Gut commensal Olsenella profusa promotes endometriosis via lysophosphatidylcholine metabolite

D.Z. and Y.M. conceived the project; Y. Chen, Z.Z., B.H., X.X., Y. Cao, S.Z., B.L., F.X., L.P., J.L., and S.C. performed and analyzed the experiments; Y. Chen, Z.Z., Y. Cao, and H.L. performed clinical investigation and analysis; Y. Chen, Z.Z., X.X., and D.Z. performed data visualization; D.Z., Y.M., H.Z., and B.L. provided funding support; D.Z., Y.M., and H.Z. supervised the experiments; Y. Chen and D.Z. wrote the original draft; Y. Chen, Z.Z., B.H., H.Z., Y.M., and D.Z. performed review and editing.

Although gut microbiota dysbiosis has been observed in women with EM, the EM-associated gut microbial signatures are not well defined. To further thoroughly determine the features of gut microbiota that may affect EM pathogenesis, we conducted 16S rRNA gene sequencing of feces from controls and EM patients, including healthy control group (HC, n = 50), benign gynecological diseases control group (BGC, n = 50; comprising patients undergoing surgery for ovarian cysts, uterine fibroids, or tubal pathologies, with no evidence of EM confirmed by laparoscopy), I/II EM group (I/II EM, n = 50) and III/IV EM group (III/IV EM, n = 64) ( Figure 1 A; Table S1 ). The demographic and clinical characteristics of all participants, including age, body mass index (BMI), and menstrual cycle, were comparable across the four groups, with no significant differences observed ( Table S1 ), thus minimizing potential confounding effects on gut microbiota analysis. All patients receiving hormone therapy were excluded from our cohort study. Menstrual cycle stage (proliferative or secretory) was assessed where possible based on last menstrual period and sampling date, and no significant difference in stage distribution was observed across groups ( p = 0.94, Table S1 ). After analyzing gut microbiota by alpha diversity indices Simpson, Pielou, and Shannon, we found that total abundance of gut bacteria in both I/II EM and III/IV EM groups was reduced relative to HC group, with I/II EM group showing the most significant reduction ( Figure 1 B). However, no significant differences emerged when compared to BGC group, though I/II EM group still demonstrated a declining trend ( Figure 1 B). As for beta diversity of gut microbiota based on unweighted UniFrac distance, the principal component analysis (PCA) revealed notable difference between groups ( Figure 1 C), while the PCoA also showed certain difference ( Figure S1 A). Specifically, significant differences were identified between I/II EM and HC group, as well as between I/II EM and III/IV EM groups ( Figure 1 C). These data suggest that more pronounced gut microbiota dysbiosis was detected in patients with stage I/II pathology. Figure 1 O. profusa is enriched in the gut of EM patients with microbiota dysbiosis (A) The schematic diagram illustrating the workflow of gut microbiota analysis in study cohort. (B) Alpha diversity analysis of gut microbiota through Simpson, Pielou, and Shannon indices, with study cohort stratified into healthy control group (HC, n = 50), benign gynecological diseases control group (BGC, n = 50), I/II endometriosis group (I/II EM, n = 50), and III/IV endometriosis group (III/IV EM, n = 64). Individual data points are overlaid on boxplots, where horizontal lines denote median values. (C) Beta diversity analysis of gut microbiota among HC, BGC, I/II EM, and III/IV EM groups through PCA based on unweighted UniFrac distances. (D) The top ten bacterial orders in terms of abundance among HC, BGC, I/II EM, and III/IV EM groups. (E) Venn diagram showing bacterial taxa distribution. Unique taxa (non-overlapping regions) and shared taxa (overlapping regions) across four groups (HC, BGC, I/II EM, and III/IV EM). (F) Correlation analysis of four candidate bacteria abundance with EM-related clinical indicators. Positive correlations are depicted in red, while negative correlations are shown in blue. (G) Relative abundance of O. profusa in feces across four groups (HC, BGC, I/II EM, and III/IV EM). O. profusa was analyzed by 16S sequencing. (H) Relative abundance of O. profusa in feces across four groups (HC, BGC, I/II EM, and III/IV EM). O. profusa was analyzed by qPCR. Data are presented as mean ± standard deviation (SD), analyzed by one-way ANOVA with Fisher’s least significant difference (LSD) post hoc test (B, C, and G), Spearman’s correlation (F), and Kruskal-Wallis test followed by Dunn’s post hoc test (H). ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001. I/II EM, stage I/II of endometriosis; III/IV EM, stage III/IV of endometriosis; ROMA_post, risk of ovarian malignancy algorithm-postmenopausal; ROMA_pre, risk of ovarian malignancy algorithm-premenopausal; AFP, alpha-fetoprotein; HE4, human epididymis protein 4; O. profusa, Olsenella profusa ; qPCR, quantitative polymerase chain reaction; CT, cycle threshold. O. profusa is enriched in the gut of EM patients with microbiota dysbiosis (A) The schematic diagram illustrating the workflow of gut microbiota analysis in study cohort. (B) Alpha diversity analysis of gut microbiota through Simpson, Pielou, and Shannon indices, with study cohort stratified into healthy control group (HC, n = 50), benign gynecological diseases control group (BGC, n = 50), I/II endometriosis group (I/II EM, n = 50), and III/IV endometriosis group (III/IV EM, n = 64). Individual data points are overlaid on boxplots, where horizontal lines denote median values. (C) Beta diversity analysis of gut microbiota among HC, BGC, I/II EM, and III/IV EM groups through PCA based on unweighted UniFrac distances. (D) The top ten bacterial orders in terms of abundance among HC, BGC, I/II EM, and III/IV EM groups. (E) Venn diagram showing bacterial taxa distribution. Unique taxa (non-overlapping regions) and shared taxa (overlapping regions) across four groups (HC, BGC, I/II EM, and III/IV EM). (F) Correlation analysis of four candidate bacteria abundance with EM-related clinical indicators. Positive correlations are depicted in red, while negative correlations are shown in blue. (G) Relative abundance of O. profusa in feces across four groups (HC, BGC, I/II EM, and III/IV EM). O. profusa was analyzed by 16S sequencing. (H) Relative abundance of O. profusa in feces across four groups (HC, BGC, I/II EM, and III/IV EM). O. profusa was analyzed by qPCR. Data are presented as mean ± standard deviation (SD), analyzed by one-way ANOVA with Fisher’s least significant difference (LSD) post hoc test (B, C, and G), Spearman’s correlation (F), and Kruskal-Wallis test followed by Dunn’s post hoc test (H). ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001. I/II EM, stage I/II of endometriosis; III/IV EM, stage III/IV of endometriosis; ROMA_post, risk of ovarian malignancy algorithm-postmenopausal; ROMA_pre, risk of ovarian malignancy algorithm-premenopausal; AFP, alpha-fetoprotein; HE4, human epididymis protein 4; O. profusa, Olsenella profusa ; qPCR, quantitative polymerase chain reaction; CT, cycle threshold. To determine the causal relationship of gut microbiota dysbiosis in EM pathogenesis, fecal microbiota transplantation (FMT) was conducted in EM mice model by using fecal bacterial suspension prepared from pooled fecal samples of healthy subjects ( n = 5) and EM patients ( n = 5) respectively. Starting from the modeling of EM, the mice were gavaged 200 μL of fecal bacterial suspension daily for 21 consecutive days, while vehicle controls received 200 μL of 20% glycerol ( Figure S2 A). The mice that received fecal microbiota from EM patients showed significant accelerated progression of EM lesions compared with those that received fecal microbiota from HC ( Figure S2 B). Specifically, the average weight, average diameter, and quantity of all lesions in FMT-EM group were markedly higher compared with both control group and FMT-HC group ( Figure S2 B). Histologically, the lesion in FMT-EM group exhibited increased glandular density, thicker glandular epithelial cells, and higher proliferation rates of endometrial cells (Ki-67 + ) compared with both control group and FMT-HC group ( Figure S2 C). These results indicate that dysbiotic gut microbiota is directly involved in EM pathogenesis. Next, we explored the potential gut bacteria that may be implicated in the pathogenesis of EM. The comparative analysis of top 10 bacterial order revealed distinct taxonomic shifts across groups in our