Identification of distinct stool metabolites in women with endometriosis for non-invasive diagnosis and potential for microbiota-based therapies

Summary

Background

Endometriosis, a poorly studied gynecological condition, is characterized by the presence of ectopic endometrial lesions resulting in pelvic pain, inflammation, and infertility. These associated symptoms contribute to a significant burden, often exacerbated by delayed diagnosis. Current diagnostic methods involve invasive procedures, and existing treatments provide no cure.

Methods

Microbiome-metabolome signatures in stool samples from individuals with and without endometriosis were determined using unbiased metabolomics and 16S bacteria sequencing. Functional studies for selected microbiota-derived metabolites were conducted in vitro using patient-derived cells and in vivo by employing murine and human xenograft pre-clinical disease models.

Findings

We discovered a unique bacteria-derived metabolite signature intricately linked to endometriosis. The altered fecal metabolite profile exhibits a strong correlation with that observed in inflammatory bowel disease (IBD), revealing intriguing connections between these two conditions. Notably, we validated 4-hydroxyindole, a gut bacteria-derived metabolite that is lower in stool samples of endometriosis. Extensive in vivo studies found that 4-hydroxyindole suppressed the initiation and progression of endometriosis, -associated inflammation and hyperalgesia in heterologous mouse and in pre-clinical models of the disease.

Conclusions

Our findings are the first to provide a distinct stool metabolite signature in women with endometriosis, which could serve as stool-based non-invasive diagnostics. Further, the gut microbiota-derived 4-hydroxyindole pose as a therapeutic candidate for ameliorating endometriosis.

Graphical Abstract

eTOC blurb

Talwar et al. utilized stool samples from women with endometriosis and healthy women and identified distinct microbiota-metabolite correlations that can be used for non-invasive diagnosis. They identify the gut microbiota-derived 4-hydroxyindole as a promising therapeutic that inhibits the growth of mouse and human (xenograft) endometriotic lesions and reduces disease-associated pain.

Introduction

Endometriosis is a chronic inflammatory disorder that afflicts up to 198 million women worldwide and is characterized by the growth of endometrium outside the uterus in the form of inflamed lesions that cause severe pain¹. The gynecological pathology is a major health concern that substantially affects the quality of life of women, and in serious cases, renders them infertile¹. Its diagnosis is achieved through invasive procedures such as exploratory laparoscopy while surgical removal of lesions for its treatment has an immediate effect on post-treatment conceptions¹. Albeit the disease is not neoplastic, the symptoms recur in up to 75% of cases². Therefore, the need to rethink the diagnosis and management of endometriosis is acknowledged.

The origin of endometriosis is most widely attributed to the flow of menstrual blood in a retrograde manner that deposits the endometrial remnants into the peritoneal cavity where they adhere and start to grow into what are known as the endometriotic lesions³. Central to the progression of these lesions is heightened inflammation resulting from macrophage infiltration and increased secretion of pro-inflammatory cytokines⁴. While retrograde menstruation is a common phenomenon in reproductive age women, the disease originates only in some women, indicating a role for peritoneal environment. Interestingly, the gut microbiota is recognized as one of the key determinants shaping the peritoneal environment.

The human gut hosts an intricate bacterial community closely linked to various host physiologies, including immunological processes⁵. These commensal bacteria convert otherwise indigestible nutrients in the diet into biologically active metabolites. Numerous gut bacterial metabolites play pivotal roles in host metabolic homeostasis and inflammation^6,7. While gut microbiota composition remains relatively stable throughout life, changes in gut microbiota are correlated with the onset of several diseases, such as inflammatory bowel disease (IBD), metabolic disorders (e.g., obesity, non-alcoholic fatty liver disease, and diabetes), and neurological diseases^8-10. Additionally, metabolites and pro-inflammatory molecules produced by gut bacteria linked to inflammatory and cardiovascular disorders¹¹. Therefore, it is plausible that metabolites derived from the gut and its microbiota may influence the development of endometriosis. Presumably, the 'leaky gut' phenotype associated with a dysbiotic microbiome exacerbates inflammation in endometriotic lesions, promoting their growth. In fact, we previously found that antibiotic treatment with metronidazole inhibits the progression of endometriosis in murine models¹². Subsequent reports also identified altered gut microbiota profiles in women with endometriosis^13-16. Several studies linking gut microbiota to endometriosis highlighted the altered (dysbiosis) microbial profiles during the disease progression^13-16. However, we found a causal role for gut microbiota in endometriosis revealing that the disease fails to progress in the absence of gut microbes¹² and that the gut microbiota-derived short-chain fatty acid (SCFA), butyrate, protects against endometriosis¹⁷. However, these studies are from murine models of endometriosis and translational relevance of these studies remains elusive.

In this report, we identified a distinctive bacteria-derived metabolite signature in stool samples from women with endometriosis, which correlated with signature of IBD, a significant comorbidity associated with endometriosis. We also found that bacteria-derived metabolites, 4-hydroxyindole is lower in stool samples of women with endometriosis and revealed its inhibitory effects on endometriosis disease progression.

Results

Women with endometriosis have a distinct signature of stool metabolites

While recent advancements have identified altered gut microbiota profiles, there are significant variations in the reported perturbations of these profiles^13-15,18. Since fecal metabolites can be considered a functional readout of the gut microbiota, exploring these metabolites provides a robust and reliable reflection of alterations in the gut microbiota. Therefore, we aimed to identify specific bacteria-derived metabolites in women with endometriosis. Toward this, we collected stool samples from women with and without endometriosis. The clinical cohort consisted of 18 women with clinically confirmed endometriosis (EMS) and 31 control subjects (Table 1). By performing a high throughput unbiased metabolomics workflow on these stool samples (Figure 1A), we identified 371 distinct metabolites belonging to different classes of compounds (Supplementary Figure S1A). Upon statistical filtering, we uncovered a subset of 61 metabolites that differentiated the two cohorts (Figure 1B-C; Supplementary Table S1). Among these metabolites, 50 were reduced in abundance (and 11 increased) in EMS than in healthy subjects. Based on the variable importance in the projection (VIP)-scores of orthogonal Projections to Latent Structures Discriminant Analysis (oPLS-DA), we identified 22 stool metabolites that are distinct in women with endometriosis (Figure 1D).

Table 1.

Characteristics of endometriosis patients and control subjects included in the study.

	Patients	Control
Number of participants	18	31
Age (years)
Mean (SD)	27.52 (2.55)	35.3 (8.3)
Median (range)	27 (25-32)	25.185 (21.3 – 31.6)
BMI
Mean (SD)	25.62 (4.29)	26.81 (3.12)
rASRM Stage of Endometriosis
Stage I	3 (Peritoneal=3)	NA
Stage II	3 (Peritoneal=1) (Peritoneal & Ovarian=2)
Stage III	2 (Peritoneal & Ovarian=2)
Stage IV	6 Ovarian, Deep infiltrating=2) (Peritoneal & Ovarian=4)
Stage not confirmed	4
Menstrual Phase	9	16
Proliferative	9	15
Secretory
Comorbidities/Family History	None identified or reported
Comorbidities		NA
Bowel issues		None
Family history of CRC		None=13 Paternal=1 Maternal=3 Not known=14
Smoking	Not available
Active		3
Never		12
Past Smokers		0
Not disclosed		16

Figure 1. Untargeted metabolomics of human stool and identification of altered gut metabolites in endometriosis.

