Reproductive tract microbiota dysbiosis in ovarian endometrioma and adenomyosis: multi-site 16S rRNA profiling and functional impact of key bacterial species on human endometrial stromal cells

The analysis of the mock community demonstrated high fidelity of our pipeline. All expected bacterial genera were detected, and their observed relative abundances showed a strong correlation (Pearson’s r  > 0.99) with the theoretical abundances (Supplementary Figure S2). No non-target genera were detected above a threshold of 0.1%, indicating minimal contamination or sample cross-talk. These results confirm that our experimental and bioinformatic workflow accurately represents microbial community structure. Furthermore, rarefaction curves for all samples approached a plateau, indicating that the sequencing depth was sufficient to capture the majority of the microbial diversity present (Supplementary Figure S1). 16S rRNA gene sequencing data were generated from samples belonging to the three study groups (CC: Ovarian Chocolate Cysts; AM: Adenomyosis; NC: Normal Control) collected from four anatomical sites (cervical canal, posterior fornix, ascites fluid, endometrium). Analysis of these data revealed differences in microbial community composition and diversity among the groups within these specific locations. (Hereafter, the suffixes ‘a, ‘b, ‘c’, and d ' appended to group names denote samples from the cervical canal, posterior fornix, ascites fluid, and endometrium, respectively.). Venn diagrams displayed the counts of Amplicon Sequence Variants (ASVs) for each group and the overlap between compared groups at each anatomical site (Fig. 1 A). Specifically, in the cervical canal (site ‘a’), the AMa group possessed 9793 ASVs and the NCa group possessed 4058 ASVs, sharing 500 ASVs; the CCa group possessed 14117 ASVs and the NCa group possessed 2826 ASVs, sharing 303 ASVs. In the posterior fornix (site ‘b’), the AMb group possessed 19290 ASVs and the NCb group possessed 1454 ASVs, sharing 216 ASVs; the CCb group possessed 9350 ASVs and the NCb group possessed 1192 ASVs, sharing 192 ASVs. Within the ascites fluid (site ‘c’), the AMc group possessed 32166 ASVs and the NCc group possessed 8430 ASVs, sharing 1101 ASVs; the CCc group possessed 29152 ASVs and the NCc group possessed 3561 ASVs, sharing 600 ASVs. Finally, in the endometrium (site d), the AMd group possessed 25046 ASVs and the NCd group possessed 8119 ASVs, sharing 888 ASVs; the CCd group possessed 24333 ASVs and the NCd group possessed 4380 ASVs, sharing 562 ASVs. Overall, 940 ASVs were shared among the AM, CC, and NC groups across all sites. These findings visually represent the alterations in microbial composition associated with AM and CC across the sampled reproductive tract locations (Fig. 1 A). Fig. 1 Comparison of ASV counts and alpha diversity across anatomical sites among AM, CC, and NC groups. A Venn diagram illustrating the number of shared and unique Amplicon Sequence Variants (ASVs) between AM vs. NC groups and CC vs. NC groups within the cervical canal, posterior fornix, ascites fluid, and endometrium. Different colored ellipses represent different sample groups; overlapping areas indicate shared ASVs, while non-overlapping areas represent unique ASVs for each group. The substantial variation in ASV counts between groups reflects both biological differences in microbial community complexity and the stringent quality control measures applied prior to diversity analyses. B Bar charts comparing alpha diversity indices (ACE, Chao1, PD_whole_tree, Shannon, Simpson) between AM vs. NC groups and CC vs. NC groups across the four anatomical sites. Statistical significance is indicated where applicable ( P  < 0.05) Comparison of ASV counts and alpha diversity across anatomical sites among AM, CC, and NC groups. A Venn diagram illustrating the number of shared and unique Amplicon Sequence Variants (ASVs) between AM vs. NC groups and CC vs. NC groups within the cervical canal, posterior fornix, ascites fluid, and endometrium. Different colored ellipses represent different sample groups; overlapping areas indicate shared ASVs, while non-overlapping areas represent unique ASVs for each group. The substantial variation in ASV counts between groups reflects both biological differences in microbial community complexity and the stringent quality control measures applied prior to diversity analyses. B Bar charts comparing alpha diversity indices (ACE, Chao1, PD_whole_tree, Shannon, Simpson) between AM vs. NC groups and CC vs. NC groups across the four anatomical sites. Statistical significance is indicated where applicable ( P  < 0.05) It is noteworthy that substantial differences in ASV counts were observed between groups (e.g., 19,290 ASVs in AMb versus 1,454 ASVs in NCb), which may reflect genuine changes in microbial community complexity associated with disease states, as well as the diversity of microbial ecological niches across different anatomical sites. All subsequent diversity analyses were conducted using stringently filtered ASVs to ensure data quality and analytical reliability. Site-specific differences in alpha diversity between the study groups were evaluated using multiple indices (Fig. 1 B). When comparing the CC group to the NC group, statistically significant differences ( p  < 0.05) were observed only in the ascites fluid (CCc vs. NCc), specifically for the richness indices Chao1 and Ace. No other significant differences in alpha diversity were found between the CC and NC groups at any other site or using other indices. Comparisons between the AM group and the NC group revealed more varied results depending on the site. In the cervical canal (AMa vs. NCa), all calculated alpha diversity indices (Chao1, Ace, PD_whole_tree, Shannon, Simpson) showed statistically significant differences ( p  < 0.05). Similarly, in the posterior fornix (AMb vs. NCb), Chao1, Ace, PD_whole_tree, and Shannon indices differed significantly ( p   0.05). In contrast, no significant