study cohort. Specifically, the EM groups exhibited significantly higher abundance of Fusobacteriales , Coriobacteriales , and Enterobacteriales as well as lower abundance of Clostridiales compared with HC and BGC groups ( Figure 1 D). In addition, the analysis results of differentially enriched bacteria across groups were also shown ( Figure S1 B). Importantly, Venn analysis revealed the significant enrichment of four bacterial taxa in I/II EM and III/IV EM groups compared with HC and BGC groups, including V119_ Olsenella profusa , V191_ Helicobacter sp . , V159_ Mycoplasma muris , and V164_ Succinivibrio sp. ( Figure 1 E). Then, these four candidate bacterial species were subjected to clinical correlation analysis. After adjusting for potential confounding factors, such as age, BMI, and hormone therapy status, significant positive correlations were still observed between the relative abundance of Olsenella profusa ( O. profusa ; phylum: Actinobacteria, order: Coriobacteriales) and EM-related clinical biomarkers, including CA125, ROMA-post, and ROMA-pre ( Figure 1 F). In contrast, V191, V159, and V164 showed no significant correlation with EM-related clinical markers, and their relative abundances were very low in almost all EM patients ( Figure 1 F; Figures S1 C–S1E). Compared with HC and BGC groups, the EM groups showed significant enrichment of O. profusa in the gut ( Figure 1 G). Meanwhile, quantitative PCR (qPCR) results consistently confirmed the 16S sequencing data ( Figure 1 H). Collectively, these findings support the hypothesis that O. profusa may be a characteristic pathogenic bacterium for EM. To confirm the pathogenic impact of O. profusa on EM progression, single O. profusa colonization was conducted in pseudo-sterile EM mice that were pre-treated with antibiotics combination ( Figure 2 A). Antibiotic treatment effectively depleted the gut microbiota ( Figures S3 A–S3C). After estrous cycle regulation by estrogen, we transplanted uterine fragments from donor mice into the peritoneal cavities of pseudo-sterile recipient mice to model ectopic lesions of EM. Then, recipient mice were subjected to daily gavage of O. profusa for three weeks, while controls received either saline (vehicle control) or E. coli 1655 (negative bacterial control) respectively. Here, qPCR was performed to confirm the successful colonization of O. profusa in the gut of model mice ( Figure S3 D). After three weeks, cystic lesions developed in the abdominal cavities of recipient mice that resembled those found in EM patients, accompanied by the presence of aged blood within the cysts ( Figure 2 B). Meanwhile, mice receiving O. profusa gavage exhibited significantly lower body weight compared with those receiving saline and E. coli 1655 ( Figure S3 E). Furthermore, O. profusa -treated mice displayed larger lesion weight and diameter ( Figure 2 C). Lesion numbers showed no significant difference between groups, although an upward trend was observed in the O. profusa group ( Figure S3 F). Histological examination revealed the thickening of glandular epithelium along with increased epithelial and stromal cell proliferation (Ki-67 + ) ( Figure 2 D). In addition, O. profusa gavage significantly induced the expressions of pro-inflammatory factors, including Il6 , Cxcl15 , Il1b , Il18 , and Tnf , in EM lesions ( Figure 2 E). To address whether the timing of colonization influences the outcome, we performed an additional independent experiment in which recipient mice were gavaged with E. coli 1655 or O. profusa for 7 days prior to lesion induction ( Figure S3 G). Notably, the pre-colonization experiment demonstrated that administration of O. profusa prior to lesion induction significantly enhanced endometriotic lesion development, as evidenced by increased lesion weight and diameter as well as lesion number compared with those in control ( Figures S3 H and S3I). Histological examination also showed the thickening of glandular epithelium ( Figure S3 J). Together, these findings suggest that O. profusa could directly promote the progression of EM lesions in mice. Figure 2 O. profusa promotes EM progression in mice (A) The schematic diagram showing experimental timeline and procedures for EM