摘要

Background: Patterns of change in ecomorphological traits have traditionally been studied using data from the fossil record. Recent advances in molecular phylogenetics created new opportunities for inferring ancestral character states and estimating the modes and rates of trait evolution from phylogenetic hypotheses of extant organisms. However, without fossil taxa useful information is potentially discarded and, in the worst case, results from extant taxa only may be misleading, in particular if extinction rates have been high and directional selection has acted.
Question: How does the integration of fossil information affect our understanding of macro-evolutionary dynamics?
Organisms: Extant species of the superfamily Musteloidea (Carnivora - weasels and allies); extinct lineages of this clade (c. 30 to 2 Ma) sampled predominantly throughout the northern hemisphere.
Experiments: We focus on the evolutionary dynamics of carnivoran ecometric traits (three index ratios associated with locomotor habit and posture calculated from the osteological measurements), and we highlight the impact of using extant-only phylogenies versus integrated analyses when evaluating modes and rates of trait evolution.
Methods: We integrated extinct taxa into a molecular phylogeny of the extant species based on taxonomic knowledge (association with a particular family/subfamily) and sampling from the estimated times of speciation and extinction of each fossil lineage. We repeated the procedure for each of 500 trees from the posterior distribution of a Bayesian phylogenetic analysis. We compared the fit of different macroevolutionary models (Brownian motion, Ornstein-Uhlenbeck, accelerating/decelerating evolution, and Brownian motion with a directional trend) for all trees (with and without fossils), using AIC regression.
Results: The integration of fossil data into the analyses of trait evolution significantly affected model selection, evolutionary rates, as well as estimated trait values at the root of the phylogeny. In the case of the metatarsal III-to-femur ratio, the integrated analyses provided strong support for a Trend model, whereas the different macroevolutionary models could not be distinguished with strong statistical support when only extant taxa were considered. Furthermore, our analyses indicated substantial variation of evolutionary rates across the phylogeny, with high rates of evolution of locomotor traits along the backbone of the Mustelidae tree and within the Lutrinae clade in particular, reconciling complex evolutionary dynamics with more realistic root node estimates than obtained under the Trend model. Thus, the integration of fossils and molecular data has implications not only for our interpretation of evolutionary dynamics, but also for ancestral character reconstructions. Importantly, even in cases where extant-only and integrated analyses are in agreement, fossils remain essential because they provide additional support for inferred patterns of trait evolution.