A role for gut mycobiome and altered fungal-bacterial interactions in women with endometriosis†

The women enrolled in the study were 18–45 years of age. All enrolled endometriosis patients ( n =18) were confirmed with laparoscopic examination and underwent surgeries for removal of endometriotic lesions. The biopsies of ectopic endometriotic lesions derived from patients were assessed to identify the disease stage to confirm the severity of the condition in each patient. Although age differed significantly between the two groups, it was not found to affect microbiome diversity in the regression analysis in both control ( p =0.65) and patient ( p =0.06) groups which suggested no significant relationship between age and microbiota profiles. Table 1 summarizes the characteristics of human subjects. The workflow for the study of gut fungal profiles in women with endometriosis ( n =18) and healthy cohorts ( n =31) is illustrated in Figure 1A . The fungal alpha diversity index based on observed OTUs did not differ significantly ( p =0.063) between endometriosis patients versus control subjects with a moderate effect size (Cohen’s d =0.53), meaning that the magnitude of difference between the two groups is not very large ( Figure 1B ). We next compared the two groups in terms of beta diversity obtained using PCoA based on unweighted Bray–Curtis distances ( Figure 1C ). The matrix computes distances based on dissimilarities between all shared and non-shared taxa. This analysis of beta diversity further revealed that the types of fungal species between the patients and control healthy women were only significantly different ( p =0.043) when a large proportion of the variation could not be explained ( Figure 1C ). A low R 2 value (0.027) also indicated a small proportion of explained variance, suggesting that there is not much differentiation in the gut fungal diversity between the healthy subjects and the patients. The F- statistic of 1.33 (PERMANOVA) indicated a moderate ratio of the between-group variability to within-group variability, i.e., the separation between the two groups is not as large at the level of gut fungal diversity. Next, we evaluated the gut mycobiota profiles at phylum and genus levels to understand these trends within the samples. At phylum level, Ascomycota, Basidiomycota, and Mucoromycota were predominant in both the control and endometriosis groups ( Figure 1D and E ). The top abundant genera in both groups included Saccharomyces, Penicillium, Candida, Aspergillus , and Cladosporium ( Figure 1F ). While the top genera in both groups were the same, we noted that the difference was slightly more apparent in the less prevalent genera, such as Dipodascus and Nakaseomyces , etc. ( Figure 1F ). We previously compared the bacterial profiles in these control and patient groups by performing 16S rRNA gene sequencing where we noticed a significantly reduced bacterial diversity among the endometriosis patients [ 15 ]. The bacterial populations of Ruminococcus , Faecalibacterium , Alistipes , and Roseburia significantly decreased in endometriosis, whereas Erysipelatoclostridium , Streptococcus , Lacticaseibacillus , Actinomyces , and Tyzzerella were increased (Mann–Whitney; p <0.05) [ 15 ]. Therefore, while the altered gut bacteria are more evident in dysbiosis, we noted that the fungal profiles do not alter drastically. Notably, in this regard, we determined a significant lower relative abundance of the less prevalent genera Talaromyces ( p =0.004) and Rhodotorula ( p =0.013) in endometriosis patients ( Figure 1G ). Since the mycobiota is largely constituted by the few abundant taxa, their functional interactions within the community might underscore the state of homeostasis. Therefore, to account for these insignificant variations determined in the two groups, we integrated the profiles of gut bacterial microbiota to determine these community interactions. We hypothesized that not the mycobiota diversity but their interactions with the gut bacteria underpin the disease states in endometriosis. Therefore, we studied the correlations between fungi and bacteria to evaluate the alterations in the bacteria–fungal interactions ( Supplementary Table 1 ). The analysis demonstrated several fungal genera in endometriosis that correlate distinctly with the gut bacteria versus the control subjects ( Figure 2 and Supplementary Table 1 ). In the order from highest to lowest number of interactions with gut bacteria, we observed genera Saccharomyces, Candida, Penicillium, Trichoderma, Aspergillus , and Talaromyces in the control group (first network in Figure 2A , SparCC>0.2). On the other hand, Penicillium formed the highest number of interactions in endometriosis followed by Saccharomyces, Cladosporium, Aspergillus , and Candida (first network in Figure 2B , SparCC>0.2). All findings were supported by the large difference (effect size=1.6) in the correlations observed within the two groups. We noticed that the coefficients of several of these co-abundances were stronger (SparCC>0.5) in endometriosis (side panel in Figure 2B ) compared with the control group (side panel in Figure 2A ),while the correlations within the two groups were clearly distinct (SparCC>0.5; effect size=1.4). Genus Saccharomyces , for instance, formed the largest number of correlation patterns with a number of bacteria in the control group ( Figure 2A ); however, it only showed a stronger negative correlation (SparCC>0.5) with Bacteroides ( Figure 2B and C ). On other hand, the co-abundance relationships of Saccharomyces were stronger with several bacterial genera in endometriosis, as it additionally showed a negative