(A) Schematic representation of the workflow of untargeted metabolomics analysis. (B) Volcano plot of all identified metabolites (n=371) in the datasets. The x-axis indicates Log2 (fold change) while the y-axis indicates −Log10 (P-value). Every single metabolite is represented as a dot. Different colors were used to represent down-regulated (red), up-regulated (green), or non-significant (black) metabolites. (C) Heatmap based on significantly altered metabolites (n=61; FC>1; FDR Adjusted p-value <0.05) in between healthy controls and EMS patients. Hierarchical clustering (Distance: Euclidean, Linkage: Ward) of the metabolite abundance is shown. (D) Orthogonal Partial Least Squares-Discriminant Analysis (oPLS-DA) of the samples from two groups. The scores are indicated by different color circles for healthy control and EMS patients (E) ROC curve performance of biomarkers (AUC>0.8) determined from metabolites data from in endometriosis patients. (F) Enriched functional pathways detected in the altered metabolites (n=61) data. Each dot denotes a functional pathway colored and sized based on p-values (lowest red to highest yellow) and enrichment ratio, respectively. (G) Bar graph showing the top enriched disease signatures in altered metabolites (n=61). The x-axis indicates the number of assigned metabolites to total metabolites while the y-axis indicates the disease signature. Each bar is color coded for p-value <0.05: Red: low; yellow: high.

Subsequently, we also identified putative fecal biomarkers for their potential as non-invasive diagnostic biomarkers in EMS. We built Receiver Operating Characteristic (ROC) curves to characterize the discriminating powers of VIP-selected gut metabolites (n=22) during EMS and found 12 metabolites with an area under the curve (AUC) >0.8 and p<0.005 (Figure 1E; Supplementary Table S2). These biomarkers include linoleic acid, 2'-deoxyadenosine, N-formyl-L-methionine, adenine, cytosine and adenosine, among others (Figure 1E). We next analyzed the perturbed metabolic functions and found that the metabolism of purines is highly impacted in EMS (Figure 1F) which is also observed in IBD. This distinct stool metabolite-derived association between endometriosis and IBD is intriguing as the latter is widely reported as one of the closely associated comorbidities in women with endometriosis. Therefore, we further confirmed this with a disease-based enrichment analysis where we queried the altered metabolite sets in EMS and found that the perturbed molecular functions in EMS closely mirror those reported in the unclassified IBD and its common manifestations: Ulcerative colitis (UC) and Crohn’s disease (CD) (Figure 1G). Together, through these analyses, we unveiled a distinct stool metabolite signature of endometriosis, which is correlated well with the other comorbidities as well.

Microbiota-derived fecal 4-hydroxyindole (4HI) is lower in women with endometriosis

Our goal was to dissect the origin of these metabolites and characterize the microbiota derived metabolites that are altered in endometriosis. Therefore, we additionally determined the gut microbiota profiles of EMS and healthy subjects. Compared with control subjects, we noticed a reduced alpha-diversity of gut microbiota in the EMS gut based on Inverse Simpson index (Figure 2A; Adj. p=0.047). The beta-diversity measures of unweighted UniFrac distance revealed that the gut microbiota in EMS deviated significantly (Figure 2B; p=0.043). Taxonomic comparisons of relative bacterial abundance revealed that the gut microbiome composition in EMS differs in a manner of ways from healthy controls. Populations of Ruminococcus, Faecalibacterium, Alistipes and Roseburia were significantly decreased in EMS (Figure 2C); whereas Erysipelatoclostridium, Streptococcus, Lacticaseibacillus, Actinomyces and Tyzzerella were increased (Mann-Whitney; p<0.05; Figure 2C; Supplementary Table S3).

Figure 2. Integration of metabolome and microbiome identified reduction in bacteria-derived 4HI levels in women with endometriosis.

(A) Analysis of alpha diversity in EMS patients compared with healthy controls based on Inverse Simpson index (Adj. P=0.047). The horizontal lines inside boxes indicate the median and the dots represent outliers. (B) Beta diversity compared between healthy control and EMS groups using Principal Component Analysis (PCoA) based on unweighted UniFrac distances (Mann-Whitney p<0.05). The centroids and dispersion of the groups are shown using thin colored lines (from each sample point to corresponding centroid). (C) The relative taxonomic abundance of bacterial genera in the two groups. Significantly altered taxa are denoted with red asterisk. (D) Spearman correlation between metabolites and microbial genera. Only the metabolites of microbiota origin that correlated significantly (p<0.05) are shown. The metabolites that are altered (FC >1) in endometriosis are denoted with asterisk. (E) Violin plots of the significantly altered metabolites from microbial driven indole pathway in endometriosis- 4-hydroxyindole (4HI) and indole-4-carbaldehyde (I4CAld) with fold change, FC >1. See also Supplementary Figure S2 for non-significantly altered indole metabolites. (F) Selected bacteria-metabolite correlation pattern in endometriosis. The shading of the squares indicates the correlation (green color metabolite denotes positive correlation; red colored metabolite denotes negative correlation). qPCR-based quantitative estimation of indicated bacterial taxon using 16S rDNA gene amplification products from control and patient cohorts (n=10 samples each). Statistical significance was determined using unpaired t-test with Welch’s correction (* denotes p-value <0.05, *** denote p-value <0.001). (G) Box plots showing upward-, downward- or unchanged trends of selected metabolites from LC-MS validation studies.

Subsequently, using a combination of biological and statistical correlations between metabolome and microbiota, we identified metabolites that significantly correlated with the microbiota species across samples (Figure 2D). Parallelly, we performed origin analysis and determined metabolites that were derived from gut microbiota. Combining these analyses, we identified 9 metabolites of microbiota origin in the EMS that displayed high correlations with the bacterial species. Of these, three metabolites were significantly altered in EMS subjects- 4-hydroxyindole (4HI), 4-indolecarbaldehyde (I4CAld) and ferulic acid (Figure 2D). Since, ferulic acid is primarily sourced from plants in the diet¹⁹, we focused on other two compounds that belonged to the class of indoles (4HI & I4CAId) that were significantly reduced in the EMS (Figure 2E). Indoles are a class of compounds largely derived from bacteria in the gut that can be produced and/or converted into other indole derivatives by different bacterial groups within the gut (Supplementary Figure S1B). Although several indole derivatives were detected in unbiased platform, many of these were not altered significantly in EMS (Supplementary Figure S1C). By comparing the correlations direction, we found that the abundance of 4HI correlated positively with populations of Faecalibacterium, Lachnospiraceae and Peptoniphilus while it displayed a negative correlation with Dorea (Figure 2D). On the other hand, I4CAld displayed positive correlation with Bacteroides and Parabacteroides (Figure 2D). Using quantitative PCR, we further confirmed high abundance of Dorea in EMS that corresponded to its negative correlation with reduced 4HI (Figure 2F). Similarly, we confirmed the reduced abundance of Bacteroides in EMS which correlated with its positive correlation with reduced I4CAld in EMS (Figure 2F). To confirm the significant reduction of 4HI and I4CAld, we performed LC-MS validation using fresh stool samples for these two compounds along with an additional metabolite- nicotinic acid, which was unaltered in the unbiased profiling (Figure 2G). While the reduction of I4CAld could not be validated, 4HI was again confirmed to be significantly reduced in stool from EMS compared to healthy controls and nicotinic acid remain unchanged in both the groups (Figure 2G). These findings demonstrate that the bacteria-derived indole metabolite, 4HI is reduced in women with endometriosis possibly due to altered gut microbiota.

