as that of Shimamura et al. However, the face of phylogenetic science itself is changing rapidly — it is becoming more objective and less inductive. For example, phylogeneticists no longer need to ask whether molecular data are superior or inferior to morphological data, because the signal-to-noise ratio in morphological, molecular and combined data sets can now be measured directly, even before examining phylogenetic trees11. In any case, the new analyses indicate that the use of retropositional events as molecular markers may define a new power of resolution in estimating phylogenies. This method could even bring to a close some of the most intense controversies2,12,13 in the field, such as whether the toothed whales2,12 are monophyletic or paraphyletic, and likewise for the rodents13.


  1. 1

    Thewissen, J. G. M. & Hussain, S. T. Nature 361, 444–445 (1993).

  2. 2

    Milinkovitch, M. C. Trends Ecol. Evol. 10, 328–334 (1995).

  3. 3

    Shimamura, M. et al. Nature 388, 666– 670 (1997).

  4. 4

    Irwin, D. M., Kocher, T. D. & Wilson, A. C. J. Mol. Evol. 32, 128– 144 (1991).

  5. 5

    Czelusniak, J. et al. Current Mammalogy 2 (ed. Genoways, H. H.) 545–572 (Plenum, New York, 1990).

  6. 6

    Graur, D. & Higgins, D. G. Mol. Biol. Evol. 11, 357–364 (1994).

  7. 7

    Gatesy, J. Mol. Biol. Evol. 14, 537–543 (1997).

  8. 8

    Milinkovitch, M. C. et al. Genetics 144, 1817– 1833 (1996).

  9. 9

    Schaeffer, B. Am. Mus. Novit. 1356, 1–24 (1947).

  10. 10

    Thewissen, J. G. M., Madar, S. I. & Hussain, S. T. Cour. Forschungsinst Senckenbg 191, 1–86 (1996).

  11. 11

    Lyons-Weiler, J., Hoelzer, G. A. & Tausch, R. J. Mol. Biol. Evol. 13, 749– 757 (1996).

  12. 12

    Milinkovitch, M. C., Ortí, G. & Meyer, A. Nature 361, 346–348 ( 1993).

  13. 13

    Hasegawa, M., Cao, Y., Adachi, J. & Taka-aki, Y. Nature 355, 595(1992).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.