correlation with Ruminococcaceae incertae sedis and positive correlation with Dorea, Actinomyces, Lactococcus, Gemella, Erysipelatoclostridium , and Lachnospiraceae UCG 008 (side panel in Figure 2B and C and Supplementary Table 1 ). Penicillium , which displayed a strong positive correlation with bacterium Akkermansia within heathy control group (side panel in Figure 2A and C and Supplementary Table 1 ), did not share the same correlation pattern in endometriosis. Instead, it formed positive correlation with Oscillospiraceae and Ruminococcus , and significant negative correlations with Actinomyces, Lactococcus, Gemella, Rothia, Erysipelatoclostridium , and Lachnospiraceae (side panel in Figure 2B and C and Supplementary Table 1 ). Further, we found Cladosporium and Aspergillus forming significant and stronger co-abundance patterns with several bacterial groups in endometriosis compared with healthy control group. Cladosporium displayed a negative correlation with Clostridium sensu stricto 1 and Streptococcus while it showed positive correlation with commensals Agathobacter and Subdoligranulum (side panel in Figure 2B ). Similarly, Aspergillus also showed a strong positive correlation with Subdoligranulum while a negative correlation with genus Rothia (side panel in Figure 2B and C ). Notably, Candida showed a strong negative correlation with Enterorhabdus only in the healthy group (side panel in Figure 2A and C and Supplementary Table 1 ).Overall, the co-occurring relationships were more significant in endometriosis versus the healthy groups, indicating the critical role of gut fungi in the disease. As the bacteria form a major component of the gut microbiome and therefore interact with other bacteria on a larger scale, we also determined the interactions among different groups of bacteria to explore how they differ in healthy versus endometriosis groups ( Supplementary Figure 1 ). The bacterial–bacterial co-abundance analysis revealed clearly distinct correlation networks in patients with endometriosis versus the controls ( Supplementary Figure 1 , SparCC r >0.2; effect size=1.2). Interestingly, bacterial “keystone” taxa, which could be defined as organisms that might play a central role in the overall intestinal microbial community stability and function in healthy versus disease states, were determined from these interactions when we carefully looked at the subset of strongly correlated taxa ( Figure 3A and B ; SparCC r >0.5; effect size=1.4). In healthy controls, this network of communities was dominated by commensals such as Lachnospiraceae, Dorea, Faecalibacterium, Bacteroides , and Erysipelatoclostridium ( Figure 3A ). While Dorea, Erysipelatoclostridium, Lachnospiraceae , and Bacteroides also dominated these bacterial interactions in endometriosis, in addition, Actinomyces , Turicibacter, Ruminococcaceae Incerate Sedis, Enterococcus, Gemella, Lactococcus, Eubacterium halii group, and Escherichia-Shigella groups dominate these bacteria interactions in endometriosis ( Figure 3B and Supplementary Table 2 ). One notable subset of interactions in endometriosis was the Turicibacter that showed significant positive correlation with Escherichia-Shigella group that is largely an opportunistic pathogenic group of bacteria ( Figure 3B and Supplementary Table 2 ). Further, it formed additional negative correlation with Eubacterium hallii group and Agathobacter . We further noted several differences in the interactions in endometriosis versus control subjects. Notably, in endometriosis patients, Actinomyces was strongly positively correlated with Enterococcus, Gemella , and Rothia , while it displayed significant negative correlation with Bacteroides , Ruminococcaceae Incerate Sedis, Collinsella , and Oscillospiraceae ( Figure 3B and Supplementary Table 2 ). On the other hand, Actinomyces correlated with altogether distinct genera such as Faecalibacterium (negative) and Eggerthella (positive) in healthy control subjects ( Figure 3A ). Similarly, Dorea correlated with Fusicatenibacter and Lachnoclostridium (positive) and with Alistipes (negative) in control group ( Figure 3A and Supplementary Table 2 ); while it correlated with Ruminococcus torques group, Blautia, Enterococcus, Erysipelatoclostridium, Lactococcus, Atopobium , and Anaerostipes (positive) and with Alistipes, Ruminococcaceae Incerate Sedis, and Bacteroides (negative) in endometriosis ( Figure 3B ). Intriguingly, although some genera were commonly observed in both control and endometriosis groups, they formed evidently distinct interactions in the control versus patients. For example, Eggerthella interacted (positive) with Actinomyces in control group ( Figure 3A ), while it interacted with Lachnospiraceae (positive) and Coprococcus (negative) in endometriosis ( Figure 3B and Supplementary Table 2 ). Similarly, Fusicatenibacter interacted with Dorea (positive) in control ( Figure 3A ), while it interacted with Agathobacter and Lachnospiraceae UCG008 (positive) and Roseburia (negative) in endometriosis ( Figure 3B ). Overall, these networks summarize the distinct microbial interactions that provide useful insights of microbial interactions in endometriosis that might be associated with the relevant phenotypic outcomes. The host genome regulates gut microbiome through several epigenetic processes while several of the microbe-associated loci are found to be expressed in the gut that are related to immune functions or metabolism. Therefore, next we were interested in determining these microbiome-associated genetic variants (MAVs) in the human genome and how they differ in healthy versus in diseased states. By mapping the