differences in any alpha diversity index were detected between the AM and NC groups in the ascites fluid (AMc vs. NCc) ( p  > 0.05). Finally, within the endometrium (AMd vs. NCd), only the phylogenetic diversity index (PD_whole_tree) showed a statistically significant difference ( p  < 0.05) between the AM and NC groups. These findings highlight complex, site-dependent alterations in microbial community richness, evenness, and phylogenetic diversity associated with AM and CC compared to controls. Beta diversity, assessed using Principal Coordinate Analysis (PCoA) based on Bray-Curtis distances, demonstrated discernible clustering patterns, suggesting differences in overall microbial community structure between the NC, CC, and AM groups at most sites (Fig. 2 A). To statistically test these differences, a permutational multivariate analysis of variance (PERMANOVA) was performed. The analysis revealed significant separation between groups in the cervical canal (PERMANOVA, p  < 0.05) and posterior fornix (PERMANOVA, p   0.05) or endometrial samples ( p  > 0.05), thus providing statistical support for the visual interpretation of the PCoA plots. Fig. 2 Beta diversity analysis and bacterial community composition at the genus level. A PCoA plot based on Bray-Curtis distances comparing the overall microbial community structure between AM, CC, and NC groups within the cervical canal, posterior fornix, ascites fluid, and endometrium. Ellipses represent the 95% confidence interval for each group. The significance of group separation was statistically assessed using PERMANOVA. B Stacked bar charts showing the relative abundance of the top bacterial genera within the cervical canal, posterior fornix, ascites fluid, and endometrium for AM, CC, and NC groups Beta diversity analysis and bacterial community composition at the genus level. A PCoA plot based on Bray-Curtis distances comparing the overall microbial community structure between AM, CC, and NC groups within the cervical canal, posterior fornix, ascites fluid, and endometrium. Ellipses represent the 95% confidence interval for each group. The significance of group separation was statistically assessed using PERMANOVA. B Stacked bar charts showing the relative abundance of the top bacterial genera within the cervical canal, posterior fornix, ascites fluid, and endometrium for AM, CC, and NC groups Analysis of relative abundance at the genus level revealed significant compositional shifts associated with AM and CC (Fig. 2 B). In cervical canal samples (CCa vs. NCa), Lactobacillus and Escherichia-Shigella were significantly more abundant in the CCa group, while Gardnerella and Staphylococcus were significantly reduced compared to NCa ( p  < 0.05). In posterior fornix samples (CCb vs. NCb), Staphylococcus , Escherichia-Shigella , and Alloscardovia were significantly elevated in the NCb group, whereas Gardnerella , Streptococcus , and Ureaplasma were more abundant in the CCb group ( p  < 0.05). Comparing ascites fluid samples (CCc vs. NCc), Dietzia and Psychrobacter showed higher relative abundance in CCc, while NCc was enriched in Escherichia-Shigella , Bacteroides , Candidatus_Solibacter , and Sodalis ( p  < 0.05). In endometrial samples (CCd vs. NCd), the NCd group exhibited significantly higher abundances of Lactobacillus , Escherichia-Shigella , Helicobacter , and Bacteroides , while the CCd group had higher abundances of Dietzia , Psychrobacter , and unclassified Bacteria ( p  < 0.05). Similar comparisons between AM and NC groups also revealed distinct patterns (Fig. 2 B). In the cervical canal (AMa vs. NCa), NCa samples had higher abundances of Lactobacillus and Gardnerella , while AMa samples were significantly enriched in Escherichia-Shigella , Haemophilus , Klebsiella , and Enterococcus ( p  < 0.05). In the posterior fornix (AMb vs. NCb), NCb again showed higher Lactobacillus and Streptococcus , whereas AMb had significantly increased abundances of Escherichia-Shigella , Klebsiella , Enterococcus , and Haemophilus ( p  < 0.05). Comparisons in ascites fluid (AMc vs. NCc) and endometrium (AMd vs. NCd) also demonstrated site-specific differences, generally indicating higher relative abundances of genera often associated with dysbiosis in AM samples compared to NC samples at several sites ( p  < 0.05). Collectively, these findings highlight that both AM and CC are associated with significant alterations in the relative abundance of key bacterial genera across different locations within the female reproductive tract. Robust Identification of Differentially Abundant Taxa in the Posterior Fornix To robustly identify specific taxa with significantly different abundances between the adenomyosis (AM) and normal control (NC) groups, we applied the compositionally-aware tool ANCOM-BC to the posterior fornix samples. The analysis revealed that four Amplicon Sequence Variants (ASVs) were significantly more abundant in the AM group… (Prevotella bivia, Gardnerella vaginalis, etc.). A volcano plot visualizing these changes is provided in Supplementary Figure S3, and a heatmap illustrating their abundance is provided in Supplementary Figure S4. Linear discriminant analysis Effect Size (LEfSe) identified specific taxa with significantly different relative abundances between groups (LDA score > 2.5, p  < 0.05), highlighting potential biomarkers at each site (Fig. 3 A). In the cervical canal (site ‘a’), the genus Lactobacillus was a distinguishing feature of the CCa group relative to NCa. Within the posterior fornix (site ‘b’), the family Enterococcaceae and the genus Enterococcus showed significant enrichment in the AMb group relative to NCb; however, no significant differences were found between the CCb and NCb groups at this site. In ascites fluid (site ‘c’), taxa distinguishing the CCc group from NCc included the genus Dietzia enriched in CCc and the family Comamonadaceae enriched in NCc. Further comparisons at this site revealed enrichment of the family Muribaculaceae in NCc relative to