Using fossils to break long branches in molecular dating: a comparison of relaxed clocks applied to the origin of angiosperms.

Long branches are potentially problematic in molecular dating because they can encompass a vast number of combinations of substitution rate and time. A long branch is suspected to have biased molecular clock estimates of the age of flowering plants (angiosperms) to be much older than their earliest fossils. This study explores the effect of the long branch subtending angiosperms in molecular dating and how different relaxed clocks react to it. Fossil angiosperm relatives, identified through a combined morphological and molecular phylogenetic analysis for living and fossil seed plants, were used to break the long angiosperm stem branch. Nucleotide sequences of angiosperm fossil relatives were simulated using a phylogeny and model parameters from living taxa and incorporated in molecular dating. Three relaxed clocks, which implement among-lineage rate heterogeneity differently, were used: penalized likelihood (using 2 different rate smoothing optimization criteria), a Bayesian rate-autocorrelated method, and a Bayesian uncorrelated method. Different clocks provided highly correlated ages across the tree. Breaking the angiosperm stem branch did not result in major age differences, except for a few sensitive nodes. Breaking the angiosperm stem branch resulted in a substantially younger age for crown angiosperms only with 1 of the 4 methods, but, nevertheless, the obtained age is considerably older than the oldest angiosperm fossils. The origin of crown angiosperms is estimated between the Upper Triassic and the early Permian. The difficulty in estimating crown angiosperm age probably lies in a combination of intrinsic and extrinsic complicating factors, including substantial molecular rate heterogeneity among lineages and through time. A more adequate molecular dating approach might combine moderate background rate heterogeneity with large changes in rate at particular points in the tree.