microbial taxa (both fungi and bacteria) identified in the samples, with the defined microbial taxa–human SNP genetic loci, we identified the host genes as MAVs that were associated with the microbes (mainly bacteria). Using this enrichment analysis, we identified several unique genes in endometriosis that might be associated with the disease. These were KDM4D, COCH, SLC5A8 , and PNPLA6 . In the control subjects, the microbiome is distinctly associated with genes— SRA1, CCDC9, CSPG4, ADAR, RAP1GAP, RP1L1 , and ADAMTS4 ( Figure 4A ). In addition, we determined the microbial taxa with which these MAVs were found to be associated. In the control subjects (i) SRA1 (steroid receptor RNA activator 1) associated positively with Lachnospiraceae, Bacteroides, Erysipelatotrichaceae ; (ii) CCDC9 (coiled-coil domain containing 9) correlated negatively with Odoribacter, Eubacterium, Dorea, and Acidaminicoccus ; (iii) CSPG4 (chondroitin sulfate proteoglycan 4) correlated positively with Blautia, Veillonella, and Haemophilus ; (iv) ADAR (adenosine deaminase RNA-specific gene) correlated negatively with Mogibacterium and Dialister ; (v) RAP1GAP (RAP1 GTPase-activating-protein) and RP1L1 (RP1 like 1) correlated negatively with Dorea and Lachnospira ; (vi) ADAMTS4 (ADAM metallopeptidase with thrombospondin type 1 motif 4) correlated negatively with Bacteroides and Megasphaera ( Figure 4A ). On the other hand, the distinct microbiome associated genes that were identified in endometriosis were (i) KDM4D (lysine demythylase 4D) which correlated positively with Veillonella, Megasphaera , and Klebsiella ; (ii) COCH (cochlin) gene positively correlated with Odoribacter, Coriobacteriales , and Ruminococcus ; (iii) SLC5A8 (Solute Carrier Family 5 Member 8) correlated negatively with Lachnospiraceae, Bacteroides, Odoribacter ; and (iv) PNPLA6 (patatin-like phospholipase domain containing 6) correlated negatively with Eubacterium, Blautia, Dialister , and Megasphaera ( Figure 4A ). We further retrieved the interactions of these bacteria while expanding their associations with the fungal taxa to identify the key fungal taxa associated with host MAVs ( Figure 4B ).As expected, given the low occurrences of defined fungal taxa in databases, we only found Penicillium and Saccharomyces to be associated with genes SLC5A8 and COCH in endometriosis ( Figure 4B ). Notably, the bacterial groups associated with gene KDM4D in endometriosis, namely, Megasphaera, Veillonella , and Klebsiella did not show any significant correlations with other bacteria or fungi ( Figure 4B ). Together, these results summarize the bacteria that are associated with host genes and the important fungal taxa that are involved in the key interactions along with bacteria to affect the host genetic loci in the pathogenesis of endometriosis. Our group previously established that antibiotic-mediated depletion of total gut microbiota protects against the development of endometriosis [ 13 , 16 ]. However, it is not yet clear if fungi play a distinct role in this protective effect independent of bacterial depletion. Given the more intricate fungal–bacterial relationships in endometriosis observed in our in silico studies, we wondered whether the depletion of gut fungi alone will have a similar protective effect against the disease, or does the fungal depletion cause opportunistic pathogenic bacteria to expand, promoting disease progression. Therefore, we used antifungal agent to deplete the gut fungi in mice (AF group) and studied the development of endometriosis ( Figure 5A ). In addition, with the aim of delineating the degree by which fungi alone versus bacteria and fungi together impact the disease, we also studied the disease progression in MD mice (bacteria and fungi) treated with ABx ( Figure 5A ). During the experiment, we observed no effect of treatment with AF or ABx on either the body weight or water consumption of the mice. The quantitative assessment of the fecal samples from mice upon treatments with antifungal agent and ABx confirmed significant loss ( p =0.005) of gut fungi in the AF group ( Figure 5B ).Similarly, effective depletion of both bacteria ( Figure 5C ; p =0.001) and fungi ( Figure 5D ; p =0.034) in the ABx group was achieved in MD mice upon treatment with antibiotic cocktail containing antifungal amphotericin-B. Importantly, the antifungal treatment of mice in AF group had no significant impact ( p >0.05) on the total bacterial populations ( Figure 5E ),confirming that the results are specific to fungal depletion. After 14 days following disease induction, we derived the endometriotic lesions from mice in all groups ( Figure 5F ) and found that the lesions that formed in the mice in AF group were significantly less in number ( Figure 5G ) and were markedly reduced in size ( Figure 5G ) than in control mice. Subsequently the total load of endometriotic lesions that progressed in fungi-depleted mice (AF) was significantly reduced ( Figure 5G ). The reduction in the numbers, size, and weights of lesions was, however, more pronounced in the MD mice ( Figure 5F and G ). In addition, lesions formed in AF group mice lacked the typical morphology with thinner epithelia and reduced stroma with a smaller number of endometrial glands ( Figure 5H ). These lesions also contained fewer proliferative cells (Ki-67-positive) and showed less infiltration of macrophages (F4/80-positive) than lesions derived from the control mice ( Figure 5H ).Again, this loss of typical morphological structures and reduced populations of actively proliferating cells and macrophages were even more significant with depletion of total microbiota (MD) ( Figure 5E ).

