Relationships among orders of placental mammals have proved difficult to resolve1. To extend the available mitochondrial (mt) sequences, a 2.6-kilobase (kb) segment containing the 12S rRNA, valine transfer RNA, and 16S rRNA genes was sequenced for nine taxa to generate a data set that is representative of 12 of the 18 placental orders and all three insectivore suborders4. Phylogenetic analyses provide strong support for well-established mammalian clades such as carnivores, hominoids, and Cetacea plus Artiodactyla (Fig. 1a). In agreement with other molecular studies7,8,9,10 that included an assortment of taxa, most interordinal associations are not resolved at bootstrap values >75%. However, the mtDNA data do provide strong support for the association of the two paenungulates (hyrax, manatee) together, and of these with elephant shrews, aardvarks and golden moles (Fig. 1a and Table 1). The association of hyraxes with proboscideans and sirenians was suggested by Cope11. A competing hypothesis is an association of hyraxes with perissodactyls12. Our results agree with earlier protein13,14 and DNA studies7,8,9,10 supporting Cope's paenungulate hypothesis. In addition to bootstrap support, T-PTP15 and Kishino–Hasegawa16 tests also support paenungulate monophyly (Table 2). Anatomical data provide some evidence that aardvarks and/or elephant shrews may be related to paenungulates17,18 but suggest other hypotheses as well: for example, six osteological features are putative synapomorphies uniting elephant shrews with lagomorphs and rodents19. All the available sequence data, including amino-acid sequences13,14, DNA sequences for three nuclear genes8,9,10, and the present mitochondrial genes, support an association of aardvarks and elephant shrews with paenungulates. What is most unexpected is that golden moles, a family of insectivores, are also part of this clade. 12S rRNA sequences earlier suggested an association of golden moles with paenungulates, but did not provide convincing bootstrap support for this hypothesis7. Our expanded data set demonstrates that insectivores are not monophyletic (Table 2) and that golden moles, elephant shrews, aardvarks and paenungulates are part of the same clade.

Figure 1: Majority-rule parsimony bootstrap trees based on mitochondrial (a), vWF (b), and A2AB (c) sequences.
figure 1

Bootstrap values and decay indices, respectively, are above and below branches. The minimum-length tree (3,836 steps; consistency index, 0.430) for the mitochondrial data is 4, 7, 43 and 66 steps shorter than trees that constrain shrew + hedgehog, rodent monophyly, insectivore monophyly, and hyrax + horse, respectively. Insectivore monophyly and a hyrax-perissodactyl association require 29 and 64 additional steps, respectively, in comparison to five minimum-length vWF trees (2,401 steps; consistency index, 0.511). Insectivore monophyly requires 26 additional steps in comparison to the shortest A2AB tree (1,469 steps; consistency index, 0.730).

Table 1 Bootstrap support for select clades based on different methods
Table 2 Significance levels of T-PTP and Kishino–Hasegawa tests

To corroborate these findings, we obtained sequences from exon 28 of the von Willebrand factor (vWF) gene for golden mole, hedgehog and pangolin. Adding these to the existing vWF data set9, we found high bootstrap support for the inclusion of golden moles with paenungulates, elephant shrews and aardvarks (Fig. 1b and Table 1). Sequences from the α-2B adrenergic receptor gene (A2AB) also support the association of golden moles with paenungulates, elephant shrews and aardvarks (Fig. 1c and Table 1). Parsimony and maximum-likelihood trees supporting the paenungulate–golden mole–aardvark–elephant shrew clade are significantly better than the best trees that constrain insectivore monophyly (Table 2).

This expanded clade, which includes five placental orders plus golden moles, has not been previously hypothesized on the basis of morphological or molecular data. Elephants, sirenians, hyraxes, golden moles, aardvarks and elephant shrews show a variety of ecological and morphological specializations and it is not surprising that morphology has not elucidated their common ancestry, now evident from DNA sequences. It is notable that all six of these groups are of probably African origin or, in the case of the aquatic sirenians, from along the margins of the former Tethys Sea5,6,7,8,9,10,11,12,13. In two cases (golden moles and elephant shrews), geographic distribution has been restricted to Africa for the complete temporal range of these taxa5. Thus geographic evidence adds to the molecular data in support of this ‘African origin’ clade. The radiation of the African clade parallels endemic radiations of other vertebrate taxa on Southern Hemisphere continents during the breakup of Gondwanaland; for example, marsupials and passerine birds in Australia20, and marsupials, edentates and notoungulates in South America5.

Paenungulate orders diverged from each other 51 to 59 million years (Myr) ago, as deduced from 12S rRNA transversions (Table 3). Deeper in the African clade, average divergence times between paenungulates and other lineages range from 67 to 80 Myr. The mean divergence time between taxa in the African clade and the other 13 orders of placental mammals is 91 Myr. These divergence times support the hypothesis that many eutherian orders arose before the extinction of dinosaurs at the end of the Cretaceous21 and imply that conventional views on the origins of the African mammal fauna5 are incorrect. Africa's window of isolation extended from the Late Cretaceous, when Africa became separated from South America, to the early Cenozoic, when tenuous connections developed between northern Africa and Europe. The window of isolation extended from at least 80 Myr (ref. 20), if not earlier, until the early Cenozoic. The traditional view is that condylarths, prosimian primates and creodont carnivores reached Africa from the north after the docking of Africa with Europe5. From the condylarth stock, groups such as proboscideans and sirenians ostensibly originated in Africa. Other elements of the African mammal fauna, including perissodactyls, artiodactyls, insectivores and living carnivore families, presumably arrived in the Neogene with the establishment of the Arabian Peninsula. Evidence for an extensive African clade, including taxa with divergence times as old as 80 Myr, is inconsistent with this view. The ancestor of the African clade probably resided in Africa before the window of isolation and did not arrive from the north in the early Cenozoic. The role of geographic isolation and continental break-up in the early diversification of placental mammals is potentially more important than previously recognized.

