Oral, Vaginal, and Stool Microbial Signatures in Patients With Endometriosis as Potential Diagnostic Non-Invasive Biomarkers: A Prospective Cohort Study

F.E.‐A. conceived the study. F.E.‐A., G.C., E.M.E.‐O. planned and designed the study protocol. M.L. and M.E. recruited participants and collected patient data and samples. F.E.‐A., C.H. and L.M.‐B. performed laboratory work. X.‐Y.C. performed bioinformatic and data analysis and generated all figures. C.H., M.L., X.‐Y.C., L.M.‐B., E.M.E.‐O., G.C. and F.E.‐A. took part in interpretation of results and writing of the manuscript.

This study was approved by the Nepean Blue Mountains Local Health District Human Research Ethics Committee (HREC), approval number LNR/18/Nepean/18.

We performed a prospective cross‐sectional pilot study involving three cohorts of participants, as summarised in Figure  1 . Informed consent was obtained for all participants. Two of the cohorts were recruited from patients receiving gynaecologic care at a single tertiary care institution Nepean Blue Mountains Hospital Local Health District (NBMLHD) in New South Wales, Australia from 2018 to 2019 (LNR/18/Nepean/18). These patients had clinical indications warranting laparoscopy and were surgically investigated for the presence of endometriosis. Histological diagnosis was performed on any lesion findings at the hospital pathology laboratory and staged according to the revised American Society of Reproductive Medicine (rASRM) classification of endometriosis [ 30 ]. These participants were separated into two cohorts: confirmed endometriosis (ENDO) and confirmed no‐endometriosis (N‐ENDO). The N‐ENDO cohort had confirmed gynaecological pathologies such as fibroids, abnormal uterine bleeding and ovarian cysts. Exclusion criteria included post‐menopausal women and women with concerns for cancer. Patients with previous surgical and/or ultrasound diagnoses of endometriosis were included in the study. Due to the gynaecological conditions prevalent in the N‐ENDO cohort, a third healthy control (HC) cohort, defined as women with no known gynaecological symptoms or infertility concerns, were recruited from the University of NSW Microbiome Research Centre MothersBabies Study (Ethics: 2019/ ETH00192 ) and were age‐ and sex‐matched to the ENDO and N‐ENDO cohorts [ 31 ]. Overview of sample collection and study cohorts. (A) All participants were provided with oral, stool and vaginal sample collection kits. (B) Healthy controls (HC) collected samples at a single time point. Endometriosis (ENDO) and non‐endometriosis (N‐ENDO) cohorts collected samples preoperatively, at a single time point. N‐ENDO and ENDO participants underwent laparoscopy and were surgically investigated for the presence of endometriosis, and histological diagnosis was performed on any lesion findings at the hospital pathology laboratory and staged according to the revised American Society of Reproductive Medicine (rASRM) classification of endometriosis. n  = 10 rASRM, stage 1; n  = 1, stage 2; n  = 2, stage 3; and n  = 7, stage 4. Stage 1 and 2 = minimal/mild, stage 3 and 4 = moderate/severe. The endometriosis core outcome set was developed in 2020 to standardise endometriosis research and includes outcomes such as overall pain, symptom improvement and quality of life [ 32 ]. The core outcome set was not considered when designing this study, as this study was designed in 2017, prior to the publication of the core outcome set. However, this study inadvertently addresses the core outcome set by aiming to identify a microbiota‐based signature that could potentially function as a diagnostic non‐invasive biomarker for endometriosis, thus reducing diagnostic delay. There was no patient or public involvement in the design of this study. All participants were provided with sample collection kits. Oral, stool and vaginal samples were self‐collected at home. Samples were collected at a single time point for healthy controls and preoperatively for participants with clinical indication for laparoscopy. Oral and vaginal samples were collected using FLOQSwab and eNAT tube collection sets (Copan, Italy). Stool samples were collected using gloves, a sterile ColOff stool