The study was approved by Institutional Review Board under a protocol (ID #: 201612127) at the Washington University in St. Louis School of Medicine. All patients gave informed consent to participate. Samples from healthy subjects were also obtained after informed consent by the Biobank Core of the Digestive Disease Research Core Center at Washington University in St. Louis under pay-per-service mode. All animal procedures were approved by the Institutional Animal Care and Use Committee under protocol [AN-8890]. Our study evaluated the stool samples from endometriosis patients and healthy women between 18 and 45 years of age. The characteristics of subjects enrolled in the study are summarized in Table 1 . All endometriosis patients enrolled in the study were laparoscopically confirmed and underwent surgeries for removal of ectopic endometriotic lesions. The biopsies of the lesions were further assessed to identify the revised American Society for Reproductive Medicine classification disease stage [ 30 ] and disease severity was confirmed in each patient ( Table 1 ). Patients were excluded if they were unable/unwilling to provide informed consent, had severe comorbidities, were pregnant, were on any hormonal treatments, or had taken antibiotics/probiotics within the previous 2 months. Individuals assigned to the healthy group had no significant medical conditions, or recent illnesses or ongoing health problems at the time of enrolment. Healthy subjects were only included if they reported to have no bowel issues or issues with fertility, as the two are potential endometriosis-associated comorbidities. The healthy subjects were not on any hormonal therapies for reproductive or other illnesses and participated at will before standard tubal ligation. Samples were collected from 18 patients with endometriosis and 31 healthy subjects. Statistical significance between the two groups was determined by using t -test for continuous variables (age, BMI) and chi-square test for nominal variables (menstrual phase). Whole stool samples were collected in sterile containers, homogenized and aliquots were frozen immediately at −80◦C for analyses of microbiota. Stool samples were subjected to total DNA extraction as previously described using QiAmp PowerFecal Pro DNA kit. The ITS2 region was amplified for sequencing using Illumina platform. The bacterial profiles were determined by sequencing the V4 region of the 16S rRNA gene as described in our previously published study [ 15 ]. The demultiplexed ITS2 raw reads were subjected to an initial quality filtering using bbduk.sh in BBMap, version 38.82 ( sourceforge.net/projects/bbmap/ ), removing Illumina adapters, PhiX reads, and reads with a Phred quality score <15 and length <100 bp after trimming. Quality controlled reads were then merged using bbmerge using the following merge parameters: maxstrict =t, qtrim =t, trimq =15. Merged reads were further filtered via VSEARCH[ 31 ],using max error rate =0.05, minlength =252, max length =254, deblur length limit =252. All reads were then combined into a single Fast-All (FASTA) file for further processing using UPARSE [ 32 ]. Taxa abundances were recovered by mapping the demultiplexed reads to the representative sequences file, creating a Feature table in biom format and removing the chimeric reads. The generated representative sequences were mapped against an optimized version of the latest SILVA database 138.1 [ 33 ] containing only sequences from the ITS2 region to determine taxonomies using the usearch70 “ usearch_global ” function and specifying the identity threshold to 97% [ 34 ]. Phylogeny information contained in the biom file was generated by aligning the centroid sequences with MAFFT [ 35 ] and creating a tree via FastTree [ 36 ]. The biom file was summarized, recording the number of reads per sample, and merged with a file that is generated for the overall read statistics, to produce a final summary file with read statistics and taxonomy information. Alpha diversity was computed based on “Observed OTUs” (Operational Taxonomic Units) and the effect size was assessed using Cohen’s d statistic as mean difference divided by pooled standard deviation. The beta diversity was analyzed by calculating the unweighted Bray–Curtis dissimilarity distance matrix, and PERMANOVA was used as a permutation test to assess the statistical significance. The results of the PERMANOVA test were then visualized using Principal Coordinates Analysis (PCoA) ordination. The bacterial and fungal taxa abundances were integrated using the inter-domain ecological networks workflow in the integrated network analysis pipeline for microbiome studies [ 37 ]. The same pipeline was used to derive all interactions among bacteria in both groups. The SparCC (Sparse Correlations for Compositional data) correlations were computed [ 38 ], and a bipartite matrix built. Only statistically significant ( p 0.2 or >0.5 SparCC correlations by filtering for significant interactions present in both the conditions and comparing their means using Cohen’s d statistic. Host gene-disease associations were determined via microbe set enrichment analysis between disease-centric microbe-sets and gene-centric microbe-sets using MicrobiomeAnalyst [ 40 ]. Taxonomic signatures were characterized by their associations with different host phenotypes using the defined taxa-SNP associations for microbial taxa