model in mice ( n = 5/group). (B) Representative images of EM lesions in mice. (C) Representative images of EM lesions along with the average weight and diameter measurements of EM lesions in mice ( n = 5/group). (D) Representative images of H&E and Ki67 staining of EM lesions in mice. Scale bars: 200 μm for low-magnification H&E images and 100 μm for high-magnification H&E images and fluorescence images ( n = 3/group). (E) The expressions of inflammatory factors in EM lesions determined by qPCR analysis ( n = 5/group). Data are presented as mean ± standard deviation (SD), analyzed by one-way ANOVA with Fisher’s LSD post hoc test (C; E- Il6 , Il1b , and Tnf ) or Kruskal-Wallis test followed by Dunn’s post hoc test (D; E- Cxcl15, and Il18 ). ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001. E2, estradiol; IP, intraperitoneal; ig, intragastric; NC, negative control; O. profusa, Olsenella profusa ; ABX, antibiotics combination. O. profusa promotes EM progression in mice (A) The schematic diagram showing experimental timeline and procedures for EM model in mice ( n = 5/group). (B) Representative images of EM lesions in mice. (C) Representative images of EM lesions along with the average weight and diameter measurements of EM lesions in mice ( n = 5/group). (D) Representative images of H&E and Ki67 staining of EM lesions in mice. Scale bars: 200 μm for low-magnification H&E images and 100 μm for high-magnification H&E images and fluorescence images ( n = 3/group). (E) The expressions of inflammatory factors in EM lesions determined by qPCR analysis ( n = 5/group). Data are presented as mean ± standard deviation (SD), analyzed by one-way ANOVA with Fisher’s LSD post hoc test (C; E- Il6 , Il1b , and Tnf ) or Kruskal-Wallis test followed by Dunn’s post hoc test (D; E- Cxcl15, and Il18 ). ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001. E2, estradiol; IP, intraperitoneal; ig, intragastric; NC, negative control; O. profusa, Olsenella profusa ; ABX, antibiotics combination. To further explore the metabolite and mechanism by which O. profusa affects EM progression, we performed untargeted metabolomics analysis on randomly selected fecal samples from four groups in our cohort by using liquid chromatography-tandem mass spectrometry (LC-MS/MS) ( Figure 3 A). As a result, PLS-DA (partial least squares discriminant analysis) revealed different clustering patterns across different groups in the cohort, which identified unique metabolomic characteristics in the feces of patients with EM ( Figure 3 B). The metabolomic characteristics of the HC group showed significant separation from the other three groups, especially from the I/II EM and the III/IV EM groups ( Figure 3 B). However, the BGC and the I/II EM groups showed partial overlap ( Figure 3 B). Next, Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis suggested that differential metabolites among the four groups were enriched in the glycerophospholipid metabolism, glycerolipid metabolism, phosphatidylinositol signaling system, long-term depression, and inositol phosphate metabolism ( Figure 3 C; Figure S4 A). Among them, the content of LPC (16:0) was significantly higher than other metabolites in the glycerophospholipid metabolism pathway ( Figure S4 B) and significantly higher in the EM group than in the HC and the BGC groups ( Figures 3 D and 3E; Figure S4 B). Consistently, the targeted mass spectrometry analysis indicated that the LPC (16:0) content in the serum was significantly higher in the EM group than in the HC and the BGC groups ( Figure 3 F). Meanwhile, the levels of the LPC (16:0) upstream metabolite phosphatidylcholine (PC) in the serum were lower in the EM group than in the HC group ( Figure S4 C). Spearman correlation analysis showed that serum LPC (16:0) was significantly positively correlated with EM-related clinical markers, including CA125 and CA199 ( Figure 3 G). These results suggest that LPC (16:0) may represent a gut-derived key metabolite driving EM progression. Figure 3 O. profusa drives EM progression by producing LPC (16:0) via its phospholipase (A) The schematic diagram showing untargeted metabolomics analysis of patient cohort (HC, n = 32; BGC, n = 32; I/II EM, n = 32; and III/IV