collection device and Stool Collection Tubes with DNA Stabiliser (Invitek Molecular GmbH, Germany). Upon collection, samples were stored at –80°C until DNA extraction. DNA was extracted from oral and vaginal samples using QIAamp DNA Mini Kit (Qiagen, Hilden, Germany), according to manufacturer's instructions. The PSP Spin Stool DNA Basic Kit (Invitek Molecular GmbH, Germany) was used to extract DNA from stool samples, according to manufacturer's instructions. Extracted DNA was delivered on ice to the UNSW Ramaciotti Centre for Genomics (University of New South Wales, Sydney, Australia) for DNA sequencing. Samples were sequenced according to the Illumina 16S Metagenomic Sequencing Library Preparation protocol. The V3–V4 hypervariable region of the 16S rRNA gene (ca. 464 bp fragment) was amplified using 341F‐805R primers [ 33 , 34 ]. The full‐length primer sequences with Illumina adapter sequence in lowercase were 341‐F: 5′‐tcgtcggcagcgtcagatgtgtataagagacagCCTACGGGNGGCWGCAG‐3′ and 805R: 5′‐gtctcgtgggctcggagatgtgtataagagacagGACTACHVGGGTATCTAATCC‐3′. The total volume of each sample was 25 μL, including 2.5 μL of microbial DNA (5 ng/μL), 5 μL of Amplicon Polymerase Chain Reaction (PCR) Forward Primer 1 μM, 5 μL of Amplicon PCR Reverse Primer 1 μM and 12.5 μL of 2× KAPA HiFi HotStart ReadyMix (KAPA Biosystems, USA). The PCR cycle began with initial denaturation at 95°C for 3 min, followed by 25 cycles of 95°C for 30 s (denaturation), 55°C for 30 s (annealing) and 72°C for 30 s (elongation). This was followed by 72°C for 5 min and then hold at 4°C. Upon completion of PCR and clean‐up, the Nextera XT Index Kit was used to add Illumina sequencing adapters and dual indices to the PCR products to enable multiplex sequencing. The SequalPrep Normalisation Plate Kit (Invitrogen, USA) was used to clean and normalise the samples, according to manufacturer's instructions. The samples were then sequenced using the MiSeq v3 reagent kit (Illumina, USA) and a 2 × 300 bp paired‐end run, using the MiSeq Sequencer, according to manufacturer's instructions (Illumina, USA). A Phi‐X spike‐in was used in each run as an internal control against low‐diversity samples with unbalanced nucleotide composition. Blanks were also included in the sampling runs. Sequenced amplicon reads were processed following the QIIME2 [ 35 ] pipeline installed via Conda: Reads were denoised, dereplicated and filtered for chimeric reads using DADA2 [ 36 ], then clustered to generate amplicon sequence variants (ASVs). Representative reads for each ASV were assigned to a taxon using a Naïve Bayes Classifier trained on the V3–V4 hypervariable region of the 16S rRNA gene from the Greengenes v13_5 database [ 37 ]. Data normalised by rarefaction to a sample depth of 49 000 reads/sample for stool, 86 000 reads/sample for oral and 46 000 reads/sample for vaginal samples were performed for diversity analysis. We performed alpha‐ and beta‐diversity comparisons to test if there were any differences between the three cohorts using R (v4.2.0, R Core Team). Alpha diversity summarises the microbial biodiversity of a single sample (or community), and this was assessed using the Shannon diversity index. Testing for cohort differences used either Wilcoxon rank‐sum test for two cohorts or Kruskal–Wallis tests when there were more than two cohorts. Bray–Curtis dissimilarity was used for assessing beta diversity, which compares the compositional similarity between two samples (or communities). Beta‐dispersion and permutational multivariate analysis of variance (PERMANOVA) [ 38 ] tests were used for assessing compositional differences between cohorts via the R vegan package [ 39 ]. Samples were also visually examined with principal coordinate analysis (PCoA) plots. The detection of differentially abundance taxa was performed using the linear discriminant analysis effect size (LEfSe) [ 40 ] method. In this study, we investigated differential microbial markers comparing the three cohorts (HC, N‐ENDO, and ENDO) and then, within the ENDO cohort, we compared minimal/mild‐stage and moderate/severe‐stage endometriosis. Plots for all analysis were generated using ggplot2 [ 41 ], ggpubr [ 42 ] and cowplot [ 43 ] in R.

