Endometriosis (EM) is a chronic inflammatory gynecological disease, with a high prevalence (~10%) of reproductive-age women, and affects approximately 176 million women worldwide (Giudice and Kao, 2004; Hickey et al., 2014). Defined as the implantation of endometrial-like tissue outside the uterus, the ectopic endometrium is commonly found in the pelvis. EM is associated with high morbidities of pelvic pain, dysmenorrhea, dyspareunia, and infertility (Nnoaham et al., 2011). The relative symptoms of EM are not exclusive, giving rise to the delay of diagnosis typically by 8–10 years (Greene et al., 2009). For many women, simply having a diagnosis of EM brings relief. Laboratory tests, ultrasonography, and magnetic resonance imaging (MRI), being parts of the initial battery of investigations, can only provide auxiliary diagnosis (Gupta et al., 2000). In order to establish a criterion-standard diagnosis and ease the debilitating syndrome, a minimally invasive laparoscopy operation will be adopted to confirm and remove the ectopic endometrium (Kiesel and Sourouni, 2019). However, patients who received primary surgery showed a high recurrence rate about 40%–45% and would engender further surgery (Chen et al., 2018), while the excision of endometrioma may elicit temporary detrimental effects on ovarian reserve (Garry, 2004; Goodman et al., 2016). To date, the pathogenesis of EM remains elusive, with current findings suggesting a multifactorial mechanism. One of the most broadly accepted hypotheses is proposed in 1927 by Sampson, suggesting that retrograde menstruation through fallopian tubes backs up into the peritoneal cavity followed by colonization of the endometrial tissue (Sampson, 1927). In addition, other theories suggest that fragmented endometrial tissue may migrate and implant through blood or lymphatic circulation, and hormone-induced metaplasia (Sampson, 1927; Rösner and Böger, 1979). Genetics, environmental factors, and immune system are also known to be correlated with the development of EM (Giudice and Kao, 2004). However, none of the theories have elucidated the pathogenesis of EM, indicating that unknown modulators may present during disease progression. Most importantly, 80% women at reproductive age experience retrograde menstruation, but the prevalence of EM is around 10%, giving us a hint that the microenvironment of peritoneal cavity may be different between EM patients and women without EM.

In the traditional concept, the upper genital tract and peritoneal cavity have been largely considered to be sterile areas. Recently, this concept has been challenged by a considerable number of studies investigating microbial composition across the human body (Chen et al., 2017; Ata et al., 2019). Microbial seeding within the peritoneal cavity may originate from three main routes, which include the lymphatic system, bloodstream, and female genital tract (Stinson et al., 2017; Baker et al., 2018). In the past decades, we have reached the concept that the microbiome plays a key role in maintaining human health. Dysbiosis can perturb the immunomodulation of the host and result in the development of several inflammatory diseases (Blaser and Blaser, 2014). With the rapidly developed technologies used in the sequencing of the human genome, a wide range of analytical tools are used to assess biological parameters, which are defined as biomarkers (Atkinson et al., 2001). If so, a desirable goal would be to seek for objective biomarkers for EM diagnosis at an early stage without surgical operation. Although a direct contact between the gut microbiota and EM has not been proven, studies on the pancreas have demonstrated that gut flora can migrate into the pancreas and influence the pancreatic microenvironment (Pushalkar et al., 2018), giving us a hint that the same pattern may exist for EM. Women with EM are at an elevated risk of developing depression and anxiety disorders, which are linked to the alteration of microbiome (Neufeld et al., 2014), compared to those without EM (Chen et al., 2016). There are increasing investigations on the correlation between microbiome and EM of murine models and women, which have provided preliminary evidence of the bidirectional effect between EM and microbiota as well (Yuan et al., 2018; Ata et al., 2019).

Up to now, the composition and variation of microbes in and on the human body remain uncertain. In a retrospective study researching the comparison of vaginal, cervical, and gut microbiota between women with stage 3/4 EM and healthy controls, a complete absence of genera of Atopobium in the vagina and cervix and Shigella/Escherichia-dominant stool in EM patients was observed (Ata et al., 2019). Nevertheless, serum cancer antigen 125 (CA125), neutrophil-to-lymphocyte ratio (NLR), cytokines, etc., were found to be potential diagnostic factors for EM (Muyldermans et al., 1995; May et al., 2010; O et al., 2018). These observations add importance to a possible infectious etiology in EM. However, few auxiliary diagnosis methods emerged from such analysis and either of the previous studies has demonstrated a predictive model with potential to separate EM patients from the control group.