4-hydroxyindole (4HI) suppresses the initiation and progression of endometriosis and associated inflammation and hyperalgesia

The reduced levels of 4HI suggest a potential protective effect of this bacteria-derived metabolite in endometriosis. To elucidate its functional relevance, we explored the impact of exogenous 4HI on the progression of endometriosis. Endometrial fragments from adult donor mice were injected into the peritoneal space of recipient mice for disease induction, and the metabolites were orally administered daily for subsequent 14 days (Figure 3A). In all in vivo assessments, we included an additional group of recipients receiving the parent compound, indole, to ensure that any observed effects were specifically attributable to 4HI and not to indole itself or its other derivatives. After 14 days of treatment, although there was no significant reduction in the number of lesions in both 4HI- and indole-fed mice, the lesions from 4HI-fed mice exhibited a notable decrease in both mass and volume (Figure 3B). In contrast, indole resulted in a non-significant reduction in the mass and volume of the lesions (Figure 3B). Furthermore, mice in the vehicle group developed typical endometriotic lesions with thick epithelia, well-formed stroma, and endometrial glands (Figure 3C). Lesions from 4HI-fed mice, however, displayed thinner epithelia and stroma and had fewer endometrial glands (Figure 3C).

Figure 3. 4-hydroxyindole suppresses the progression of endometriosis in vivo.

(A) Schematic representation of experimental timeline and procedures in mouse model of endometriosis employed for testing suppressive effect of 4-hydroxyindole on disease establishment. (B) Representative images of ectopic endometriotic lesions from the vehicle and metabolite treated groups, 14 days after induction of endometriosis. Lesion numbers, volumes, and masses of ectopic endometriotic lesions from the indicated treatment groups are shown. Data are presented as mean ± SE (n = 5). *lndicates significance between groups, *p-value <0.05; **p<01. (C) Representative hematoxylin and eosin (H&E) stained sections of lesions and sections stained with anti-Ki67 (proliferation) and anti-F4/80 (macrophages and inflammation). Scale bar=50μm. E, epithelium; S, stroma; G, glands; GE, glandular epithelium. Bar graphs on the right show the percentages of Ki-67-positive cells in epithelium and stroma and the F4/80-positive cells in the stroma of endometriotic lesions. (D) Flow cytometric dot plots for cell sorting on the peritoneal lavages on day 14 using the Cytek Aurora (see also Supplementary Figure S2). The MHCII⁺ CD11b⁺ F4/80 hi mac (M1-like macrophage) and CD206⁺ CD11b⁺ F4/80 hi mac (M2-like Macrophage) in indicated groups are shown. The overlay graphs on the right show the peaks area and the percentages of cells are shown for indicated groups. (E) Assessment of endometriosis-associated hyperalgesia through hind paw withdrawal-based Von Frey test on Days 0 and 14 after endometriosis induction. Data are presented as mean ± SE (n=4). *lndicates significance between groups, **p-value <0.05.

Additionally, fewer proliferative (Ki-67-positive) cells were observed in lesions treated with either 4HI or indole (Figure 3C). Reduced positive staining for macrophages (F4/80-positive cells) further confirmed that lesions from 4HI- or indole-treated mice exhibited reduced infiltration by macrophages (Figure 3C). Flow cytometric analysis of peritoneal lavages using the gating strategy (as shown in Supplementary Figure S2), we identified a reduction in both M1-like and M2-like macrophages upon treatment with 4HI or indole, in the peritoneum compared to vehicle-treated mice (Figure 3D). While 4HI reduced M1-like (pro-inflammatory) macrophages by a greater extent than indole; indole showed more pronounced inhibitory impact on M2-like (anti-inflammatory) macrophages. This was consistent with the higher degree of inhibitory effect of 4HI on lesion progression and macrophage infiltration into lesion (Figure 3C) than indole. Finally, we measured the impact of 4HI on endometriosis-associated hyperalgesia as chronic pain is one of the clinical hallmarks of endometriosis. After 14 days of induction, lesion-bearing animals withdrew from lighter stimuli compared to sham mice (Figure 3E). On the other hand, 4HI-treated mice withdrew their hind paws similarly to sham mice by day 14 (Figure 3E). Based on these results, we also studied the impact of 4HI solely on the initiation of endometriotic lesions by treating the mice with the metabolite prior to the induction of endometriosis (Supplementary Figure S3A). We noticed a significant reduction in lesions number and sizes that similarly lacked the typical morphology of endometriotic lesions by day 14 (Supplementary Figure S3B-C). The macrophages derived from peritoneal lavages showed reduced levels of both M1-like and M2-like macrophages as evidenced from reduced expression of functional marker genes for both M1-like (NF-kB and IL-6) and M2-like (Arg1) macrophages (Supplementary Figure S3D). These data indicate that 4HI reduces the progression of lesions, inflammation, and associated pain in murine endometriosis models.

4HI regresses the established lesions in murine and pre-clinical models of the disease

Given the ability of 4HI to inhibit lesion progression, we speculated whether 4HI could also induce regression in well-developed endometriotic lesions. For this, we employed a similar 14-day treatment strategy, however, mice were treated with metabolites after the lesions are fully developed and inflammation is at its peak mimicking severe disease state (Figure 4A) [12]. As expected, vehicle treated mice had well-developed lesions and 4HI treatment after lesion development resulted in significantly smaller lesions compared to the vehicle-treated group (Figure 4B). A marked reduction in lesion numbers, masses, and sizes noted with 4HI treatment, demonstrating a more substantial effect than indole on lesion volumes, although both treatments exhibited similar outcomes (Figure 4B). Histologically, lesions from 4HI- and indole-treated mice lacked typical control-like structures, displaying fewer glands and thinner epithelia (Supplementary Figure S4A). Importantly, 4HI exerted a much stronger inhibitory effect on lesion growth compared to indole. The stroma of lesions from 4HI-fed mice was extensively reduced compared to the lesions from indole-fed mice that had denser stromal compartments (Supplementary Figure S4A). Proliferative (Ki-67+) cells and macrophages (F4/80+) were also less abundant in lesions from metabolite-treated mice compared to the vehicle group (Supplementary Figure S4A).

