The Ancestry of Whales Fossils

S CIENCE’ S C OMPASS

65 4. L. van der Pijl, C. Dodson, Orchid Flowers: Their Polli- 9. R. E. Ricklefs, D. Schluter, in Species Diversity in Eco- 14. R. B Foster, S. P Hubbell, in Four Neotropical Forests 64 nation and Evolution (Univ. of Miami Press, Coral logical Communities (Univ. of Chicago Press, Chicago (Yale Univ. Press, New Haven, CT, 1990), pp. 85–98. 63 Gables, FL, 1966). and London, 1993), pp. 350–363. 15. R. Condit et al., J. Tropical Ecol. 12, 231 (1996). 5. J. Haffer, Science 165, 131 (1969). 10. A. H. Gentry, C. H. Dodson, Ann. Missouri Bot. Gard. 16. K. Romoleroux et al., in Estudios Sobre Diversidad y 62 6. J. E. Richardson et al., Science 293, 2242 (2001). 74, 205 (1987). Ecología de Plantas (Pontificia Universidad Católica 61 7. E. Zuckerkandl, L. Pauling, in Evolving Genes and Pro- 11. J. Fjeldså, Biodivers. Conserv. 3, 207 (1994). del Ecuador, Quito, 1997), pp. 189–215. 60 teins (Academic Press, New York, 1965), pp. 97–165. 12. Reviewed in C. Moritz et al., Annu. Rev. Ecol. Syst. 31, 17. We thank R. Condit, S. Lao, R. Valencia and R. Fos- 8. R. E. Latham, R. E. Ricklefs, in Species Diversity in Eco- 533 (2000). ter for providing unpublished forest plot data, 59 logical Communities (Univ. of Chicago Press, Chicago 13. A. H. Gentry, in Tropical Forests (Academic Press, and N. Smith, M. Ashton and J. Patton for helpful 58 and London, 1993), pp. 294–314. New York, 1989), pp. 113–134. comments. 57 56 PERSPECTIVES: EVOLUTION 55 from the shape and orientation of joint 54 surfaces of several ankle bones in the new 53 The Ancestry of Whales fossils. These specialized features, typical- 52 ly associated with adaptation to running, 51 Kenneth D. Rose have only been observed in artiodactyls 50 and are widely considered diagnostic of 49 hales are mammals that moved to phological evidence that whales are not just the order (see the first figure). Their pres- 48 the sea about 50 million years related to, but descended from artiodactyls ence in an animal that was probably better 47 Wago. Exactly how they are related rather than mesonychians, thus bringing the adapted for aquatic than terrestrial loco- 46 to other mammals has long been one of the morphological evidence into accord with motion strongly suggests common her- 45 most vexing questions facing mammalogists molecular data, at least at the ordinal level. itage rather than convergent evolution. 44 and paleontologists. In the last decade, The most important evidence comes Ankles from primitive an- 43 mounting evidence that whales are highly cient whales have previously 42 specialized ungulates (hoofed mammals) 5 been reported (7), but the new 6 6 41 has been bolstered by the discovery of an 5 specimens are the first that 40 impressive array of previously unknown are directly associated with 39 fossil whales in Pakistan, India, and Egypt, whale skeletons and that are 38 which largely fill the morphological gulf well enough preserved to pro- 37 between land mammals and ocean-dwelling vide important clues to the re- 36 cetaceans (whales, dolphins, and porpoises). lationship between cetaceans 35 The move to the ocean required many and artiodactyls. The special- 34 adaptations to living in water, but the earli- 1 1 ized ankle characters men- 33 est whales still closely resembled land ani- tioned above corroborate a 32 mals. One of the most spectacular transi- Mesonychians Primitive whales Artiodactyls close alliance with artio- 31 tional forms is the “walking whale” Ambu- dactyls, but the new skeletons 30 locetus from the middle Eocene (about 47 also exhibit several primitive 29 to 48 million years ago). This species had placental traits lost in all 28 relatively well-developed limbs, paraxonic 3 known artiodactyls or present 27 feet (where the plane of symmetry passes only in the most primitive 26 between the third and fourth digits), and fossil artiodactyls (see the 25 hooflike terminal toe bones (1). first figure). They thus seem 24 But fossils have failed to provide con- 2 to superimpose artiodactyl 23 clusive indications of the whales’ closest traits on a skeletal anatomy 22 relatives. Instead, they have sparked new that is in some respects more 21 controversy. Most recent morphological 4 primitive than that of any 20 analyses suggest that mesonychians, an ex- known artiodactyl. 4 19 tinct group of terrestrial carnivorous ungu- For example, the forefoot 18 lates, form the sister group of cetaceans (2, in one of the new fossils 17 3). But molecular systematists maintain (Rhodocetus) is mesaxonic 16 that cetaceans belong to the artiodactyls (the plane of symmetry pass- 15 (even-toed ungulates such as sheep, cows, Fossil comparison. Ankle bones of mesonychians, primitive fossil es through the large third dig- 14 pigs, camels, deer, and hippos) and are in whales, and early Eocene artiodactyls; astragali above, calcanei be- it). This is also the case in two 13 fact the sister group of hippopotami (4, 5). low. Diagnostic artiodactyl traits present in early whales include a of the most primitive groups 12 On page 2239 of this issue, Gingerich et trochlea (grooved joint surface) for the navicular bone (1), modified of fossil artiodactyls—the 11 al. (6) report important new fossil evi- shape and orientation of articular surfaces between the astragalus early Eocene artiodactyl Dia- 10 dence—skeletons of two very primitive an- and calcaneus [(2) also present on underside of astragalus, not visi- codexis and some anthra- ble here; see supplemental fig. 3 in (6)], and a narrow calcaneus 9 cient whales with well-developed limbs cotherioids (8, 9)—but almost with an elongate heel process (3) and a large, convex fibular articu- 8 from the middle Eocene of Pakistan—that lation (4). Primitive mesonychid-like traits present in ancient all other artiodactyls (and 7 goes a long way toward resolving the con- whales, but not in any known artiodactyl, include a shallower tibial mesonychians) have paraxon- 6 flict. The fossils provide compelling mor- trochlea with more rounded trochlear ridges (5) and retention of a ic forefeet. In addition, the 5 remnant of the astragalar foramen (6), the opening of a canal new ancient whale fossils re- 4 through which a nerve and vessels pass in primitive mammals. Al- tain a clavicle and a third 3 The author is in the Program for Functional Anatomy trochanter on the femur, ves- and Evolution, Johns Hopkins University School of though mesonychids also have a navicular trochlea, it is much shal- 2 Medicine, Baltimore, MD, 21205, USA. E-mail: lower and offset from the tibial trochlea at a greater angle than in tiges of which are found in

