Resolution of the Early Placental Mammal Radiation Using Bayesian Phylogenetics

See allHide authors and affiliations

Science  14 Dec 2001:
Vol. 294, Issue 5550, pp. 2348-2351
DOI: 10.1126/science.1067179


Molecular phylogenetic studies have resolved placental mammals into four major groups, but have not established the full hierarchy of interordinal relationships, including the position of the root. The latter is critical for understanding the early biogeographic history of placentals. We investigated placental phylogeny using Bayesian and maximum-likelihood methods and a 16.4-kilobase molecular data set. Interordinal relationships are almost entirely resolved. The basal split is between Afrotheria and other placentals, at about 103 million years, and may be accounted for by the separation of South America and Africa in the Cretaceous. Crown-group Eutheria may have their most recent common ancestry in the Southern Hemisphere (Gondwana).

Deciphering higher level relationships among mammalian orders is a difficult problem in systematics (1–6) and has important ramifications for evolutionary biology, genomics, and biomedical sciences (7, 8). Studies based on different, multikilobase molecular data sets (5, 6) independently resolved placental mammals into four superordinal groups: Afrotheria, Xenarthra, Laurasiatheria, and Euarchontoglires. However, hierarchical relationships within these groups and at deeper levels in the placental tree remain unclear. A precise resolution of the relationships among the major groups and elucidation of the root of the placental tree are critical for interpreting biogeographic patterns and evolutionary processes involved in the early diversification of placental mammals.

We combined and expanded the large data sets of Madsen et al. (5) and Murphy et al.(6) to yield a 16,397–base pair molecular data set that includes 19 nuclear and 3 mitochondrial gene sequences for 42 placentals (representing all major lineages) and 2 marsupial outgroups (9). The data set is dominated by nuclear exons. Among genes that have been evaluated for resolving deep level mammalian relationships, nuclear exons have more power than mitochondrial genes on a per-residue basis (10). Large molecular data sets have the potential to resolve longstanding controversies in systematics (5, 6, 11, 12), especially when they are analyzed with appropriate models of DNA sequence evolution and with statistically robust estimation procedures that extract the maximum amount of information from molecular sequence data (13).

We used a general-time-reversible + gamma + invariants (GTR + Γ + I) model of sequence evolution (14) and likelihood-based inferential techniques (15), including parametric bootstrap tests (13,16, 17) and Bayesian methods (18, 19) with Markov chain Monte Carlo (MCMC) sampling to assess phylogenetic relationships and examine alternative positions for the root of the placental tree. Given enough data and a correct model of sequence evolution, likelihood methods are statistically consistent and have been shown to be powerful tools for resolving complex phylogenetic problems (13).

Figure 1 shows an evolutionary tree with Bayesian posterior probabilities for individual branches. Three independent MCMC runs, each starting with random trees for each of four simultaneous chains, resulted in concordant joint posterior probability distributions for the topology and the estimated parameters of the model of sequence evolution (15). This result suggests that the chains were run for a sufficient number of generations and sampled the same posterior probability landscape. With the exception of two nodes, the entire placental superordinal tree is resolved with posterior probabilities greater than 0.95. All additional Bayesian analyses that varied taxon and gene sampling (15) resulted in well-resolved trees with high posterior probabilities (15). Furthermore, an identical topology was also obtained with maximum likelihood (ML), strengthening our confidence in the Bayesian results. For some nodes, nonparametric ML bootstrap values were lower than Bayesian posterior probabilities (Fig. 1), consistent with the suggestion of Hillis and Bull (20) that nonparametric bootstrap support may be too conservative.

Figure 1

Phylogeny of living placental mammals reconstructed using a Bayesian phylogenetic approach. An identical topology was obtained with maximum likelihood [–ln L = 211110.54; see (15) for methodological details]. The number above each branch refers to the Bayesian posterior probability (shown as percentages; i.e., 95 represents a posterior probability of 0.95) of the node derived from 26,250 MCMC sampled trees on the basis of the complete 16.4-kb data. Additional analyses with the full data set and with data sets that varied taxon sampling (i.e., jackknifing single outgroup taxa) and character sampling (nuclear only and nuclear coding loci only) produced similarly high posterior probabilities (15). Values below branches represent percent support in maximum likelihood (GTR+ Γ + I) nonparametric bootstrap. An asterisk indicates nodes constrained in the ML nonparametric bootstrap analysis. (A) Bifurcation between Afrotheria and Xenarthra + Boreoeutheria at approximately 103 million years, which corresponds to the vicariant event that separated Africa and South America (Fig. 2B). (B) Branch where dispersal from South America to Laurasia is hypothesized to have occurred (15). Blue, monophyletic Northern Hemisphere group (i.e., Boreoeutheria); red, paraphyletic Southern Hemisphere group (i.e., Xenarthra + Afrotheria); black, outgroups.

Our results firmly place Laurasiatheria and Euarchontoglires as sister taxa that together constitute a clade named Boreoeutheria, with a Northern Hemisphere origin according to the available fossil record (15, 21). Deeper in the placental tree, Xenarthra and Boreoeutheria are sister taxa. The basal split among crown-group placentals is between Afrotheria versus Xenarthra + Boreoeutheria (Fig. 1). Previous molecular studies (5,6) had suggested three most likely positions for the root of the placental tree: (i) the base of Afrotheria, (ii) the base of Xenarthra, or (iii) the branch that separates Xenarthra and Afrotheria from Boreoeutheria. Bayesian results and parametric bootstrap tests reject both the Xenarthra root (posterior probabilities < 0.01; parametric bootstrap P ≤ 0.01) and the root between Xenarthra + Afrotheria and Boreoeutheria (posterior probabilities < 0.04; parametric bootstrap P < 0.01) (Fig. 1) (15).