present in the samples, through enrichment analysis against the manually curated human SNP associated taxon library in MicrobiomeAnalyst [ 40 ]. The identified bacterial taxa for each host gene were then used to retrieve all SparCC significant correlating interactions from the total bacterial–fungal bipartite networks and the subnetworks were later visualized in Cytoscape v3.10.1 [ 41 ]. Wild-type C57BL/6 mice (9 weeks old) were purchased from Taconic Biosciences Inc. and housed in the animal facility at Baylor College of Medicine, Houston, TX, USA. The animals were maintained in standard 12-h light/dark conditions and provided ad libitum access to food and water. The animals were handled after a 1-week period of acclimatization. Mice in the antifungal (AF) group were given autoclaved water supplemented with 0.5 mg/ml fluconazole (Sigma-Aldrich PHR1160) for 2 weeks for depletion of fungi before endometriosis induction. For depletion of both gut bacteria and fungi, mice in the microbiota-depleted (MD) group were orally gavaged every 12 h for 7 days with a cocktail of antibiotics (ABx) containing ampicillin (100 mg/kg), vancomycin (50 mg/kg), neomycin (100 mg/kg), metronidazole (100 mg/kg), and amphotericin-B (1 mg/kg). A gavage volume of 10 ml/kg body weight was delivered with a 20-gauge stainless steel feeding needle [ 16 , 42 ]. The antibiotic cocktail was prepared fresh every 48 h. The antifungal and antibiotic compounds were discontinued upon endometriosis induction. The mice in the control group were gavaged with a similar volume of sterile water. Fresh feces were collected from mice in all groups before starting and at the end of treatments prior to induction of endometriosis. Total DNA was extracted using QIAamp PowerFecal Pro DNA kit (Qiagen, Catalog #51804). The total fungal and bacterial DNA in the samples were quantified with SYBR green PCR amplification (Thermo Fisher, Catalog #4309155) with universal primers against 5.8S rRNA gene (5.8S-F, 5’-GGTGGATCACTCGGCTCGT-3’ and 5.8S-R, 5’-GCAAGTGCGTTCGAAGTGTC-3’) [ 43 ] and 16S rRNA gene (UniF, 5’-GTGSTGCAYGGYYGTCGTCA-3’ and UniR, 5’-ACGTCRTCCMCNCCTTCCTC-3’) [ 44 ], respectively. The Ct values for each mouse were determined on the day before starting any treatment (Day −14 for +Fluc and Day −8 for +Abx mice) and on the last day of treatment (Day 0 in each case). For each group and time point, the Ct values were determined also from the Vehicle group of mice (no Abx or Fluc). We then calculated the ∆Ct for each sample as ∆Ct=Ct(sample)−mean Ct(vehicle). From this, relative expression was calculated as 2 (-∆Ct) that were used to perform statistical comparison within the treated group using paired Student’s t -test with significance at p <0.05. A heterologous model of endometriosis was used for in vivo studies where endometrium from donor mice was injected into the recipients to induce the disease. For this, the donor mice were subcutaneously injected with estradiol benzoate (100 μ g/kg) on Day −7 [ 15 , 21 , 45 ]. On Day 0, donor mice were euthanized (one uterine horn for each recipient), and uteri were removed in warm saline, cut open longitudinally to expose the endometrium [ 46 , 47 ]. Endometrial tissue from each uterine horn was scraped and disrupted with micro-scissors to produce fine suspension in saline [ 14 ]. This suspension was immediately injected into the recipient mice (0.4 ml/mouse) intraperitoneally [ 48 ] with a 1-ml syringe and a 18 gauge needle [ 49 ]. The endometrial cells were allowed to adhere and progress into endometriotic lesions in the abdominal cavity in the recipient mice. After 14 days, mice were euthanized by cervical dislocation and the abdominal cavity was opened. The endometriotic lesions were carefully located and excised, trimmed of excess fat, measured for weights and volumes, and fixed and later processed for histology and immunofluorescence [ 21 , 45 ]. Endometriotic lesions were fixed in 4% paraformaldehyde, processed, and embedded in paraffin. These tissues were then sectioned at 0.5 μ thickness, deparaffinized and were stained with hematoxylin and eosin ( n =5 per group) as described previously [ 15 , 50 ]. For immunofluorescence, tissue sections ( n =5 per group) were deparaffinized, rehydrated, and boiled for antigen retrieval as described previously [ 50 , 51 ]. Sections were blocked with PBS containing 2.5% goat-serum (Vector Laboratories) for 1 h, then incubated overnight in primary antibodies against Ki67 (1:100, Abcam, ab16667) and F4/80 (1:100, ThermoFisher Scientific, Cat#53–4801-82). After washing with PBS, sections were incubated with Alexa Fluor 488-conjugated or Alexa Fluor 594-conjugated secondary antibodies (Life Technologies) for 1 h at room temperature and mounted with ProLong Gold Antifade Mountant with DAPI (Thermo Scientific # P36962 ). PERMANOVA was used as a permutation test to assess the statistical significance of the alpha diversity measures and effect sizes were determined using Cohen’s d statistic. The relative abundances among fungal taxa were compared based on Mann–Whitney statistic ( p value significance <0.05). In mouse experiments, a one-way ANOVA statistic for multiple comparisons was employed to determine the significance. P <0.05 was considered significant. All data are presented as mean±SE. Statistical significance in mouse fecal qPCR validation studies was determined using Student’s t -test at p <0.05.

