Résumé
Reconstructing the chronology of mammalian evolution is a debated issue
between molecule- and fossil-based inferences. A methodological limitation
of molecules is the evolutionary rate variation among lineages, precluding
the application of the global molecular clock. We considered 2422 first
and second codon positions of the combined ADRA2B, IRBP, and vWF nuclear
genes for a well-documented set of placentals including an extensive
sampling of rodents. Using seven independent calibration points and a
maximum-likelihood framework, we evaluated whether molecular and
paleontological estimates of mammalian divergence dates may be reconciled
by the local molecular clocks approach, allowing local constancy of
substitution rates with variations at larger phylogenetic scales. To
handle the difficulty of choosing among all possible rate assignments for
various lineages, local molecular clocks were based on the results of
branch-length and two-cluster tests. Extensive lineage-specific variation
of evolutionary rates was detected, even among rodents. Cross-calibrations
indicated some incompatibilities between divergence dates based on
different paleontological references. To decrease the impact of a single
calibration point, estimates derived from independent calibrations
displaying only slight reciprocal incompatibility were averaged. The
divergence dates inferred for the split between mice and rats
(approximately 13-19 Myr) was younger than previously published molecular
estimates. The most recent common ancestors of rodents, primates and
rodents, boreoeutherians, and placentals were estimated to be,
respectively, approximately 60, 70, 75, and 78 Myr old. Global clocks,
local clocks, and quartet dating analyses suggested a Late Cretaceous
origin of the crown placental clades followed by a Tertiary radiation of
some placental orders like rodents.