There were a total of 64 subjects in this study. n =  19 were healthy controls (HC); n =  24 underwent laparoscopy with confirmed no‐endometriosis (N‐ENDO) and n =  21 underwent laparoscopy with confirmed endometriosis (ENDO) ( n =  11 rASRM, stage 1; n =  1, stage 2; n =  2, stage 3; and n =  7 stage 4). Stages 1 and 2 were further grouped as minimal/mild endometriosis and stages 3 and 4 grouped as moderate/severe endometriosis. The mean age of participants was 34.6 years (SD = 7.0, range 22–49) and there were no significant differences in the age between the cohorts (Table  1 ). Participants had a mean weight of 74.5 kg (SD = 19.1, range 47–156), with no significant differences among the three cohorts. Characteristics of the study cohorts and number of samples collected. Fisher's exact test p  = 0.4 In total, there were 44 million reads sequenced for 192 samples, and after quality control, 28 million reads remain with a mean of 146 719 reads/sample (SD = 75 277). The number of taxon detected across the three body sites were 204 genera (18 phyla) in stool; 190 genera (20 phyla) in oral and 296 general (19 phyla) in vagina samples. Figure  2 shows the relative abundances of the top seven phyla within each body site, and samples are further grouped based on the cohorts (Figure  S1 shows the top seven genera that are detected in each sample). Relative abundances of the top seven most abundance phyla sequenced using 16S rRNA. Samples are charted horizontally by body site: (A) oral, (B) stool and (C) vaginal samples and separated in vertical panels according to study cohorts: Healthy controls (HC), patients with confirmed no‐endometriosis (N‐ENDO) and patients with confirmed endometriosis (ENDO). All other detected phyla not in the top seven are cohorted together and labelled as ‘Remaining’. Alpha‐diversity comparisons revealed borderline differences between HC and ENDO group in oral samples (adjusted p  < 0.1, estimate = 0.11, 95% CI 7.4e−5–0.22) and significant differences in stool samples (adjusted p  < 0.05, estimate = 0.27, 95% CI 0.09–0.44), Figure  3 full statistical results listed in Tables  S2 and S3 . Generally, HC had higher α‐diversity compared to N‐ENDO and ENDO. There were no statistically significant differences between the α‐diversity of N‐ENDO and ENDO. There were no significant differences observed among vaginal samples between cohorts. Comparison of alpha diversity, measured by Shannon diversity index (based on features at genus rank), between study cohorts of healthy controls (HC), patients with confirmed no‐endometriosis (N‐ENDO) and confirmed endometriosis (ENDO). Comparisons were assessed using Kruskal–Wallis test and then followed by Wilcoxon rank‐sum tests for pairwise comparisons adjusted using the FDR approach (text annotation). There were borderline differences (adjusted p  < 0.1, estimate = 0.11, 95% CI 7.4e−5–0.22) between HC and ENDO group observed in oral samples (A) and significant differences in stool (adjusted p  < 0.05, estimate = 0.27 and 95% CI 0.09–0.44) (B) samples. There was no difference in alpha diversity based on the vaginal samples (C). Significant microbial composition differences between the three groups ( p  < 0.05) were observed in stool and oral samples with R 2  = 3.9% explained variance in stool, and R 2  = 4.3% explained variance in oral samples; while no significant differences between groups were observed in the vaginal samples ( R 2  = 2.6% explained variance, Figure  4 ). This indicated that at least one of the three cohorts was different to the others. Subsequent pairwise PERMANOVA analysis revealed differences between HC and N‐ENDO were significant in stool (adjusted p  = 0.003) and the same in oral (adjusted p  = 0.042). The compositional differences between study cohorts in stool were no longer significant after aggregating the ASV features to genus rank for comparison ( p  = 0.229) suggesting that the differences in composition may be contributed by the species. However, the compositional differences in oral samples remained significant when using genus features for comparison ( p  = 0.042, Figure  S2 and Tables  S4 and S5 summarise the PERMANOVA test results). Principal coordinate analysis (PCoA) plots of beta diversity measured using Bray–Curtis dissimilarity between study cohorts of oral samples (A), stool samples (B) and vaginal samples (C). Results from PERMANOVA tests are annotated in the top right corner where significant differences between cohorts were observed in stool and oral samples, and there were no significant differences observed in the vaginal samples. HC, Healthy controls; N‐ENDO, Patients with confirmed no endometriosis; ENDO, Patients with confirmed endometriosis. We also tested the beta dispersions for any differences in variability between cohorts. Significant differences in stool samples were observed ( p  = 0.044, Figure  S3 and Table  S6 ) where HC has lower dispersion compared to N‐ENDO and ENDO. This suggests