EM, n = 35). (B) Partial least-squares discriminant analysis (PLS-DA) of untargeted metabolomic results across four groups (HC, BGC, I/II EM, and III/IV EM). (C) KEGG enrichment analysis of differential metabolites in feces across four groups (HC, BGC, I/II EM, and III/IV EM). Top 20 most significant metabolic pathways ranked by p value are displayed. Rich factor represents the ratio of the number of differentially abundant metabolites in a metabolism pathway to the total number of metabolites annotated to that pathway. Dot size represents the count of significantly enriched metabolites in each pathway. (D) Multiple volcanic diagram of differential metabolite among groups displaying the significant altered metabolites identified in feces, highlighting LPC (16:0). log 2 FC > 1. p < 0.05. (E) Violin plots showing the relative abundance of LPC (16:0) in feces across four groups (HC, BGC, I/II EM, and III/IV EM). (F) The concentration of LPC (16:0) in serum across four groups (HC, BGC, I/II EM, and III/IV EM). Targeted mass spectrometry was employed to quantify the levels of LPC (16:0). (G) Correlation analysis between LPC (16:0) levels in feces/serum and characteristic clinical indicators of EM. (H and I) Correlation analysis between the relative abundance of O. profusa in feces and the levels of LPC (16:0) in feces (H) and serum (I). (J) The concentration of LPC (16:0) in mouse serum across different groups. Targeted mass spectrometry was employed to quantify the levels of LPC (16:0) ( n = 5/group). (K) KEGG enrichment analysis of O. profusa metabolites. (L) Genomic annotation of O. profusa identified a patatin-like phospholipase family protein, designated Op-PLA2. (M) The expression levels of LPC (16:0) in bacterial culture supernatant. The Op-PLA2 gene was cloned and expressed in E. coli , and E. coli -vector, E. coli -PLA2, and O. profusa were co-incubated with PC respectively. The concentrations of LPC (16:0) in culture supernatant were quantified by targeted mass spectrometry ( n = 5/group). (N) The representative images of EM lesions along with the average weight and diameter measurements of EM lesions in mice ( n = 10/group). (O) Serum LPC (16:0) concentrations in the three groups of mice (NC, E. coli -vector, and E. coli -PLA2) measured by mass spectrometry ( n = 10/group). Data are presented as mean ± standard deviation (SD), analyzed by one-way ANOVA with Fisher’s LSD post hoc test (E) or Kruskal-Wallis test followed by Dunn’s post hoc test (F, J, and M‑O) or Spearman’s correlation (G‑I). ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001. I/II EM, stage I/II of endometriosis; III/IV EM, stage III/IV of endometriosis; ROMA_post, risk of ovarian malignancy algorithm-postmenopausal; ROMA_pre, risk of ovarian malignancy algorithm-premenopausal; AFP, alpha-fetoprotein; HE4, human epididymis protein 4; O. profusa, Olsenella profusa; PC, phosphatidylcholine; E. coli -vector, E. coli BL21 carrying the empty pET-21b(+) vector (negative control); E. coli -PLA2, E. coli BL21 heterologously expressing Op-PLA2 gene. O. profusa drives EM progression by producing LPC (16:0) via its phospholipase (A) The schematic diagram showing untargeted metabolomics analysis of patient cohort (HC, n = 32; BGC, n = 32; I/II EM, n = 32; and III/IV EM, n = 35). (B) Partial least-squares discriminant analysis (PLS-DA) of untargeted metabolomic results across four groups (HC, BGC, I/II EM, and III/IV EM). (C) KEGG enrichment analysis of differential metabolites in feces across four groups (HC, BGC, I/II EM, and III/IV EM). Top 20 most significant metabolic pathways ranked by p value are displayed. Rich factor represents the ratio of the number of differentially abundant metabolites in a metabolism pathway to the total number of metabolites annotated to that pathway. Dot size represents the count of significantly enriched metabolites in each pathway. (D) Multiple volcanic diagram of differential metabolite among groups displaying the significant altered metabolites identified in feces, highlighting LPC (16:0). log 2 FC > 1. p < 0.05. (E) Violin plots showing the relative abundance of LPC (16:0) in feces across four groups (HC, BGC, I/II EM, and III/IV EM). (F) The concentration of