Thus, to address the confounding variables associated with EM, we initially recruited women with EM relative symptoms under strict inclusion and exclusion criteria and combined 16S rRNA gene sequencing and manifold forms of bioinformatic analysis trying to solve the puzzle for the first time. The results show that microbial signatures of the gut and peritoneal cavity of EM individuals significantly differ from those of the controls, whereas the cervix harbors a low-density and low-richness microbiome and is dominated by Lactobacillaceae Lactobacillus. Moreover, we demonstrated the microbial network and detected specific fecal microbiota that were correlated with P. Pseudomonas in peritoneal fluid. Overall, this study comprehensively characterizes the microbiome of the gut, cervix, and peritoneal cavity in the control group and EM patients and elucidates the potential of microbial markers as non-invasive diagnostic tools for identifying EM patients from the control group. The bioinformatic analysis of the gut microbiome has provided stirring glimpses into the putative role of the gut microbiota in directing personalized medical therapy and patient management.

By profiling the peritoneal fluid microbiomes of women with endometriosis at the reproductive age, the current study provides a new clue that microbes in the peritoneal cavity may contribute to establish a pathogenic microenvironment. It should be noted that the retrograde menstruation theory raised by Sampson in 1927 was broadly accepted (Sampson, 1927). To date, 80% women at the reproductive age experience retrograde menstruation, but the prevalence of endometriosis is around 10% (Hickey et al., 2014). It gave us a hint that there may be unknown factors present in the peritoneal cavity that could possibly affect the local environment. Previous studies also demonstrated that the microenvironment of the peritoneal cavity in endometriosis appears to be pro-inflammatory (Osteen et al., 2005) and the peritoneal cavity was proven to be not sterile (Chen et al., 2017). However, investigations into microbes of the peritoneal cavity had few progresses for the difficulty of obtaining peritoneal fluid specimen. Collecting peritoneal fluid from healthy control would be considered highly inappropriate and unapproachable. Most importantly, women with endometriosis may have no symptom, and empirical diagnosis cannot rule out endometriosis. Instead, recruiting women with moderate gynecological diseases, with a root cause not involving peritoneal microenvironment, as control subjects, had allowed the investigation of the difference of the peritoneal microenvironment between EM patients and non-endometriosis subjects. On the other hand, symptoms of endometriosis often onset at the early age followed with a 4–11-year delayed diagnosis was constantly reported worldwide (Nnoaham et al., 2011; Hudelist et al., 2012; Agarwal et al., 2019). Performance of auxiliary diagnosis methods, including but not limited to transvaginal ultrasound (TVS), transabdominal ultrasound (TUS), and magnetic resonance imaging (MRI), varies between studies depending on multiple factors, such as radiologists’ experience and the fundamental structures of different facilities (Guerriero et al., 2013; Wildenberg et al., 2016; Guerriero et al., 2018). A solid diagnosis of endometriosis requires surgical visualization with histological verification (Rösner and Böger, 1979); however, it is unrealistic for all women with suspected endometriosis to receive an invasive surgery while the recurrence rate is high and the surgery could possibly damage the reproductive system (Ahn et al., 2017). In this regard, searching for a valid but non-invasive biomarker is an urgent need and of great benefit to women who do not wish to undergo a surgical operation. Studies more recently suggested that endometriosis may originate with immunology perturbations (Symons et al., 2018), which is in association with microbiome (Opdenakker et al., 2016). Microbiota, proved to play multiple critical roles within a host (Fan et al., 2021), is also implicated to be correlated with endometriosis (Yuan et al., 2018; Ata et al., 2019; Chadchan et al., 2019; Ni et al., 2020). Accordingly, establishing the relationship between the microbiota of different body sites for women with or without endometriosis and interrelationship of microbiota with different endometriosis stages may give us an extra knowledge on the susceptibility and aggravation of endometriosis. Based on the next-generation sequencing (NGS) technique and/or bacterial cultivation, investigators suggested that peritoneal cavity and female genital tract (FGT) harbor unique microbial communities (Chen et al., 2017; Akiyama et al., 2019; Ata et al., 2019). Moreover, microbial colonization of the peritoneal cavity could originate from the gastrointestinal tract or the lower female genital tract via bloodstream or ascension (Rescigno et al., 2001; Baker et al., 2018). Herein, using matched sampling, we collected feces, cervical mucus, and peritoneal fluid from 20 non-endometriosis and 21 endometriosis individuals aiming to reveal the correlation between endometriosis and microbiota from different body sites. Primarily, we demonstrated systemic analysis and provided clues in accessing the association between microbiota and endometriosis across three body sites. In addition, we have elucidated the aberrant taxonomic composition in feces and peritoneal fluid of endometriosis versus that of the controls, but conversely, cervical microbiota differs slightly among individuals. Most importantly, we have identified two microbial biomarkers that mightily contributed in differentiating endometriosis and complex co-occurrence network elucidating potential coexistence patterns of microbial communities. Eventually, robust endometriosis classifiers were constructed based on endometriosis-associated bacteria of feces and peritoneal fluid, with AUC = 0.840 and AUC = 0.886, respectively. Moreover, the results suggest that gut microbiota performs better than cervical microbiota in early diagnosis of endometriosis. The idea of electing endometriosis-relative taxa for endometriosis classifier establishment may allow us to diminish the confounding factors and precisely identify endometriosis patients by evaluating personal gut microbiota.

