Endobiota-Estrobolome Profiles in Reproductive-Aged Women With Ovarian Endometriosis

Informed consent was obtained from all subjects involved in the study. The study was reviewed and approved by the institutional review board of the Human Investigation and Ethical Committee of Chang Gung Medical Foundation (CGMHIRB No. 202000462A3).

This research works was supported in part under Chang Gung Memorial Hospital Research Grant (CMRPG3G1981, CMRPG3K0782).

Of the 38 Taiwanese women included, 24 had pathologically confirmed ovarian endometriosis, while the other 14 had final pathology diagnoses of benign ovarian tumors (e.g., teratoma, corpus luteum, serous adenofibroma, and mucinous cystadenoma) and served as controls. The descriptive characteristics of study participants are summarized in Table  1 . There was no significant difference in average age (38.5 ± 5.1 vs. 35.3 ± 6.6 years, p  = 0.129) and BMI (24.5 ± 4.4 vs. 23.4 ± 4.3 kg/m 2 , p  = 0.450) between the control and ovarian endometriosis groups. Comparison of the pre‐operative CA‐125 level showed, as expected, significantly higher level in the ovarian endometriosis group than that of the control (67.4 ± 63.4 vs. 27.4 ± 4.6 U/mL, p  = 0.007). There were also significant differences in gravida (0 vs. 2, p  = 0.004) and parity (0 vs. 2, p  < 0.001) between the two groups. Baseline characteristics of control and endometriosis group. Note: The endometriosis stage was according to American Fertility Society Revised Classification [ 34 ]. Gravida and parity are presented as median (interquartile range) and compared using the Mann–Whitney U test. Other values are presented as mean ± SD. Abbreviations: BMI, body mass index; rAFS score, revised American Society for Reproductive Medicine score. * p ‐value(s) with statistical significance. Gut microbial β‐glucuronidase (gmGUS) is involved in the metabolism of estrogen [ 35 ]. It has been shown that stable interaction between gmGUS and estrogen maintains physiological balance while disruption can lead to estrogen‐related diseases [ 35 ]. While β‐glucosidase, functioning as a control enzyme, plays a key role in the final step of cellulose hydrolysis and aids in nutrient digestion [ 36 ], β‐glucuronidase is responsible for deconjugating biologically inactive conjugated estrogen, converting it into its active, deconjugated form [ 35 ]. However, enzymatic activity assays conducted on fecal samples from both the control and ovarian endometriosis group did not show significant differences in the average levels of β‐glucosidase activity (−14.82 U/L vs. 23.15 U/L, p  = 0.76) or β‐glucuronidase activity (5851.95 U/L vs. 5122.04 U/L, p  = 0.70) (Figure  2 ). It should be noted that β‐glucosidase activity values were calculated after background subtraction during the enzymatic assay. When enzyme activity levels were close to the detection limit, normal assay variability occasionally produced slightly negative values, which therefore represent activity at or near the detection threshold. Enzymatic assay with results given in U/L. Boxplots: Box indicate the 1st and 3rd quartiles, dashed lines the upper and lower whiskers, cross indicates mean, horizontal bold lines the median, and dots signify outliers. Analyses of 14 estrogen metabolites showed nominal differences between patients with ovarian endometriosis and controls. In vaginal samples, lower fold changes were observed for 4‐methoxyestrone ( p  = 0.046), 2‐methoxyestrone ( p  = 0.043), and 2‐hydroxyestrone‐3‐methyl ether ( p  = 0.006) in patients with ovarian endometriosis compared with controls. In urine samples, 16α‐hydroxyestrone showed a higher fold change in the endometriosis group ( p  = 0.032). However, these associations did not remain statistically significant after Benjamini–Hochberg FDR correction. No apparent differences were observed in fecal samples. Detailed results are presented in Table  2 . Estrogen metabolites in fecal, urinary, and vaginal samples of patients with or without ovarian endometriosis. Note: Fold change represents the ratio of metabolite levels in the endometriosis group relative to the control group. Values are presented as p ‐values with Benjamini–Hochberg false discovery rate (FDR)–adjusted p ‐values in parentheses. FDR correction was applied separately within each biological compartment (stool, urine, and vaginal samples). Abbreviations: N.A., not applicable; N.D., not detected (below the detection limit). Nominal p  < 0.05 before FDR correction. Paired‐end 16S amplicon sequencing generated raw reads from stool of 14 control women and 24 patients with ovarian endometriosis. Following quality control, effective tags were used for operational taxonomic unit (OTU) clustering and species annotation. Species composition and abundance were analyzed across samples, with a