More than 33 million people are currently infected with HIV-1, resulting in 2–3 million deaths every year. Natural immunity to the virus is virtually nonexistent; hence, the creation of a vaccine to combat this global pandemic is an international public-health priority 12. In 2003, the results were released for a phase III HIV-1 vaccine efficacy trial conducted in North America and The Netherlands (VAX004) 3. This study tested the efficacy of bivalent vaccines containing recombinant HIV-1 envelope glycoprotein 120 (rgp120) antigens, the major antigen on the surface of the virus 4. Overall, the vaccine candidate did not seem to reduce the incidence of HIV-1 infection, but an interesting trend was noted in the analysis of the different self-described ethnic groups [white (non-Hispanic), blacks, Hispanic, Asian, and "others"]. When only the non-white volunteers (17% of the total study population) were considered, the vaccine seemed to confer a slight benefit (P = 0.012). After adjustment for multiple tests, this difference was not significant (P = 0.13) 3. Despite a lack of statistical support, this is not a trivial result. If this trend was to be confirmed, it would imply that non-whites developed protective immunity to HIV-1 or, even more important, that rgp120 immunogens could protect against HIV infection under certain circumstances. Here we explored this possibility further.

HIV-1 evolution is driven, to a significant extent, by the immune response. If viruses isolated from non-white patients are in fact under a stronger selection pressure either because of genetic differences in the magnitude, specificity, or potency of the natural immune response, or because of differences in factors affecting virus replication, we should expect higher ratios of nonsynonymous (amino acid changing) to synonymous nucleotide substitutions (d_N/d_S) 5678 than in viral samples isolated from vaccinated and placebo (non-vaccinated) white individuals.

To test whether levels of selection were significantly different between vaccinated and placebo individuals in different races, we analyzed 3 clones per individual from 345 infected North Americans from the VAX004 study (Table 1; data available at http://www.gsid.org). Full-length HIV-1 subtype B gp120 sequences were amplified as described in Gilbert et al. 9. Since, as expected, viruses isolated from individuals from the same race did not form monophyletic groups 10, viral samples for each patient were analyzed separately. In each case, individual clones were aligned in MAFFT v5.7 11, and d_N/d_S ratios were estimated using Nei and Gojobori's method 6 in SNAP 12, model M0 (one-ratio) in PAML v3.14 13, and Fixed Effects Likelihood (FEL) with tree branch correction in HYPHY 14. In the latter case, we took recombination into account by first detecting recombination breakpoints with GARD 15, and then estimating the d_N/d_S ratios independently for each fragment.

Mean d_N/d_S ratios across races and treatments were compared using ANOVA, linear models (lm) and pairwise t-tests. Because treating all non-whites as a single unit is unrealistic considering their own genetic differences 3, we tested for differences in selection pressure on a race-by-race basis. Multiple significance in the pairwise t-tests was corrected using the Benjamini and Hochberg's procedure 16.

The estimates of d_N/d_S obtained with SNAP, PAML and HYPHY were all significantly correlated among the different estimators used (correlation coefficient > 0.85; P < 0.001). Importantly, the mean d_N/d_S ratios varied across races (Table 1), and these differences were globally significant (ANOVA;P < 0.02) for SNAP and PAML estimates. Blacks (vaccinees and placebo combined) showed higher d_N/d_S ratios for all the estimators than individuals from other ethnicities (Table 1). Significant differences (pairwise t-tests; P < 0.05) between black and white, Hispanic and "others" viral samples were observed for all the estimators before corrections, but only the comparisons involving SNAP and PAML estimates remained significant after the Benjamini and Hochberg's adjustment (Table 2). The higher d_N/d_S ratios observed for blacks suggest that the rate of virus evolution is greater in this group than in other volunteers. Differences in immune response to HIV-1 infection have been pointed out by the rgp120 HIV Vaccine Study Group 3 as one of the potential factors to explain vaccine efficiency differences between white and non-white volunteers in the VAX004 trial. Ethnic differences in immune response have been also reported for other viruses such as the hepatitis C virus 17.

Does the greater virus adaptive variation presumed in black participants reflect genetic differences in the intrinsic (no-preconditioned) immune response to HIV-1, or is it a consequence of the conditioned immune response induced by vaccination with rgp120? Comparison of vaccine and placebo recipients showed different results based on the d_N/d_S estimators used. No significant differences were observed among vaccinees, but significant differences (ANOVA; P = 0.025) in SNAP d_N/d_S ratios were detected among placebo individuals. Moreover, black-placebo patients showed significantly (lm coefficients; P < 0.02) higher mean d_N/d_S ratios (1.3, 1.38 and 1.66, for SNAP, PAML and HYPHY, respectively) than the other races (0.65–1.01, 0.84–1.14 and 0.73–1.56, respectively). These results might indicate that natural differences in the immune response may have increased viral rgp120 adaptation in blacks.

In North America, blacks correspond to 42% of all newly diagnosed HIV/AIDS cases, while white (non-Hispanic) and Hispanic individuals represent approximately 40% and 17%, respectively 18. If more data including both placebo and vaccinated recipients confirm that selection pressure differs between viruses infecting these three races, deciphering the genetic determinants of these differences should become a public-health priority. Indeed, our results highlight the need for selecting a broader representation of volunteers, based on ethnicity, in the design of future HIV-1 vaccine studies and trials 19.

The authors declare that they have no competing interests.

MPL, DP, and KC developed the genetic and statistical strategies implemented in this work. MPL, DP, and MA performed the genetic and statistical analyses. DVJ, FS, and PWB carried out the molecular genetic studies and immunoassays. All authors participated in the design of the study and helped to draft the manuscript. All authors read and approved the final manuscript.