Characterization of vaginal microbiomes in clinician-collected bacterial vaginosis diagnosed samples

Abstract

Bacterial vaginosis (BV) is a type of vaginal inflammation caused by bacterial overgrowth, upsetting the healthy microbiome of the vagina. Existing clinical testing for BV is primarily based upon physical and microscopic examination of vaginal secretions. Modern PCR-based clinical tests target panels of BV-associated microbes, such as the Labcorp NuSwab test that targets Atopobium (Fannyhessea) vaginae, Megasphaera-1, and Bacterial Vaginosis Associated Bacterium (BVAB)-2. Remnant clinician-collected NuSwab vaginal swabs underwent DNA extraction and 16S V3-V4 rRNA gene sequencing to profile microbes in addition to those included in the Labcorp NuSwab test. Community state types (CSTs) were determined using the most abundant taxon detected in each sample. PCR results for NuSwab panel microbial targets were compared against the corresponding microbiome profiles. Metabolic pathway abundances were characterized via metagenomic prediction from amplicon sequence variants (ASVs). 16S V3-V4 rRNA gene sequencing of 75 remnant vaginal swabs yielded 492 unique 16S V3-V4 ASVs, identifying 83 unique genera. NuSwab microbe quantification was strongly concordant with quantification by sequencing (P < 0.01). Samples in CST-I (18 of 18, 100%), CST-II (three of three, 100%), CST-III (15 of 17, 88%), and CST-V (one of one, 100%) were largely categorized as BV-negative via the NuSwab panel, while most CST-IV samples (28 of 36, 78%) were BV-positive or BV-indeterminate. BV-associated microbial and predicted metabolic signatures were shared across multiple CSTs. These findings highlight robust sequencing-based quantification of Labcorp NuSwab BV microbes, accurate discrimination of vaginal microbiome CSTs dominated by distinct Lactobacilli, and expanded the identification of BV-associated bacterial and metabolic biomarkers.IMPORTANCEBacterial vaginosis (BV) poses a significant health burden for women during reproductive years and onward. Current BV diagnostics rely on either panels of select microbes or on physical and microscopic evaluations by technicians. Here, we sequenced the microbiome profiles of samples previously diagnosed by the Labcorp NuSwab test to better understand disruptions to the vaginal microbiome during BV. We show that microbial sequencing can faithfully reproduce targeted PCR diagnostic results and can improve our knowledge of healthy and BV-associated microbial and metabolic biomarkers. This work highlights a robust, agnostic BV classification scheme with potential for future development of sequencing-based BV diagnostic tools.

Keywords: 16S rRNA; bacterial vaginosis; vaginal microbiome.

Fig 1

16S V3–V4 rRNA sequencing of remnant clinician-collected vaginal swabs previously analyzed via the Labcorp NuSwab BV PCR test. (a) Example process of determining the bacterial vaginosis (BV) status of a vaginal swab using the NuSwab three-microbe panel composed of Atopobium vaginae, BVAB-2, and Megasphaera-1. Each microbe is quantified and given a score of 0 (low), 1 (moderate), or 2 (high). The total score is then interpreted as BV-POS (3–6), BV-IND (2), or BV-NEG (0–1). (b) 16S relative abundances of NuSwab panel microbes stratified by their scores. Statistical comparisons of 16S relative abundances were made between samples with NuSwab scores of 2 and those with scores of 0 or 1 using Wilcoxon rank-sum tests with FDR control.

Fig 2

Taxonomic and phylogenetic analysis of 16S V3–V4 amplicon sequence variants (ASVs). (a) Stacked bar plot of class relative abundances colored consistently with (b) and with samples stratified by BV status. (b) ML phylogeny of all ASVs, with branches colored by class. The Bacilli class was highlighted and a phylogeny constructed. (c) ML phylogeny of ASVs within the Bacilli class shown in (b) with branches colored by genus and tips labeled with the lowest taxonomic classifications. Neighboring tips with the same label were aggregated into a single label.

Fig 3

CST analysis of vaginal microbiomes using CST definitions described by Ravel et al. (14) (a) Stacked bar plot of key genera and species’ relative abundances across CSTs I, II, III, IV, and V. Samples are stratified by CSTs (solid vertical lines) and then by BV status (dashed vertical lines). (b) Multidimensional scaling scatter plot of vaginal microbiome Bray–Curtis dissimilarities with data points colored by CST and shaped by BV status. (c) Boxplots of vaginal microbiome diversities as measured by Shannon index, stratified first by CST, and then by BV status with data points shaped and colored consistently with those in (b).

Fig 4

Differential abundance (DA) analysis and modularized co-occurrence network analysis of BV-POS and BV-NEG samples. (a) Volcano plot of the differential abundance results, depicting the log2 fold-change (L2FC) (x-axis, BV-POS vs BV-NEG), and the -log10(P) (y-axis) for each taxa assessed. FDR control for multiple testing was used to calculate P values. Data point colors represent statistical significance of taxa, with red representing enriched taxa (P < 0.05, L2FC >= 1), blue representing depleted taxa (P < 0.05, L2FC <= −1), and yellow representing all other taxa without significant changes. (b) Boxplots comparing the relative abundances of select taxa between BV-POS and BV-NEG samples, with statistical significance indicated to the right (*** P < 0.001). Unique identification of taxa is represented by only a single boxplot shown (e.g., Parvimonas exclusively detected in BV-NEG samples). (c) Clustering results from C3NA (32), with the first column representing clustering among the BV-NEG samples (n = 25 clusters), followed by the modular preservation ribbon linked to the colored BV-POS clusters (n = 19). The arc on the right represents the Spearman correlations above 0.2 among each BV-POS cluster, with each node colored by the DA result (enriched, neutral, and depleted) and the edge colored by the correlation magnitude. All testing was performed on the species level with any nonspeciated taxa labeled at the highest resolved taxonomic levels, i.e., "g_DNF000809" represents the ASVs that resolved to the DNF000809 genus without species assignment.

Fig 5

Heatmap of differentially abundant (DA) predicted metabolic pathways between BV-POS and BV-NEG samples found in CST-I, CST-III, and CST-IV. Rows represent MetaCyc (35) metabolic pathways and are ordered via hierarchical clustering, while columns (samples) are stratified first by CST (solid vertical lines) and then by BV status (dashed vertical lines). The heatmap combines the results of four pairwise comparisons, namely (i) CST-I BV-NEG vs CST-III BV-NEG; (ii) CST-I BV-NEG vs CST-IV BV-POS; (iii) CST-III BV-NEG vs CST-IV BV-POS; and (iv) CST-IV BV-POS vs CST-IV BV-NEG. The top 20 differentially abundant pathways of each comparison with the greatest variance were selected and merged into the list of significant pathways across all pairs of sample groups (combined set of 45 pathways shown).