AMc, and Comamonadaceae enrichment in AMc relative to NCc. Lastly, within the endometrium (site d), the family Lactobacillaceae was characteristic of the NCd group when compared to both CCd and AMd, whereas unclassified Bacteria were significantly enriched in the AMd and CCd groups relative to NCd. Fig. 3 Differential abundance analysis (LEfSe) and predicted functional pathway enrichment. A Linear discriminant analysis Effect Size (LEfSe) plots identifying differentially abundant taxa (from family to genus level) characterizing each group (AM vs. NC; CC vs. NC) within the specified anatomical sites (cervical canal, posterior fornix, ascites fluid, endometrium). Bars indicate taxa enriched in the corresponding group. B Predicted KEGG pathway enrichment analysis comparing AM vs. NC groups within the posterior fornix and endometrium. Plots show the mean proportion of sequences assigned to significantly different pathways ( P < 0.05, corrected) and the difference between proportions with 95% confidence intervals Differential abundance analysis (LEfSe) and predicted functional pathway enrichment. A Linear discriminant analysis Effect Size (LEfSe) plots identifying differentially abundant taxa (from family to genus level) characterizing each group (AM vs. NC; CC vs. NC) within the specified anatomical sites (cervical canal, posterior fornix, ascites fluid, endometrium). Bars indicate taxa enriched in the corresponding group. B Predicted KEGG pathway enrichment analysis comparing AM vs. NC groups within the posterior fornix and endometrium. Plots show the mean proportion of sequences assigned to significantly different pathways ( P < 0.05, corrected) and the difference between proportions with 95% confidence intervals To explore the potential functional implications of the observed microbial changes, we performed predictive analysis of community metabolic potential using PICRUSt2 based on KEGG orthology (Fig. 3 B). Importantly, these predicted results represent computationally inferred metabolic possibilities based on 16 S marker genes, not direct measurements of functional activity. The predictive analysis suggested potential enrichment ( p  < 0.05) in pathways related to the ‘Metabolism of terpenoids and polyketides’ within the posterior fornix communities of the AM group (AMb) compared to the NC group. Additionally, for endometrial communities (AMd vs. NCd), the algorithm predicted a higher potential for pathways including ‘Infectious diseases: Viral,’ ‘Cancers: Specific types,’ and ‘Cardiovascular diseases’ in the AM group. These predictions, while speculative, provided a rationale to further investigate the direct functional impact of these distinct microbial communities on host cells. To investigate the functional impact of altered microbiota, several bacterial taxa were identified as different between groups. Excluding those potentially unobtainable for culture, the following representative strains were selected for co-culture experiments with human endometrial stromal cells (T-HESC): Lactobacillus helveticus (NC-associated), Enterococcus faecalis (AM-associated), and Enterobacteriaceae (CC-associated). Morphological assessment via brightfield microscopy revealed alterations and apparent reductions in cell density when T-HESC cells were co-cultured with AM- and CC-associated bacteria, compared to the NC-associated bacterium or the T-HESC only control (Fig. 4 A). Cell viability, quantified by CCK8 assay, was significantly decreased ( p  < 0.05) in T-HESC cells co-cultured with E. faecalis (AM) or Enterobacteriaceae (CC) relative to those co-cultured with L. helveticus (NC) (Fig. 4 B). Viability in the NC group did not differ significantly from the T-HESC only control ( p  > 0.05). Fig. 4 Effects of CC-associated bacteria (Enterobacteriaceae) on T-HESC cells and transcriptomic analysis compared to the NC group (L. helveticus). A Brightfield microscopy image showing T-HESC morphology under different co-culture conditions (T-HSC only, NC, AM, CC). B Statistical comparison of T-HESC viability (measured by CCK-8 assay) among various treatment groups. Data presented as mean ± SD or similar; significance indicated. Principal Component Analysis (PCA) plot of transcriptomic data from T-HESCs co-cultured with NC-, AM-, and CC-associated bacteria. D Volcano plot (left) showing differentially expressed genes (DEGs) between CC and NC co-culture groups (Red: upregulated, Blue: downregulated); Heatmap (right) visualizing the expression patterns of key DEGs clustered by expression profile across samples. E KEGG pathway enrichment analysis of DEGs between CC and NC groups. F GO term enrichment analysis (enriched terms for Biological Process - BP, Cellular Component - CC, Molecular Function - MF) of DEGs between CC and NC groups Effects of CC-associated bacteria (Enterobacteriaceae) on T-HESC cells and transcriptomic analysis compared to the NC group (L. helveticus). A Brightfield microscopy image showing T-HESC morphology under different co-culture conditions (T-HSC only, NC, AM, CC). B Statistical comparison of T-HESC viability (measured by CCK-8 assay) among various treatment groups. Data presented as mean ± SD or similar; significance indicated. Principal Component Analysis (PCA) plot of transcriptomic data from T-HESCs co-cultured with NC-, AM-, and CC-associated bacteria. D Volcano plot (left) showing differentially expressed genes (DEGs) between CC and NC co-culture groups (Red: upregulated, Blue: downregulated); Heatmap (right) visualizing the expression patterns of key DEGs clustered by expression profile across samples. E KEGG pathway enrichment analysis of DEGs between CC and NC groups. F GO term enrichment analysis (enriched terms for Biological Process - BP, Cellular Component - CC, Molecular Function - MF) of DEGs between CC and NC groups RNA sequencing (RNA-seq) was performed to characterize the transcriptomic response of T-HESC cells to co-culture with the different representative bacteria. Principal Component Analysis (PCOA) of the resulting expression data demonstrated clear separation between the transcriptomic profiles of cells exposed to AM-, CC-, and NC-associated bacteria, indicating distinct gene expression programs induced by each condition (Fig. 4 C). Differential gene expression analysis between the CC and NC co-culture conditions identified 9155 DEGs (FDR  2), comprising 4344 upregulated and 4811 downregulated genes (Fig. 4 D). Notably upregulated genes included ADH1B, MYOM1, and PTGFR, whereas ACLY was significantly downregulated ( p  < 0.05). KEGG pathway analysis revealed significant enrichment ( p  < 0.05) of these DEGs in pathways such as Mismatch repair, Insulin signaling, Human papillomavirus infection, Metabolic pathways, Autophagy, and DNA replication (Fig. 4 E). Gene Ontology (GO) analysis showed enrichment ( p  < 0.05) in Biological Processes (BP), including DNA replication and cell cycle regulation; Cellular Components (CC) like nuclear membrane and Golgi apparatus; and Molecular Functions (MF) such as ATP hydrolysis activity and transcription factor binding (Fig. 4 F). Similarly, comparison between the AM and NC co-culture conditions yielded 8517 DEGs (4244 upregulated, 4273 downregulated) (Fig. 5 A). Among the most highly upregulated genes were CEBPB and SERPINE1, while S100A1 and CKS1B featured prominently among the downregulated genes ( p  < 0.05) (Fig. 5 B). KEGG pathway analysis indicated significant enrichment ( p  < 0.05) in Metabolic pathways, Pathways in cancer, Human papillomavirus infection, and the MAPK signaling pathway (Fig. 5 C). GO terms enriched ( p  < 0.05) in the AM group relative to NC included BP associated with small GTPase signaling, Wnt signaling, and embryonic development; CC, like Golgi apparatus and lysosomal membrane; and MF related to actin binding and GTPase binding (Fig. 5 D). Fig. 5 Transcriptomic analysis of T-HESC cells co-cultured with AM-associated bacteria (E. faecalis) compared to the NC group (L. helveticus). A Volcano plot showing DEGs between AM and NC co-culture groups (Red: upregulated, Blue: downregulated). B Heatmap visualizing the expression patterns of key DEGs clustered by expression profile across AM and NC samples. CKEGG pathway enrichment analysis of DEGs between the AM and NC groups. D GO term enrichment analysis (enriched terms for BP, CC, MF) of DEGs between AM and NC groups Transcriptomic analysis of T-HESC cells co-cultured with AM-associated bacteria (E. faecalis) compared to the NC group (L. helveticus). A Volcano plot showing DEGs between AM and NC co-culture groups (Red: upregulated, Blue: downregulated). B Heatmap visualizing the expression patterns of key DEGs clustered by expression profile across AM and NC samples. CKEGG pathway enrichment analysis of DEGs between the AM and NC groups. D GO term enrichment analysis (enriched terms for BP, CC, MF) of DEGs between AM and NC groups To test for a functional relationship between the reproductive tract microbiota and host endometrial cell responses, we performed a multi-omics correlation analysis. A Mantel test revealed a highly significant and strong negative correlation between the microbial community composition (Bray-Curtis distance) and the host gene expression profiles (Euclidean distance) (Mantel’s r = −0.9929, p  < 0.001)(Supplementary Figure S5). This indicates that samples with similar microbial profiles also tend to have similar host gene expression patterns, suggesting a strong association between the two. Key DEGs identified by RNA-seq were selected for validation at both the mRNA and protein levels using qRT-PCR and ELISA, respectively. The qRT-PCR results substantiated the transcriptomic data (Fig. 6 A-B). Specifically, relative to the NC group, mRNA levels of ADH1B, MYOM1, and PTGFR were significantly elevated ( p  < 0.05) in the CC group, while ACLY mRNA was significantly reduced ( p  < 0.05). In parallel, mRNA levels of CEBPB and SERPINE1 were significantly increased ( p  < 0.05) in the AM group compared to NC, whereas S100A1 and CKS1B expression were significantly downregulated ( p  < 0.05). Fig. 6 Validation of key differentially expressed genes (DEGs) at mRNA and protein levels. A-B qRT-PCR validation of relative mRNA expression levels for selected genes (CEBPB, SERPINE1, S100A1, CKS1B, ADH1B, MYOM1, PTGFR, ACLY) in T-HESCs comparing AM vs. NC and CC vs. NC co-culture conditions. Expression normalized to GAPDH. C-D ELISA validation of protein concentrations for corresponding key molecules in cell culture supernatants, comparing AM vs. NC and CC vs. NC co-culture conditions. Data presented as mean ± SD or similar; significance indicated ( P  < 0.05) Validation of key differentially expressed genes (DEGs) at mRNA and protein levels. A-B qRT-PCR validation of relative mRNA expression levels for selected genes (CEBPB, SERPINE1, S100A1, CKS1B, ADH1B, MYOM1, PTGFR, ACLY) in T-HESCs comparing AM vs. NC and CC vs. NC co-culture conditions. Expression normalized to GAPDH. C-D ELISA validation of protein concentrations for corresponding key molecules in cell culture supernatants, comparing AM vs. NC and CC vs. NC co-culture conditions. Data presented as mean ± SD or similar; significance indicated ( P  < 0.05) ELISA measurements of protein concentrations in the cell culture supernatants further confirmed these expression changes (Fig. 6 C-D). Compared to the NC group, protein levels of ADH1 B, MYOM1, and PTGFR were significantly higher ( p  < 0.05) in the CC group supernatant, while ACLY protein levels were significantly reduced ( p  < 0.05). Correspondingly, CEBPB and SERPINE1 protein concentrations were significantly higher ( p  < 0.05) in the AM group supernatant relative to NC, whereas S100A1 and CKS1B protein levels were significantly lower ( p  < 0.05). These validation data confirm that exposure to bacteria representative of AM and CC conditions elicits robust and consistent alterations in the expression of specific genes and their corresponding proteins in endometrial stromal cells.