Figure 4. 4-hydroxyindole regresses the established lesions in mouse and human xenograft models of endometriosis.

(A) Schematic representation of experimental timeline and procedures employed for testing therapeutic action of 4-hydroxyindole in heterologous and human xenograft mouse models of the disease. (B) Representative images of ectopic endometriotic lesions from the vehicle and metabolite treated groups, 28 days after induction of endometriosis (See also Supplementary Figure S4). Lesion numbers, volumes and masses of ectopic endometriotic lesions from the indicated treatment groups are shown. Data are presented as mean ± SE (n=5). *Indicates significance between groups, *p-value <0.05, **p<.01, ***p<0.001. (C) MTT cell viability assays of iHEECs/Luc treated with different concentrations of 4-hydroxyindole (4HI) and indole. Results are shown as mean±SE (n=3) and experiment was repeated three times. ***p<0.001, ****p<0.0001. (D) Representative bioluminescence images of lesions from mice transplanted with human endometriotic cells at indicated days and lesion data (same as in B) collected on day 28 from indicated groups. (E) Representative hematoxylin and eosin (H&E) stained- and immunofluorescence in sections of lesions derived from mice induced with human endometriotic cells. The sections were stained with anti-Ki67 (proliferation) and anti-F4/80 (macrophages and inflammation). Scale bar=50μm. E, epithelium; S, stroma; G, glands; GE, glandular epithelium. The bar graphs on the right show the percentages of Ki-67-positive cells in epithelium and stroma and the F4/80-positive cells in the stroma of endometriotic lesions are shown.

Based on our findings in murine model, we characterized the translational relevance of 4HI in endometriosis. To test this, we first examined the impact of 4HI on the viability of immortalized human endometriotic epithelial cells expressing luciferase (iHEECs/Luc). Treatment with 4HI resulted in reduced cell viability at lower dose (~250μM) and indole could affect cell growth only at higher dose (1mM) (Figure 4C). At cellular level, we examined the changes in the expression of genes implicated in endometriosis in these cells upon treatment with 4HI (Supplementary Figure S4B). 4HI notably reduced the expression of IGF1, NF-Κβ, TGF-β1 and GATA3 that augment endometriosis (Supplementary Figure S4B). In contrast, other genes such as GREB1, CCND1 and GATA6 remained unchanged or altered modestly (Supplementary Figure S4B).

Next, we evaluated in vivo efficacy of 4HI in a pre-clinical xenotransplant model of the disease using human derived endometriotic cells. In this model, iHEECs/Luc and immortalized human endometrial stromal cells expressing luciferase (iHESCs/Luc) were injected into the peritoneal space of immunocompromised mice (Figure 4D). Upon 14 days, we confirmed lesions establishment using bioluminescent imaging and mice were treated either with vehicle or 4HI or indole for the next 14 days. Following treatments, vehicle treated mice had robust bioluminescent signal, indicating that the lesions progressed well. However, 4HI treatment markedly regressed the well-formed lesions as evident from low intensity of bioluminescent signal (Figure 4D). Measures of the extracted lesions on day 28 further confirmed the regression of lesions with 4HI treatment, in terms of lesion numbers, volume, and masses compared to vehicle-treated mice (Figure 4D). Like murine studies, we noticed that the indole had much lesser impact on the lesions’ volume compared to 4HI. Both 4HI and indole treatments produced regressed lesions lacking typical structures (Figure 4E). However, stromal compartments in 4HI-fed mice derived lesions are much less dense than indole-fed mice lesions (Figure 4E). This was accompanied by fewer proliferative (Ki-67+) cells and macrophages (F4/80+) compared to the controls, with a more pronounced reduction in 4HI-fed mice lesions (Figure 4E). Together, these results indicate that the microbiota-derived metabolite- 4HI regresses the well-developed lesions in both mouse model of the disease and in the mice xenotransplanted with human endometriotic cells.

Discussion

Endometriosis is a debilitating health condition that adversely impacts the quality of women. Unfortunately, there is no cure for this disease and there is significant latency to diagnose the disease. The endometriosis-associated inflammation is an important risk factor for chronic pain that affects 65% of the patients^20,21. The gut microbiome governs the inflammatory milieu by regulating the integrity of gut barrier. As the gut microbiome likely plays a role in gut-brain axis²², there is a significant interest to find specific bacteria or their produced metabolites that can be leveraged to combat the disease with chronic pain and inflammation. In this study, we identified a distinct signature of stool metabolites in women with endometriosis that are clearly perturbed during the disease. Several of these altered metabolites impact systemic inflammation. Importantly, a majority of the perturbed metabolites in endometriosis are also perturbed in IBD, including, 4-dodecylbenzenesulfonic acid²³, piperine, and niacinamide^24-31. Additionally, several altered metabolic pathways in endometriosis such as metabolism of purines, linoleic acid, taurine and hypotaurine are also perturbed in IBD^32-36. This is interesting because while endometriosis has several associated comorbidities such as systemic lupus erythematosus, multiple sclerosis, atopic diseases, and neurological disorders³⁷, the condition is most commonly misdiagnosed as IBD. Women with endometriosis have a two- to three-fold increased risk of having IBS³⁸ and a significant overlap is seen in the symptoms of two diseases- chronic low-grade inflammation, dysbiosis, pain receptors stimulation, and leaky gut³⁹. We also reveal 12 metabolites identified as putative confounding biomarkers for preclinical diagnosis of endometriosis in women. These markers are also linked with inflammation, and some are also perturbed in IBD. For example, fMet acts as a neutrophil attractant and mediates several autoinflammatory conditions⁴⁰. Thus, our findings provide evidence that the progression of endometriosis mimics that of IBD and we posit that gut bacteria links the onset of these two diseases. This is intriguing, as endometriosis only develops in a fraction of women that experience retrograde menstruation, a theory well described for its origin³. Therefore, these findings further explain the role of perturbed gut microbiota in the onset of the disease. It is conceivable that the gut microbiota of women with endometriosis may be altered, especially considering the potential influence of pre-existing inflammatory gut disorders. On the contrary, endometriosis associated peritoneal inflammation alters gut microbiota, which could accentuate inflammatory gut disorders. Currently, our limited ability to non-invasively diagnose the disease severely delays the effective treatment plans². Thus, the metabolite markers identified in this study could provide options for cost-effective and non-invasive prognosis of endometriosis.

