381    

Background: The development of candidate HIV vaccines demands a thorough investigation into the consistency of the selective environment - presumed primarily to be due to the host immune response - between divergent HIV lineages. This study aims to explore whether there are differences in the location and intensity of selection among the different HIV groups and subtypes.

Methods: Detection of positive selection involves the comparison of synonymous (d_S) and nonsynonymous (d_N) substitutions between protein-coding DNA sequences. The d_N/d_S ratio, w, is then used to measure the difference between these two rates of substitution such that an w< 1 corresponds to purifying (negative) selection, an w = 1 corresponds to neutral evolution (absence of selection), and an w > 1 indicates adaptive evolution (positive selection). Estimations were carried out on alignments of sequences in a maximum likelihood framework that accounts for phylogenetic structure, biases in codon usage and the transition/transversion (T_S/T_V) rate ratio. A Bayesian approach further allows statistical significance to be attached to the assignment of each site to the conserved, neutral or positively selected classes. Positive selection was tested for in gag, pol, and env genes sequences alignments of HIV-1 group M subtypes A to D, group O, and HIV-2 subtype A. The location and strength of positive selection were compared between clades and analysed with respect to epitopes and glycosylation sites. Non-parametric tests and Monte Carlo simulations were used to determine significant results.

Results: Positive selection was strongest in the envelope gene, particularly in the gp120 sub-unit, irrespective of the clade examined. In env, positive selection tends to occur at the same sites in different clades. Assuming that positive selection on this gene is primarily due to immune pressure, this result suggests that the immune response tends to target the same sites on the HIV envelope glycoprotein, irrespective of the HIV clade. Areas involved in CD4 and chemokine binding are apparently not under positive selection between individuals, presumably to maintain efficient interaction with these cellular molecules. Sites under positive selection significantly tend to include N-glycosylation sites, which are not necessarily associated with predicted CTL and Ab sites.

Conclusions: The results have implications in understanding HIV-evasion mechanisms and potential practical applications for vaccine design.

Keywords: vaccination; positive selection; glycosylation