that the stool microbial diversity in patients with and without endometriosis is more variable compared to HC. There were no differences in beta dispersion for the vaginal or oral samples ( p  > 0.05). In total, 36 microbial features were detected to be differentially abundant between the three cohorts and across the three body sites using LEfSe analysis. The detected features are summarised in Figure  5 (with LDA effect size and p ‐values output summarised in Table  S7 , while a comparison with other studies are included in Table  S1 ). More specifically, in stool samples, there were 13 genera of which eight were increased in HC, two in N‐ENDO, and three in ENDO. In oral samples, there were 14 genera of which eight were increased in HC, three in N‐ENDO, and three in ENDO. In vaginal samples, there were 12 genera of which four were increased in HC, four in N‐ENDO and four in ENDO. Overview of taxonomy features detected as differentially abundant using LEfSe analysis. Colours denote the cohorts the corresponding taxon is higher in relative abundance. Samples from the three body sites are separated into vertical panels. HC, Healthy controls; N‐ENDO, Patients with confirmed no endometriosis; ENDO, Patients with confirmed endometriosis. The three body sites largely displayed unique profiles for the three conditions with little overlap. Only three genera appeared in multiple body sites, Lactobacillus , Haemophilus and Prevotella , but the cohort in which they were discriminatory was inconsistent. For instance, Lactobacillus in stool samples was enriched in ENDO, but in oral samples, it was enriched in N‐ENDO; and similarly for the other two genera. This is most likely due to the actual species present in the different body sites. There were eight taxonomic features detected as different when comparing patients with minimal/mild versus moderate/severe endometriosis, which is summarised in Figure  6 (and Table  S8 consist the results from LEfSe). There were no overlaps in detected taxon with each body site generating its own profile. In stool, Actinomyces was enriched in patients with minimal/mild endometriosis and Paraprevotellaceae was enriched in patients with moderate/severe endometriosis. In oral samples, Cardiobacterium was enriched in patients with minimal/mild endometriosis and Fusobacterium was enriched in patients with moderate/severe endometriosis. In vaginal samples, Blautia , Dorea , Collinsella and Eubacterium were enriched in patients with moderate/severe endometriosis. Overview of taxonomy features detected as differentially abundant using LEfSe analysis between patients ( n  = 12) with minimal/mild stage endometriosis (pink) and patients ( n  = 9) with moderate/severe endometriosis (purple). Samples from the three body sites are separated into vertical panels.

Overall, diversity analyses revealed compositional differences in the oral and stool microbiota between HC, N‐ENDO and ENDO cohorts. LEfSe analysis revealed differentially abundant microbial taxa within each cohort. Study strengths include the robust diagnostic capabilities and intraoperative assessment of ENDO and N‐ENDO by clinicians, permitting correct classification of disease presence or absence, and severity, as well as the comprehensive evaluation across three anatomical sites (oral, stool and vaginal), within three distinct cohorts (ENDO, N‐ENDO and HC). Establishing a definitively disease‐free HC group is difficult due to the at times asymptomatic presentation of endometriosis. It is also ethically challenging to confirm the absence of disease in a HC cohort with laparoscopy, making this is a limitation of our study. However, this limitation reinforces the heterogeneity of the disease and underscores the complexities of endometriosis research. Other limitations in our study include a moderate sample size, lack of data on menstrual cycle phase during sample collection, potential confounders such as other gynaecologic pathologies as well as data on the oral administration of hormones, antibiotics, probiotics or supplements. While we did collect information regarding hormone, antibiotic, probiotic and supplement use upon preoperative sample collection, we were unable to include these features in analysis due to the high rate of missing data in the ENDO group (9/21, > 40%) without introducing other potential biases via imputation. Moreover, variables such as dietary behaviours, smoking habits and recent sexual activity were not captured in this pilot study but warrant future considerations in light of their impact on the microbiome. Although differentially abundant microbial taxa were identified, inconsistent findings across existing literature in relation to individual taxa makes it challenging to determine what constituents are relevant as potential biomarkers of endometriosis. Table  S1 summarises our findings within the context of previous research. Notably, we found that Escherichia was enriched in ENDO vaginal