Bacterial vaginosis (BV) increases transmission of HIV, enhances the risk of preterm labour, and is associated with malodour. Clinical diagnosis often relies on microscopy to presumptively detect a loss of lactobacilli and acquisition of anaerobes.
This diagnostic may not reflect the microbiota composition accurately as lactobacilli can assume different morphotypes, and assigning BV associated morphotypes to specific organisms can be challenging. Using an untargeted metabolomics approach we identify novel biomarkers for BV in a cohort of 131 Rwandan women, and demonstrate that metabolic products in the vagina are closely associated with bacterial diversity.
Metabolites associated with high diversity and clinical BV includes 2-hydroxyisovalerate and γ-hydroxybutyrate (GHB), but not the anaerobic end-product succinate, while low diversity is characterized by lactate and amino acids. These biomarkers are independent of pregnancy status, and were validated in a blinded replication cohort from Tanzania (n=45), in which we predicted clinical BV with 91% accuracy. Correlations between the metabolome and microbiota identified Gardnerella vaginalis as a putative producer of one of these compounds, GHB, and we demonstrate production by this species in vitro. This work provides novel insight into the metabolism of the vaginal microbiota and identifies highly specific biomarkers for a common condition.