tree graph displaying species annotations for each group presented in Figure  3 . In this present study, the predominant bacterial phyla in the gut microbiomes of Taiwanese women with endometriosis were Bacteroidetes (49%), Firmicutes (41%), and Proteobacteria (8%) (Figure  3a ). The distribution histogram shows the relative abundance of the top‐10 major phyla of gut microbiomes between two groups (Figure  3b ). The compositions of microbial species at each hierarchical level are illustrated using heat trees (Figure  3c,d ). Both groups displayed similar relative bacterial abundances. Abundance and composition of gut microbiomes between control and patients with ovarian endometriosis. (a) The top‐5 major phyla of bacteria in gut microbiota of Taiwanese women. (b) Relative abundance of the top‐10 major phyla of gut microbiota in two groups are shown in distribution histogram. (c, d) Circular heat trees represent the sequence abundance of different hierarchical taxa, displayed from the center outwards. The innermost circle represents the highest taxonomic level, the bacterial domain (Bacteria). Moving outward from the center, the taxonomic levels decrease, and the number of sequences annotated to different taxonomic levels decreases accordingly. Sequence abundance is represented by node size, branch thickness, and color. Species with higher abundance are indicated by larger nodes, thicker branches, and colors closer to brown. The legend in the lower‐right corner shows the sequence count along with their corresponding colors and node sizes. (c) Control and (d) Ovarian Endometriosis. CES, control; ES, ovarian endometriosis. With its various indicators, alpha diversity, which is a measure of microbial diversity, richness, and evenness within samples, showed no significant differences in Shannon–Wiener diversity index (Figure  4a ), Simpson diversity index (Figure  4b ), and species richness (Figure  4c ) between the two groups. Meanwhile, beta diversity, as assessed by unweighted (Figure  4d ) and weighted UniFrac distance (Figure  4e ), showed similar microbial composition between the two groups. Comparison of gut microbiota composition between the two groups. (a–c) Alpha diversity analysis on a single sample reflects the richness, evenness, and diversity of microbial communities within that sample. Shown with box plots, there were no significant differences observed using Shannon (a), Simpson (b), and species richness (c) indices. This signified that the alpha diversity of gut microbiota was similar between the two groups based on t ‐test analysis. (d–e) Beta diversity studies use OTU representative sequences to construct phylogenetic trees and calculate unweighted UniFrac and weighted UniFrac (when relative abundance of species within sample was considered) distances in order to evaluate differences between samples. Smaller values indicate less differences in species diversity between samples. Results are represented with box plots: (d) unweighted UniFrac and (e) weighted UniFrac. T ‐test was used to calculate for significant differences. (f) F/B (Firmicutes/Bacteroidetes) ratio of the control and ovarian endometriosis group shown in box plot. (g) LEfSe (linear discriminant analysis (LDA) Effect Size) was performed to define microbial biomarkers for ovarian endometriosis group with statistical differences. The representative microbiotas for each group are shown with the LDA scores (left panel). The phylogenetic trees of dominant microorganisms are revealed in cladogram (right). (h) Histogram of abundance comparison of species at genus level. CES, control; ES, ovarian endometriosis. A balanced gut microbiota that compose mostly of the Bacteroidetes and Firmicutes phyla supports the gut epithelial barrier and overall homeostasis [ 11 ]. Higher Firmicutes/Bacteroidetes ratio (F/B) in the gut suggests dysbiosis [ 37 ]. As depicted in the box plot (Figure  4f ), the median F/B ratio of the ovarian endometriosis group was marginally higher than that of the control group (0.98 vs. 0.93, p  = 0.88), but it did not reach statistical significance. To better understand the characteristic gut microbiota associated with patients with ovarian endometriosis, LEfSe analysis was conducted on the phylogenetic trees of the dominant microorganisms in the two groups. We used a LDA threshold > 2 to identify significant characteristic microbiota at the genus level. The LDA scores and cladogram highlighted Rothia genus within the Micrococcales lineage as more prevalent in the control group (Figure  4g ). Genera such as Megamonas [ Eubacterium ] coprostanoligenes_group , Allisonella , Ruminiclostridium_5 [ Eubacterium ] hallii_group , Negativibacillus were significantly more abundant in the ovarian endometriosis group (Figure  4h ). Similar to gut microbiota, different