Quantifying differences in the tempo of human immunodeficiency virus type 1 subtype evolution

Abstract

Human immunodeficiency virus type 1 (HIV-1) genetic diversity, due to its high evolutionary rate, has long been identified as a main cause of problems in the development of an efficient HIV-1 vaccine. However, little is known about differences in evolutionary rate between different subtypes. In this study, we collected representative samples of the main epidemic subtypes and circulating recombinant forms (CRFs), namely, sub-subtype A1, subtypes B, C, D, and G, and CRFs 01_AE and 02_AG. We analyzed separate data sets for pol and env. We performed a Bayesian Markov chain Monte Carlo relaxed-clock phylogenetic analysis and applied a codon model to the resulting phylogenetic trees to estimate nonsynonymous (dN) and synonymous (dS) rates along each and every branch. We found important differences in the evolutionary rates of the different subtypes. These are due to differences not only in the dN rate but also in the dS rate, varying in roughly similar ways, indicating that these differences are caused by both different selective pressures (for dN rate) and the replication dynamics (for dS rate) (i.e., mutation rate or generation time) of the strains. CRF02_AG and subtype G had higher rates, while subtype D had lower dN and dS rates than the other subtypes. The dN/dS ratio estimates were also different, especially for the env gene, with subtype G showing the lowest dN/dS ratio of all subtypes.

FIG. 1.

Total substitution rate and 95% CI for the posterior distribution estimated in the Bayesian MCMC analysis of each of the seven analyzed subtypes/CRFs for pol (white squares) and env (gray diamonds). Only internal branches of the tree were included in this analysis. Substitution rate is presented as the number of expected substitutions per site per year. The P values of significantly different distributions are presented for pairwise comparisons in the table below the plot. P values for the env data set comparisons correspond to the upper right part of the matrix, while pol P values are presented in the lower left part of the matrix. AG, CRF02_AG; AE, CRF01_AE; A1, sub-subtype A1; B, subtype B; B full, subtype B full-genome data set; C, subtype C; C full, subtype C full-genome data set; D, subtype D; G, subtype G; NS, not significant.

FIG. 2.

Synonymous (a) and nonsynonymous (b) substitutions in pol (white squares) and env (gray diamonds) for the analyzed subtypes and CRFs and 95% CI of the posterior distribution estimated in the Bayesian MCMC analysis. Only internal branches of the tree were included in this analysis. Substitution rate is presented as the number of expected substitutions per site per year (subs/site/year). The P values of significantly different distributions are presented for pairwise comparisons in the table below the plot. P values for the env data set comparisons correspond to the upper right part of the matrix, while pol P values are presented in the lower left part of the matrix. AG, CRF02_AG; AE, CRF01_AE; A1, sub-subtype A1; B, subtype B; B full, subtype B full-genome data set; C, subtype C; C full, subtype C full-genome data set; D, subtype D; G, subtype G; NS, not significant.

FIG. 3.

(a) dN/dS ratios based on all branches of the tree for all subtypes/CRFs in pol (white squares) and env (gray diamonds) and standard deviation of each estimate. (b) dN/dS ratios based only on internal branches of the tree for all subtypes/CRFs in pol (white squares) and env (gray diamonds). dN/dS ratios were calculated from the global dN and dS absolute estimates of each of 200 analyzed trees.