Comparative study of adaptive molecular evolution in different human immunodeficiency virus groups and subtypes

Abstract

Molecular adaptation, as characterized by the detection of positive selection, was quantified in a number of genes from different human immunodeficiency virus type 1 (HIV-1) group M subtypes, group O, and an HIV-2 subtype using the codon-based maximum-likelihood method of Yang and coworkers (Z. H. Yang, R. Nielsen, N. Goldman, and A. M. K. Pedersen, Genetics 155:431-449, 2000). The env gene was investigated further since it exhibited the strongest signal for positive selection compared to those of the other two major HIV genes (gag and pol). In order to investigate the pattern of adaptive evolution across env, the location and strength of positive selection in different HIV-1 sequence alignments was compared. The number of sites having a significant probability of being positively selected varied among these different alignment data sets, ranging from 25 in HIV-1 group M subtype A to 40 in HIV-1 group O. Strikingly, there was a significant tendency for positively selected sites to be located at the same position in different HIV-1 alignments, ranging from 10 to 16 shared sites for the group M intersubtype comparisons and from 6 to 8 for the group O to M comparisons, suggesting that all HIV-1 variants are subject to similar selective forces. As the host immune response is believed to be the dominant driving force of adaptive evolution in HIV, this result would suggest that the same sites are contributing to viral persistence in diverse HIV infections. Thus, the positions of the positively selected sites were investigated in reference to the inferred locations of different epitope types (antibody, T helper, and cytotoxic T lymphocytes) and the positions of N and O glycosylation sites. We found a significant tendency for positively selected sites to fall outside T-helper epitopes and for positively selected sites to be strongly associated with N glycosylation sites.

FIG. 1.

Mean ω ratios in gag, pol, env, env-gp120 and env-gp41 for HIV-1 group M subtypes A, B, and C, and HIV-2 subtype A (data sets A to K and N to V in Table 1). The mean ω ratios are calculated by averaging the results over all of the sites and are obtained from model M0. The numbers above the bars indicate the number of sequences and the number of codons in each data set. For example, “11/404” above the first gag bar indicates that there were 11 sequences and 404 codons in the gag HIV-1 group M subtype A data set (called data set A in Table 1).

FIG. 2.

Positions of positively selected sites across env for HIV-1 group M subtypes A, B, C, and D; group O; and HIV-2 subtype A (data sets I to N in Table 1). Each data set analyzed by CODEML is represented by one sequence, with the sites included in the analysis indicated with boldface type. Sites identified as being positively selected with a posterior probability of more than 95% are shaded. Notations above the sequences divide env into the gp120 and gp41 subunits and show the position of vpu and the second rev exon with the beginning and end of regions. The positions of the five variable regions V1 to V5 are indicated. Sites critical for CD4 binding are identified by a vertical bar, and sites implicated in the CXCR4 to CCR5 receptor switch are indicated with an * above the sequences. The numbers 2, 3, and 4 below the sequences indicate the number of data sets for which positive selection was identified at that site. The representative sequences for HIV-1 group M subtypes A, B, C, and D; group O; and HIV-2 subtype A are MA246, MBC18, BU910112, 84ZR085, ANT70, and CBL21, respectively.

FIG. 3.

The weighted mean ω ratio greater than 1 at each codon position in the env data sets comparing HIV-1 group M subtypes A, B, C, and D; group O; and HIV-2 subtype A (data sets I to N in Table 1). The weighted mean ω value for each site is calculated by multiplying ω by the posterior probability for each class under M8 and summing the results (see Materials and Methods). The positions of the five variable regions V1 to V5 are indicated.