sections of the female reproductive tract exhibit distinct microbial distributions. The female reproductive tract is divided into a higher segment, consisting of the endocervix and uterus, and a lower segment, comprising of the vaginal canal and ectocervix. Vaginal samples were analyzed for this study in order to evaluate the microbiota of the lower female reproductive tract. The lower female reproductive tract has been shown to be primarily populated by Lactobacillus spp., which protects it against pathogens by creating an acidic environment and producing bacteriocins and hydrogen peroxide [ 38 ]. Compatibly, Lactobacillus was the predominant genus observed in the vaginal microbiota of both the endometriosis and control groups. In the ovarian endometriosis group, Lactobacillus accounted for 64.4%, Prevotella 6.6%, and Gardnerella 5.5%. In contrast, the control group had 48.5% Lactobacillus , 6.8% Gardnerella , and 5.8% Prevotella (Figure  5a,b ). The compositions of microbial species at each hierarchical level are illustrated with heat trees (Figure  5c,d ). No significant differences were seen in alpha diversity, as indicated by the Shannon‐Wiener diversity index (Figure  6a ), Simpson diversity index (Figure  6b ), and species richness (Figure  6c ). However, lower beta diversity was found in the vaginal samples of patients with ovarian endometriosis, as shown by the unweighted UniFrac distance (Figure  6d , p  < 0.001) and weighted UniFrac distance (Figure  6e , p  = 0.006). The results signified lower species annotation and OTU abundance in the vaginal samples of patients with endometriosis. Abundance and composition of vaginal microbiomes between control and patients with ovarian endometriosis. (a) The top‐5 major genera of bacteria in vaginal microbiota of Taiwanese women. (b) Relative abundance of the top‐10 major genera of vaginal microbiota in two groups is shown in a distribution histogram. (c, d) Circular heat trees represent the sequence abundance of different hierarchical taxa, displayed from the center outwards. (c) Control and (d) Ovarian endometriosis. CEV, control; EV, ovarian endometriosis. Comparison of vaginal microbiota composition between the two groups. Shown with box plots, no significant differences were observed using Shannon (a), Simpson (b), and species richness (c) indices, which suggested that the alpha diversity of vaginal microbiota was similar in the endometriosis and control group. (d–e) Beta diversity studies use OTU representative sequences to construct phylogenetic trees and calculate unweighted UniFrac (d) and weighted UniFrac (e) distances in order to evaluate differences between samples. Smaller values indicate less differences in species diversity between samples. T ‐test was used to calculate significant differences. (f) LEfSe (linear discriminant analysis Effect Size) was performed to define microbial biomarkers for the endometriosis group with statistical differences. The representative microbiotas for each group are shown with the LDA scores (left panel). The phylogenetic trees of dominant microorganisms are revealed in a cladogram (right). CEV, control; EV, ovarian endometriosis. LEfSe analyses on the phylogenetic trees of the dominant microorganisms, using a LDA threshold > 2, defined significant characteristics of the microbiota between the two groups. With LDA scores and cladogram, the present study revealed that, at the family level, the Ruminococcaceae ( p  = 0.047), Eggerthellaceae ( p  = 0.009) and Beijerinckiaceae families ( p  = 0.047) were more dominant in the control group when compared to the ovarian endometriosis group (Figure  6f ). We further explored whether menstrual cycle phase influenced the microbiome or metabolite profiles. No significant differences were observed between samples collected during the proliferative and secretory phases. To explore potential interactions across datasets, we performed integrative correlation analyses between microbiota composition, enzyme activity, and estrogen metabolites (Tables  S3–S6 ). Given the large number of microbial taxa identified, we selected the top 10 genera of gut and vaginal microbiota in both the case and control groups for further analysis (Tables  S1 and S2 ). Several microbiota–metabolite associations were identified at nominal significance levels ( p  < 0.05), particularly involving genera such as Megamonas and Parabacteroides (Table  S3 ). In addition, a trend toward association was observed between β‐glucuronidase activity and estriol levels in urine samples ( p  = 0.006, FDR‐adjusted q  = 0.082) (Table  S5 ). Furthermore, correlations between gut and vaginal microbiota suggested potential cross‐site microbial interactions, although these did not remain significant after multiple testing correction (Table  S6 ).

