Nonadaptive Amino Acid Convergence Rates Decrease over Time.

Convergence is a central concept in evolutionary studies because it provides strong evidence for adaptation. It also provides information about the nature of the fitness landscape and the repeatability of evolution, and can mislead phylogenetic inference. To understand the role of adaptive convergence, we need to understand the patterns of nonadaptive convergence. Here, we consider the relationship between nonadaptive convergence and divergence in mitochondrial and model proteins. Surprisingly, nonadaptive convergence is much more common than expected in closely related organisms, falling off as organisms diverge. The extent of the convergent drop-off in mitochondrial proteins is well predicted by epistatic or coevolutionary effects in our "evolutionary Stokes shift" models and poorly predicted by conventional evolutionary models. Convergence probabilities decrease dramatically if the ancestral amino acids of branches being compared have diverged, but also drop slowly over evolutionary time even if the ancestral amino acids have not substituted. Convergence probabilities drop-off rapidly for quickly evolving sites, but much more slowly for slowly evolving sites. Furthermore, once sites have diverged their convergence probabilities are extremely low and indistinguishable from convergence levels at randomized sites. These results indicate that we cannot assume that excessive convergence early on is necessarily adaptive. This new understanding should help us to better discriminate adaptive from nonadaptive convergence and develop more relevant evolutionary models with improved validity for phylogenetic inference.

Analysis of host-parasite incongruence in papillomavirus evolution using importance sampling.

The papillomaviruses (PVs) are a family of viruses infecting several mammalian and nonmammalian species that cause cervical cancer in humans. The evolutionary history of the PVs as it associated with a wide range of host species is not well understood. Incongruities between the phylogenetic trees of various viral genes as well as between these genes and the host phylogenies suggest historical viral recombination as well as violations of strict virus-host cospeciation. The extent of recombination events among PVs is uncertain, however, and there is little evidence to support a theory of PV spread via recent host transfers. We have investigated incongruence between PV genes and hence, the possibility of recombination, using Bayesian phylogenetic methods. We find significant evidence for phylogenetic incongruence among the six PV genes E1, E2, E6, E7, L1, and L2, indicating substantial recombination. Analysis of E1 and L1 phylogenies suggests ancestral recombination events. We also describe a new method for examining alternative host-parasite association mechanisms by applying importance sampling to Bayesian divergence time estimation. This new approach is not restricted by a fixed viral tree topology or knowledge of viral divergence times, multiple parasite taxa per host may be included, and it can distinguish between prior divergence of the virus before host speciation and host transfer of the virus following speciation. Using this method, we find prior divergence of PV lineages associated with the ancestral mammalian host resulting in at least 6 PV lineages prior to speciation of this host. These PV lineages have then followed paths of prior divergence and cospeciation to eventually become associated with the extant host species. Only one significant instance of host transfer is supported, the transfer of the ancestral L1 gene between a Primate and Hystricognathi host based on the divergence times between the upsilon human type 41 and porcupine PVs.