While fecal metabolites can unveil the composition of the intestinal microbiota, conducting an integrated analysis of both the metabolome and microbiome within the same cohort is essential for obtaining concrete functional outcomes. Indeed, by combining metabolite and microbiota profiles, we correlated 9 microbiota derived metabolites in endometriosis with bacterial species and identified three that were significantly altered during disease. Of these, we validated 4HI using LC-MS, thus making 4HI a unique bacteria-derived metabolite that is downregulated in women with endometriosis. Unlike other indole derivates, which are generated by bacteria from dietary sources, we have limited knowledge of the bacterial species that produce 4HI in human gut. However, we found a strong correlation of 4HI with bacterial species in endometriosis patients such as Faecalibacterium, Lachnospiracae spp. and Dorea. A strong positive correlation of 4HI with Faecalibacterium and Lachnospiracae and their downregulation in endometriosis suggests that these bacteria might help in increased 4HI production although other bacteria could also generate this metabolite. Remarkably, 4HI exhibits a robust negative correlation with the commensal bacterium Dorea formicigenerans. Interestingly, Dorea, a member of the commensal bacteria community, has been associated with dysbiosis in several diseases^41,42. Therefore, studying the specific functions of Dorea, under normal and dysbiotic states would help in understanding its role in endometriosis, which is of our immediate interest.

Despite recent advancements, effective treatments for endometriosis remain elusive, and existing strategies do not offer a cure. While the FDA has recently approved Myfembree^® to manage moderate to severe endometriosis pain in premenopausal women, it does not address the underlying drivers of disease onset and progression. Ongoing clinical trials that aim to treat endometriosis ignore the potential impact of these compounds on the normal endometrium that raises serious concerns. This is crucial as endometriosis predominantly affects women of reproductive age, many of whom aspire to conceive. Consequently, there is a pressing need for innovative therapies that not only target disease progression but also preserve the fertility of women with endometriosis. Our findings highlight 4HI as a gut microbiome-derived metabolite that is abundant in healthy gut and that has the potential to prevent the progression of the disease. We observed that 4HI not only inhibits the formation of lesions but also induces significant regression of fully developed lesions in vivo as observed from the lesion analysis upon 14 and 28 days of disease induction, respectively. Importantly, 4HI also remarkably prevented lesion formation upon pre-treatment prior to disease induction in mice. To note, the disease establishes through the pro-inflammatory macrophages that populate the peritoneal cavity within the first 14 days after donor endometrial implantation^4,43. These pro-inflammatory macrophages are produced to mediate acute inflammation for establishment of the disease. While after 14 days, the macrophage phenotype switches to anti-inflammatory that contribute to tissue remodeling for furthering the disease^4,43. Interestingly, 4HI demonstrated the ability to prevent the infiltration of M1-like pro-inflammatory macrophages while a lesser reduction was observed for M2-like macrophages that can mediate immunosuppression and neuroangiogenesis⁴⁴. Alternatively, a lesser reduction of these immunosuppressive M2-like macrophages in 4HI group might also reflect an overall decrease in the inflammation, and thus their requirement for immunosuppression. Of translational significance, 4HI severely affected the viability of endometriotic cells derived from women with endometriosis. In cultures, 4HI significantly reduced the expression of genes that augment endometriosis through proliferative and inflammatory signaling such as IGF1, NF-Κβ, TGF-β1 and GATA3^45-50. While other genes implicated in disease progression remained unchanged, such as GREB1 and CCND1⁵⁰. On the contrary, although GATA6 expression which activates estrogen synthesis and promotes endometriosis was reduced at first but increased at later time point⁵¹. This suggests that 4HI acts through distinct molecular pathways that need to be further elucidated.

In the pre-clinical model of mice xenotransplanted with human derived endometriotic cells, 4HI induced lesion regression, confirming its clinical relevance in endometriosis. Considering that 4HI produced naturally in humans by bacteria under homeostasis and could be well tolerated. Thus, 4HI presents itself as an attractive therapeutic target for women with endometriosis.

Collectively, of clinical relevance, we found a distinct stool metabolite signature in women with endometriosis, which could pave the way for stool-based non-invasive diagnostic testing. Importantly, we found a striking correlation between the gut metabolome of endometriosis with that of IBD, a comorbidity that the disease is most misdiagnosed for. Therefore, the biomarkers identified in this work provide a novel diagnostic metabolomics paradigm that could improve the prediction of disease burden and associated comorbidities. Specifically, these fecal biomarkers, could be utilized in non-invasive stool-based diagnostic tests, like Cologuard^® for colon cancer.

Limitations of Study

This work addressed the hypothesis that gut microbiota dysbiosis plays a pivotal role in endometriosis in affected women and that it does this, in part, by way of the produced metabolites. Hence, we utilized an untargeted metabolomics approach to uncover the global profiles of gut metabolites in patients and healthy subjects. Participants from different ethnicities and ages were included in this study. However, this is a gut microbiome-based study and is subjected to differences in the individual microbiota profiles due to several factors such as host genetics, lifestyle, diet, hormonal status, and other environmental factors. The metabolic capabilities of the host also vary with these factors. The key exclusion criteria included pregnancy, and any other gynecological pathologies for the patients’ cohort. However, information of participants’ socioeconomic status was not collected. Therefore, a more comprehensive view of these profiles would come from larger size of the cohorts. Our findings suggest that underlying mechanisms that help progress the disease might be similar as in the inflammatory bowel disease. Simultaneous comparisons of IBD patients and endometriosis patients will comprehensively address the overlaps between the two pathologies. We intend to address this limitation in the future work. The patient-derived endometriotic cell lines used in this study were immortalized and the tests have not been performed with primary endometriotic cells. We used the athymic nude mice model to study the metabolites’ effect on growth of xenotransplanted human derived endometriotic cells in vivo. Nude mice exhibit higher activity of natural killer (NK) cells that may prevent ectopic implantation resulting in varying transplantation outcomes, although complete absence of NK cells has also been postulated as a contributing factor for disease initiation and progression⁵².

STAR Methods

Resource availability

Lead contact

Further inquiries and requests should be directed to the lead contact: Dr. Ramakrishna Kommagani (Ramakrishna.Kommagani@bcm.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• This study did not generate or report the original code.

• The 16Sv4 sequence data generated in this study is available on NCBI under BioProject PRJNA1145097. The metabolomics data generated in this study is deposited in National Metabolomics Data Repository (NMDR, ID 5155).

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental models and Study Participant Details

Human

Stool samples from confirmed endometriosis patients and control healthy subjects were collected at Washington University in St. Louis upon informed consent. The protocol was approved by the Institutional Review Board (IRB) [ID #: 201612127]. Additional stool samples from healthy subjects were also provided by Biobank Core of the Digestive Disease Research Core Center (DDRCC) at Washington University in St. Louis under pay-per-service mode. Patients with confirmed disease-stage level diagnosis was the key inclusion criterion for patient enrolment. The key exclusion criteria included pregnancy for both cohorts, and any other gynecological pathologies for the patients’ cohort. Participants’ information on age and race was self-reported at the time of enrollment, however, information on socioeconomic status was not collected. The characteristics of subjects enrolled in the study are summarized in Table 1. The stool samples were homogenized to permit efficient recovery of microbial and metabolite components and aliquoted for use before storage at −80 °C.

