Bursts of nonsynonymous substitutions in HIV-1 evolution reveal instances of positive selection at conservative protein sites

Abstract

The fixation of a new allele can be driven by Darwinian positive selection or can be due to random genetic drift. Identifying instances of positive selection is a difficult task, because its impact is routinely obscured by the action of negative selection. The nature of the genetic code dictates that positive selection in favor of an amino acid replacement should often cause a burst of two or three nucleotide substitutions at a single codon site, because a large fraction of amino acid replacements cannot be achieved after just one nucleotide substitution. Here, we study pairs of successive nonsynonymous substitutions at one codon in the course of evolution of HIV-1 genes within HIV-1 populations inhabiting infected individuals. Such pairs are more numerous and more clumped than expected if different substitutions were independent and than what is observed for pairs of successive synonymous substitutions. Bursts of nonsynonymous substitutions in HIV-1 evolution cannot be explained by mutational biases and must, therefore, be due to positive selection. Both reversals, exact or imprecise, of fixed deleterious mutations and acquisitions of amino acids with new properties are responsible for the bursts. Temporal clumping is strongest at codon sites with a low overall rate of nonsynonymous evolution, implying that a substantial fraction of replacements of conservative amino acids are driven by positive selection. We identified many conservative sites of HIV-1 proteins that occasionally experience positive selection.

Fig. 1.

Distribution of codon sites by their d_N/d_S values (the values of d_S are assumed to be gene-specific; see SI Supporting Text for details).

Fig. 2.

Distributions of nonsynonymous and synonymous substitutions on the phylogenetic trees. For each possible number of substitutions at a codon site (within the range 0–50), we present the number of sites with this number of substitutions (Top, empty bars); the fractions of leaves of the phylogenetic tree, for which the number of substitutions at these sites, on the path from the leaf to the root, is two (Top, solid line) or three or more (Top, dotted line) within each sliding window of length 30; the actual average per-site number of substitutions with at least one descendant substitution (Bottom, filled bars); and the average per-site number of substitutions with at least one descendant substitution obtained in simulations of independent substitutions (Bottom, solid line). Data on only nonsynonymous (top row of graphs) and on only synonymous (bottom row of graphs) substitutions are shown. Numbers of substitutions that were not encountered at any site are marked by crosses on the x axis.

Fig. 3.

Distances between successive nonsynonymous (Upper) and synonymous (Lower) substitutions on the phylogenetic trees. The mean distance between successive nonreversing substitutions is shown for codon sites with each total number of substitutions on the tree (dots). Solid lines present mean distances between successive substitutions within each sliding 30-site window. Dashed lines show mean distances between independent substitutions obtained in simulations.

Fig. 4.

Amino acid sites inferred to be under positive selection in HIV-1 gp120. (A) Amino acid sites with >80 replacements. (B) Rapidly evolving sites previously inferred to be under positive selection (27). (C) Conservative sites (<80 replacements) with strongly clumped substitutions. Protein structure was visualized with the VMD package (28).