Background: It is unclear whether immune selection is involved in PR evolution as few CTL epitopes are known in PR. Tests using sequences obtained after the introduction of protease inhibitors (PIs) are confounded by drug-related selection pressure. We analyzed genetic variation in PR sequences obtained prior to the introduction of PIs in order to elucidate the possible role of immune selection.

Methods: Data were obtained as part of the Merck 035 trial, which compared AZT+3TC, IDV, and AZT+3TC+IDV. Clonal sequences (median = 6 /patient) and viral load measurements were obtained from the plasma of 90 patients at baseline. The ratio of nonsynonymous to synonymous substitution rates, w, was estimated using maximum likelihood; w > 1 indicates positive selection.  Phylogenetic reconstruction of within-host sequences was performed using statistical parsimony.

Results: There was strong evidence that selection pressure on PR varied between amino acid sites (p<0.001).  Although most were under negative ‘stabilizing’ selection (overall w=0.27), amino acids 10, 37 and 63 in PR were found to be under positive selection (w =3.8).  These sites were located within regions of high genetic diversity, but the neighboring variable sites were not positively selected.  We hypothesize that this variation was due to a “bystander effect,” whereby positive selection resulted not only in divergence at selected amino acids, but also at closely linked sites. With a mathematical model, we show that this can occur when there is spatial clustering of infected cells within the body, which limits the effect of recombination, and a global, fluctuating selection pressure such as that which can be exerted by HIV-specific immune responses. The pattern of within-host diversity observed was consistent with our model, which predicts no correlation between viral load and within-host viral diversity, and ‘star-like’ viral phylogenies within the host.

Conclusions: We have detected evidence of positive selection in PR in the absence of drug therapy and propose that HIV-specific immune responses are responsible, thus identifying the PR domain as a potential vaccine target. Positive selection can increase diversity at the population level not only at selected amino acids, but also at closely linked sites.