Animals

All animal studies were approved by the Institutional Animal Care and Use Committee (IACUC) of Baylor College of Medicine, Houston, TX, USA. Eight to eleven-weeks old C57BL/6 mice (Taconic Biosciences Inc.) and athymic nude mice (#088NU/NU, Charles River) were used for heterologous mouse and human xenograft disease models, respectively. All animals were maintained in standard 12-h light/dark conditions and provided ad libitum access to food and water. Animals were handled according to an approved IACUC protocol number [AN-8890].

Cell Lines

Immortalized Human Endometriotic Epithelial Cells expressing Luciferase (iHEECs/Luc) and Immortalized Human Endometrial Stromal Cells/Luciferase (iHESCs/Luc) cells (generously provided by Dr. Sang Jun Han from Baylor College of Medicine, were maintained in DMEM/F12 containing 10% FBS, 100 U/ml penicillin, 100 mg/ml streptomycin, and 2.5 mg/ml amphotericin- B in humidified condition with 5% CO₂ and 95% air at 37°C, as described previously⁵³. The iHEECs/Luc cells are derived from an ovarian endometrioma lesion as described previously⁵⁴. The medium was changed every other day.

Method details

Unbiased metabolome profiling and data analyses

For all analyses, a 100 mg aliquot of fecal sample was mixed with 1000 μL of ice-cold 50/50 methanol/acetonitrile (v/v), homogenized, and centrifuged. The 700 μL of the homogenized sample was spiked with a standard internal mix containing representative standards for all metabolic classes. The samples were vortexed for 5 minutes and kept at −20°C for 10 minutes. Samples were then centrifuged for 10 minutes at 4°C and 15,000 rpm. Twenty microliters (20 μL) of each supernatant were pooled and separated in aliquots for quality controls (QC). Samples were then further twice aliquoted and evaporated until dryness using a GeneVac EZ-2 Plus SpeedVac (Ipswich, United Kingdom). For HILIC separation, aliquoted samples were reconstituted in 100 μL acetonitrile/water (50/50; v/v), while for reversed-phase separation, samples were reconstituted in 100 μL of methanol/water (20/80; v/v).

Separation was performed on a Thermo Scientific Vanquish Duo UHPLC. A Waters ACQUITY HSS T3 column (1.8 μm, 2.1 mm x 150 mm) was used for reversed phase separation and a Waters ACQUITY BEH amide column (1.7 μm, 2.1 mm x 150 mm) was used for HILIC separation. For reversed phase the gradient was from 99% mobile phase A (0.1% formic acid in H2O) to 95% mobile phase B (0.1% formic acid in methanol) over 16 minutes. For HILIC separation the gradient was the same with solvent A: 0.1% formic acid, 10 mM ammonium formate, 90% acetonitrile, 10% H₂O, and solvent B: 0.1% formic acid, 10 mM ammonium formate, 50% acetonitrile, 50% H₂O. Both columns were run at 45 °C with a flow rate of 300 μL/min with an injection volume of 1 μL. Thermo Scientific Orbitrap Exploris 480 was used for data collection with a spray voltage of 3500 V for positive mode (reverse phase separation) and 2500 V for negative mode (HILIC separation) using the H-ESI source. Vaporizer temperature and ion transfer tube were both 350°C. Compounds were fragmented using data-dependent MS/MS with HCD collision energies of 20, 40, and 80%.

The peak area of each metabolite was log2 transformed and normalized by an isotopically spiked internal standard for each method. The differential metabolites were determined using a threshold of 1 log fold change (FC) and p-values of t-test following the Benjamini-Hochberg method with less than 0.25 false discovery rate (FDR). Classification of total metabolites, and quantitative metabolite set enrichment analysis (qMSEA) of pathways and disease signatures in altered metabolites were conducted in MetaboAnalyst v5.0⁵⁵ using log-transformed peak area as data input to compute based on KEGG and 44 a priori defined sets of disease associated metabolites reported in human feces. For the latter, metabolite sets that contained at least two compounds were used. Heatmaps were generated using pheatmap package in R (R Development Core Team). The area under curve (AUC) of ROC curves as well as sensitivity and specificity values were calculated to determine the diagnostic effectiveness of important metabolites using MetaboAnalyst 5.0⁵⁵.

Microbiota profiling

Total DNA was isolated using QiAmp PowerFecal Pro DNA kit and V4 region of the 16S rRNA gene sequence were amplified for sequencing using Illumina platform. Demultiplexed read pairs 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 following merge parameters: maxstrict=t, qtrim=t, trimq=15. Merged reads were further filtered via VSEARCH⁵⁶, using max error rate=0.05, min length=252, max length=254, deblur length limit=252. All reads were then combined into a single FASTA file for further processing using UPARSE⁵⁷. 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 138.1⁵⁸ containing only sequences from the V4 region of the 16S rRNA gene to determine taxonomies using the usearch70 ‘usearch_global’ function and specifying the identity threshold to 97%⁵⁹. Phylogeny information contained in the biom file was generated by aligning the centroid sequences with MAFFT⁶⁰ and creating a tree via FastTree⁶¹. 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 calculated based on Inverse Simpson indices.

Integrative analysis of metabolome-microbiome

Metabolites-microbiota correlation analysis and origin analysis was performed using MetOrigin⁶² to identify metabolites of microbial origin. The data was normalized based on microbial relative abundances (percentage) and log-transformed metabolites. Correlation between microbiota and metabolite features was computed based on Spearman coefficients (p-value<0.05). For validation studies, selected bacterial taxa that correlated with the two compounds were quantified using real-time quantitative PCR using group-specific primers by normalizing with universal bacterial primers^63-65 (Supplementary Table S4).

LC-MS validation assay

For validation of the selected metabolites, the sample extraction, binary gradient, solvents, and LC columns were the same as described for unbiased metabolomics. The injection volume was 10 μL. The data were acquired via multiple reaction monitoring (MRM) with positive and negative electrospray ionization (ESI) mode using a 6495C Triple Quadrupole mass spectrometry through Agilent Mass Hunter Software. Identified peaks and retention time were carefully reviewed using Agilent Mass Hunter Quantitative Analysis Software (Agilent Technologies, Santa Clara, CA). Pooled quality control samples were monitored for the overall quality of the extraction and mass spectrometry analyses. The relative peak area was log2 transform, followed by internal standard normalization for each method. The differentially expressed analysis was applied with the Benjamini-Hochberg method for false discovery rate (FDR<0.25) correction accounting for multiple comparisons.