Building on our previous work, this study further explored the association between endobiota—encompassing both urogenital and gastrointestinal microbiota—and the estrobolome in ovarian endometriosis. Overall, our findings suggest that ovarian endometriosis is not characterized by overt alterations in gut enzymatic activity. However, modest, compartment‐specific differences in microbial composition and estrogen metabolite profiles were observed across urogenital and gastrointestinal samples. Although gut β‐glucuronidase activity did not differ significantly between groups, patients with ovarian endometriosis exhibited nominal differences in estrogen metabolite patterns, including lower levels of methoxylated estrogens in vaginal samples and increased urinary 16α‐hydroxyestrone. However, these associations did not remain statistically significant after FDR correction for multiple comparisons. In parallel, differences in endobiota composition were observed. Several gut bacterial genera, including Megamonas [ Eubacterium ] coprostanoligenes_group , Allisonella, Ruminiclostridium_5 [ Eubacterium ] hallii_group , and Negativibacillus , were more abundant in patients with ovarian endometriosis. In contrast, vaginal samples from these patients demonstrated reduced bacterial abundance and diversity, whereas controls showed higher relative abundances of the families Ruminococcaceae and Beijerinckiaceae. Taken together, these observations suggest that ovarian endometriosis may be accompanied by compartment‐specific differences in microbial composition and estrogen metabolism rather than generalized dysbiosis. To further explore potential interactions across datasets, we performed integrative correlation analyses incorporating microbiota composition, microbial enzyme activity, and estrogen metabolites (Tables  S3–S6 ). Although no associations remained statistically significant after multiple testing correction, several consistent patterns were observed. Specifically, genera such as Megamonas and Parabacteroides were among the top‐ranked taxa associated with estrogen metabolites. In addition, β‐glucuronidase activity demonstrated a trend toward association with estriol levels in urine samples. Given the known role of β‐glucuronidase in deconjugating estrogen metabolites and regulating enterohepatic recirculation [ 35 , 36 ], this observation is biologically plausible and suggests a potential functional link between microbial enzymatic activity and systemic estrogen metabolism. Furthermore, correlations between gut and vaginal microbiota were identified, suggesting a possible cross‐compartment interaction along a gut–vaginal axis. Although these findings should be interpreted cautiously due to limited statistical power, they provide preliminary integrative support for the hypothesized estrobolome–endobiota axis. Our previous study explored associations between gut microbiota and endometriosis using bacterial profiling, enzymatic assays, and targeted estrogen metabolite quantification [ 2 ]. Although no significant differences were observed in microbial richness, diversity, β‐glucuronidase activity, or urinary estrogen metabolites, fecal samples from patients with endometriosis demonstrated enrichment of specific bacterial taxa and increased levels of selected estrogen metabolites [ 2 ]. Consistent with these findings, several human studies have reported associations between gut microbiota and endometriosis. Svensson et al. identified higher overall gut microbial diversity in controls compared with women with endometriosis, along with differences in multiple bacterial taxa [ 39 ]. Other studies have reported reduced alpha diversity and altered Firmicutes/Bacteroidetes ratios in patients with moderate‐to‐severe endometriosis, accompanied by differential abundance of specific genera [ 32 ]. In contrast, the largest metagenomic analysis to date did not identify distinct gut microbial or estrobolome‐related functional profiles between women with and without endometriosis [ 40 ]. These discrepancies may reflect differences in study design, sample size, populations, and analytical methods. The estrobolome–endometriosis axis is complex and remains incompletely understood. Gut microbial β‐glucuronidase plays a role in estrogen deconjugation and enterohepatic recirculation, potentially influencing systemic estrogen exposure [ 41 , 42 ]. Animal studies further support a bidirectional relationship between endometriosis and microbial alterations [ 19 , 35 , 36 , 39 , 41 , 43 , 44 , 45 , 46 ]. However, these findings do not establish causality, and the clinical utility of antibiotics as a non‐hormonal treatment for endometriosis remains uncertain [ 47 ]. Previous studies analyzing the role of the reproductive tract microbiome had demonstrated that opportunistic pathogens, such as Streptococcaceae and Staphylococaceae, were enriched in the cystic fluid of females with ovarian endometrioma [ 16 ]. Studies from Brazil and China noted lower Lactobacillus abundance in endometriosis patients compared to controls in vaginal samples [ 48 , 49 , 50 ]. Additionally, Ata et al. reported differences in vaginal microbiota between patients with severe endometriosis and healthy individuals, including the absence of Gemella and Atopobium spp. in the endometriosis group [ 51 ]. Perrotta et al. further incorporated vaginal community state types to construct a random forest model predicting r‐ASRM endometriosis stages, highlighting the