Decades of research into the etiology and pathogenesis of endometriosis have partially elucidated the underlying mechanisms involved in disease occurrence and progression, whereas endometriosis-related risk factors are still under exploration. Although ectopic endometrium tissue is mainly located in the peritoneal cavity, the microbiome from the gut and cervix might still have a role in modulating endometriosis (Ata et al., 2019; Perrotta et al., 2020); however, there is no study ever investigating the “cross talk” between the microbiome from the gut, cervix, and peritoneal fluid. Thus, by comparing α- and β-diversity as well as microbial profiling between microbiomes derived from different body sites, we identified a distinct inter-habitat microbial composition and potential correlation between different body sites. Species of the Enterobacteriaceae Klebsiella, especially K. pneumoniae, were reported to be opportunistic pathogens through producing extended-spectrum beta-lactamase and have the potential to cause spontaneous bacterial peritonitis (Jae et al., 2014). Another gram-negative bacterium of the family of Enterobacteriaceae, Salmonella, was able to reach the spleen after oral administration, suggesting the existence of penetration of the gut mucosa by pathogens (Rescigno et al., 2001). Interestingly, in the current study, Enterobacteriaceae Klebsiella was enriched in the cervix of endometriosis compared to the control group, indicating that E. Klebsiella may migrate to the peritoneal cavity and modulate the microenvironment by causing inflammation. On the other hand, Prevotella spp. are common oral, vaginal, and gut commensals, reported to be a potential mediator in modulating the body mass index and hormone biosynthesis. Bacteroidales S24-7 (recently classified as Muribaculaceae) were identified as potential mucus degraders in several previous studies (Berry et al., 2015; Lagkouvardos et al., 2019; Lee et al., 2019), therefore impeding pathogens that utilize mucus-derived sugars as crucial nutrients (Pereira et al., 2020). In this study, we observed pronounced abundance of Bacteroidales S24-7 in peritoneal fluid of the control group. Likewise, Enterobacteriaceae Klebsiella, Prevotellaceae Prevotella, and Bacteroidales S24-7 together with the remaining 16 microbial taxa shared by feces, cervical mucus, and peritoneal fluid may also have a role in modulating the microenvironment of the peritoneal cavity via metabolic function and/or penetration depending on the milieu. Thus, the current study provides a clue for future investigation into the association between the key commensal microbes at different body sites in the progression of endometriosis.

