39J. Paleont., 78(1), 2004, pp. 39–44Copyright q 2004, The Paleontological Society0022-3360/04/0078-39$03.00MOLECULAR CLOCK DIVERGENCE ESTIMATES AND THE FOSSIL RECORDOF CETARTIODACTYLAJESSICA M. THEODORDepartment of Geology, Illinois State Museum, Springfield 62703, ,[email protected]—Molecular clock estimates of divergence times for artiodactyls and whales vary widely in their agreement with the fossilrecord. Recent estimates indicate that the divergence of whales from artiodactyls occurred 60 Ma, a date which compares well with thefirst appearances of fossil whales around 53.5 Ma, and artiodactyls at 55 Ma. Other estimates imply significant gaps in the fossil record.A date of 65 Ma for the divergence of Suidae and Ruminantia predates the appearance of Ruminantia by over 10 million years, andan estimate of 58 Ma for the divergence of Suidae from Cetacea implies a gap of over 20 million years. Further, although a molecularclock estimate has not been reported, the hypothesis that hippos are the closest living relatives of the whales implies a potential ghostlineage for hippos of over 40 million years. There are only two living species of hippos, and their fossil record is sparse, while cetaceansand other artiodactyls are speciose and have rich fossil records. A 40-million-year gap in the fossil record of hippos could be explainedby several possibilities: inadequate biogeographic sampling, taphonomic biases, or undifferentiated primitive morphology. Similarly, anumber of possible problems may exist in the molecular data: rate variation in the genes sampled, the low numbers of genes examined,and insufficient age calibrations. In addition, there are potential problems in molecular phylogeny estimation, such as long branchattraction and inappropriate taxonomic sampling. Additional estimates of divergence times among living taxa should provide a broaderframework for comparison with the fossil record and provide information to help identify which of these factors are causing conflict.INTRODUCTIONRECENT MOLECULARphylogenetic work on the relationships ofplacental mammals show strong support for close relation-ship between whales (Cetacea) and artiodactyls, with numerousstudies nesting Cetacea deep within a paraphyletic Artiodactyla,in most cases sister taxon to the Hippopotamidae (Gatesy, 1997,1998; Gatesy et al., 1996, 1999; Graur and Higgins, 1994; Mil-inkovitch et al., 1998; M. A. Nikaido et al., 1999). New fossildiscoveries of postcranial material of early whales show theyshare what had been thought to be a key synapomorphy of ar-tiodactyls, the double-trochleated astragalus (Gingerich et al.,2001; Thewissen et al., 2001). Although these finds provide ev-idence supporting the close relationship of whales and artiodac-tyls, there is as yet no fossil evidence that supports a sister-taxonrelationship between the whales and the family Hippopotamidae.However, if correct, this hypothesis has radical implications forthe fossil record of cetaceans and artiodactyls and the history ofdivergences within the Cetartiodactyla, because it implies thatthere are several ghost lineages of considerable length, eitheryet to be found in fossil deposits, or lurking in museum drawersand undetectable by morphological methods. In order to evaluatethat possibility, molecular clock estimates of divergence timescould point to lineages in need of further study, but the reli-ability of such estimates for Cetartiodactyla is unknown. There-fore, molecular clock estimates for more nodes within Cetar-tiodactyla must be calculated and compared with the known fos-sil record.The fossil record for Artiodactyla and Cetacea accords wellwith published molecular clock estimates for the divergence ofthe two groups: the oldest known whale, Himalayacetus, is earlyEocene in age, 53.5 Ma (Bajpai and Gingerich, 1998), and theoldest artiodactyl, Diacodexis, known from the earliest Eocene,55 Ma (Gingerich, 1989), are both slightly younger than themolecular clock estimate of 60 Ma (Arnason and Gullberg,1996). The most recent molecular clock estimate of the diver-gence of odontocete and mysticete whales around 34–35 Maalso agrees well with the fossil record (M. Nikaido et al., 2001).However, estimates of divergence times for other subgroups arescarce, and some estimates of suid divergence from ruminantspredate the oldest records of either order (Kumar and Hedges,1998). The hippopotamid/cetacean relationship is especially dif-ficult to reconcile, given that whales appear in the early Eoceneand the oldest hippopotamid fossil is known from the mid-Mio-cene (Behrensmeyer et al., 2002), 15.6–15.8 Ma, a gap of 40million years. This gap is somewhat lessened, to about 10 mil-lion years, if the fossil anthracotheres are closely related to thehippopotamids (Colbert, 1935), as they have a long fossil recordgoing back to the middle Eocene. However, the morphologicalevidence linking anthracotheres and hippos is not strong, and ithas been suggested that hippopotamids may be derived fromother lineages, including the tayassuids (Pickford, 1983) and thecebochoerids (Pearson, 1927). Such a relationship would meanthat whales diverged genetically from hippopotamids or theirclose relatives before the first appearance of whale fossils, andhence over 53 million years ago.This study was undertaken to examine additional molecularclock estimates of divergence times within the Artiodactyla andCetacea, and compare those divergence times with the fossil re-cord, to better understand the causes of conflict between the mo-lecular data and the fossil record.METHODSThe data used include the cytochrome b (cytb) mitochondrialgene sequence data (1,143 base pairs, 63 taxa) and the kappacasein (kcas) nuclear gene sequence data (679 base pairs, 33 taxa)from the WHIPPO-2 data matrix (Gatesy et al., 1999).Each data set was tested for clocklike behavior using relativerate tests (Fitch, 1976; Sarich and Wilson, 1967; Tajima, 1993)and likelihood ratio tests (Felsenstein, 1981; Huelsenbeck andCrandall, 1997; Huelsenbeck and Rannala, 1997; D. L. Swoffordet al., 1996). The substitution models selected for the likelihoodratio tests were selected using hierarchical likelihood ratio teststo select an appropriate model, as implemented in Modeltest 3.06(Posada and Crandall, 1998). The likelihood ratio tests were per-formed on three trees, using PAUP*: the best likelihood tree cal-culated using the selected substitution model, the pruned shortesttopology and
View Full Document
Unlocking...