Sixty participants were enrolled: 20 with ovarian chocolate cysts (CC), 20 with adenomyosis (AM) (diagnoses confirmed pathologically/surgically), and 20 controls (NC) undergoing surgery for benign conditions. Exclusion criteria included systemic diseases, recent hormone/antibiotic use, pregnancy, or lactation. Clinical data (age, BMI, menarche age, menstrual volume, dysmenorrhea VAS, infertility duration) were recorded. The study was approved by the Yunnan First People’s Hospital Clinical Research Ethics Committee (Approval No. KHLL2025-KY114), and written informed consent was obtained from all participants. Pre-operatively, posterior fornix and endocervical canal secretions were collected using sterile swabs. Intra-operatively, endometrial tissue and pelvic peritoneal fluid (via sterile saline lavage) were obtained aseptically. Samples were immediately stored at −80 °C. Total genomic DNA was extracted from all samples using a standardized commercial kit following the manufacturer’s protocol. The V3-V4 hypervariable regions of the bacterial 16S rRNA gene were amplified using PCR with region-specific primers tagged with unique barcode sequences. Amplification products were purified, quantified, and pooled in equimolar ratios to construct sequencing libraries. Library quality was verified, and libraries meeting quality standards were sequenced on an Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA). The raw sequencing data for both the 16S rRNA and transcriptome analyses described in this study have been deposited in the NCBI BioProject database under accession number PRJNA1268694( https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1268694/ ). To assess the accuracy of our sequencing and bioinformatics pipeline, a commercially available mock microbial community (ZymoBIOMICS Microbial Community Standard, Zymo Research) with a known theoretical composition was included as a positive control. This mock community DNA was processed, amplified, and sequenced in parallel with all experimental samples using the exact same protocols. Raw reads were quality-filtered (Trimmomatic v0.33) and primers removed (Cutadapt v1.9.1). ASVs were generated using DADA2 (QIIME 2 v2020.6). Taxonomy was assigned against the SILVA database (v138). Prior to diversity analyses, low-abundance ASVs were filtered from the dataset; specifically, any ASV that was not present in at least two samples or did not have a total abundance of at least 10 reads across all samples was removed. Alpha (Chao1, ACE, Shannon, Simpson, Faith’s PD) and beta diversity (Bray-Curtis, PCoA) were calculated. Prior to beta diversity analysis, the ASV table was rarefied to an even sequencing depth of 5,000 reads per sample to normalize for differences in library size. Principal Coordinate Analysis (PCoA) was then performed on the resulting Bray-Curtis dissimilarity matrix to visualize clustering between groups. To identify differentially abundant taxa while addressing the potential for compositional bias in microbiome data, we employed a robust consensus approach combining three distinct methods. In addition to the standard Linear Discriminant Analysis Effect Size (LEfSe) (LDA score > 2.5), we utilized two compositionally-aware tools: Analysis of Compositions of Microbiomes with Bias Correction (ANCOM-BC) and Anova-Like Differential Expression tool (ALDEx2). ALDEx2 transforms relative abundance data using the centered log-ratio (CLR) transformation, while ANCOM-BC includes a bias correction framework for differential abundance testing. An ASV or taxon was considered a robust differential feature only if it was identified as statistically significant by at least two of these three methods. This multi-method validation ensures higher confidence in our findings and minimizes the risk of false positives arising from the inherent properties of compositional data. Functional profiles were predicted using PICRUSt2 against the KEGG database. It is important to note that PICRUSt2 functional predictions infer metabolic potential based on marker gene data rather than direct measurement of functional activity. This method has several inherent limitations: (1) prediction accuracy depends on the completeness of reference genome databases; (2) it cannot capture horizontal gene transfer events or environment-specific gene expression regulation; (3) prediction reliability may be reduced for uncultured microorganisms commonly found in complex clinical samples. Therefore, functional prediction results in this study should be considered as hypothesis-generating tools rather than direct evidence of functional activity. Detailed ASV filtering statistics are provided in Supplementary Table S2. The human endometrial stromal cell line, T-HESC, was procured from the American Type Culture Collection (ATCC; LGC Standards, Milan, Italy). Cells were maintained in Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F12) supplemented with 10% fetal bovine serum (FBS), 50 U/mL penicillin, 50 µg/mL streptomycin, 2 mM L-glutamine, and 500 ng/mL puromycin. Cultures were incubated at 37 ∘ °C in a humidified atmosphere containing 5% CO 2 . Based on the genus-level LEfSe analysis, representative strains were selected from the key differential taxa identified as enriched in each group. The selected strains, sourced from CGMCC/CMCC, included Lactobacillus helveticus to represent the NC-associated genus Lactobacillus , Enterococcus faecalis to represent the AM-associated genus Enterococcus , and a representative strain from the CC-associated family Enterobacteriaceae . All strains were subsequently cultured to the logarithmic growth phase for use in co-culture experiments. Cells were washed and resuspended in PBS to ~ 1 × 10 8 CFU/mL (OD₆₀₀ verified by plating). Co-culture experiments were performed using a Transwell system featuring 0.4 μm pore size inserts (Corning Inc., Corning, NY, USA) in 96-well plates. T-HESC cells were seeded into the lower chambers at a density of 5 × 10 3 cells/well. Prepared bacterial suspensions (1 × 10 8 CFU/mL) were added to the upper inserts, allowing for soluble factor exchange while preventing direct contact. Experimental groups included: T-HSC control (cells only); NC group (T-HESC co-cultured with L. helveticus ); AM group (T-HESC co-cultured with E. faecalis ); CC group (T-HESC co-cultured with Enterobacteriaceae ). Co-cultures were maintained for 48 h at 37 °C under 5% CO 2 . T-HESC viability post-co-culture was assessed using CCK-8 (Dojindo, Cat. No NU679). Absorbance (450 nm) was measured, and viability was expressed relative to the control. Total RNA from T-HESCs was isolated (TRIzol, Invitrogen). RNA integrity (Agilent 2100) was confirmed. Sequencing libraries were prepared (Novogene) and sequenced (Illumina platform). Reads were aligned (e.g., STAR) to the human genome (GRCh38/hg38). Differential gene expression (DEG) analysis (DESeq2; FDR  1) compared co-culture groups. Gene Ontology (GO) and KEGG pathway enrichment analyses were performed using tools such as clusterProfiler in R or the ClueGO plugin in Cytoscape, applying a significance threshold of p  < 0.05. To corroborate RNA-seq findings, the relative mRNA expression levels of selected key DEGs were quantified via qRT-PCR. Total RNA, extracted using Trizol (Thermo Fisher Scientific; Cat. No. 15596026), was reverse-transcribed into cDNA using a synthesis kit (CWbio, Beijing, China; Cat. No. CW2569). qRT-PCR amplification was performed using UltraSYBR Mixture (CWbio; Cat. No CW2601) on a PikoReal 96 Real-Time PCR System (Thermo Fisher Scientific). Relative mRNA abundance was calculated using the 2 − ΔΔCt method, normalizing target gene expression to the internal control gene, GAPDH. Primer sequences are detailed in Supplementary Table S1. The concentrations (typically in pg/mL) of specific proteins corresponding to validated DEGs secreted into the cell culture supernatants were measured using