1 [email protected] primitive whales and artiodactyls. Scale bars, 1 cm. only the most primitive CREDIT: E. KASMER

2216 21 SEPTEMBER 2001 VOL 293 SCIENCE www.sciencemag.org S CIENCE’ S C OMPASS

65 known Eocene artiodactyls Whales and their rela- slightly older whales from Pakistan that 64 (8, 10, 11). Loss of these tions. (Top) Current hy- have ankle bones similar to those illustrated 63 features was long consid- pothesis of relationships here, providing further evidence that this 62 ered diagnostic of Artio- based on morphology (3, ankle morphology, shared with artiodactyls, 61 dactyla. Some of them Cetacea Mesonychia Artiodactyla Outgroup 4). (Bottom) Possible rela- was primitive for whales. Two other evolu- 60 could represent character tionships compatible with tionary transitions vital to our understand- 59 reversals or independent fossil and molecular data, ing of the relationship between whales and 58 losses, but taken together based on the new fossil artiodactyls beg for elucidation: the precise 57 they suggest very early di- data (6). A, ankle special- ancestry of hippopotami and the origin of 56 vergence of cetaceans izations of ancient whales artiodactyls themselves. The answers seem 55 from artiodactyls. and artiodactyls; B, other likely to come only from an improved fossil 54 Despite this evidence artiodactyl specializations. record—perhaps from the same region that 53 that cetaceans evolved has yielded fossils showing that whales 52 from artiodactyls, substan- more primitive ankle evolved from artiodactyls. 51 tial discrepancies remain. than any known artio- 50 If cetaceans belong to ar- dactyl, including Dia- References and Notes 49 tiodactyls, then similarities codexis (the oldest artio- 1. J. G. M. Thewissen, S. I. Madar, S. T. Hussain, Cour. Fourschungsinst. Senckenb. 191,1 (1996).

