Introduction The aetiology and pathogenesis of bacterial vaginosis (BV) are unclear which has impacted greatly on efforts to improve the efficacy of current treatment approaches. We examined the microbial composition of the vaginal microbiota and factors associated with the development of BV, in women-who-have-sex-with-women (WSW) who were participating in a two year cohort study, in order to gain insights into the microbial changes that occur around the development of BV.

Methods 298 women self-collected high vaginal swabs and completed questionnaires detailing behavioural practices and symptoms three monthly for 24 months or until incident BV, whichever occurred first. BV was diagnosed by the Nugent method and women could only enrol in the cohort if they were BV negative on 3 weekly vaginal samples at screening. Fifty-one cases of incident BV occurred over 24 months (BV incidence rate, 9.75/100 woman-years). Vaginal swabs were stored at - 80°C. Available longitudinal vaginal specimens from the 51 cases who developed BV and 51 age-matched controls who did not, were included in this study to examine the vaginal microbial composition by 16S rRNA gene sequencing; 47 case participants and 50 control participants met the requirements for specimen submission and sequencing quality (353 swabs). Microbial factors associated with the development of BV were determined by multivariable analysis, adjusting for sexual behaviours. Microbial diversity and stability were assessed by the Shannon diversity index and Bray-Curtis dissimilarity scores between consecutive paired longitudinal samples.

Results For each 1% increase in Gardnerella vaginalis abundance there was a 2% increased risk of developing BV (Adjusted Hazard Ratio [AHR]=1.02, 95% CI 1.01–1.03, p0.001). Detection of BVAB TM7 (uncharacterised bacterium of candidate division TM7) was associated with a 6 fold increase in risk of developing BV (AHR=6.06, 95% CI: 1.99, 18.43, p=0.002). In contrast for each 1% increase in Lactobacillus crispatus abundance there was a 1% reduction in the risk of developing BV (AHR=0.99 95% CI 0.098–1.00, p=0.038). The vaginal microbiome of women who developed BV was characterised by high microbial diversity and less stability compared to controls (p=0.04).

Conclusion In a cohort that was designed carefully to study incident BV, lower abundance of L. crispatus, increased abundance of G. vaginalis, and detection of BVAB TM7 were significantly associated with development of BV, after adjusting for bacterial species and sexual behaviour. Increased vaginal microbial diversity, decreased stability and exposure to new sexual partners were also associated with the development of BV in WSW. Incident BV may result from sexual exchange of key BV-associated bacteria such as G. vaginalis which could destabilise the microbial ecology through displacement of beneficial bacteria such as L. crispatus.