Vaginal microbiome-hormonal contraceptive interactions associate with the mucosal proteome and HIV acquisition

Abstract

Alterations to the mucosal environment of the female genital tract, such as genital inflammation, have been associated with increased HIV acquisition in women. As the microbiome and hormonal contraceptives can affect vaginal mucosal immunity, we hypothesized these components may interact in the context of HIV susceptibility. Using previously published microbiome data from 685 women in the CAPRISA-004 trial, we compared relative risk of HIV acquisition in this cohort who were using injectable depot medroxyprogesterone acetate (DMPA), norethisterone enanthate (NET-EN), and combined oral contraceptives (COC). In women who were Lactobacillus-dominant, HIV acquisition was 3-fold higher in women using DMPA relative to women using NET-EN or COC (OR: 3.27; 95% CI: 1.24-11.24, P = 0.0305). This was not observed in non-Lactobacillus-dominant women (OR: 0.95, 95% CI: 0.44-2.15, P = 0.895) (interaction P = 0.0686). Higher serum MPA levels associated with increased molecular pathways of inflammation in the vaginal mucosal fluid of Lactobacillus-dominant women, but no differences were seen in non-Lactobacillus dominant women. This study provides data suggesting an interaction between the microbiome, hormonal contraceptives, and HIV susceptibility.

Fig 1. Probability of HIV seroconversion by microbiome type among women using different hormonal contraceptives.

Probability of HIV-seroconversion was assessed in women using DMPA relative to other hormonal contraceptives (NET-EN and/or COC) using logistic regression. Comparisons were performed as a stratified analysis of women who had either Lactobacillus dominant (n = 407) or non-Lactobacillus dominant (n = 287) vaginal microbiome types, or across all participants (n = 685). A significant association between DMPA and HIV-risk was only observed in the subset of Lactobacillus dominant women. Full statistical results of the main comparisons can be found in S5 Table.

Fig 2. MPA-associated cervicovaginal proteome alterations in Lactobacillus-dominant women.

(A) Volcano plots display cervicovaginal host proteins differentially abundant in women with “High” MPA (>300 pg/ml, n = 190) vs “Low” MPA (≤50 pg/ml-299 pg/ml, n = 253) for each microbiome type (LD and non-LD) using two-tailed independent t tests (FDR = 5%; 576 proteins; participant numbers: LD-High MPA = 120, LD-Low MPA = 140, non-LD-High MPA = 70, non-LD-Low MPA = 113). Full statistical results and protein descriptions are available in S1 Data. (B) Proteins differentially abundant with MPA levels were annotated to functional pathways using the DAVID Gene Ontology database and the Ingenuity Pathway Analysis database (-log₁₀(P value) displayed). MPA = Medroxyprogesterone acetate, LD = Lactobacillus dominant, non-LD = Non-Lactobacillus dominant, ns = non-significant.

Fig 3. Cervicovaginal protein signatures associated with MPA levels of women with vaginal Lactobacillus dominance.

A LASSO algorithm identified 17 proteins that best classified LD women with low MPA levels from LD women with high MPA levels, as well as all non-LD women. (A) An area under the curve (AUC) plot shows the protein signature was accurately able to classify women with an LD vaginal microbiome and low serum MPA from other groups (AUC = 0.827, P<0.0001). (B) A PLSDA plot shows LD/low MPA women separated from other groups along X-variates 1 and 2. (C) A loadings plot shows protein contributions to the identified signature. Orange bars indicate proteins increased in the LD/low MPA group, while purple bars indicate those increased in all other groups. Box and whisker plots display the hazard ratios (HR) estimated from cox proportional hazards models between MPA-associated glucose metabolism biomarker levels and HIV acquisition risk across all women in the study (n = 685). (D) MPA-risk biomarker expression in microbiome-MPA groupings. LASSO = Least Absolute Shrinkage and Selection Operator, PLSDA = Partial Least Squares Discriminant Analysis, MPA = Medroxyprogesterone acetate, LD = Lactobacillus dominant, non-LD = Non-Lactobacillus dominant, HR = hazard ratio, CI = 95% confidence interval.