Distinct cervical tissue-adherent and luminal microbiome communities correlate with mucosal host gene expression and protein levels in Kenyan sex workers

Abstract

Background: The majority of studies characterizing female genital tract microbiota have focused on luminal organisms, while the presence and impact of tissue-adherent ectocervical microbiota remain incompletely understood. Studies of luminal and tissue-associated bacteria in the gastrointestinal tract suggest that these communities may have distinct roles in health and disease. Here, we performed a multi-omics characterization of paired luminal and tissue samples collected from a cohort of Kenyan female sex workers.

Results: We identified a tissue-adherent bacterial microbiome, with a higher alpha diversity than the luminal microbiome, in which dominant genera overall included Gardnerella and Lactobacillus, followed by Prevotella, Atopobium, and Sneathia. About half of the L. iners-dominated luminal samples had a corresponding Gardnerella-dominated tissue microbiome. Broadly, the tissue-adherent microbiome was associated with fewer differentially expressed host genes than the luminal microbiome. Gene set enrichment analysis revealed that L. crispatus-dominated tissue-adherent communities were associated with protein translation and antimicrobial activity, whereas a highly diverse microbial community was associated with epithelial remodeling and pro-inflammatory pathways. Tissue-adherent communities dominated by L. iners and Gardnerella were associated with lower host transcriptional activity. Tissue-adherent microbiomes dominated by Lactobacillus and Gardnerella correlated with host protein profiles associated with epithelial barrier stability, although with a more pro-inflammatory profile for the Gardnerella-dominated microbiome group. Tissue samples with a highly diverse composition had a protein profile representing cell proliferation and pro-inflammatory activity.

Conclusion: We identified ectocervical tissue-adherent bacterial communities in all study participants of a female sex worker cohort. These communities were distinct from cervicovaginal luminal microbiota in a significant proportion of individuals. We further revealed that bacterial communities at both sites correlated with distinct host gene expression and protein levels. The tissue-adherent bacterial community could possibly act as a reservoir that seed the lumen with less optimal, non-Lactobacillus, bacteria. Video Abstract.

Keywords: 16S rRNA gene; Biofilm; Cervix; Ectocervix; Luminal; Microbiota; Protein profiling; Tissue; Tissue-adherent; Transcriptomics.

Fig. 1

Characterization of a highly diverse microbiome in cervicovaginal (luminal) samples from Kenyan sex workers. Cervicovaginal lavage (luminal) and ectocervical tissue study samples were assessed by 16S rRNA sequencing, gene expression, and protein profiling. a Schematic drawing depicting the sampling scheme and the resulting omics datasets. b Bar plots of alpha diversity indices and taxonomy profiles for each individual luminal sample. Color-coded squares above the stacked bar plots indicate bacterial vaginosis (BV, binned Nugent’s scores): Gray: negative, orange: intermediate, red: positive; and HIV diagnosis: Gray: HIV seronegative, red: HIV seropositive. Two12-SNN graphs were constructed using: c Louvain community detection algorithm, and d Uniform Manifold Approximation (UMAP). The two graphs were overlayed in color with the predefined luminal study groups. The undirected edges are included in gray connecting the nodes

Fig. 2

Identification of bacterial communities and functional profiles in the luminal samples. The luminal study samples were assessed for bacterial communities and functional profiles. a Bacterial community embedding of 5-SNN graph clustered using Louvain community detection algorithm based on luminal bacterial abundances. b Wet smear counts of the genera Lactobacillus, Mobiluncus, and Gardnerella with corresponding 16S read counts. c Differential expression analysis was applied to PICRUSt2 predicted KO terms across the five luminal study groups. Resulting significant (FDR < 1 × 10⁻⁵) KO terms were divided into seven modules by hierarchical agglomerative clustering using inverse Pearson’s correlation as distance measure and Ward’s method (“ward. D2”) for linkage. Enrichment analysis was performed on each module and the three most significant KEGG pathways were included in the heatmap. d Uniform Manifold Approximation (UMAP) of the predicted KO terms

Fig. 3

Identification of a distinct ectocervical tissue-adherent microbiome. Ectocervical tissue samples were assessed for presence of a tissue-adherent microbiome. a Bar plots of alpha diversity indices and taxonomy profiles for each individual tissue sample. Color-coded squares above the stacked bar plots show bacterial vaginosis (BV, binned Nugent’s scores) and HIV diagnosis, respectively. Gray: negative, orange: intermediate, red: positive BV; Gray: HIV seronegative, red: HIV seropositive. b Total relative abundance in the luminal and tissue microbiome datasets. All taxa with a total relative abundance < 0.55 are included in the “other” category. c Microbiome profile shift between luminal and tissue samples

Fig. 4

Characterization of the host transcriptome as stratified by the luminal microbiome study groups. The luminal samples were assessed for differential gene expression across the study groups. a Differential gene expression analysis was applied across the five luminal study groups. Significant DEGs (p-value < 0.01) were divided into six modules by hierarchical agglomerative clustering using inverse Pearson’s correlation as distance measure and Ward’s method (“ward. D2”) for linkage. Enrichment analysis was performed on each module using both the KEGG and GO databases. The three most significant terms were included in the heatmap. b Pairwise enrichment analysis of protein-protein interactions of transcription factors (TF-PPI). Top-10 up- and downregulated transcription factors with p-value < 0.01 were included in the bar plots

Fig. 5

Characterization of the host transcriptome as stratified by the tissue-adherent microbiome study groups. The tissue samples were assessed for differential gene expression across the study groups. a Differential gene expression analysis was applied across the five tissue study groups. Significant DEGs (p-value < 0.01) were divided into five modules by hierarchical agglomerative clustering using inverse Pearson’s correlation as distance measure and Ward’s method (“ward. D2”) for linkage. Enrichment analysis was performed on each module using both the KEGG and GO databases. The three most significant terms were included in the heatmap. d Pairwise enrichment analysis of protein-protein interactions of transcription factors (TF-PPI). Top-10 up- and downregulated transcription factors with p-value < 0.01 were included in the bar plots

Fig. 6

Characterization of the host protein profile as stratified by the luminal and tissue-adherent microbiome study groups. All study samples were assessed for significant differences in protein levels across the study groups. a,b The five luminal study groups and c,d the five corresponding tissue groups. a, c Proteins with a p-value > 0.05 were omitted from the heatmap. Proteins with significantly different levels across the groups were clustered by hierarchical agglomerative clustering using inverse Pearson’s correlation as distance measure and Ward’s method (“ward. D2”) for linkage. Color-coded rows below the heatmaps show clinical diagnosis of bacterial vaginosis (BV, binned Nugent’s scores) and HIV diagnosis, respectively. Color coding for BV: Gray: negative, orange: intermediate, red: positive; and for HIV: Gray: HIV seronegative, red: HIV seropositive. b, d Violin plots of log₂ transformed cytokine levels across the luminal and tissue study groups (as indicated on the x-axis), respectively. The asterisk and lines indicate statistically significant results of Dunn’s test with Benjamini Hochberg’s correction analysis. Adjusted p-values: * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001