commercially available sandwich ELISA kits following the manufacturers’ protocols precisely. Kits used included those for: ADH1B (MyBioSource, Cat. No. MBS4503934); MYOM1 (Thermo Fisher Scientific, Cat. No. PIPA5100626; or Proteintech, Cat. No. 22084-1-AP); PTGFR (ABclonal, Cat. No. RK07240; or MyBioSource, Cat. No. MBS9714736); ACLY (RayBiotech, Cat. No. ELH-ACLY; Biomatik, Cat. No. EKF60626 ; BPS Bioscience, Cat. No. 79904; or ABclonal, Cat. No. RK12483); CEBPB (Antibodies-online, Cat. No. ABIN989871); SERPINE1 (Proteintech, Cat. No. KE00109; MyBioSource, Cat. No. MBS175945; R&D Systems, Cat. No. DSE100; ABclonal, Cat. No. RK00133; or Sino Biological, Cat. No. KIT10296 ); S100A1 (Thermo Fisher Scientific, Cat. No. EH402RB; RayBiotech, Cat. No. ELH-S100A1; or MyBioSource, Cat. No. MBS4503934); and CKS1B (MyBioSource, Cat. No. MBS7207214; or Biomatik, Cat. No. EKA50789 ). Data analysis was performed using R software and GraphPad Prism version 9.5 (GraphPad Software, San Diego, CA, USA). Quantitative data were presented as mean ± standard deviation (SD). Statistical significance between two groups was determined using Student’s t-test. Comparisons among three or more groups were conducted using one-way analysis of variance (ANOVA), followed by appropriate post-hoc tests if necessary. Qualitative data (e.g., clinical characteristics) were presented as counts (percentages) and compared using the Chi-square test or Fisher’s exact test where appropriate. A p -value < 0.05 was universally considered indicative of statistical significance. Where multiple comparisons were performed, such as for the alpha diversity analyses across different sites and groups, p -values were adjusted using the Benjamini-Hochberg procedure to control the false discovery rate (FDR). Specific statistical thresholds employed in bioinformatic analyses, such as LDA score > 2.5 for LEfSe and FDR 1 for DEG identification, defined significance for those particular metrics. To investigate the relationship between microbial community structure and host gene expression patterns, we performed a Mantel test. A Bray-Curtis dissimilarity matrix was calculated from the 16S rRNA ASV abundance table, and a Euclidean distance matrix was calculated from the normalized host gene expression data. The Mantel test was performed to assess the correlation between these two distance matrices, with significance determined by 999 permutations.

In conclusion, this study not only demonstrates distinct, site-specific microbial dysbiosis in women with ovarian chocolate cysts and adenomyosis but, more importantly, provides direct experimental evidence that key bacteria associated with these conditions are not merely bystanders but active participants that reprogram endometrial stromal cells. The induction of pro-fibrotic and pro-inflammatory pathways by E. faecalis (AM) and pathways related to metabolic stress and DNA repair by Enterobacteriaceae (CC) offers a novel mechanistic framework for how microbial dysbiosis may contribute to the distinct pathophysiologies of these diseases by altering endometrial cell function. Elucidating these precise mechanisms and exploring the therapeutic potential of targeting the microbiome represent critical avenues for future research.

This study aimed to characterize the microbiota across different sites of the reproductive tract in patients with ovarian chocolate cysts (CC) and adenomyosis (AM), and to evaluate the in vitro effects of representative differential bacteria on human endometrial stromal cells (T-HESC). Our findings reveal significant alterations in the microbial community structure and diversity at multiple reproductive tract sites, from the cervical canal to the endometrium, in patients with CC and AM compared to the control group (NC). These results support the hypothesis that microbial dysbiosis may be involved in the pathological processes of these common gynecological diseases. We observed complex and site-specific changes in microbial composition in CC and AM patients. For instance, in endometrial samples, the NC group showed significant enrichment of Lactobacillaceae, whereas unclassified bacteria had higher relative abundance in the CCd and AMd groups. This aligns with views suggesting that a Lactobacillus-dominated endometrial environment is beneficial for reproductive health. Notably, in posterior fornix samples, we found a significant enrichment of the genus Enterococcus specifically in the AM group (AMb), with no significant difference observed between the CC (CCb) and NC (NCb) groups. E. faecalis is often considered an opportunistic pathogen. Its specific enrichment in the posterior fornix of AM patients might suggest a unique role in the pathogenesis of AM or potential as a biomarker, although this requires further investigation. A recent study by Rahbar Saadat et al. (2020) has shown that Lactococcus lactis modulates the expression of miR-21, miR-200b, and TLR-4 in CAOV-4 cells, demonstrating that microbial dysbiosis may influence immune and gene expression pathways relevant to cancer progression and immune response, which is consistent with our findings in endometriosis and adenomyosis [ 15 ]. We observed complex and site-specific changes in microbial composition in CC and AM patients. For instance, in endometrial samples, the NC group showed significant enrichment of Lactobacillaceae , whereas unclassified bacteria had higher relative abundance in the CCd and AMd groups. This aligns with views suggesting that a Lactobacillus -dominated endometrial environment is beneficial for reproductive health [ 16 , 17 ] and implies that CC and AM might be associated with a reduction or alteration of the protective microbiota locally within the endometrium. Notably, in posterior fornix samples, we found a significant enrichment of the genus Enterococcus specifically in the AM group (AMb), with no significant difference observed between the CC (CCb) and NC (NCb) groups. E. faecalis is often considered an opportunistic pathogen [ 18 ]. Its specific enrichment in the posterior fornix of AM patients might suggest a unique role in the pathogenesis of AM or potential as a biomarker, although this requires further investigation. Furthermore, microbial differences between CC, AM, and NC groups were also noted in ascites fluid, such as the enrichment of Dietzia in CCc, while taxa like Escherichia-Shigella were enriched in NCc. Changes in the ascites microbiota could reflect alterations in the pelvic microenvironment and potentially influence disease progression [ 19 ]. The substantial differences in ASV counts observed between disease and control groups warrant further discussion. Disease groups, particularly the AM group, displayed higher ASV numbers compared to controls, which may reflect several biological phenomena. First, disease states may disrupt microbial homeostasis, leading to increased microbial diversity and the emergence of rare taxa. Second, inflammatory environments characteristic of endometriosis and adenomyosis may provide conditions for more diverse microbial ecological niches. Third, differences in the microenvironments across anatomical sites may influence microbial community complexity. Importantly, we implemented stringent quality control measures, including mock community validation and low-abundance ASV filtering, to ensure that these differences represent genuine biological variation rather than technical artifacts. While our multi-site microbiota analysis revealed strong correlations and predictive profiling suggested altered functional potential, we recognized the critical need to move beyond in silico prediction to establish a direct functional link. Therefore, we employed an in vitro co-culture model to directly test the hypothesis that disease-associated bacteria actively modulate host endometrial cell behavior. This approach, while not a direct multi-omics integration from a single