microbiota. A positive correlation between ENDO and Escherichia/Shigella in the vaginal microbiome has been previously identified [ 7 ]. Ata et al. [ 8 ] found that Escherichia/Shigella was enriched in the vagina and cervix of ENDO patients after excluding Lactobacillus. Escherichia/Shigella was also found to be enriched in an adenomyosis/ENDO cohort, compared to ENDO and HC groups individually [ 12 ]. Finally, Escherichia coli was enriched in the endometrium, as well as the menstrual blood of patients with ENDO [ 16 ]. Additionally, Enterococcus was enriched in ENDO vaginal samples. Previously, Enterococcus was enriched in cervical mucus of ENDO but depleted in the endometrium [ 22 ]. Tepidimonas was also increased in vaginal samples and previously has been shown to be enriched in the endometrial microbiota of ENDO [ 23 ] samples. The increased abundance of Escherichia , Enterococcus and Tepidimonas in ENDO vaginal microbiota may be supportive of the ‘bacterial contamination hypothesis’. This theory postulates that the bacterial endotoxin lipopolysaccharide activates the Toll‐like receptor 4 cascade, causing inflammation that facilitates lesion development [ 16 , 44 ]. Further investigation is required to validate this hypothesis; however, this theory holds potential as both a diagnostic tool and a possible therapeutic target. Within the stool microbiota, Phascolarctobacterium was enriched in the ENDO cohort. Previous studies have identified an increased abundance of Phascolarctobacterium in peritoneal fluid following induction of endometriosis in female baboons [ 17 ], and it was also found to be more prevalent in the peritoneal fluid of patients with endometriosis compared to controls. Given the proximity of the gut to the peritoneal cavity, the detection of Phascolarctobacterium in both the gut and peritoneal fluid raises the possibility of intestinal bacterial translocation into the peritoneum. This idea has recently been postulated in the case of peritoneal fibrosis in chronic kidney disease, in which it is suggested that translocation of gut bacteria causes intraperitoneal chronic low‐grade inflammation, thereby promoting fibrosis [ 45 ]. Future work should aim to investigate this concept within endometriosis. Moreover, Lactobacillus was more prevalent in ENDO stool microbiota samples. Interestingly, multiple Lactobacillus species are constituents of the ‘estrobolome’, a term used to describe microbes that inhabit the gastrointestinal tract and encode β‐glucuronidase and β‐galactosidase enzymes. These enzymes have the capacity to increasing circulating oestrogen levels, and therefore, the estrobolome is suggested to be implicated in endometriosis disease progression [ 45 ]. Although the contribution of the estrobolome to disease within our study cohort is undetermined, exploring this concept in a larger cohort is warranted, given its potential as a therapeutic target. Future work should also aim to correlate environmental factors such as air pollution, which is known to impact the function and composition of the gut microbiota [ 45 ]. Finally, Fusobacterium was enriched within the oral samples of moderate/severe endometriosis. Interestingly, a recent study found Fusobacterium was enriched in the endometrium and endometrial lesions of endometriosis participants, compared to controls [ 27 ]. Fusobacterium infection facilitated endometriosis lesion development and antibiotic therapy reduced lesions in a mouse model [ 27 ]. While we detected Fusobacterium within oral samples, it is well established that Fusobacterium is an opportunistic pathogen that can cause periodontal disease [ 46 ], and interestingly, women with endometriosis have higher incidences of periodontal disease, compared to those without endometriosis [ 47 , 48 ]. Although our sample size for the analysis of ENDO stages was limited, our finding, coupled with the findings of Muraoka et al., and the association of endometriosis and periodontal disease warrant deeper investigation into the role of Fusobacterium in causing inflammation at both local and systemic levels. Future work should aim to increase sample size to validate a consistent microbial signature for endometriosis diagnosis and treatment. Furthermore, investigating host–microbiome interactions would enable greater understanding on the complex relationship between the microbiome and endometriosis. Exploring the potential mechanisms of intestinal bacterial translocation in endometriosis, as well as the estrobolome presents an intriguing avenue for future research. Additionally, elucidating the impact of Fusobacterium enrichment in oral samples on local and systemic inflammation could provide valuable insights into disease aetiology. While our study provides an overview of the microbiome dynamics in endometriosis, addressing identified limitations, validating a microbial signature and exploring novel hypotheses are essential for advancing our understanding of this disease.

