Protungulatum, Confirmed Cretaceous Occurrence of an Otherwise Paleocene Eutherian (Placental?) Mammal

Abstract

Neither pre-Cenozoic crown eutherian mammals (placentals) nor archaic ungulates (“condylarths”) are known with certainty based on the fossil record. Herein we report a new species of the Paleocene archaic ungulate (“condylarth”) Protungulatum from undisputed Late Cretaceous aged rocks in Montana USA based on an isolated last upper premolar, indicating rare representatives of these common early Tertiary mammals appeared in North America a minimum of 300 k  years before the extinction of non-avian dinosaurs. The other 1200 mammal specimens from the locality are characteristic Late Cretaceous taxa. This discovery overturns the current hypothesis that archaic ungulates did not appear in North America until after the Cretaceous/Tertiary (K/T) boundary and also suggests that other reports of North American Late Cretaceous archaic ungulates may be correct. Recent studies, including ours, cannot determine whether Protungulatum does or does not belong to the crown clade Placentalia.

Appendix

A reanalysis of the Nexus formatted data matrix compiled by Wible et al. (2009), using the program Mr.BAYES accessed through the Cyberinfrastructure for Phylogenetic Research (CIPRES) portal 2.0 website, was preformed to determine relationships and quantify support values for alternate topologies of the crown group Placentalia, specifically to examine the placental affinities of the latest Cretaceous and/or early Paleocene taxa Protungulatum, Oxyprimus, and Purgatorius. A constrained treesearch was used to force the taxa Protungulatum and Oxyprimus into a clade with the placental ungulates Gujaratia, Hyopsodus, Meniscotherium, and Phenacodus; while the genus Purgatorius was forced into a clade with the primatomorph taxa Plesiadapis, Northarctus, and Adapis. A second treesearch was run completely unconstrained, and the likelihood values of these topologies were compared using a Bayes factors test (Kass and Raftery 1995). All treesearches used the same matrix of 69 taxa and 408 unordered characters and were run for ten million generations with sampling every 1000 generations, generating a stable distribution of parameter estimates. Likelihood values were saved, after a 10% burn in, and used an Mk model of character change with alpha values (rates of evolution) estimated by a Gamma distribution binned into four categories (Lewis 2001; Prieto-Marquez 2010). This distribution of alpha values was chosen over one assuming strictly equal rates of evolution after both types of models were run using the constrained and unconstrained data described above; in both cases the Mk+G models generated significantly higher likelihood values compared to their constant rate counterparts.

The harmonic means of the log likelihood values for both the constrained and unconstrained analyses are, respectively:

• Constrained + Gamma harmonic mean= −8520.87

• Unconstrained + Gamma harmonic mean= −8499.48

The large difference between these two values when compared with a Bayes factors test (difference = 42.78) suggests that the unconstrained analysis is truly a better explanation of the data given the model parameters, and that the taxa Purgatorius, Protungulatum, and Oxyprimus cannot confidently be placed inside the crown group Placentalia.