potential relevance of specific microbial taxa across menstrual phases [ 52 ]. Although our results did not demonstrate a causal relationship between endobiota and ovarian endometriosis, the observed variations in bacterial genera and families suggest a potential association between urogenital and gastrointestinal microbiota and disease status. Consistent with prior research linking gut microbial imbalance—particularly an increased Firmicutes/Bacteroidetes ratio—to metabolic and inflammatory conditions [ 53 , 54 , 55 ], our study revealed higher abundances of several Firmicutes‐associated genera, including Megamonas [Eubacterium] coprostanoligenes_group , Allisonella , Ruminiclostridium_5 [Eubacterium] hallii_group , and Negativibacillus , in the gut microbiota of patients with ovarian endometriosis. The Megamonas genus has previously been associated with obesity and metabolic disorders [ 56 , 57 ]. Additionally, in previous vaginal microbiota studies, Kunaseth et al. [ 58 ] reported that Alloscardovia , Oscillospirales , Ruminococcaceae, Oscillospiraceae, Enhydrobacter , Megamonas , Selenomonadaceae, and Faecalibacterium all exhibited increased abundance in the vaginal microbiota of patients with adenomyosis. However, the functional metabolic pathways contributed by these taxa in the context of endometriosis remain to be elucidated. In contrast, the Ruminococcaceae family—found at higher levels in vaginal samples from our control group—has been proposed as a protective factor against endometriosis, potentially through the production of short‐chain fatty acids [ 59 ]. Given evidence that enterohepatic recirculation influences systemic estrogen levels and reproductive health [ 24 , 60 , 61 , 62 ], we analyzed parent estrogens and multiple estrogen metabolites across fecal, urinary, and vaginal samples. We observed lower levels of methoxylated estrogen metabolites in vaginal samples and higher levels of 16α‐hydroxyestrone in urinary samples from patients with ovarian endometriosis. However, these associations did not remain statistically significant after Benjamini–Hochberg FDR correction for multiple comparisons. Similar to previous reports [ 62 , 63 ], no significant differences in fecal estrogen metabolites were detected between groups. Notably, whereas our prior study identified increased fecal 16α‐hydroxyestrone in patients with endometriosis [ 2 ], the current study similarly demonstrated higher levels of this metabolite in urine, suggesting that estrogen metabolism may vary across biological compartments. These findings underscore the importance of evaluating multiple sample types when investigating estrobolome‐related mechanisms and should be interpreted as exploratory and hypothesis‐generating. Taken together, these findings suggest a potential link between microbial composition and estrogen metabolism across different biological compartments. Alterations in gut microbiota have been proposed to influence systemic estrogen metabolism through estrobolome‐related enzymatic activity, whereas changes in vaginal microbiota may affect the local hormonal microenvironment within the reproductive tract. In the present study, integrative correlation analyses further revealed consistent, albeit non‐significant, patterns linking specific microbial taxa, enzyme activity, and estrogen metabolites across datasets. These findings provide preliminary support for a functional interplay between microbial composition and estrogen metabolism, beyond compartment‐specific observations. Although no associations remained statistically significant after multiple testing correction, the observed trends suggest that microbial–hormonal interactions may operate differently across gastrointestinal and urogenital environments. Future studies integrating microbiome composition, microbial enzyme activity, and estrogen metabolite profiling will be essential to better elucidate these multi‐layer interactions in endometriosis. This study has several limitations. First, an imbalance in parity between the endometriosis and control groups may represent a potential confounding factor, as pregnancy and reproductive history are known to influence both gut and vaginal microbiota composition. Because only one participant in the control group was nulliparous, subgroup analysis restricted to nulliparous individuals was not feasible and could not provide a statistically meaningful comparison. Menstrual cycle phase at the time of sampling was recorded for all participants. Exploratory analysis comparing samples collected during the proliferative and secretory phases did not reveal significant differences in microbiota composition or metabolite profiles; however, subtle cycle‐related variations cannot be completely excluded given the modest sample size. Individual variation in microbiota composition may be influenced by numerous confounding factors, a challenge further compounded by the analysis of gut, urinary, and vaginal microbiota simultaneously. Although the Shannon diversity index is well suited for environments with high microbial richness, such as stool, it may be less informative in low‐diversity sites such as the vagina and cervix [ 64 ]. To mitigate variability, we restricted enrollment to women of similar age and ethnicity and excluded participants with restrictive diets, systemic diseases, malignancy, or recent antibiotic or probiotic use. Hormone therapy was not an exclusion criterion, as participants underwent surgery