LPC (16:0) in serum across four groups (HC, BGC, I/II EM, and III/IV EM). Targeted mass spectrometry was employed to quantify the levels of LPC (16:0). (G) Correlation analysis between LPC (16:0) levels in feces/serum and characteristic clinical indicators of EM. (H and I) Correlation analysis between the relative abundance of O. profusa in feces and the levels of LPC (16:0) in feces (H) and serum (I). (J) The concentration of LPC (16:0) in mouse serum across different groups. Targeted mass spectrometry was employed to quantify the levels of LPC (16:0) ( n = 5/group). (K) KEGG enrichment analysis of O. profusa metabolites. (L) Genomic annotation of O. profusa identified a patatin-like phospholipase family protein, designated Op-PLA2. (M) The expression levels of LPC (16:0) in bacterial culture supernatant. The Op-PLA2 gene was cloned and expressed in E. coli , and E. coli -vector, E. coli -PLA2, and O. profusa were co-incubated with PC respectively. The concentrations of LPC (16:0) in culture supernatant were quantified by targeted mass spectrometry ( n = 5/group). (N) The representative images of EM lesions along with the average weight and diameter measurements of EM lesions in mice ( n = 10/group). (O) Serum LPC (16:0) concentrations in the three groups of mice (NC, E. coli -vector, and E. coli -PLA2) measured by mass spectrometry ( n = 10/group). Data are presented as mean ± standard deviation (SD), analyzed by one-way ANOVA with Fisher’s LSD post hoc test (E) or Kruskal-Wallis test followed by Dunn’s post hoc test (F, J, and M‑O) or Spearman’s correlation (G‑I). ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001. I/II EM, stage I/II of endometriosis; III/IV EM, stage III/IV of endometriosis; ROMA_post, risk of ovarian malignancy algorithm-postmenopausal; ROMA_pre, risk of ovarian malignancy algorithm-premenopausal; AFP, alpha-fetoprotein; HE4, human epididymis protein 4; O. profusa, Olsenella profusa; PC, phosphatidylcholine; E. coli -vector, E. coli BL21 carrying the empty pET-21b(+) vector (negative control); E. coli -PLA2, E. coli BL21 heterologously expressing Op-PLA2 gene. Intriguingly, we found that the levels of LPC (16:0) in both feces and serum were significantly positively correlated with the relative abundance of O. profusa in feces in our cohort ( Figures 3 H and 3I). To verify the potential causality of this association, targeted mass spectrometry was performed on the serum of mice from the aforementioned O. profusa gavage experiments. The results showed that serum LPC (16:0) levels in mice of O. profusa group were significantly increased compared with the control group ( Figure 3 J). To further explore the direct contribution of O. profusa to LPC (16:0) production, we performed untargeted metabolomic analysis of O. profusa culture supernatant ( Figure S4 D). Surprisingly, we found that the differential metabolites of O. profusa culture supernatant were also significantly enriched in the glycerophospholipid metabolism pathway ( Figure 3 K). Then, the culture supernatant of O. profusa and PC co-incubation was detected by targeted mass spectrometry. The results showed that the content of LPC (16:0) in O. profusa supernatant was significantly increased ( Figure 3 M), indicating the direct metabolic production of LPC (16:0) from PC by O. profusa . To elucidate the enzymatic mechanism underlying this conversion, we annotated the genome of O. profusa and identified a gene encoding a patatin-like phospholipase family protein (designated Op-PLA2; GenBank: GCF_030811115.1), which belongs to the family of phospholipases that hydrolyze PC to LPC ( Figure 3 L). The Op-PLA2 gene was cloned into the pET-21b(+) vector and heterologously expressed in E. coli BL21 (hereinafter, E. coli -PLA2). When incubated with PC, E. coli -PLA2 but not control E. coli- vector efficiently converted PC into LPC (16:0) ( Figure 3 M). Next, we administered E. coli -PLA2 or control E. coli -vector to pseudo-sterile mice in an EM mouse model. Colonization with E. coli -PLA2 promoted EM lesion growth, as evidenced by increased lesion weight, diameter, and number as well as aggravated lesion pathology ( Figure 3 N; Figures S5 A and S5B). Moreover, mice colonized with E. coli -PLA2 