Endometriosis, a prevalent gynecological disease, is a serious cause of concern with implications such as infertility. The disease also manifests in a wide range of symptoms, including dysmenorrhea, severe pelvic pain, pain during intercourse, and bowel issues, that compromise the quality of life of affected women [ 12 ]. The role of gut microbiome, particularly bacteria, is widely established [ 13 , 14 , 16 , 18 , 20 , 21 ]. In evidence, we previously showed that depletion of microbiota (bacteria and fungi) halts the progression of endometriosis in mice [ 14 ] and that the perturbed gut bacteria, and derived metabolites play a role in the pathogenesis of endometriosis in both murine models and in women with endometriosis [ 14 – 16 , 52 ]. Several advancements have also been made towards identifying bacterial species that are possibly linked with endometriosis. For instance, we showed that the microbiota-derived metabolite, 4-hydroxyindole, that is, particularly correlated with higher populations of specific commensal bacteria in the gut of women with endometriosis and inhibits the progression of endometriosis in mice [ 15 ]. Noteworthy, the fungi dwelling within the gut also form a crucial component of gut microbiome that cannot be ignored. Therefore, utilizing the stool samples from women with endometriosis, we integrated the data on the gut bacterial microbiota that we made previously available [ 15 ] and the fungal microbiota sequenced in this work, to gain insights of the role of fungi in endometriosis that are summarized in Figure 6 . We identified that mycobiota structure is significantly changed in women with endometriosis, and like the reduced diversity of gut bacteria, the mycobiota diversity is also slightly reduced in endometriosis patients. Intriguingly, these changes are not clearly distinguished at phylum level as with the bacteria. The overall profiles of both the control and patients’ groups showed Ascomycota, Basidiomycota, and Mucoromycota as the top abundant phyla. This is true even for the abundant fungal genera, such as Saccharomyces, Penicillium, Candida, Aspergillus , and Cladosporium . Although, it is worthy to note that several fungal taxa are opportunistic pathogens that are only identified at species level and may belong to the same genus [ 53 ]. Therefore, the mycobiota structures may not be clearly distinct at genus level. Furthermore, our results demonstrate that the less prevalent genera, such as Dipodascus and Nakaseomyces , are more important and contribute to the distinct attributes of the mycobiota in healthy and disease states. Therefore, it is the less prevalent fungal taxa that might underscore dysbiosis. Specifically, we determined Talaromyces and Rhodotorula, that might be more relevant in pathogenesis of endometriosis than the prevalent taxa. Both these taxa are low occurrence genera in human gut, although their altered abundances are linked with several clinical manifestations including gastrointestinal symptoms and immune dysfunction, which are common during the pathogenesis of endometriosis [ 54 – 58 ]. Although these taxa are understudied, it will be interesting to evaluate the underlying roles of these less prevalent genera that also contribute to the distinctive profiles of the mycobiota in endometriosis. Our findings demonstrate that although the top abundant taxa profiles are similar in healthy and disease states, their interactions with the bacteria within the community might underline the pathogenesis of endometriosis. Fungi and bacteria commensally cohabit the human gut and the bacterial–fungal interactions underpin the host metabolism and immunity [ 59 ]. More importantly, the interactions between gut bacteria and fungi (bipartite microbe–microbe interactions), and the microbe–host interactions might determine the occurrence of disease. Our co-abundance analysis suggests that a complex homeostatic co-regulation network between fungi and bacteria exists in a healthy gut, which is perturbed in endometriosis. Our results indicated reduced overall interactions of Saccharomyces and Candida with other bacteria in endometriosis. The commensal species, such as Saccharomyces cerevisiae and Candida albicans are shown to exhibit immune modulatory properties and alleviate gut mucosal injuries from dextran sodium sulfate-induced colitis in mice [ 60 ], suggesting the mycobiome’s ability to maintain gut homeostasis. In contrast, other spp. of Saccharomyces can also exacerbate gut inflammation and increase gut permeability in Inflammatory bowel disease (IBD) through overproduction of uric acid [ 61 ]. Consistent with this, we noticed that Saccharomyces still showed a significant number of stronger correlations with other bacteria in endometriosis. Importantly, it displayed negative correlations with two important gut bacteria, Bacteroides and Ruminococcaceae incertae sedis , both of which are associated with endometriosis and were reduced in endometriosis groups in our datasets [ 15 ]. On the contrary, Saccharomyces shared positive correlations with several commensal bacteria (such as Erysipelatoclostridium , Lactococcus , Dorea , Actinomyces , and Lachnospiraceae ). This suggests that Saccharomyces spp. might play protective or detrimental roles in the disease based on interactions with co-colonizing bacteria. The genus Aspergillus consisting of several mold species that regularly colonize human gut commonly produce aflatoxins and can cause opportunistic infections in humans [ 62 ] and may activate a gut immune response that indirectly affects systemic inflammation. This gut response was demonstrated in a study which showed that prolonged treatment with antifungal drug fluconazole decreased the levels of Candida in the gut, but increased Aspergillus amstelodami , resulting in elevated colitis severity [ 63 ]. These results put forth the fungal–bacterial interactions that potentially play important roles in regulating the microbial ecosystem in endometriosis and extend our current knowledge about the role of the gut microbiome in endometriosis. Overall, we demonstrate that the extensively studied probiotic fungal species function by interacting with bacteria and eliciting protective functions by possibly favoring the growth of commensal bacteria. In addition to the fungal–bacterial interkingdom interactions, the interactions among bacteria are enriched and important as the bacteria constitute a large component of the gut microbiome. Hence, we also determined how these bacterial–bacterial relationships vary in healthy versus in women affected with endometriosis. The results revealed useful insights into bacterial interactions relating to the pathogenesis of endometriosis. (1) We found that