Table 3 Divergence times (Myr) based on 12S rRNA transversions


Amplification and sequencing. 12S rRNA and tRNA genes were amplified and sequenced as described22. 16S rRNA genes were amplified using primers for valine-tRNA (for example, 5′-tacaccyaraagatttca-3′) and leucine-tRNA (for example, 5′-agaggrtttgaacctctg-3′) and sequenced. Accession numbers for the new mitochondrial sequences (Echymipera kalubu (bandicoot); Dromiciops gliroides (monito del monte); Sorex palustris (shrew); Manis sp. (pangolin); Amblysomus hottentotus (golden mole); Procavia capensis (hyrax); Trichechus manatus (manatee); Orcyteropus afer (aardvark); Elephantulus rufescens (elephant shrew)) are U97335–U97343. 12S rRNA sequences for several of these taxa have been deposited in GenBank (M95108 (golden mole), U61073 (monito del monte), U61079 (pangolin), U61083 (manatee), U61084 (hyrax)). Accession numbers for additional mitochondrial sequences are as follows: cow (J01394); blue whale (X72204); fin whale (X61145); horse (X79547); cat (U20753) harbour seal (X63726); grey seal (X72004); human (J01415); gorilla (D38114); orang-utan (D38115); guinea-pig (L35585); hedgehog (X88898); rat (X14848); mouse (J01420); opossum (Z29573); platypus (U33498; X83427). Exon 28 of the vWF gene was amplified and sequenced as described9. Accession numbers for Manis sp., Erinaceus europaeus (hedgehog), and Amblysomus hottentotus vWF sequences are U97534–U97536. Additional vWF sequences are from ref. 9. Part of the single-copy, intronless A2AB gene was amplified using the primers A2ABFOR (5′-asccctactcngtgcaggcnacng-3′) and A2ABREV (5′-ctgttgcagtagccdatccaraaraaraaytg-3′). PCR products were cloned into a T/A cloning vector (Promega) and both strands were sequenced for at least two clones using the Thermo Sequenase fluorescent-labelled primer cycle sequencing kit (Amersham). Accession numbers for the new A2AB sequences (Elephas maximus (elephant); Orycteropus afer (aardvark); Macroscelides proboscideus (elephant shrew); Amblysomus hottentotus (golden mole); Procavia capensis (hyrax); Erinaceus europaeus (hedgehog); Talpa europaea (mole)) are Y12520–Y12526. Additional α-2 adrenergic sequences are M34041 (human); M32061 (rat); (L00974) (mouse), and U25722–U25724 (guinea-pig).

Sequence alignment and phylogenetic analysis. Sequences were aligned using CLUSTAL W (ref. 23). rRNA alignments were modified in view of secondary structure24,25. Ambiguous regions were omitted from subsequent analyses26; this resulted in 2,152, 1,261 and 1,132 nucleotide positions, respectively, for the mt, vWF and A2AB genes. The mt, vWF and A2AB data sets contain 810, 497 and 393 informative sites, respectively. Phylogenetic analyses (parsimony, minimum evolution with Tamura–Nei27 and Logdet28 distances, maximum likelihood under the HKY85 (ref. 29) model) were conducted with PAUP 4.0d52-54, written by D. L. Swofford. The mitochondrial tree was rooted using platypus and marsupial sequences. The vWF tree was rooted with the sloth7; alternatively, rooting with either hedgehog or rodents supports the ‘African’ clade and contradicts insectivore monophyly. For the A2AB tree, sequences with the suffix B are from the A2AB subfamily. GuineaPigA and GuineaPigC sequences are from other subfamilies in the α-2 adrenergic receptor family and were used as outgroups. Bootstrap analyses used full heuristic searches with 500 replications for parsimony and minimum evolution and 100 replications for maximum likelihood. Shape parameters for the gamma distribution were estimated from minimum length trees26 and were 0.32 (mtDNA), 0.59 (vWF) and 0.52 (A2AB).

Divergence times. 12S rRNA transversions accumulated linearly as far back as the eutherian–metatherian split24. Nine independent cladogenic events were selected based on 12S rRNA sequence availability and paleostratigraphic data10,24,30 (for example, Rattus to Mus (14 Myr); Sus to Tayassu (45 Myr); ruminants to Cetacea (60 Myr); Erinaceus to Metatheria (130 Myr)). Relative rates were calculated in reference to xenarthrans. Tamura–Nei transversion distances (transversions only) were adjusted for relative rate differences30 against the xenarthran standard. Rate-adjusted estimates of sequence divergence were regressed against paleostratigraphic divergence estimates for each of the nine calibration points (origin forced through zero; r2= 0.97; P = 0.0000002). The resulting equation (divergence time (in Myr) = sequence divergence/0.00063) was used to estimate interordinal divergence times after making similar adjustments for relative rates. Additional details will be presented elsewhere (M.S., manuscript in preparation).