patient sample, provides the first crucial step in establishing a causal chain from a specific microbial shift to a measurable host transcriptomic response, addressing a key gap in the field. To explore the functional significance of these microbial alterations, we performed in vitro co-culture experiments. The results demonstrated that representative differential bacteria identified from AM and CC patients, specifically E. faecalis and Enterobacteriaceae , significantly reduced the viability of T-HESCs compared to the NC-associated representative, L. helveticus . This finding directly suggests that disease-associated bacteria may negatively impact endometrial function and receptivity by compromising the viability of endometrial stromal cells. The neutral effect of L. helveticus , which was identified as characteristic of the NC group in some analyses (e.g., cervical canal) and used here as the control representative, on T-HESC viability contrasted sharply with the detrimental effects observed with E. faecalis and Enterobacteriaceae . More importantly, our transcriptomic analysis revealed distinct regulatory effects of these bacteria on T-HESC cells. Co-culture with Enterobacteriaceae (CC-associated) significantly altered the gand and and enzyme expression profile of T-HESCs, upregulating genes such as ADH1B, MYOM1, and PTGFR, and downregulating ACLY, while enriching KEGG pathways like Mismatch repair and Insulin signaling. Upregulation of ADH1B (Alcohol Dehydrogenase 1B) might be related to inflammation and oxidative stress [ 20 ]; changes in PTGFR (Prostaglandin F Receptor), a key component of prostaglandin signaling pathways implicated in inflammation and uterine contractility [ 21 ], were also observed; downregulation of ACLY (ATP Citrate Lyase), identified as a characteristic gene in endometriosis, may indicate alterations in cellular metabolism, particularly lipid synthesis [ 22 ]. Furthermore, the enrichment of the Mismatch repair pathway KEGG terms, known to be crucial for maintaining genomic fidelity by correcting DNA replication errors, suggests that cellular DNA maintenance processes were affected in T-HESCs exposed to Enterobacteriaceae. In contrast, co-culture with E. faecalis (AM-associated) induced a different set of unique gene expression changes, including the upregulation of CEBPB and SERPINE1, and downregulation of S100A1 and CKS1B. CEBPB (CCAAT/Enhancer Binding Protein Beta) is recognized as a crucial transcription factor involved in regulating inflammation, cell differentiation, and proliferation across various cellular contexts [ 23 , 24 ]. The observed upregulation of SERPINE1 (Plasminogen Activator Inhibitor-1), a well-established inhibitor of plasminogen activation widely implicated in processes such as tissue remodeling, angiogenesis, and fibrosis [ 25 ], was also a notable finding in the AM group co-culture. Furthermore, KEGG analysis revealed a significant enrichment of terms related to the MAPK signaling pathway. This pathway serves as a central hub for mediating cellular responses to a diverse array of external stimuli and fundamentally regulates processes including inflammation and cell proliferation [ 26 ], suggesting its potential involvement in the T-HESC response observed in the AM co-culture group. Additionally, the downregulation of S100A1, encoding a calcium-binding protein with diverse functions, and CKS1B, encoding a known cell cycle regulator, points towards possible disturbances in related signaling and cell cycle control pathways within these cells following exposure to AM-associated bacteria. Together, these specific transcriptomic alterations induced by E. faecalis in vitro, validated at the mRNA and protein level, highlight potential mechanisms by which AM-associated bacteria could interfere with endometrial stromal cell function, thereby possibly contributing to the pathophysiology of adenomyosis, although the precise roles of these specific pathways and genes in AM pathogenesis warrant further investigation. Crucially, a major finding of this study is the strong and significant correlation we observed between the reproductive tract microbiota and host endometrial stromal cell gene expression, as demonstrated by the Mantel test. This provides compelling statistical evidence for a functional link, suggesting that the shifts in microbial composition are closely linked to the host’s transcriptional state. The strong negative correlation (Mantel’s r = −0.9929) is particularly intriguing. It suggests a complex inverse relationship where distinct microbial community structures may drive suppressive or opposing transcriptional programs in endometrial cells, a dynamic that warrants further investigation. By integrating microbial profiling with in vitro functional transcriptomics, this study forges a mechanistic hypothesis linking specific dysbiotic bacteria to distinct, disease-relevant cellular responses. We demonstrate that bacteria are not passive bystanders but active participants that can reprogram host cells. However, we acknowledge the limitations of this model. A full multi-omics integration would require the simultaneous analysis of the microbiome, transcriptome, and perhaps metabolome from the same patient’s endometrial tissue to confirm these associations in vivo. Moreover, epidemiological studies linking the reproductive tract microbiome to ovarian endometrioma and adenomyosis are inherently subject to major confounders, including hormonal status, menstrual cycle phase, prior antibiotic or probiotic exposure, diet, sexual activity, and environmental influences, all of which may substantially alter microbial composition and host responses, thereby complicating causal inference [ 27 ]. Our study uses representative strains, and the complex interactions within a polymicrobial community may yield different outcomes. Future research should focus on validating these pathways in patient tissues, exploring the specific bacterial metabolites responsible for these effects, and investigating whether targeting this microbial dysbiosis could offer a novel therapeutic avenue for endometriosis and adenomyosis. An important limitation of this study concerns the interpretation of PICRUSt2 functional prediction results. Although we observed differences in predicted pathways between groups, these findings are derived from computational inference based on 16S rRNA gene sequences rather than direct measurements of gene expression or enzymatic activity. The accuracy of such predictions is particularly constrained in complex clinical samples containing diverse uncultured microorganisms, as they rely heavily on the availability of closely related reference genomes. To partially overcome this limitation, we performed in vitro co-culture experiments that directly assessed the effects of representative bacteria on endometrial stromal cells. Nevertheless, future research should incorporate metagenomics, metatranscriptomics, or metabolomics to generate direct evidence of microbial functional activity and clarify its mechanistic role in the pathogenesis of endometriosis and adenomyosis. Beyond methodological issues, it should also be acknowledged that most current investigations in this field, including our own, are observational in nature, which limits the ability to establish causality. Moreover, epidemiological studies linking the reproductive tract microbiome to ovarian endometrioma and adenomyosis are susceptible to major confounding factors such as hormonal status, menstrual cycle phase, prior antibiotic or probiotic exposure, diet, sexual activity, and environmental influences. These confounders complicate causal inference and underscore the need for rigorously designed longitudinal and interventional studies that integrate multi-omics analyses with mechanistic models to better delineate the causal pathways by which microbial dysbiosis contributes to disease development.