for benign gynecological conditions after unsuccessful medical treatment or severe symptoms. In addition, the control group consisted of patients undergoing surgery for benign ovarian tumors rather than completely healthy individuals. Although this design allowed standardized perioperative sampling and surgical confirmation of the absence of endometriosis, underlying gynecologic conditions or surgical indications may still influence microbiota composition and represent a potential source of bias. Nevertheless, all patients had discontinued hormonal medications for at least 6 months before enrollment. Other factors known to influence microbiota composition, including detailed dietary habits, sexual activity, and vaginal hygiene practices, were not systematically recorded in this cohort and may represent additional sources of residual confounding. A formal power calculation was not performed prior to this study. Given the multi‐omics design and the lack of prior effect size estimates across combined microbiome, metabolite, and enzymatic datasets, accurate sample size estimation was challenging. Therefore, the present findings should be interpreted primarily as exploratory and hypothesis‐generating, and larger, well‐powered studies will be required to validate these observations. In conclusion, our findings suggest that ovarian endometriosis may involve compartment‐specific differences in microbial composition and estrogen metabolite profiles rather than generalized dysbiosis. These integrative findings further suggest that microbial functional activity, rather than taxonomic composition alone, may play a role in modulating estrogen metabolism across biological compartments. These observations highlight the potential importance of localized microbial–hormonal interactions within the reproductive tract environment. Although causal relationships cannot be inferred from this exploratory study, our results provide a framework for future investigations integrating microbiome composition, microbial enzymatic activity, and estrogen metabolite profiling across multiple biological compartments. Larger, well‐controlled studies are warranted to further elucidate endobiota–estrobolome interactions and to clarify the directionality of these associations.

The authors have nothing to report.

Genetic material produced by the vast number of microorganisms inhabiting the human body constitutes the human microbiome [ 1 , 2 ]. Accumulating evidence indicates that microbiota play essential roles in maintaining physiological homeostasis across multiple organ systems, including the skin [ 3 ], gastrointestinal tract [ 4 ], and female reproductive tract [ 5 ]. Disruptions in microbial composition or diversity, commonly referred to as dysbiosis, have been associated with a wide range of pathological conditions, such as autoimmune diseases, malignancies, metabolic disorders, and inflammatory bowel disease [ 6 ]. Endometriosis is a chronic, estrogen‐dependent disorder characterized by the presence of endometrial glands and stroma outside the uterine cavity. Affecting approximately 6%–10% of reproductive‐aged females worldwide [ 7 ], it commonly inflicts clinical symptoms like dyspareunia, infertility, dysmenorrhea, and severe pelvic discomfort. Retrograde menstruation, also known as the Sampson's theory [ 8 ], is the most widely accepted pathogenesis. However, with 90% of women experiencing retrograde menstruation and only 10% developing the condition [ 9 ], other possible mechanisms relating to genetic, anatomical, endocrine, inflammatory, and environmental factors, are being explored. The estrobolome refers to the collection of microbial genes capable of metabolizing estrogens [ 10 ]. Microbial enzymes such as β‐glucuronidase can deconjugate estrogen metabolites, allowing reabsorption into the circulation and subsequent interaction with estrogen receptors [ 10 ]. Recent studies have highlighted the bidirectional estrogen–microbiome axis linking the gut microbiome to the host hormonal milieu [ 11 ]. Perturbations of the estrobolome have been proposed to influence estrogen‐dependent conditions, including endometriosis and endometrial cancer [ 12 ]. The female reproductive tract harbors distinct microbial communities and accounts for approximately 9% of the body's total microbial population [ 13 , 14 ]. Previous studies have reported increased bacterial colonization in the endometrial tissues and menstrual blood of women with endometriosis compared with healthy controls [ 15 , 16 , 17 , 18 ]. Altered gut microbiota profiles have also been observed in animal models of endometriosis [ 19 ]. However, clinical data examining the relationship between endometriosis and the endobiota—encompassing both gastrointestinal and female reproductive tract microbiomes—remain limited. Endometriosis is a heterogeneous disease that includes superficial peritoneal lesions, ovarian endometriomas, and deep infiltrating endometriosis, each potentially representing distinct pathophysiological entities [ 20 ]. Pooling different phenotypes may therefore obscure phenotype‐specific microbial alterations. Accordingly, the present study focused on women with ovarian endometriosis. By comprehensively analyzing gut and vaginal microbiota, microbial enzyme activity, and estrogen metabolites, this study aimed to characterize endobiota–estrobolome profiles and to explore their potential association with altered estrogen metabolism in ovarian endometriosis.

