Microbiome Composition and Function Drives Wound-Healing Impairment in the Female Genital Tract

Abstract

The mechanism(s) by which bacterial communities impact susceptibility to infectious diseases, such as HIV, and maintain female genital tract (FGT) health are poorly understood. Evaluation of FGT bacteria has predominantly been limited to studies of species abundance, but not bacterial function. We therefore sought to examine the relationship of bacterial community composition and function with mucosal epithelial barrier health in the context of bacterial vaginosis (BV) using metaproteomic, metagenomic, and in vitro approaches. We found highly diverse bacterial communities dominated by Gardnerella vaginalis associated with host epithelial barrier disruption and enhanced immune activation, and low diversity communities dominated by Lactobacillus species that associated with lower Nugent scores, reduced pH, and expression of host mucosal proteins important for maintaining epithelial integrity. Importantly, proteomic signatures of disrupted epithelial integrity associated with G. vaginalis-dominated communities in the absence of clinical BV diagnosis. Because traditional clinical assessments did not capture this, it likely represents a larger underrepresented phenomenon in populations with high prevalence of G. vaginalis. We finally demonstrated that soluble products derived from G. vaginalis inhibited wound healing, while those derived from L. iners did not, providing insight into functional mechanisms by which FGT bacterial communities affect epithelial barrier integrity.

Fig 1. Mass spectrometry and 16S sequence analysis of cervicovaginal secretions show two distinct bacterial groups dominated by Lactobacillus (G1) or Gardnerella (G2) in two distinct cohorts.

(A) Bacterial genera distributions detected in Cohort 1 using MS. Bacterial abundances from the MS data were calculated by summing normalized total spectral counts for all proteins associated with a genus/species. Clustering of samples was performed using unsupervised hierarchical linkage with average Euclidean distances of the proportional bacterial abundance in each sample. (B) Bacterial genera distributions for Cohort 1 detected using 16S rRNA gene sequencing shows close concordance to MS, and clustered into two groups based on Lactobacillus (G1) or Gardnerella (G2) dominance. Bacterial genera detected in Cohort 2 using MS (C) showed similar grouping as Cohort 1. Nugent scores (top) and bacterial alpha diversity (Shannon’s H index, bottom) are also shown.

Fig 2. Cervicovaginal bacterial profiles display diverging functional traits.

Bacterial proteome functions for both cohorts were determined by KEGG ontology. For both cohorts, G1 profiles (Lactobacillus dominant) showed increased abundance of proteins related to Transport and Catabolism and Energy Metabolism while G2 (Gardnerella dominant) showed increased abundances of proteins associated with Membrane Transport. Student’s t-tests were used to compare the abundances of each category between G1 and G2.

Fig 3. Vaginal epithelial barrier proteins vary significantly between microbial groups and are considerably lower in Gardnerella (G2) vs. Lactobacillus (G1) dominant profiles.

(A) Heatmap of differentially abundant host proteins found in common between Cohort 1 (longitudinal) and Cohort 2 (cross-sectional) based on cervicovaginal microbiome profiles. (B) Functional annotation of differentially expressed proteins based on their gene ontology. Fisher’s exact test was used to compare annotated protein functions between G1 and G2 across both cohorts. (C) Lowered cornified envelope proteins (Involucrin and Small Proline-Rich Protein 1A) strongly correlate with G. vaginalis (G2) profiles, regardless of the presentation of clinical BV symptoms for both cohorts. Student’s t-test was used to compare the abundances of INVO and SPR1A between G1 and G2. BP: Biological process; CC: Cellular component.

Fig 4. Bacterial proteins significantly correlate with vaginal epithelial barrier markers.

Spearman correlations between (A) Lactobacillus proteins vs. INVO (B) Lactobacillus proteins vs. SPR1A; (C) G. vaginalis proteins vs. INVO; (D) G. vaginalis proteins vs. SPR1A. (SPR1A: Small proline-rich protein 1A; INVO, Involucrin) for both cohorts. Only proteins that passed multiple comparison correction testing in at least one comparison are shown (Bonferonni, P < 0.0003).

Fig 5. G. vaginalis-derived soluble products inhibit wound healing of cervical cell monolayers.

(A) Representative light micrographs show that scratches in HeLa cell monolayers healed fully in the presence of NYCIII medium and L. iners, culture supernatants while HeLa cell monolayers exposed to G. vaginalis culture supernatants did not heal fully or expanded. (B) Statistical analysis of replicate co-culture experiments. Statistical significance was calculated using ANOVA with Bonferonni correction. ***P<0.001.