Cell viability assays and qRT-PCR

Cell viability at different concentrations of metabolites was determined on the iHEECs/Luc cells by performing the MTT assay (Promega, #G401A) according to the manufacturer's instructions. For this, cells were plated in 96-well plates. After 24 h, cells were treated with vehicle (DMSO) or 0 μM, 50 μM, 100 μM, 250 μM, or 500 μM or 1 mM concentrations of metabolites for 0h, 24h, 48h and 72h and relative cell viability was determined by the MTT assay. In all cases, 15 μl of MTS (dye solution) reagent was added to each well and incubated for another 2h. After addition of 100 μl of solubilization solution, absorbance was measured at 570 nm with 630 nm as a reference wavelength in a 96-well plate reader. The experiments were performed three times each with three technical replicates. In vitro assays to study the effect of 4HI on gene expression changes in iHEECs/Luc cells were performed using real time-quantitative PCR from RNA derived from cells treated with 4HI for 0h, 3h and 6h compared with untreated (DMSO) controls in duplicates. The relative gene expression was measured using 18S rRNA gene as reference and normalized using measures of untreated controls at corresponding time points. The statistical significance of each time point compared with the 0h expression of each gene target was determined using student’s t-test.

Induction of endometriosis in mice and metabolite treatments

Donor mice were subcutaneously injected with estradiol benzoate (3 μg/mouse or 100 μg/kg) on day −7^66,67. On day 0, donor mice were euthanized (one donor mouse for every two recipients), and uteri were removed, placed in a petri dish containing warm saline, and cut longitudinally^68,69. Endometrial tissue from each uterine horn was mechanically disrupted with micro-scissors to produce suspensions in which the maximal diameter of any piece of endometrial tissue was less than 1 mm⁷⁰. The suspension was intraperitoneally injected⁷¹ into recipient mice (0.4 ml/mouse) with a 1-ml syringe and a 25 ga needle⁷². Following induction, the recipient mice were orally gavaged with metabolites 4-HI, I4CAld or indole starting either from Day 1 through Day 14 or from Day 14 through Day 28 to study the effect of each of these metabolites on establishment of endometriotic lesions and their progression, respectively. Each metabolite was gavaged at 15 mg/Kg concentration and a total gavage volume of 200 μl was used per mouse. Upon 14 or 28 days, mice were sacrificed for harvesting body fluids and endometriotic lesions. For collecting peritoneal lavages, the mice were first injected intraperitoneally with 5 ml of peritoneal macrophage (PM) isolation media (1X PBS supplemented with 3% Fetal Bovine Serum) to flush immune cells into the peritoneal space. The mice were then euthanized by cervical dislocation and the abdominal cavity was immediately opened to collect as much peritoneal lavage using a 1ml syringe. The endometriotic lesions were carefully located and excised, trimmed of excess fat and then measured for weights and volumes, and processed for histology and immunofluorescence^66,67. To study the effect of 4HI on initiation of endometriosis, mice were pre-treated with metabolite 7 days prior to disease induction. The mice were then sacrificed upon 14 days from induction and lesions were examined as detailed above.

Induction of human ectopic lesion and in vivo analysis

This model of endometriosis was generated in athymic nude mice (#088NU/NU, Charles River) as described previously⁵³. Briefly, 2 d before the day of transplantation, mice were ovariectomized, and a sterile 60-d release pellet containing 0.36 mg of 17-β estradiol (Innovative Research of America) was implanted. On the day of transplantation, iHESCs/Luc and iHEECs/Luc cells were trypsinized with 0.05% trypsin-EDTA, and 2 × 10⁶ iHESCs/Luc and 2 × 10⁶ iHEECs/Luc cells were combined in 10 ml of DMEM/F12, pelleted, washed, resuspended in 100 μl of DMEM/F12, and mixed with 100 μl of Matrigel (BD Biosciences). The cell suspension/matrigel mixture (200 μl) was intraperitoneally injected into the mice on the midventral line just caudal to the umbilicus. The mice were administered either vehicle or metabolites 4HI, I4CAld, or Indole at 15 mg/kg concentration using oral gavage starting from D14 through D28. Bioluminescence images of each mouse were collected with an in vivo image analysis system at Mouse Phenotyping Core at Baylor College of Medicine at D0, D14 and D28 upon injection with D-Luciferin (Sigma, #L2916). Mice were then euthanized on Day 28 and endometriotic lesions were collected, measured, and processed for histology and immunofluorescence.

H&E staining and immunofluorescence

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 as described previously⁷³. For immunofluorescence, tissue sections (n=5 per group) were deparaffinized, rehydrated, and boiled for antigen retrieval as described previously^73,74. 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), F4/80 (1:100, ThermoFisher Scientific, #53-4801-82) and CD31 (1:100, CST, #77699S). After washing with PBS, sections were incubated with Alexa Fluor 488-conjugated secondary antibodies (Life Technologies) for 1 h at room temperature and mounted with ProLong Gold Antifade Mountant with DAPI (Thermo Scientific #P36962).