With a larger sample size and improved control of confounders, this pilot study provides evidence to support the potential development of a non‐invasive microbial‐based diagnostic biomarker for endometriosis. Our study is the first to include three cohorts: healthy controls, non‐endometriosis and endometriosis, while also examining the microbiota of multiple body sites, including the mouth, gut and vagina. The variability observed in the existing literature demonstrates the necessity for establishing standardised methods in endometriosis microbiome research. Further research is needed to understand the delicate interactions and mechanisms that underpin the dynamic role of the microbiome pathogenesis of endometriosis.

Endometriosis is a chronic, inflammatory, oestrogen‐dependent condition that affects approximately 10% of women of reproductive age [ 1 ]. It is defined as the presence of endometrial‐like tissue outside of the uterine cavity and symptoms include dysmenorrhoea, dysuria, pelvic pain as well as subfertility or infertility [ 2 ]. The gold standard for diagnosis is laparoscopic visualisation of lesions, and on average, women with endometriosis experience a 6.7‐year delay in diagnosis [ 3 ]. As such, there is a need for the discovery of a non‐invasive diagnostic biomarker to facilitate earlier diagnosis and therefore early intervention [ 1 ]. The cause of endometriosis remains unknown [ 4 ]. The microbiome, defined as the genetic material of bacteria, fungi, archaea and viruses that reside within or on the body, are known to regulate several pathophysiological functions and are involved in the development of several inflammatory diseases [ 5 , 6 ]. Both human and animal studies have shown that there are differences in the microbiome composition of individuals with and without endometriosis [ 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 ]. Chen et al. [ 12 ] described the first report of distinct bacterial communities along the reproductive tract between women with or without endometriosis. A systematic review of studies examining the relationship between endometriosis and the microbiome concluded that endometriosis appears to be associated with elevated levels of Proteobacteria , Enterobacteriaceae , Streptococcus and E. coli across various microbiome sites [ 26 ]. Interestingly, a recent study in mice found that Fusobacterium infection may contribute to the pathogenesis of endometriosis and treatment with antibiotics reduced endometriosis lesions [ 27 ]. To date, most studies characterising the microbiota in endometriosis have investigated a single body site and have compared endometriosis and non‐endometriosis cohorts, which were validated by examining the presence of endometriosis via laparoscopy [ 28 ]. The symptom heterogeneity of endometriosis combined with the invasive nature of surgical diagnosis poses ethical challenges in establishing a confirmed endometriosis‐free cohort [ 1 ]. To our knowledge, no study has incorporated a healthy control cohort that is free from gynaecological disease. The role of the microbiome in the development of endometriosis is unknown. In this study, we aimed to describe the oral, stool and vaginal microbiota‐based signatures in patients with and without endometriosis, including self‐reported healthy controls. A microbial signature is defined as the unique composition of microbial communities within a specific body site and/or niche that correlates with the host phenotype [ 29 ]. We hypothesised that microbiota diversity would be different among the three cohorts across three body sites: mouth, gut and vagina, and that these differences could stratify the diagnosis of endometriosis. This study was performed with the aim of identifying a microbiota‐based signature that could potentially function as a novel diagnostic non‐invasive biomarker for endometriosis.

Mathew Leonardi reports grants from Australian MRFF, AbbVie, CanSAGE, Hamilton Health Sciences, Hyivy, Pfizer; honoraria for lectures/writing from AIUM, GE Healthcare, Bayer, AbbVie, TerSera, consultancy work with Hologic, Chugai, Roche Diagnostics, AIMA, affiliations with Imagendo, outside the submitted work. George Condous reports grants from Australian MRFF, ASUM and Endometriosis Australia, Honorarium from GE Healthcare, Samsung, affiliated with IMAGENDO, outside the submitted work.

Data S1