Understanding the mechanism for HIV-1 superinfection is crucial to the development of an HIV vaccine in order to prevent HIV-1 acquisition in a naïve host, since HIV-1 superinfection calls into question the idea that a robust primary immune response to HIV-1 infection provides some immunological protection from re-infection with a heterologous HIV-1. Superinfections occurring in heterosexual cohabiting pairs have not been studied in detail or longitudinally, despite the fact that most primary infections occur in HIV-1 discordant couples [40
]. Though discordant couples are considered 'high risk,' they are not typically thought to be exposed to as many different viruses as sex workers or intravenous drug users [31
]. For this reason, it might be predicted that superinfection would be observed at lower frequency and would occur predominantly between individuals within a couple; however, this was not the case. In the 22 newly infected partners, who had acquired HIV outside the marriage, we observed a frequency of superinfection in these individuals in the first year of follow up that was similar to that of primary infection (13.6% vs. 7.8%, p > 0.05). Even though we excluded couples in which the chronically infected partners had viral loads lower than 1000 from this study, all of the superinfections in the seroconverting partner originated from a non-spousal partner. Thus, it is clear that these couples are a higher risk subset of the cohort with exposure to HIV-1 infection outside the main partnership. The very limited frequency of intra-couple superinfection (1/4) studied here in part reflects continued safe sexual practices within the couple, since greater than 95% of reported sexual activity was with a condom. Nevertheless, significant numbers of non-condom exposures did occur (104/2146) in 19 of the 22 newly infected partners who did not become superinfected. Although self-report of external sexual activity is clearly underreported [37
], it seems unlikely that for each of the three superinfected individuals, the number of unprotected extra-marital exposures would exceed the number within the marriage. Moreover, with an adult seroprevalence rate of ~20% in Zambia, extramarital exposures should in a majority of instances be with seronegative individuals and therefore present less of a risk of potential superinfection than with the known seropositive partner. Interestingly, a similar lack of intra-couple superinfection has been observed in a recent study of 11 seroconcordant couples infected with disparate viruses in Uganda [48
One factor that could influence susceptibility to superinfection is the presence of sexually transmitted diseases. Since genital infections and ulcers break down mucosal barriers and contribute to increased risk for primary HIV infection [42
]. In the analysis of behavioral characteristics and clinical signs, the factors that trended toward significance were the presence of GUD on physical exam and RPR positivity in the superinfected group as compared to the non-superinfected group, although 7/19 non-superinfected individuals did have GUD. Previous studies in this Zambian cohort have shown a 2-3 fold increase in risk of HIV-1 infection in uninfected partners with GUD, after correction for viral load in their chronically infected partner [51
]. In contrast to this higher-risk group, longitudinal gag, pol
, and nef
gene sequence data from 80 epidemiologically linked transmission pairs in the cohort (where transmission was from the cohabiting spousal partner) in the first two years of follow-up have not demonstrated any evidence of superinfection from non-spousal partners (data not shown), consistent with a lower frequency of extra-marital sexual activity in this cohort subset.
Despite the fact that a majority of the acutely infected individuals in this study of ZEHRP transmission pairs have > 2 years of follow-up, HIV-1 superinfection was observed within the first year of follow-up in each of the 3 acutely superinfected individuals. This is consistent with recent studies of intra-subtype superinfection in subtype B infected individuals, where in one case mathematical modeling indicated a 21-fold reduction of superinfection after 1 year of infection [25
], and in a second case, a retrospective analysis of individuals in the San Diego and Los Angeles Acute HIV Infection and Early Disease Research Programs demonstrated 3 cases of superinfection within 13 months of seroconversion [17
]. In contrast, the timing of superinfection in a subtype A commercial sex worker cohort appears less constrained, with superinfection detected as late as 5 years after primary infection [26
The analysis of longitudinal env
sequences, amplified by the SGA approach, for each of the individuals identified through degenerate base analysis allowed the definitive resolution of both the timing and nature of superinfection. In each of the three recent seroconversion cases a distinct superinfecting genetic variant could be identified which segregated independently on the phylogenetic tree (Figure ). Recombination between the primary infection variant and the superinfecting variant was observed in each case; and at some time points, consistent with the Highlighter analyses of population sequences, these recombinants became the dominant variant in the circulating virus population. Interestingly, we observed the conservation of recombination break-points within different variants in an individual over several months, suggesting that recombinant viruses with these particular sequence mixes possess fitness benefits over either the initial or the superinfecting strain. This is consistent with the observation of Streeck et al., [36
], who showed that recombination between initial and superinfecting viruses could accelerate immunological escape from cellular immune responses. In a more global sense, the selection of mixed genotypes with enhanced population fitness is evidenced by the numerous circulating recombinant forms of HIV-1 resulting from dual infection of individuals [9
], which clearly contribute to the overall diversity of a virus population. Additional studies will be required to fully characterize the basis of recombinant virus selection in the subtype C infected individuals under study here.
The SGA analysis of viral sequences bolstered our interpretation that ZM211M was superinfected from his spousal partner, ZM211F, during her acute seroconversion. At the time of her seroconversion, ZM211M has two dominant and distinct quasispecies with limited evidence for recombination between them. In contrast at month 3, a distinct population of recombinant viruses arises. This is consistent with superinfection of ZM211M during his spouse's acute viremia (viral load greater than 750,000), followed by the emergence of recombinants. Moreover, shortly after the probable superinfection, the viral load of ZM211M increased 10-fold and he is deceased within 6 months.
Determining why HIV-1 does or does not superinfect an exposed individual will be crucial to understanding the nature of an immune response that is capable of preventing de novo
infection. Given the considerable antigenic dissimilarities between subtypes, we might not expect that initial infection by one subtype of HIV-1 would provide significant immune protection against other subtypes; on the other hand we might expect there to be some protection from reinfection of infected patients by more closely related HIV-1 strains of the same subtype. This does not appear to be the case during the first year of infection in the subtype C infected individuals studied here, where rates of intra-subtype superinfection in the first year of study were similar to those of primary infection [39
]. However, it is of interest that in the three individuals that are superinfected, little variation in env
sequences is observed in the period prior to the superinfection event, suggesting that there may be limited neutralizing antibody pressure on the founder virus. Indeed, preliminary studies indicate the absence of potent neutralizing antibody responses to the founder virus at the visit prior to superinfection (D. Basu et al., unpublished). It will be of interest to determine whether there is a more potent neutralizing antibody response in the non-superinfected individuals who also report extra-marital contact. Moreover, given that in this study each partner in the couple is infected with a different strain of subtype C HIV-1, it is possible that repeated exposure to a partner's HIV-1 strain could stimulate the development of HIV-1 specific immune responses and that this might have provided protection against intra-couple superinfection. This type of immune stimulation with boosting of the cellular immune response has been reported to occur in subtype B infected men who have sex with men [23
The existence of HIV-1 superinfection presents an obstacle to develop a vaccine to prevent primary infection with HIV-1. With technologies such as next-generation sequencing being employed to detect HIV-1 superinfection [48
], the detection of very small viral sub-populations at a given time point will increase resolution. There are behavioral and clinical aspects (e.g. circumcision, genital ulcers) that influence this phenomenon but there are likely immunologic correlates that render some individuals more susceptible to superinfection. Continued study of HIV-1 superinfection within cohabiting heterosexual couples can provide insights into such correlates in the context of a potentially highly susceptible and relatively low-risk cohort type.