FACS analysis

The peritoneal lavages were collected from endometriotic mice treated with metabolites from day 1 through day 14 before euthanasia by injecting 1ml sterile PBS into the peritoneal space. Red blood cells were removed by incubating with RBC lysis buffer (ThermoFisher Scientific #J62150.AK) for 5 minutes on ice and filtered through 40μM filters. All the cells of respective groups were pooled together, blocked with 0.25μg anti-CD16/CD32 (clone 93 eBioscience) for 30 minutes and then stained with an antibody from a panel of conjugated-antibodies- anti-CD16/CD32, anti-CD45, anti-CD11b, anti-MHCII, anti-CD206, anti-Ly6G, anti-Ly6c as described in the Key Resources Table. Super bright complete staining buffer (eBioscience #SB-4401-75) was included when required. Ultra-compensation beads were used as single stain controls along with unstained controls to validate the gating strategies. Samples were analyzed using an Aurora with Flowjo v9software (Flowjo v.9). Analysis was performed in single live cells using a forward scatter versus negative for live/dead cell (Near-IR dead cell stain). For peritoneal populations absolute counts were analyzed by the total absorption of cells/volume. Final volume for cytofluorimetric analysis performed was 350μl.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-Ki-67 (Dilution 1:100)	Abcam	Catalog #ab16667; RRID: AB_302459
anti-F4/80 (Dilution 1:100)	Thermo Fisher Scientific	Catalog #53-4801-82; RRID: AB_469915
Goat anti-Rabbit IgG (H+L) Alexa Fluor™ Plus 488 (Dilution 1:200)	Thermo Fisher Scientific	Catalog #A32731; RRID: AB_2633280
Goat anti-Rat IgG (H+L) Alexa Fluor™ Plus 488 (Dilution 1:200)	Thermo Fisher Scientific	Catalog #A11006; RRID: AB_2534074
anti-CD16/CD32	eBioscience	Catalog #14-0161-82
anti-CD45 - FITC (Dilution 1:600)	eBioscience	Catalog #11-0451-82
anti-CD11b- BV650 (Dilution 1:200)	eBioscience	Catalog #416-0112-82
anti-MHCII- SB702 (Dilution 1:200)	eBioscience	Catalog #67-5321-82
anti-CD206 - PE-CYN7 (Dilution 1:200)	eBioscience	Catalog #25-2061-82
anti-Ly6G - PCP-E710 (Dilution 1:200)	eBioscience	Catalog #46-5931-82
anti-Ly6c - APC-E780 (Dilution 1:200)	eBioscience	Catalog #47-5932-82
Biological samples
Human Stool Samples	Dr. Patricia Jimenez and Dr. Scott Biest, Washington University School of Medicine, St. Louis, MO, 63110, USA	IRB ID #201612127
Chemicals, peptides, and recombinant proteins
4-hydroxyindole	Sigma	Catalog #219878
Indole-4-carbaldehyde	Sigma	Catalog #632422
Indole	Sigma	Catalog #I3408
Nicotinic acid	Sigma	Catalog #N4126
Amphotericin-B	Sigma	Catalog #171375
D-Luciferin	Sigma	Catalog #L2916
Matrigel	BD Biosciences	Catalog #356234
0.36 mg pellets of 17-β estradiol	Innovative Research of America	Catalog #E-121
β-Estradiol-3-benzoate	Sigma	Catalog #1251000
ProLong Gold Antifade Mountant with DAPI	ThermoFisher Scientific	Catalog #P36962
RBC lysis buffer	ThermoFisher Scientific	Catalog #J62150.AK
Super bright complete staining buffer	eBioscience	Catalog #SB-4401-75
Critical commercial assays
CellTiter 96® Non-Radioactive Cell Proliferation Assay (MTT)	Promega	Catalog #G401A
QIAamp PowerFecal Pro DNA Kit	Qiagen	Catalog #51804
Deposited data
16Sv4 Sequence Reads	This Study	NCBI BioProject: PRJNA1145097
Metabolomics data	This Study	National Metabolomics Data Repository: ID 5155
Experimental models: Cell lines
Immortalized Human Endometriotic Epithelial Cells expressing Luciferase (iHEECs/Luc)	Dr. Sang Jun Han, Baylor College of Medicine (Han et al., 2015)	N/A
Immortalized Human Endometrial Stromal Cells/Luciferase (iHESCs/Luc)	Dr. Sang Jun Han, Baylor College of Medicine (Han et al., 2015)	N/A
Experimental models: Organisms/strains
C57BL/6 Mice	Taconic Biosciences Inc.	Catalog #B6-MPF-F
Athymic nude mice	Charles River Laboratories International, Inc.	Catalog #088NU/NU
Oligonucleotides – Refer to Supplementary Table S4
Software and algorithms
MetaboAnalyst v5.0	(Pang et al., 2021)	http://www.metaboanalyst.ca
Pheatmap	R Development Core Team	https://github.com/raivokolde/pheatmap.git
BBMap version 38.82	N/A	sourceforge.net/projects/bbmap/
VSEARCH	(Rognes et al., 2016)	N/A
UPARSE	(Edgar, 2013)	N/A
MAFFT	(Katoh and Standley, 2013)	N/A
FastTree	(Price et al., 2009)	N/A
MetOrigin	(Yu et al., 2022)	https://metorigin.met-bioinformatics.cn/home/
Bruker Molecular Imaging	Bruker Corporation	N/A
GraphPad Prism version 9.5.0	N/A	https://www.graphpad.com/
FlowJo	BD Biosciences	N/A
ImageJ	(Schneider et al., 2012)	https://imagej.nih.gov/ij/
Agilent Mass Hunter Quantitative Analysis Software	Agilent Technologies, Santa Clara, CA	N/A

Macrophage isolation and qRT-PCR

Peritoneal macrophages were cultured from the peritoneal lavages after RBC removal using RBC lysis buffer. The cells were cultivated in RPMI medium (ThermoFisher Scientific #11875135) supplemented with 10% FBS, 1% Penicillin and Streptomycin. After 6h, cells were washed with PBS and replenished with fresh RPMI medium. After 24h, cells were lysed with RNA Lysis buffer to extract RNA. RT-qPCR was performed for M1-like and M2-like macrophage gene markers using 18S rRNA as the reference gene.

Von Frey Tests

Von Frey test as a non-invasive technique to determine mechanical pain sensation in mice was carried out at Baylor College of Medicine Intellectual and Developmental Disabilities Research Center. Animals were placed on a wire mesh platform in a transparent Plexiglas chamber and allowed to habituate for 1 hour, with dim lighting (150 Lux) and 60 dB background white noise. A series of stiff filaments ranging from 0.02 to 8 grams (Stoelting Co., Wood Dale, IL, USA) were applied through the wire mesh onto the plantar surface of both hind paws in ascending order starting with the finest fiber. Each mouse was subjected to 2 trials on each paw and the average hind paw withdrawal response was calculated. The tester was blind to the treatment information.

Quantification and statistical analyses

For unbiased metabolomics, the peak area of each metabolite was log₂ transformed and normalized by an isotopically spiked internal standard for each method. The differential metabolites were determined using p-values of t-test following the Benjamini-Hochberg method with less than 0.25 false discovery rate (FDR). To select metabolites that highly contributed to the clear separation behaviors of healthy and EMS groups, the variable importance in the projection (VIP)-scores of orthogonal Projections to Latent Structures Discriminant Analysis (oPLS-DA) statistical approach were calculated. The VIP quantitatively estimates the importance of each variable in the projection for its discriminatory power. Metabolites with a VIP score of >1 were considered important features driving the alterations. Significance of the altered microbial taxa abundance was determined using Mann-Whitney statistic (p<0.05). Measures of correlation between pairs of metabolite and bacterial genera were determined using Spearman coefficients and p-value thresholds of <0.05 were considered significant. In experiments with two groups, a two-tailed paired t-test was employed while ANOVA by nonparametric alternatives was used for multiple comparisons to analyze data from in vivo experiments. P<0.05 was considered significant. All data are presented as mean ± SE. For quantitative estimation of Ki67 and F4/80 staining positive cells, we randomly imaged five regions in the stained lesions sections from each group and counted the total cells and positively stained cells using ImageJ⁷⁵. Counting was performed by three researchers blinded to the groups and expressed as percentages of Ki67 and F4/80 -staining positive cells.

Highlights.

1. Women with endometriosis have a distinct stool metabolome for non-invasive diagnosis.

2. Bacteria-derived metabolites in endometriosis are associated with those in IBD.

3. Bacteria-derived 4-hydroxyindole level is lower in stool from women with endometriosis.

4. 4-hydroxyindole inhibits the onset and progression of endometriosis.

Context and Significance.

Clinical management of endometriosis, a gynecological disease that is a leading cause of infertility in women, remains a challenge with no known cure. Scientists from Baylor College of Medicine combined the gut metabolomic and microbiota signatures in the stool samples from women with endometriosis to identify a gut bacteria-derived therapeutic intervention candidate, 4-hydroxyindole. They demonstrate that 4-hydroxyindole prevents the formation of endometriotic lesions, augments their regression, and reduces disease-associated pain. Their work also identified altered bacteria-derived distinct metabolites that could be potentially used for non-invasive diagnosis of the disease.