in endometriosis, other than the common gut commensals, the interactions were enriched and formed by the taxa that have been shown to be involved in disease onset and progression, for example, Escherichia-Shigella [ 17 , 64 ] and Enterococcus [ 65 ]. Additionally, genera such as Actinomyces , Dorea , Gemella , and Turicibacter formed important taxa making significant number of interactions in endometriosis group compared with the healthy group. Recently, Wang et al. determined that Actinomyces could potentially trigger or exacerbate endometriosis-associated infertility [ 66 ]. While it is typically not pathogenic in humans, its altered abundances can cause immune dysfunction that might aggravate inflammation induced damage as is seen in endometriosis [ 67 , 68 ]. (2) Another striking feature was the interactions of the genus Dorea. Essentially, it formed relationships with different sets of bacteria in healthy compared with the endometriosis groups, that were more enriched in the latter. This highlights that the same genus can potentially interact with different bacteria in eubiosis versus in endometriosis, to elicit different responses. In homeostasis, Dorea showed positive correlation with commensals such as Lachnoclostridium while in endometriosis, Dorea formed positive relationships with Blautia and Enterococcus that are key taxa identified in the pathogenesis [ 19 , 65 ]. Dorea also showed negative correlation with Alistipes, a genus that has been shown to protect against inflammation [ 69 ] and inflammatory gut diseases like colitis [ 70 ] thereby highlighting a potential promoting role of this interaction in endometriosis-associated inflammation. Recently, we demonstrated that Dorea is an important genus that should be focused on in endometriosis for its possible role in pathogenesis as we found it in increased abundance in the gut of endometriosis patients [ 15 ]. The genus has been linked with dysbiosis in several diseases [ 71 , 72 ]. Certain species of Dorea induce IFNγ [ 73 ] and metabolize sialic acids and degrade mucin [ 74 ] acting as pro-inflammatory. Therefore, the genus is viewed as a microbial member of the gut that plays different roles depending on the surrounding microenvironment [ 75 ]. Accordingly, our results unveil the important interactions that can be further studied to identify the specific functions of Dorea , based on its interactions under normal and dysbiotic states. The interactions within the human gut are not only limited to the microbes but also extend to their interactions with the host acting as a key underlying mechanism by which the microbiome affects the host by altering the expression of host genes. In this study, we identified several MAVs of host genetic loci that might be associated with disease risk and altered relative abundances of microbial taxa during pathogenesis. These microbe–host relationships are independently observed and are concordant with associations reported in case–control studies. In the healthy group, genes like SRA1 , RAP1GAP , CCDC9 , CSPG4 were among the ones identified as MAVs. The gene SRA1 acts as a co-activator of steroid receptors that modulate the production of steroid hormones [ 76 ]. SRA1 and the other genes including CCDC9, CSPG4 , ADAR , RAP1GAP , RP1L1 , and ADAMTS4 correlated with the common gut commensals such as Lachnospiraceae, Bacteroides, Erysipelatotrichaceae, Odoribacter, Eubacterium, Dorea, Acidaminicoccus, Blautia, Veillonella, Haemophilus, Mogibacterium , etc. As expected, these genera have not been extensively linked with pathogenesis of endometriosis, although they might or might not be involved in homeostasis in healthy individuals. Among the identified MAVs in endometriosis, the gene KDM4D , when overexpressed, promotes the proliferation and angiogenesis in several cancers [ 77 – 79 ]. Its inhibition is shown to significantly suppress the tumor growth in colon cancer by transcriptionally activating hypoxia-inducible factor (HIF)-1 β promotor [ 77 ]. KDM4D is also reported to be capable of promoting liver cancer formation and development by antagonizing tumor suppressor gene p53 and activating Wnt/ β -catenin signaling pathway [ 78 ]. Given the cancer-like, but benign pathology of endometriotic lesions growth and the enhanced proliferation and extensive angiogenesis that occurs during endometriosis [ 80 ], the gene poses as an important target of clinical significance. Another host gene identified to be significantly correlated with endometriosis gut microbiota is SLC5A8 , that codes for a plasma membrane transporter that facilitates the entry of histone deacetylase inhibitor—butyrate into the colonic cells [ 81 ]. Therefore, SLC5A8 promotes growth arrest, differentiation, and apoptosis while functioning as a tumor suppressor. In endometriosis, histone deacetylase inhibition reactivates E-cadherin, attenuates invasion, decreases proliferation of endometriotic cells, and causes lesion regression [ 82 ]. We additionally identified PNPLA6 gene which is implicated in Boucher-Neuhauser syndrome, a condition that is accompanied with female infertility [ 83 ]. In summary, the commensal gut bacteria were strongly positively correlated with the MAVs associated with cell growth arrest, and apoptosis ( Lachnospiraceae and Erysipelotrichaceae with SLC5A8 ), while the opportunistic pathogenic gut bacteria were positively correlated with the MAVs promoting cell proliferation ( Megasphaera and Klebsiella with KDM4D ). These findings unveil the role of gut bacteria in the endometriosis progression through microbe–host interactions. Our group previously established that antibiotic-mediated depletion of total gut microbiota significantly reduces the disease onset and progression in mouse model of endometriosis [ 13 , 16 ]. However, there has been no effort to understand how the depletion of fungi alone would affect its pathogenesis. Moreover, the results from our study of fungal–bacterial interactions above revealed stronger interkingdom interactions during the disease. However, due to fungal interactions with both commensal and opportunistic pathogenic bacteria, we explored the overall role of the fungi in vivo in fungi-depleted mice with endometriosis. We reveal that treatment with antifungal agents alone can significantly prevent the pathogenesis of endometriosis by preventing growth of lesions, although not as dramatically as is observed with depletion of total microbiota. This demonstrates that absence of fungal microbiota directly impacts bacterial microbiota, thereby suppressing the disease in similar manner as observed with microbiota-depletion. The fungal depletion-specific reduction in disease progression was further confirmed as we did not