Analyses of evolutionary dynamics in viruses are hindered by a time-dependent bias in rate estimates

Abstract

Time-scales of viral evolution and emergence have been studied widely, but are often poorly understood. Molecular analyses of viral evolutionary time-scales generally rely on estimates of rates of nucleotide substitution, which vary by several orders of magnitude depending on the timeframe of measurement. We analysed data from all major groups of viruses and found a strong negative relationship between estimates of nucleotide substitution rate and evolutionary timescale. Strikingly, this relationship was upheld both within and among diverse groups of viruses. A detailed case study of primate lentiviruses revealed that the combined effects of sequence saturation and purifying selection can explain this time-dependent pattern of rate variation. Therefore, our analyses show that studies of evolutionary time-scales in viruses require a reconsideration of substitution rates as a dynamic, rather than as a static, feature of molecular evolution. Improved modelling of viral evolutionary rates has the potential to change our understanding of virus origins.

Keywords: evolutionary rates; molecular clock; phylogenetics; virus evolution.

Figure 1.

(a) Meta-analysis of temporal rate variation among viral lineages. Each data point corresponds to a nucleotide rate estimate as reported in this and other studies. The original data are in nucleotide substitutions per site per year and have been log-transformed in this figure. Colours distinguish DNA (red) and all RNA viruses (black), which were analysed separately. (b) Temporal variation in ENV (blue) and POL (green) genes in Lentivirus. Data points are mean rate estimates for different time-scales of primate Lentivirus evolution.

Figure 2.

Temporal variation of (a) selective constraints as the ratio of non-synonymous and synonymous substitutions per site (d_N/d_S); (b) separate estimation of non-synonymous (d_N) and synonymous variation (d_S); (c) saturation according to the ratio of observed saturation (I) and full saturation (I_c) and (d) model adequacy, as the Z-score of expected and observed model performance. Note that the x-axis displays time on a logarithmic scale for comparison with figure 1; however, the regression lines correspond to the non-transformed data. The colours correspond to analyses of genes ENV and POL. The symbols represent d_N and d_S in (b), and first and second, and third codon positions in (c), as indicated in the legends. (Online version in colour.)