Pattern and timing of diversification of the mammalian order Carnivora inferred from multiple nuclear gene sequences

Abstract

The mammalian order Carnivora has attracted the attention of scientists of various disciplines for decades, leading to intense interest in defining its supra-familial relationships. In the last few years, major changes to the topological structure of the carnivoran tree have been proposed and supported by various molecular data sets, radically changing the traditional view of family composition in this order. Although a sequence of molecular studies have established a growing consensus with respect to most inter-familial relationships, no analysis so far has included all carnivoran lineages (both feliform and caniform) in an integrated data set, so as to determine comparative patterns of diversification. Moreover, no study conducted thus far has estimated divergence dates among all carnivoran families, which is an important requirement in the attempt to understand the patterns and tempo of diversification in this group. In this study, we have investigated the phylogenetic relationships among carnivoran families, and performed molecular dating analyses of the inferred nodes. We assembled a molecular supermatrix containing 14 genes (7765 bp), most of which have not been previously used in supra-familial carnivoran phylogenetics, for 50 different genera representing all carnivoran families. Analysis of this data set led to consistent and robust resolution of all supra-familial nodes in the carnivoran tree, and allowed the construction of a molecular timescale for the evolution of this mammalian order.

Figure 1.

Maximum likelihood phylogram depicting the evolutionary relationships of major extant lineages of the order Carnivora. Nodes are numbered sequentially for cross-reference with Tables 2 – 6. Numbers in bold, underlined types indicate nodes supported by >90% bootstrap values (or posterior probabilities) for all likelihood-based phylogenetic methods (see Table 4 for detailed results). Hatched arrowheads indicate the phylogenetic position of the 21 fossil constraints used in our final divergence dating analysis with the program Divtime (see Table 2 for details): right-pointing arrowheads are maximum ages for the subsequent node, while left-point arrowheads are minimum ages for the previous node. Black circles on nodes indicate the eight calibrations used in the BEAST divergence dating analysis that presented the best posterior probability (Run 2 – see text for details).

Figure 2.

Timescale of carnivoran diversification, based on the Thorne/Kishino relaxed molecular clock method, incorporating 21 fossil calibrations (similar results were obtained with BEAST - see Figure 3, Table 6 and Supplementary Information). The tree is a cladogram with node depths (branch lengths) drawn proportional to time, with a timescale shown at the bottom, including the paleontological epochs of the Cenozoic Era (Paleoc.=Paleocene; Pi=Pliocene; Pe=Pleistocene). Family names are indicated on the right, as are the major suborders Feliformia and Caniformia. Numbers above nodes are divergence ages in MYA (millions of years ago); only nodes defining families or higher clades are labeled (the only exception is the New World sub-clade of the Mephitidae, mentioned in the main text); see Table 6 for point estimates and credibility intervals for node ages estimated with Divtime and BEAST.

Figure 3.

A) Comparison of point estimates for the age of 27 carnivoran familial and supra-familial nodes obtained from two different relaxed molecular clock methods, implemented in the programs Divtime and BEAST, respectively. The Divtime analysis assumed a single model for the full concatenated data set, whereas BEAST allowed each of the 14 genes to have independent substitution and molecular clock models. The Divtime approach incorporated 21 different fossil constraints (minimum or maximum) applied to nodes throughout the Carnivora tree, while the BEAST run used calibrations for eight nodes (see Figure 1 and text for details). B) Estimated age for the crown-group of each terrestrial carnivoran family, inferred with the Divtime and BEAST approaches. Pinniped dates are not shown here as these dates do not seem to have been as robustly estimated (given the observed discrepancy between the two methods - see Table 6), along with the fact that the base of Otariidae was not sampled (see text for details). Only point estimates are shown; see Table 6 for Credibility Intervals.