Up to now, little attention has been paid to key microorganisms and their function related to endometriosis. Specific negative associations with metabolic health problems, e.g., cardiovascular or inflammatory diseases, were detected for several genera, including Roseburia (Clostridiales order, Clostridia class), a genus of butyrate-producing colonic bacteria (Louis and Flint, 2009; Pichler et al., 2020). Reduction of butyrate eventually results in a disturbed integrity of epithelial cells and consequently in an increased colonic permeability (Kinoshita et al., 2002). In this study, the pattern of gut microbiota has been comprehensively elucidated and Clostridiales Clostridia, Clostridiales Lachnospiraceae, and Lachnospiraceae Ruminococcus were significantly reduced in FE compared with FN. This indicates that certain alterations of microbiome and metabolites in the gut may be involved in the regulation of physiological function. Unfortunately, no studies have demonstrated the interaction between these microorganisms and endometriosis, which are worth more focus and further exploration. Moreover, several of the taxa identified to be distinct between groups in this study are known to exert negative effects in intestinal homeostasis. Patients with Crohn’s disease are reported to harvest a decreased abundance of several genera, including Ruminococcus and Dorea, within the families of Ruminococcaceae and Lachnospiraceae (Gevers et al., 2014). Interestingly, in this study, Ruminococcaceae Ruminococcus and Lachnospiraceae Dorea were remarkably reduced in FE versus FN. Others, such as Eggerthella lenta, were significantly increased in FE, while being reported to be potentially harmful (Fattorusso et al., 2019), reminding us that the loss of protective microbes could be the triggering factor in exacerbating the proliferation of certain potential pathogens (Looft and Allen, 2012), inducing systemic inflammation. In conclusion, feces of endometriosis individuals were aberrant from that of the control subjects and beta diversity analysis has pinpointed endometriosis status as a potent factor in distinguishing FE from FN. Although the ectopic endometrium is mostly located in the peritoneal cavity, the microbial difference of peritoneal fluid between groups is not as prevalent as that of feces in our study. Endometriosis is characterized to be an inflammatory-like disease, and pro-inflammatory cytokines were enriched in PF of women with endometriosis (Dziunycz et al., 2009). Beta diversity analysis implies that PFE is significantly different from PFN, but the EM group did not separate from the control group completely. One possible explanation could be that peritoneum mesothelial cells continuously produce peritoneal fluid, followed by reabsorption through the large surface area of the peritoneum and removal of pathogens and cells ascending from the female genital tract (Blackburn and Stanton, 2014; Capobianco et al., 2017), resulting in a rather stable microenvironment. In the endometriosis group, we observed increased abundance of Gram-negative bacteria, including Pseudomonadaceae Pseudomonas and Prevotellaceae Prevotella, with the potential of releasing lipopolysaccharide (LPS) (Raia et al., 2005), which can activate macrophages and further induce secretion of a variety of local products, enhancing the proliferation of endometriotic stromal cells (Sakamoto et al., 2003).

Knowing that much difference exists in the gut microbiota and peritoneal microenvironment, we presume there might be microbial markers of endometriosis. Animal studies have elucidated that the gut microbiota might have a role in promoting endometriosis disease progression, and communication between the host and gut microbiome might be bidirectional (Yuan et al., 2018; Chadchan et al., 2019; Ni et al., 2020). In numerous human studies, noninvasive peripheral markers have been explored in blood, tissue, feces, cervical mucus, peritoneal fluid, and urine from endometriosis patients compared with non-endometriosis subjects: imaging, proinflammatory cytokines, miRNAs and metagenome, etc. (Upson et al., 2014; Wessels et al., 2015; Nematian et al., 2018). However, these research results are inconsistent possibly due to differences in study design, biologic sample collection, storage, data analysis workflow, and defecation time. Owing to ethical constraints, we adopted the least interferential protocol in collecting peritoneal fluid, which is internally consistent with Chen et al. (Chen et al., 2017). Moreover, all fecal samples were collected 1 to 3 days before the laparoscopy procedure in order to eliminate the variation which could be possibly caused by invasive surgery. Notably, we highlighted the unique bacterial differences with biomarker potential that are found in the feces and peritoneal fluid. The reduced level of Lachnospiraceae Ruminococcus in the gut and increased abundance of Pseudomonadaceae Pseudomonas in peritoneal cavity are closely associated with endometriosis, providing a hint on future diagnosis and therapy. Interestingly, Ruminococcus faecis was reported to be positively correlated with SCFA production, especially the biosynthesis of butyrate (Jin et al., 2019), believed to counteract gastrointestinal cancer and inflammation. In addition, formation of propionate, a type of SCFAs with potential health-promoting effects, was accomplished by Ruminococcus obeum (in the family of Lachnospiraceae) in the human large intestine (Jin et al., 2019). Taking into account the aforementioned specificities of SCFAs, we hypothesize that the reduction of L. Ruminococcus may result in decreased concentration of protective metabolites in the human intestine, leading to the development of diseases. On the other hand, arylsulfatase from Pseudomonas aeruginosa was able to enhance the steroid sulfatase activity (Uduwela et al., 2018), which is capable of catalyzing sulfated steroid precursors to the free steroid. Enhanced steroid sulfatase activity included catalyzing estrone sulfate to estrone, a weak estrogen. In previous results, prolonged exposure to estrogen was reported to be an environmental toxin correlated with the occurrence of endometriosis (Huhtinen et al., 2012). In summary, enriched P. Pseudomonas in the peritoneal cavity of endometriosis patients may be a result in the increased level of estrogen in the microenvironment, thereby increasing the estradiol concentration locally and thus exacerbating the proliferation of the ectopic endometrium (Huhtinen et al., 2012).

