Our data provide empirical evidence that differences in genital tract concentrations of HIV-1 influence the transmission risk of HIV-1 infection, and we found that this relationship was independent of plasma HIV-1 concentration. Our large sample size of heterosexual African HIV-1 serodiscordant couples and prospective follow-up with collection of genital samples prior to HIV-1 transmission permitted analyses demonstrating that the concentration of HIV-1 RNA in endocervical and seminal samples from HIV-1 infected individuals was strongly correlated with risk of HIV-1 transmission to their HIV-1 susceptible sexual partners. Genomic analysis of HIV-1 isolates to confirm HIV-1 transmission within the study partnerships further strengthens our findings. These data support the concentration of HIV-1 RNA in genital secretions as a marker of HIV-1 sexual transmission risk.
The first studies of genital HIV-1 using viral culture provided qualitative evidence for infectious virus in genital secretions as a mechanism for HIV-1 transmission [20
]. Subsequent studies have used nucleic acid amplification to quantify genital HIV-1, with results suggesting that higher genital HIV-1 levels are likely to be a measure of increased HIV-1 infectiousness [1
]. Higher plasma HIV-1 levels, genital tract infections, and advanced HIV-1 disease have been associated with increased genital HIV-1 levels [2
]. In prospective interventional studies with pre- and post-treatment genital tract samples, cure of sexually transmitted infections and initiation of ART significantly reduced genital HIV-1 RNA concentrations [7
]. However, to demonstrate that genital HIV-1 levels predict risk of HIV-1 sexual transmission required longitudinal studies of HIV-1 infected persons and their initially uninfected partners. The establishment of such cohorts has been logistically challenging [1
]. Only one previous case-control study, among men who have sex with men, assessed the relationship between genital HIV-1 RNA concentrations and risk of HIV-1 sexual transmission. This study reported that plasma and seminal fluid HIV-1 RNA concentrations in 15 transmitting partners were significantly higher than in 32 non-transmitting partners [23
We found a step-wise association between genital HIV-1 levels and HIV-1 incidence, with an approximately two-fold increased risk for each one log10
increase in genital HIV-1. This was comparable to the association between endocervical HIV-1 RNA and female-to-male HIV-1 transmission, and seminal HIV-1 RNA and male-to-female transmission. We also found that plasma HIV-1 RNA quantity predicted HIV-1 transmission risk in a similar step-wise manner. This linear risk relationship between log10
HIV-1 RNA concentrations and HIV-1 outcomes has been previously reported for systemic HIV-1 concentrations versus both sexual and perinatal HIV-1 transmission [4
], including in the Ugandan study that first demonstrated that higher blood HIV-1 concentrations resulted in increased heterosexual infectiousness [4
], as well as for plasma HIV-1 levels as a predictor of the risk of HIV-1 clinical progression to AIDS. The consistency of this relationship raises the question of whether the log10
quantity is a fundamental pathogenic property of the virus, although discerning the precise biologic mechanisms is not possible with the samples we tested for this study.
