Possible protective effect of HIV-2 against incident HIV-1 infection: review of available epidemiological and in vitro data

In 1995, an article by Travers et al. appeared in Science demonstrating a significantly lower incidence of HIV-1 infection in HIV-2-infected than in HIV-uninfected female sex workers in Senegal, with an adjusted incidence rate ratio (IRR) of 0.32 suggesting a protective effect of 68%[1]. As pointed out both by the authors and an accompanying editorial [2], these findings had potential implications for HIV vaccine development efforts. In an exchange of letters to the Editor in Science in 1996, our group questioned whether a systematic bias had been introduced in the selection of the HIV-uninfected comparison population in Senegal [3], but in a reanalysis of a larger group of uninfected women, a protective effect of 64% was still found [4]. Consequently, this discussion stimulated various research teams working in West Africa, the only region in the world where HIV-2 is epidemic [5], to examine existing longitudinal datasets to assess whether a similar protective effect could be identified. A summary of their published reports [6–10], including the most recent by van der Loeff et al. in this issue of AIDS, appears in Table 1.Table 1: Published epidemiological studies evaluating the possible protective effect of HIV-2 infection upon subsequent HIV-1 infection. The possible protective effect of HIV-2 has been assessed in six additional populations, including female sex workers, urban and rural adults, postpartum women, and police officers; these studies were conducted in Guinea-Bissau, the Gambia, and Côte d'Ivoire from 1987 through 1998. In each of these studies, and in contrast to the initial findings from Senegal, the incidence of HIV-1 was actually found to be higher, and in two of the studies significantly higher, in HIV-2-infected than in HIV-uninfected persons, with unadjusted IRR values ranging from 1.6 to 19.8. It is unlikely that the discrepancy between the original study from Senegal and the subsequent studies can be attributed to differences in HIV-2 viral subtypes, since HIV-2 subtype A is predominant throughout the region. Table 1 is a meta-analysis of all published reports, which includes more than 21 000 person-years of observation. The overall weighted IRR was 2.47 (95% confidence interval, 1.76–3.46], demonstrating a significantly higher HIV-1 incidence in HIV-2-infected than in HIV-uninfected persons. Concurrent with the publication of these epidemiological studies have been several elegant manuscripts reporting that HIV-2 has an inhibitory effect on HIV-1 replication in vitro[11–13], and one demonstrating that peripheral blood mononuclear cells from HIV-2-infected subjects can resist in vitro challenge with certain HIV-1 viruses [14]. However, in a study of female sex workers in Côte d'Ivoire, median HIV-1 RNA viral load levels were not found to be significantly different among HIV-1-infected and dually (HIV-1 and HIV-2) infected women, and a positive correlation was found between HIV-1 and HIV-2 viral load in dually infected women; these data provide strong evidence that HIV-2 does not have a demonstrable inhibitory effect on HIV-1 replication in vivo[15]. The methodological and epidemiological challenges of truly demonstrating a protective effect of HIV-2 upon incident HIV-1 infection in longitudinal studies are formidable, since it is extremely difficult to control for sexual risk-taking behavior between HIV-2-infected and HIV-uninfected populations. Some of the groups attempted to do so by adjusting for age [1,9,10], duration of sex work and gonorrhea [1], urethral discharge, genital ulcer disease and Treponema pallidum hemagglutination assay (TPHA) status [9], and by comparing lifetime number of sexual partners [8]. However, ultimately these factors are only surrogates for sexual risk behavior and are not accurate markers of known exposure to HIV-1-infected persons. In addition, these studies appear to be retrospective analyses of data that were primarily collected for other scientific purposes and, therefore, were not originally designed to control for exposure to HIV-1. Simply put, it is reasonable to assume that HIV-2-infected persons are, as a group, more likely than HIV-uninfected persons to engage in high-risk sexual behaviors that both led to them becoming infected with HIV-2 in the first place and put them at greater risk of subsequent HIV-1 infection. The data from the meta-analysis could be construed to indicate that HIV-2 infection facilitates HIV-1 infection, but this is doubtful especially in light of the in vitro data. It is also possible that there are other differences between HIV-2-infected and HIV-uninfected persons that could not be controlled for in these analyses, such as increased biological or genetic susceptibility. It is unlikely that there will be many more data forthcoming to address this issue, since the number of large cohort studies in West Africa is limited, and the cost and feasibility of conducting additional studies that would adequately address the question of protection would pose considerable challenges. In summary, the available epidemiological data do not support a protective effect of HIV-2 infection upon incident HIV-1 infection. In six of seven populations studied to date, the HIV-1 incidence is actually higher in HIV-2-infected than in HIV-uninfected persons, with a meta-analysis showing this effect to be statistically significant overall. Lastly, although HIV-2 appears to inhibit HIV-1 replication in vitro, this does not appear to have any demonstrable effect on HIV-1 viral load levels in vivo. Acknowledgments The author would like to thank Dr John Karon for his statistical support, and Dr John Nkengasong and Dr Ida Onorato for their scientific comments.