Despite the emerging evidence for a correlation of P. Pseudomonas and endometriosis (Akiyama et al., 2019; Hernandes et al., 2020), we cannot ignore the importance of bacterial migration and microbial communities themselves as a disease promoter. To peek into the relationship across gut flora and the potential pathogen, P. Pseudomonas, co-occurrence networks were built, in which both L. Ruminococcus and P. Pseudomonas were shown to be within certain communities. The commensal community, instead of an independent taxon, is also an intriguing possibility in disease occurrence. This observation emphasizes the need to unite opportunistic commensals in future research. The knowledge of the community composition of microbiome may precipitate strategies in mitigating the debilitating symptom of endometriosis and lead to the development of probiotics which has the potential to be an additional treatment for endometriosis patients.

Attempting to construct novel endometriosis classifiers, a hypothesis-free machine-learning strategy was adopted based on the taxa selected in 10 trials of RFCV from three body sites respectively, from which we concluded that the endometriosis classifier based on fecal or peritoneal fluid microbiota performs much better and steadily compared with that of cervical mucus. Interestingly, a similar trend of accuracy was shown in models built on the indicator species, with a mass overlap with random forest-selected taxa, indicating that microbiota in feces and peritoneal fluid is of greater importance in predicting endometriosis. Moreover, either obtaining cervical mucus or peritoneal fluid is not friendly to traditional concept and would acquire the guidance of professional physicians or surgeons. This suggests that fecal microbiota is of great importance in the classification of endometriosis compared to that of cervical mucus and peritoneal fluid, laying the basis for the usage of the endometriosis classifier model based on RFCV-selected taxa in subsequent studies. However, a limitation for the accuracy of models may be the handful of population included in the study; thus, validation in larger cohorts is of great necessity and urgent need. In previous studies, microbiota of cervical mucus was reported to have the potential in identifying endometrial cancer, endometritis, or infertility due to endometriosis (Cicinelli et al., 2008; Walther-António et al., 2016; Chen et al., 2017). Considering that the endometriosis classifier model based on the microbiota of cervical mucus presents low predictive validity in our study, we cannot exclude the possibility that disease cohorts, regional and individual differences in microbiota profiles, and methodological differences in the study design contributed to the taxonomic results.

Taken together, dysbiosis of the gut microbiota and altered microbial composition in the peritoneal cavity may together contribute to the development of endometriosis. The presence of P. Pseudomonas may be a potential risk factor in peritoneal cavity, and the depletion of L. Ruminococcus in the gut might be a biomarker for endometriosis. Whether endometriosis-related taxa and their intracorrelation or intercorrelation structure are causally linked to different clinical stages of endometriosis remains to be studies, but these findings clearly urge us to consider microbiota alteration in gut flora for novel diagnostic and prognostic research. Most importantly, investigation of the potential effect of P. Pseudomonas in the peritoneal microenvironment and elucidation of the underlying mechanism are emerging.

The original contributions presented in the study are publicly available in NCBI under accession number PRJNA753920.

The studies involving human participants were reviewed and approved by the Southern Medical University Clinical Research Ethics Committee. The patients/participants provided their written informed consent to participate in this study.

LX and YM designed the experiment and acquired grants. LH, BL, ZL, and ZC collected and analyzed the data. LH drafted the manuscript. YW, DP, XF, and HZ collected samples. All authors contributed to the article and approved the submitted version.

This work was supported by the ‘GDAS’ Project of Science and Technology Development (Grant Nos. 2021GDASYL-20210102003, 2018GDASCX-0102) to LX, National Natural Science Foundation of China (Grant No. 81701418), the College Students’ Innovative Entrepreneurial Training (Grant No. S202112121053), and the Scientific Research Enlightenment Program of Southern Medical University to YM. This work was also supported by the Guangdong Basic and Applied Basic Research Foundation (Grant No. 2020B1515020046) and by the National Natural Science Foundation of China (Grant Nos. 81900797, 82072436) to LX.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.