The authors declare no conflicts of interest.

This case–control study included 38 women of reproductive age (20–48 years old) who underwent surgery for benign gynecological diseases at Linkou Chang Gung Memorial Hospital from August 1, 2020, to March 31, 2022. (Figure  1 ) The control group consisted of patients undergoing surgery for benign ovarian tumors, which allowed intraoperative confirmation of the absence of endometriosis. Postmenopausal women, pregnant individuals, or those with past or current malignancies were excluded. Relevant clinical variables, including parity, BMI, and medical history, were not used as exclusion criteria. Due to the referral nature of our center, prior hormone exposure could not be completely excluded; however, all participants had discontinued hormonal therapy for at least 6 months prior to enrollment. Menstrual cycle phase (proliferative or secretory) at the time of sample collection was recorded based on patient history. Patients with ovarian endometriosis, defined as endometriosis characterized by the presence of endometrial‐like tissue within or on the surface of the ovaries [ 21 ], were required to have pathologically confirmed specimens. The stage of endometriosis for each patient was classified according to the revised guidelines from American Society for Reproductive Medicine: stage I as minimal, stage II as mild, stage III as moderate, and stage IV as severe [ 22 ]. Categorization of deeply infiltrated endometriosis was based on the Enzian score [ 23 ]. Participants who planned for surgical intervention for endometriosis or benign ovarian tumor were provided a toilet attached pouch (Protocult, Rochester, MN), from which they collected aliquots of stool, as well as a simultaneous urine specimen. Study design and participant screening. Urine samples (20–50 mL) were collected in sterile screw‐top containers without preservatives, immediately chilled on frozen gel packs (4°C), frozen in liquid nitrogen within 3 h, and later aliquoted (1 mL) before storage at −80°C for estrogen and estrogen metabolite analysis. Participants collected 16 aliquots, half in RNAlater (QIAGEN Inc., Valencia, CA) and half in sterile phosphate buffer saline (PBS), from various parts of a single stool. As with the urine, all fecal aliquots were chilled immediately on frozen gel packs (4°C) and frozen in liquid nitrogen within 3 h. The fecal aliquots were stored at −80°C until used for DNA and protein extraction. Vaginal samples were collected mid‐vagina at the vaginal introitus by research doctors before surgery, without speculums or lubricants. Swabs were immediately immersed in MoBio PowerBead tube lysis buffer (MoBio Laboratories Inc., Carlsbad, CA, USA), pressed against the tube walls three times (20 s each) to ensure biological material transfer, and transported on wet ice to a clinical lab within 2 h before storage at −80°C until extraction. This study complied with the Helsinki Declaration and received approval from the Institutional Review Board of the Human Investigation and Ethical Committee of Chang Gung Medical Foundation (CGMHIRB No. 202000462A3). Protein extraction was conducted using a previously published protocol [ 24 ]. Approximately 0.5 g of thawed feces was placed into 10 mL conical tube with 5 mL of extraction buffer and vortexed for 1 min for homogenization. Bacterial cell lysis was performed by sonication at maximum power in 30 s intervals, with all procedures carried out in an ice bath. The lysates were centrifuged at 4°C, and the supernatant containing the extracted proteins was collected for enzymatic activity assays. The activities of β‐glucuronidase and β‐glucosidase were measured using commercial assay kits (Pierce, Rockford, IL, USA) according to the manufacturers' instructions. Briefly, 100 μL of fecal lysate was added to each well of a 96‐well microplate. After adding the provided substrate mix (Abcam, Cambridge, UK), the reaction mixture was incubated at 37°C for 0–60 min. Fluorescence was measured (Ex/Em = 330/450 nm) using a multi‐well microplate reader (Tecan Infinite M200 Plate Reader, Ramsey, MN, USA). Enzyme activity values were calculated after background subtraction. When activity levels were close to the assay detection limit, minor variability occasionally resulted in slightly negative values. For analyzing urine, fecal, and vaginal estrogens and their metabolites, samples were transferred on dry ice to Biotools Laboratory in New Taipei City, Taiwan. The lab used advanced liquid chromatography and mass spectrometry for compound separation and analysis. This followed US FDA guidelines for detection and quantification limits, measuring 14 target metabolites including estrone, estradiol, and estriol variations (as listed in Table  2 ), with final concentrations reported in ng/mL. Metabolites with concentrations below the detection limit of the assay were recorded as “not detected (N.D.)”