We observed a small number of HIV-1 transmission events (annual incidence <1%) among couples in which the HIV-1 infected partner had genital HIV-1 levels below the limit of quantification. Importantly, plasma HIV-1 was detectable for all 11 persons with undetectable genital HIV-1 concentrations who transmitted HIV-1 to their partners. The reason for this could be that a single assessment of genital HIV-1 burden may miss intermittent shedding of genital virus [16
In our study, as in multiple previous studies, plasma and genital HIV-1 concentrations were only modestly correlated [1
]. We found that genital HIV-1 concentrations remained independently associated with HIV-1 transmission risk after adjustment for plasma HIV-1 levels, as well as other clinical and behavioral factors. Genital HIV-1 levels display greater variability than do plasma HIV-1 levels [13
]; greater variability in the measurement of genital versus plasma HIV-1 would not alter the accuracy of our findings (i.e., the point estimate of risk of HIV-1 transmission versus log10
genital HIV-1 levels) but would contribute to the precision of the estimate (i.e., the width of the confidence intervals). Recent work suggests that genital HIV-1 levels, like those in plasma, establish a relatively stable set point after acute infection [18
]. Thus, a single measurement, as done in this study, may provide a useful biomarker of HIV-1 infectiousness, particularly given the challenges of obtaining repeat genital HIV-1 measurements in large studies. HIV-1 replication may be different at genital mucosal sites compared to other sites that contribute virus to the blood, potentially due to genital tract infections or local immunological factors [1
]. Thus, genital HIV-1 levels, as potentially the most relevant and proximate marker of HIV-1 exposure for sexual HIV-1 transmission, may predict HIV-1 risk as well as or better than plasma HIV-1 concentrations alone. We found that only genital HIV-1 levels in men were statistically related to HIV-1 transmission risk in a model that included both genital and plasma HIV-1 RNA concentrations, whereas for women both blood plasma and genital HIV-1 RNA were independently predictive. These findings could reflect the biology of menses and the contribution of blood HIV-1 to the female genital tract, which is not a consideration for men. More limited statistical power for our analysis of male-to-female HIV-1 transmission (given a smaller number of HIV-1 infected men compared to women in our study population) may also explain these findings. Future studies of genital HIV-1 should explore characteristics of those variants that are transmitted, including genetic sequence differences, viral fitness, and whether the source of transmitted virus is cell-free or cell-associated HIV-1 [1
We found that acyclovir reduced genital HIV-1 levels by ~0.3 log10
, a result that was statistically significant and similar to prior studies of HSV-2 suppressive therapy [13
]; in our trial, acyclovir reduced plasma HIV-1 levels by 0.25 log10
copies/mL but did not reduce HIV-1 transmission [19
]. We recently estimated that a nearly 0.75 log10
copies/mL reduction in plasma HIV-1 RNA would be necessary to decrease HIV-1 transmission by 50% [24
]. Thus, interventions that greatly reduce HIV-1 levels, like ART, are likely to have more substantial effects on HIV-1 transmission risk than interventions that reduce HIV-1 concentrations minimally.
We only collected one sample per study participant for genital HIV-1 quantification. Repeat measurements might have increased precision in our regression estimates as the variability in HIV-1 concentrations is greater in genital samples than in plasma samples [2
]. However, in spite of this potential for improvement in analytical precision, we still observed a strong relationship between genital HIV-1 levels and HIV-1 transmission risk. Previous studies of genital HIV-1 have collected a single or a small number of genital samples per individual to measure the effect on genital HIV-1 shedding of interventions aimed at decreasing HIV-1 infectiousness, including HSV-2 suppressive therapy [9
], treatment of curable STIs [7
], and initiation of ART [15
]. Our results confirm that a single measurement of genital HIV-1 quantity is a strong surrogate marker of HIV-1 transmission risk, and suggest that the potential impact of new interventions aimed at reducing HIV-1 transmission can be assessed through studies of genital HIV-1 RNA. With >2500 participants, this is the largest study of genital HIV-1 in African persons.
A limitation of this study is that some HIV-1 transmission events occurred before or several months after collection of the genital sample. However, the median time from acquiring the genital sample to HIV-1 seroconversion was less than six months and for the majority of HIV-1 transmission events the genital sample was collected before or at the time of seroconversion. Sensitivity analyses assessing the timing of genital sample collection relative to HIV-1 transmission and the collection of a plasma sample for HIV-1 RNA quantification generated results similar to those from the analysis of all participants. Etiological screening for sexually transmitted infections was done at study enrollment and not when genital HIV-1 RNA samples were collected. Finally, HIV-1 infected partners were also HSV-2 seropositive. HSV-2 is common among persons with HIV-1 (seroprevalence 50–90%) [31
], and thus this is unlikely to limit the generality of our findings.
Understanding the relationship between genital HIV-1 replication and the risk of HIV-1 transmission is central to describing the fundamental biological mechanisms underlying HIV-1 transmission. ART and other potential new interventions (such as HIV-1 vaccines) that reduce systemic and genital HIV-1 replication, and interventions that reduce genital HIV-1 concentrations alone (such as treatment of genital tract infections and antiretroviral-based microbicides) should continue to be evaluated for their potential to reduce HIV-1 transmission. Genital sampling should be used to quantify the potential reduction in HIV-1 transmission risk of interventions that are directed at reducing the infectiousness of persons with HIV-1.