Common cervicovaginal sequencing methods result in discordant molecular diagnoses of bacterial vaginosis and reveal strain level effects of Gardnerella vaginalis.

Abstract

Bacterial vaginosis (BV) is associated with HIV transmission and pre-term birth, yet the etiology of BV remains unknown. Our analysis addressed that knowledge gap by comparing diagnostic techniques and using Bayesian inference to find species-specific associations with clinical indicators. We also assessed the effect of sequencing methodology on the results of molecular BV profiling. We observed significant differences in microbial diversity within BV-associated CSTs based on clinical diagnosis. CST assignments were substantially influenced by sequencing methodology, with concordance between methods as low as 59% for metatranscriptomic and metataxonomic-based CST assignment. We also found that Gardnerella has a strain-dependent association with individual Amsel's criteria, and that Dialister micraerophilus and Parvimonas micra are positively associated with Amsel's criteria while Lactobacillus is negatively associated. These results highlight the challenge of characterizing a condition without a single etiological agent, reinforcing the need for more granular diagnoses and treatments that are sensitive to BV variability.