Within each of the four major clades, relationships that were previously unresolved or controversial are also well resolved. There is now strong support for a basal split between paenungulates versus other afrotherians (aardvark, elephant shrews, and afrosoricidans). Among the latter, afrosoricidans and elephant shrews were well supported as sister taxa. Within Euarchontoglires, our molecular results are the first to render robust support for the monophyly and internal structure of Euarchonta (3). Euarchonta is similar to the morphology-based Archonta hypothesis, but bats are excluded. Inside Laurasiatheria there is now strong support for the basal position of Eulipotyphla and a carnivore + pangolin clade.

The resolution of the placental root and the pattern of basal divergences lead to a plausible biogeographic inference regarding the origin and diversification of this group. Afrotheria and Xenarthra have Gondwanan origins in Africa and South America, respectively. Given their basal positions in the placental tree, the hypothesis that crown-group eutherians have their most recent common ancestry in Gondwana demands consideration. This view is at odds with a prevalent and long-held view that crown-group eutherians have their most recent common ancestry in the Northern Hemisphere (22–24). We tested this hypothesis by estimating molecular divergence dates using the quartet dating (25) and linearized tree methods (26) for specific nodes in the placental tree (Fig. 2A) (15). Our point estimates for the basal split among living placentals range from 101 to 108 million years ago (Mya) (Fig. 2A) (15), in agreement with independent molecular estimates of the split between Afrotheria-Boreoeutheria representatives at 105 Mya (27). It is striking that this date coincides with the vicariant event that separated South America and Africa approximately 100 to 120 Mya (Fig. 2B) (28, 29). We suggest a causal relationship between the sundering of Africa and South America and basal cladogenesis among crown-group eutherians, placing their origin in Gondwana. Subsequently, a trans-hemispheric dispersal event from Gondwana to Laurasia was of fundamental importance in the early history of crown-group eutherians (15). Given the sister-group relationship between Xenarthra and Boreoeutheria, the directionality of this event was likely from South America to Laurasia. Molecular dates for the Xenarthra-Boreoeutheria split (88 to 100 Mya) and for basal divergences within Boreoeutheria (79 to 88 Mya) suggest a window during the Late Cretaceous during which dispersal occurred (Fig. 2A) (15).

Figure 2

Biogeographic scenario for the basal divergence among crown-group placental mammals. (A) Maximum likelihood molecular divergence estimates for the early radiation of placental mammals, estimated with the quartet-dating (QD) and linearized tree (LT) methods (25, 26). Open squares, point estimates based on LT; open circles, median point estimates based on QD; gray bars, range of 95% confidence intervals based on QD. A summary of QD and LT methods and results can be found in supplemental material (15). (B) Final vicariant separation of Africa and South America, approximately 100 to 120 Mya (28, 29), isolates Afrotheria in Africa and the common ancestor of Xenarthra and Boreoeutheria in South America. Reprinted with permission from Cambridge University Press (28).

The suggestion that crown-group placentals have their most recent common ancestry in Gondwana does not imply that stem eutherians also have their origins in Gondwana. Kumar and Hedges (27) and Penny et al. (30) suggested that marsupials and placentals split 173 to 176 Mya on the basis of molecular data. At this time, Gondwana and Laurasia remained connected and stem eutherians may have established a Pangaean distribution before the vicariant separation of Gondwana and Laurasia 160 to 170 Mya (28, 29). Early Cretaceous eutherians in the Northern Hemisphere, such as Prokennalestes(31), may be representatives of an extinct evolutionary radiation in Laurasia that predates Boreoeutheria.

Foote et al. (32) argued that Cretaceous molecular dates for the early evolutionary history of crown-group placentals are incongruent with the fossil record. One explanation for the discrepancy between molecular dates and paleontological data is the “Garden of Eden” hypothesis (32), which postulates early placental diversification in regions with a poorly known fossil record. Notably, the Cretaceous fossil record for most of Gondwana fits this description (32). A second explanation is the phylogenetic placement of fossil taxa (32). Foote et al.(32) assumed that Cretaceous eutherians were stem taxa rather than crown-group eutherians. In contrast, the cladistic analysis of Archibald et al. (33) suggests that 85- to 90-million-year-old zalambdalestids and zhelestids, both from Laurasia, are members of crown-group Eutheria and have affinities with Glires and archaic ungulates, respectively. In the context of our molecular results, zalambdalestids may be early representatives of Euarchontoglires, and zhelestids may be an early branch within Laurasiatheria. Reliable phylogenetic placements for these and other Cretaceous taxa are critical for evaluating whether the available fossil record is compatible with Cretaceous molecular dates. The well-resolved phylogeny that we have established for living orders of placental mammals provides a molecular scaffold (34) that may be used in cladistic analyses of morphological characters to examine inter-relationships among living and fossil forms. Understanding the phylogenetic and biogeographic patterns connecting extinct and living mammalian lineages remains one of the major challenges ahead.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: mark.springer{at} (M.S.S.) or obrien{at} (S.J.O.)


View Abstract

Navigate This Article