ANZA MAPEMA KWA AFYA JOZI (AMkAJ)

Significance:       BV, GUD, and genital epithelial trauma are common, and are strong, independent risks for HIV acquisition and transmission. MMC is not 100% effective, 67% of men worldwide are uncircumcised, and not all men will become circumcised. Antiretroviral regimens that can achieve >90% risk reduction will be difficult to implement on a population level in sub-Saharan Africa, which accounts for ~70% of new HIV infections worldwide. Because of this, we need to continue to develop simple and cost-effective interventions which can be scaled up to a population level. Defining how the penile microbiome increases risk of BV, GUD, and genital epithelial disruption has major implications for preventing these conditions, whereby pathogenic penile bacteria could be viewed as a treatable condition. Identifying specific bacteria that lead to these conditions and whether they are caused by different bacteria – is necessary for determining the appropriate classes of antimicrobials to disrupt acquisition and transmission. Interventions to improve penile microbiome health may include systemic or topical antimicrobials, microbicides incorporating cytokine modulators, or behavioural or hygiene practices that reduce bacteria associated with BV, GUD, and genital epithelial disruption. Such interventions would (1) reduce HIV acquisition and transmission through reduction of BV, GUD, and genital epithelial trauma as co-factors for infection, and (2) improve pregnancy outcomes and genitourinary health.

Design:     Cross-sectional exploratory study measuring penile and rectal bacteria and cytokines, among 40 MSM and map these to urinary and rectal cytokine profiles. We hypothesize: (1) behavioral practices will affect bacterial community structure differentially for MSM vs. men who have sex with women (MSW); (2) MSM will have greater genital mucosal inflammation than MSW, and this will be associated with differences in bacterial community structure.

Study procedures include personal interview, medical history, physical examination, and collection of urine, penile and rectal specimens. Evaluations and data collection will occur one time only

Results:  We determined the predictive capacity of bacteria recovered from two penile sites for incident BV in female sex partners: meatus (MEA) and glans-coronal sulcus-distal shaft (GCSS). Incident BV was defined as Nugent score (7-10) at a follow-up visit, following a Nugent score of 0-6 at baseline. We applied four classification algorithms on selected and non-selected features: Random Forest, Logistic regression, SVM and k-Nearest Neighbor. To evaluate the classification performance, 10-Fold Cross validation was conducted. Finally, majority vote classifier was conducted to combine the decisions of four classifiers. Among 96 couples, 20 women had prevalent BV at baseline and were excluded from subsequent analysis. Among 76 couples, the incidence of BV was 28%. The prediction accuracies of the two penile sites were comparable (majority vote): GCSS, prediction error=22.9, [95% CI: 21.5-24.0]; MEA, prediction error=24.7, [95% CI: 20.4-25.9]. Combined information from the meatus and GCSS and inclusion of circumcision status in the models did not improve predictive performance. Predictive performance was driven by high specificity.

Preliminary analysis indicates inflammatory markers are decreased with increasing abundance of vaginal Lactobacillus, and this is in keeping with findings from the literature. Of note, we also observe IL-10 and TNF-α are higher for women with uncircumcised male partners, independent of Lactobacillus level.