Endometriosis and adenomyosis represent prevalent, estrogen-dependent gynecological disorders that markedly compromise the quality of life for reproductive-aged women, primarily manifesting as chronic pelvic pain and infertility [ 1 – 4 ]. Pathologically, endometriosis involves the ectopic implantation and proliferation of endometrial-like tissue outside the uterine cavity, whereas adenomyosis is characterized by the infiltration of such tissue into the myometrium [ 1 , 5 ]. A common clinical presentation of endometriosis is the ovarian endometrioma (OE), colloquially termed a “chocolate cyst” (CC), the specific cellular microenvironment of which is crucial for understanding disease pathogenesis [ 6 ]. The etiology of these conditions remains incompletely elucidated but is understood to be multifactorial, encompassing retrograde menstruation, hormonal dysregulation, genetic predispositions (including specific mutations like KRAS), epigenetic modifications, and, notably, immune system dysfunction [ 1 , 3 , 4 , 7 , 8 ]. Indeed, detailed analyses have identified alterations in immune cell populations, such as macrophages and NK cells, alongside aberrant cellular proliferation within endometriotic lesions [ 6 ]. The association between these conditions and infertility is particularly significant; proposed underlying mechanisms include distorted pelvic anatomy, altered endocrine signaling, compromised gamete or embryo quality, chronic inflammation, and impaired endometrial receptivity, potentially stemming from progesterone resistance [ 1 , 4 , 8 ]. Reflecting its broad impact, endometriosis is increasingly recognized as a systemic condition [ 3 ], while adenomyosis presents analogous challenges to reproductive health [ 2 ]. Emerging research increasingly implicates the host microbiome in the pathophysiology of diverse diseases, including those affecting the female reproductive system [ 9 ]. The reproductive tract normally harbors a complex microbial ecosystem, often dominated by Lactobacillus species, which contribute significantly to maintaining local homeostasis [ 9 ]. Disruptions to this delicate microbial equilibrium, termed dysbiosis, have been progressively linked to various gynecological disorders [ 9 , 10 ]. Specifically, alterations in the composition and function of the microbiota within the reproductive tract—encompassing the vagina, cervix, and endometrium—and potentially the gut, have been documented in women with endometriosis and adenomyosis when compared to healthy controls [ 4 , 9 – 14 ]. These microbial shifts frequently involve changes in diversity and the relative abundance of key bacterial taxa, potentially resulting in diminished Lactobacillus dominance and a concurrent rise in opportunistic or pro-inflammatory bacteria [ 4 , 9 , 11 – 14 ]. Such microbial dysbiosis is hypothesized to fuel the chronic inflammation, immune dysregulation, and altered estrogen metabolism characteristic of endometriosis and adenomyosis, thereby potentially influencing disease initiation, progression, and associated infertility [ 4 , 9 , 10 ]. Despite these recognized associations, a comprehensive understanding of specific microbial signatures across multiple distinct anatomical locations within the reproductive tract (cervical canal, posterior fornix, endometrium, and ascites fluid) remains lacking, particularly when comparing well-defined patient cohorts—specifically, those with ovarian endometrioma (CC) versus adenomyosis (AM). Although previous studies have compared microbiota at individual sites or between different conditions [ 11 – 13 ]and characterized the cellular milieu within lesions [ 1 ], a simultaneous, multi-site microbial analysis directly comparing these specific patient groups is warranted. Furthermore, while compositional differences in microbiota are increasingly identified [ 14 ], the direct functional repercussions of specific bacteria associated with these conditions on endometrial cellular physiology remain largely uninvestigated. Elucidating whether and how key differential bacterial taxa influence endometrial stromal cell viability and gene expression is critical for establishing a mechanistic link between microbial dysbiosis and the pathogenesis of these debilitating diseases. Therefore, this study was designed with two principal objectives. Firstly, to comprehensively characterize and compare the microbial communities inhabiting the cervical canal, posterior fornix, ascites fluid, and endometrium among patients diagnosed with ovarian chocolate cysts (CC), adenomyosis (AM), and control individuals (NC). Secondly, to investigate the direct functional consequences of in vitro exposure to representative bacterial taxa, identified as differentially abundant between the study groups, on the viability and transcriptomic landscape of human endometrial stromal cells (T-HESC). By addressing these objectives, we aim to furnish novel insights into the role of site-specific reproductive tract microbiota in the complex pathophysiology of endometriosis and adenomyosis.

Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2. Supplementary Material 3. Supplementary Material 3. Supplementary Material 4. Supplementary Material 4.