observe a significant change in the total bacterial populations upon antifungal treatment. These results provide evidence that fungi play a key role in pathogenesis, although these effects may be indirectly mediated through bacteria. There remains much to be uncovered regarding finding the specific roles of each fungal taxon. Considering the drastic effects that antibacterials have on the gut microbiome and homeostasis, these findings open a new avenue for evaluation of antifungals as treatment alternatives for endometriosis. Whether the antifungals are able to suppress pathogenesis by primary changes in the fungal community, or through secondary impacts on bacterial populations requires further evaluation. Essentially, we conclude that in terms of pathogenesis, fungal taxa may act protective or detrimental based on the populations of surrounding bacteria and their interactions with these bacterial species. This study addresses a crucial gap in our understanding of the profiles of gut fungi and their underlying roles that might be linked with the pathogenesis of endometriosis. We utilized integrated approach to understand the complex relations in between the fungal and bacterial components of the gut microbiome in women affected with endometriosis. However, this study is subjected to the biases pertaining to the individual heterogeneity arising from factors such as ethnicity and genetics, lifestyle, geographical location, diet, hormonal status, and other environmental factors. A more comprehensive view of these profiles would come from larger cohorts. This work summarizes microbial co-abundance relationships based on sequence data. These methods are quick to derive the patterns of interactions in the microbial communities but require further functional validation. In this study, we applied the established SparCC method to establish microbial co-abundance networks, while other methods of estimating correlation may yield different pictures of community interactions. Our findings also reveal the host genetic variants that are observational from microbiome sequence data and might be linked to the pathogenesis of endometriosis in women. Further analyses of the transcription level data deposited in the databases from larger populations, as future studies on patients are conducted, will further suggest the functional and translational relevance of the gene variants in larger populations. Overall, our findings show that dysbiosis of the gut microbial ecosystem in endometriosis can not only be assessed by the altered abundance level of certain microbiome components but must be seen at the level of microbial–microbial community and microbe–host interactions.

Endometriosis is a common gynecological pathology in which the inner uterine wall (endometrium) grows outside as ectopic lesions in reproductive age women [ 1 ].These ectopic implants are accompanied by chronic inflammation that is aggravated with increased estrogen [ 2 ]. The principal symptoms include severe chronic pain and associated infertility that markedly compromise the quality of life of women [ 3 – 6 ].Endometriosis is observed in 50%–80% of women with pelvic pain and up to 50% of women with infertility [ 7 , 8 ]. Accordingly, about 30%–40% of women with endometriosis are subfertile [ 9 ]. The prevalence is reported to be equally high in patients with chronic pain and in asymptomatic women [ 10 ]. The time of diagnosis from disease onset ranges from 4 to 11 years and about 65% of women are misdiagnosed during initial stages [ 11 ]. This delayed diagnosis represents a barrier to the treatment of the disease. In the United States, ∼6%–10% of all women are affected and the economic burden exceeds US$22 billion [ 12 ]. Although a range of treatments are available that can help manage the symptoms, no current option offers a complete cure. The disease follows an enigmatic pathogenesis such that the symptoms overlap several of those in functional bowel disorders, and that indicate a possible involvement of gut microbiota in the disease pathogenesis. Our group previously demonstrated that gut microbiota is intricately linked to disease onset as the disease fails to develop in complete absence of gut microbiota [ 13 ] and their microbiota-derived metabolites may either promote [ 14 ] or protect against the disease [ 15 , 16 ]. Therefore, it is only accurate to infer that the disease is mediated by an imbalanced (dysbiosis) gut microbial composition. We and several other groups deduced that abacterial dysbiosis that is characterized by a reduced diversity is common in endometriosis [ 14 , 17 – 21 ]. However, microbes that colonize the human gut other than bacteria have not been investigated. The diverse population of fungi residing in the gastrointestinal tract is collectively termed the gut mycobiome. It accounts for ∼0.1% of gut microbiota and is ubiquitous in all human populations [ 22 , 23 ]. Although the fungi co-colonize with bacteria, mycobiome–bacteria and mycobiome–host interactions have not been studied in endometriosis. Several studies suggest mutual or competitive relationships between gut mycobiome and bacteria [ 24 ]. The mycobiome produces alcohol, antimicrobial peptides, and other metabolites that affect bacterial colonization [ 25 , 26 ], the bacteria, in turn, generate fatty acids that regulate fungal germination and hyphal growth [ 27 , 28 ]. Together, these bacterial–fungal interactions also modulate the host genome via their metabolites and functions, targeting the microbiome-associated genetic variants (MAVs) [ 29 ]. Therefore, while bacteria are the most abundant class of gut microbes, the presence of other microorganisms such as fungi, and their interactions with the bacteria may produce distinct outcomes than can be expected with the bacterial communities alone. Here, we characterized the fungal microbiota in women with well-phenotyped endometriosis compared with control (healthy) women using ITS2 marker gene sequencing. We also determined the bacterial gut microbiota in the same patient cohort to identify the fungal–bacterial interkingdom interactions and the bacteria–bacteria interactions in endometriosis. Moreover, we identify the microbiome-associated host genetic variants and the associated bacteria that are implicated in endometriosis. By studying the interactions of these bacterial taxa with fungi, we reveal the microbial candidates that possibly drive the onset and progression of the disease. We also dissect the independent roles of fungi versus both bacteria and fungi in disease progression in vivo. Overall, the study presents a rationale for the role of gut fungi in endometriosis and paves the way for further intervention studies on fungal microbiota.