48 in the cranial and dental Artiodactyl Outgroup Hippopotamidae Anthracotherioidea Cetacea Other Artiodactyls dactyl) and anthracothe- 2. M. A. O’Leary, J. H. Geisler, Syst. Biol. 48, 455 (1999). 47 morphologies of mesony- rioids. The latter may be 3. Z. Luo, P. D. Gingerich, Univ. Mich. Pap. Paleontol. 31, 46 chians and cetaceans (2, 3) ancestral to hippopotami 1 (1999). B 4. M. C. Milinkovitch, M. Bérubé, P. J. Palsbøll, in The 45 must be a result of conver- B (13, 14). Emergence of Whales,J.G.M.Thewissen, Ed. 44 gent evolution or must It is thus conceiv- (Plenum, New York, 1998), pp. 113–131. 43 5. J. E. Gatesy, M. C. Milinkovitch, V. G. Waddell, M. J. have been lost in artio- able that hippopotami Stanhope, Syst. Biol. 48, 16 (1999). 42 dactyls. Furthermore, A and cetaceans are the 6. P. D. Gingerich et al., Science 293, 2239 (2001). 41 molecular data favor a sis- only living members of 7. J. G. M.Thewissen, S. I. Madar, Syst. Biol. 48, 21 (1999). 8. J. G. M. Thewissen, S. T. Hussain, Anat. Histol. 40 ter-group relationship between whales and a clade that has been separate from other Embryol. 19, 37 (1990). 39 hippopotami (5). This conflicts with the artiodactyls since before the Eocene (see 9. W. B. Scott, Trans. Am. Philos. Soc. 28, 363 (1940). 38 conventional view based on morphology the second figure). Such a scenario im- 10. K. D. Rose, J. Paleontol. 59, 1203 (1985). 11. J. Erfurt, Hallesches Jahrb. Geowiss. B 12, 57 (2000). 37 that hippopotami are closer to other artio- plies that some advanced artiodactyl fea- 12. M. C. McKenna, S. K. Bell, Classification of Mammals 36 dactyls than they are to whales (12). tures evolved more than once: in the an- Above the Species Level (Columbia Univ. Press, New 35 Can a special affinity between whales thracotherioid-hippopotamid clade (after York, 1997). 13. E. H. Colbert, Am. Mus. Novitates 799,1 (1935). 34 and hippopotami be reconciled with the the cetaceans diverged) and indepen- 14. W. P. Coombs Jr., M. C. Coombs, Neues Jahrb. Geol. 33 fossil record? The existing evidence sug- dently in other artiodactyls. Palaeontol. Monatsh. 1977 584 (1977). 32 gests that cetaceans branched very early We are rapidly filling the gaps in the 15. J. G. M. Thewissen et al., Nature 413, 277 (2001). 16. I thank P. D. Gingerich for providing casts of the new 31 from artiodactyls, emerging from an un- cetacean transition from land to water. Also fossils, E. Kasmer for preparing the first figure, and J. 30 known basal artiodactyl that had a slightly this week, Thewissen et al. (15) report Mussell and S. Zack for helpful discussion. 29 28 PERSPECTIVES: APPLIED PHYSICS 27 The localization of the light waves results 26 in the formation of evanescent waves, 25 Optics in the Nano-World which have an imaginary wave number and 24 decay exponentially in space (in contrast to 23 S.W. Koch and A. Knorr conventional light waves, which propagate 22 freely). The intensity of an evanescent 21 pplications of optical microscopy this field. The authors describe a technique wave thus decays rapidly as the distance 20 are generally limited by the standard that combines the high spatial resolution of from the aperture increases. Therefore, the 19 Aresolution limit set by the wave- NSOM with the high spectral resolution of aperture has to be close to the object, often 18 length of visible light. The invention of coherent nonlinear optical spectroscopy. only a fraction of the wavelength away. 17 near-field scanning optical microscopy Optical measurements at the nanometer This is the regime of near-field optics. 16 (NSOM) first enabled this limit to be over- scale require a light source with an illumi- NSOM techniques have many applica- 15 come, opening up many systems, from nation spot in the nanometer range. For tions in solid state physics, where substan- 14 physics to biology, to investigation by opti- visible-light frequencies, where the wave- tial efforts are made to design electronic 13 cal microscopy. NSOM offered greatly im- length is a few hundred nanometers, con- devices with features on the nanometer 12 proved spatial resolution compared with ventional optical microscopy fails because scale. Electrons can be confined in 11 conventional optical microscopy, and the the resolution is restricted to half the nanometer-scale structures, called quantum 10 use of tunable excitation sources allowed wavelength of the used light (2). To over- dots (4). In these structures, the matter- 9 basic spectroscopic information to be ob- come this problem, the light must be local- wave properties of the electrons are 8 tained. On page 2224 of this issue, Guest et ized in a spot with a diameter much small- changed drastically because the spatial 7 al. (1) report the next major step forward in er than the wavelength of the light. Ideally, confinement of the electrons approaches 6 the spot should have nanometer-scale di- the deBroglie wavelength. Their electronic 5 S. W. Koch is in the Department of Physics, Philipps mensions. This can be done by applying and optical properties therefore differ qual- 4 University, 35032 Marburg, Germany. E-mail: small apertures (3). itatively from those of the bulk material. 3 [email protected] A. Knorr is at The price for this high resolution is that The atomic landscape encountered by the Institute for Theoretical Physics, Technical Univer- 2 sity, 10623 Berlin, Germany. E-mail: andreas.knorr@ the character of the light changes drastical- electrons in a quantum dot can be mapped 1 physik.tu-berlin.de ly when it propagates through the aperture. and analyzed with tunneling spectroscopy