Reversions to consensus are positively selected in HIV-1 and bias substitution rate estimates

Abstract

AbstractHIV-1 is a rapidly evolving virus able to evade host immunity through rapid adaptation during chronic infection. The HIV-1 group M has diversified since its zoonosis into several subtypes at a rate of the order of 10−3 changes per site per year. This rate varies between different parts of the genome and its inference is sensitive to the time scale and diversity spanned by the sequence data used. Higher rates are estimated on short time scales and particularly for within-host evolution, while rate estimates spanning decades or the entire HIV-1 pandemic tend to be lower. The underlying causes of this difference are not well understood.We investigate here the role of rapid reversions toward a preferred evolutionary sequence state on multiple time scales. We show that within-host reversion mutations are under positive selection and contribute substantially to sequence turnover, especially at conserved sites. We then use the rates of reversions and non-reversions estimated from longitudinal within-host data to parametrize a phylogenetic sequence evolution model. Sequence simulation of this model on HIV-1 phylogenies reproduces diversity and apparent evolutionary rates of HIV-1 in gag and pol, suggesting that a tendency to rapidly revert to a consensus-like state can explain much of the time dependence of evolutionary rate estimates in HIV-1.