. Samples with N.D. values were excluded from fold‐change calculations and statistical comparisons for that metabolite. Microbial DNA was extracted from fecal and vaginal samples using the Genomic DNA Isolation Kit (FairBiotech Corp., Taoyuan, Taiwan). The hypervariable regions (V3‐V4) of the 16S ribosomal RNA (rRNA) sequence were amplified by PCR, and the detailed procedure and sequences of the universal primers have been described in a previous study [ 2 ]. Fourteen samples from the control group and 24 samples from patients with ovarian endometriosis exhibited clear amplified DNA fragments and were subjected to library construction. Amplicon sequencing was performed on the Illumina paired‐end platform. Raw paired‐end reads were merged using FLASH and quality‐filtered using the QIIME pipeline to obtain high‐quality clean reads [ 25 , 26 ]. Chimeric sequences were identified and removed using UCHIME, resulting in the final set of effective tags [ 27 ]. The effective tags were clustered into operational taxonomic units (OTUs) at 97% sequence similarity using the UPARSE algorithm implemented in the USEARCH pipeline (v7.0.1090) [ 28 ]. After quality filtering and chimera removal, an average of approximately 110,000 high‐quality sequences per sample (range: 58,026–152,906) were retained for downstream analyses. To normalize sequencing depth across samples, reads were rarefied prior to diversity analyses. Alpha diversity indices, including species richness, Shannon index, Simpson index were calculated to evaluate microbial diversity within samples. Good's coverage exceeded 99.9% for all samples, indicating that sequencing depth was sufficient to capture the majority of microbial diversity. Statistical analyses were performed using QIIME v2 for microbiome diversity analysis, including alpha diversity (Shannon, Simpson, and Good's coverage) and beta diversity (weighted and unweighted UniFrac distances) [ 29 ]. In order to emphasize the difference of the dominant species between endometriosis and control groups in each taxonomic rank, the top 10 species were selected and distribution histograms were drawn based on relative abundance. To visualize microbial genomes and metagenomes and illustrate evolutionary relationships, species richness, and annotations for a comprehensive phylogenetic tree, GraPhlAn (Graphical Phylogenetic Analysis) was used to construct tree graphs of fecal and vaginal species annotations for the endometriosis group [ 30 ]. To identify potential biomarkers with statistical differences between the groups, metagenomic features were analyzed using Linear Discriminant Analysis Effect Size (LEfSe) [ 31 ]. In this study, taxa with a linear discriminant analysis (LDA) score greater than 2 (log 10) were considered significant as LDA > 2 is commonly used as a threshold for statistical significance in many microbiome studies [ 31 , 32 , 33 ]. Clinical variables and estrogen metabolite levels were compared using unpaired t ‐tests or Wilcoxon rank‐sum tests, as appropriate based on data distribution. A two‐sided p  < 0.05 was considered statistically significant. To account for multiple comparisons, false discovery rate (FDR) correction was performed using the Benjamini–Hochberg method. Partial Spearman correlation analyses were conducted to assess associations between microbiota composition, enzyme activity, and estrogen metabolites while accounting for potential interdependencies among variables (Tables  S3–S6 ).

Table S1: Top 10 genera of gut microbiota in the case and control groups. Table S2: Top 10 genera of vaginal microbiota in the case and control groups. Table S3: Top‐ranked correlations between gut microbiota and estrogen metabolites. Table S4: Correlation between enzyme activity and estrogen metabolites in stool samples. Table S5: Correlation between enzyme activity and estrogen metabolites in urine samples. Table S6: Top‐ranked correlations between gut and vaginal microbiota. Data S1: rmb270061‐sup‐0002‐DataS1@Supplementary Data.xlsx.