exhibited significantly elevated LPC (16:0) levels in serum compared with controls ( Figure 3 O). These results demonstrate that O. profusa produces LPC (16:0) through its PLA2, which is sufficient to drive EM progression. To assess whether the increase in gut LPC (16:0) may be also derived from host cells, we detected the expressions of its key metabolic enzyme PLA2 in intestinal tissue of aforementioned O. profusa gavage mice. Interestingly, we found that O. profusa did not stimulate PLA2 expression in colon, but rather inhibited its expression ( Figures S6 A–S6C). In summary, we revealed that O. profusa drives EM progression by producing LPC (16:0) via its phospholipase PLA2. To investigate the potential role of gut-derived LPC (16:0) in the pathogenesis of EM, we established an EM mice model to observe the direct impact of LPC (16:0) gavage on the progression of endometriotic lesions ( Figure 4 A). As expected, compared with control group, the average weight, average diameter, and numbers of endometriotic lesions were significantly increased after LPC (16:0) treatment ( Figures 4 B and 4C; Figure S7 ). Histologically, the glandular epithelium was thickened and showed a transition from monolayer to stratified epithelium after LPC (16:0) gavage, with more proliferating ESCs (Vimentin + Ki-67 + ) in the LPC (16:0) group ( Figure 4 D). In addition, LPC (16:0) gavage significantly upregulated the expressions of inflammatory factors, including Il1b , Il18 , and Tnf , in EM lesions ( Figure 4 E). These findings indicate that gut-derived LPC (16:0) could directly promote the progression of EM. Figure 4 LPC (16:0) facilitates the progression of endometriotic lesions through stimulating endometrial stromal cell proliferation and migration (A) The schematic diagram showing experimental procedures for assessing the pathogenic effect of LPC (16:0) on EM progression in mice ( n = 6/group). (B) Representative images of EM lesions in mice. (C) Representative images of EM lesions ( n = 6/group) along with the average weight and diameter measurements of EM lesions in mice ( n = 16/group). (D) Representative images of H&E and Ki67 staining of EM lesions in mice. Vimentin is used to label ESCs. Scale bars: 200 μm for low-magnification H&E images, 50 μm for high-magnification H&E images, and 100 μm for fluorescence images ( n = 3/group). (E) The expressions of inflammatory factors in EM lesions by qPCR analysis ( n = 6/group). (F) CCK-8 cell viability assay showing the time- and concentration-dependent effect of LPC (16:0) on ESCs proliferation ( n = 5/group). G) Transwell assay determining the time- and concentration-dependent effect of LPC (16:0) on ESCs migration activity ( n = 3/group). Scale bars, 100 μm. (H) Wound healing assay demonstrating the time- and concentration-dependent effect of LPC (16:0) on ESCs migration activity ( n = 3/group). Scale bars, 100 μm. Data are presented as mean ± standard deviation (SD), analyzed by unpaired t test (D; E- Il1b , Il18 , Tnf , and Cxcl15 ), Mann-Whitney test (C; E- Il6 ), and two-way ANOVA followed by Fisher’s LSD post hoc test (F–H). ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001. E2, estradiol; IP, intraperitoneal; ig, intragastric; O. profusa, Olsenella profusa ; ABX, antibiotics combination. LPC (16:0) facilitates the progression of endometriotic lesions through stimulating endometrial stromal cell proliferation and migration (A) The schematic diagram showing experimental procedures for assessing the pathogenic effect of LPC (16:0) on EM progression in mice ( n = 6/group). (B) Representative images of EM lesions in mice. (C) Representative images of EM lesions ( n = 6/group) along with the average weight and diameter measurements of EM lesions in mice ( n = 16/group). (D) Representative images of H&E and Ki67 staining of EM lesions in mice. Vimentin is used to label ESCs. Scale bars: 200 μm for low-magnification H&E images, 50 μm for high-magnification H&E images, and 100 μm for fluorescence images ( n = 3/group). (E) The expressions of inflammatory factors in EM lesions by qPCR analysis ( n = 6/group). (F) CCK-8 cell viability assay showing the time- and concentration-dependent effect of LPC (16:0) on ESCs proliferation ( n = 5/group). G) Transwell assay determining the time- and concentration-dependent effect of LPC (16:0) on ESCs migration activity ( n = 3/group). Scale bars, 100 μm. (H) Wound healing assay demonstrating the time- and concentration-dependent effect of LPC (16:0) on ESCs migration activity ( n = 3/group). Scale bars, 100 μm. Data are presented as mean ± standard deviation (SD), analyzed by unpaired t test (D; E- Il1b , Il18 , Tnf , and Cxcl15 ), Mann-Whitney test (C; E- Il6 ), and two-way ANOVA followed by Fisher’s LSD post hoc test (F–H). ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001. E2, estradiol; IP, intraperitoneal; ig, intragastric; O. profusa, Olsenella profusa ; ABX, antibiotics combination. Next, we investigated whether LPC (16:0) could promote the proliferation and migration of ESCs in vitro . Firstly, primary ESCs were isolated from endometrial tissue of healthy women and identified through vimentin immunofluorescence staining ( Figures S8 A and S8B). Subsequently, CCK8 cell proliferation assay demonstrated that LPC (16:0) treatment could enhance ESCs viability in a concentration-dependent and time-dependent manner ( Figure 4 F). Then, both transwell assay and wound healing assay showed that LPC (16:0) could significantly promote the migration of ESCs ( Figures 4 G and 4H). These data suggest that LPC (16:0) promotes the progression of endometriotic lesions through enhancing the proliferative and migratory capacity of ESCs. To elucidate the molecular mechanism underlying the effect of LPC (16:0) on ESCs proliferation and migration, we performed transcriptome profiling of ESCs after LPC (16:0) treatment ( Figures S9 A–S9C). The RNA-seq analysis identified a total of 41 differentially expressed genes (DEGs), comprising 26 upregulated and 15 downregulated transcripts ( Figure 5 A). Among them, SPP1 showed the most significant upregulation ( Figure 5 A). SPP1 is a multifunctional secreted phosphorylated glycoprotein also known as osteopontin (OPN), named for its initial discovery in the bone matrix. 27 LPC (16:0)-induced expressions of SPP1 in ESCs were further verified by qPCR ( Figure 5 B) and immunofluorescence ( Figure 5 C). To confirm whether LPC (16:0) promotes ESCs proliferation and migration via SPP1, the knockdown of SPP1 was performed in ESCs through SPP1 siRNA. Three candidate siRNA targeting SPP1 were constructed and validated by qPCR and western blot ( Figures S10 A and S10B), and siRNA3 was finally selected for subsequent intervention experiments. The immunofluorescence results further showed that SPP1 siRNA could significantly interfere with the expression of SPP1 in ESCs, even upon LPC (16:0) stimulation ( Figure 5 C). The CCK8 assay indicated that SPP1 siRNA could inhibit ESCs proliferation and also suppress the proliferation ability of ESCs induced by LPC (16:0) ( Figure 5 D). In wound healing assay and transwell assay, the migration ability of ESCs was also significantly impaired after SPP1 knockdown, and SPP1 siRNA could also suppress the ESCs migration stimulated by LPC (16:0) ( Figures 5 E and 5F). Moreover, SPP1 siRNA could significantly suppress the expression of ESC marker vimentin, while also inhibit its expression induced by LPC (16:0) ( Figure 5 G). It is reported that the PI3K-AKT signaling axis is a classic pathway involved in the progression of EM, and SPP1 can promote the proliferative and migratory capacity of ESCs by modulating the PI3K-AKT pathway. 28 , 29 , 30 , 31 Here, we confirmed that LPC (16:0) treatment could activate PI3K-AKT pathway in ESCs, as shown by enhanced phosphorylation of both PI3K and AKT, while SPP1 knockdown could suppress the activation of PI3K-AKT pathway induced by LPC (16:0) ( Figure 5 H). These results suggest that LPC (16:0) induces the proliferation and migration of ESCs via SPP1-PI3K-AKT pathway. Figure 5 LPC (16:0) induces the proliferation and migration of endometrial stromal cells via SPP1-PI3K-AKT pathway (A) RNA sequencing demonstrating 41 differentially expressed genes (DEGs) in ESCs after LPC (16:0) treatment, comprising 26 upregulated and 15 downregulated transcripts. log 2 FC > 1. p < 0.05. (B) The validation of LPC (16:0) inducing SPP1 expression in ESCs by qPCR ( n = 3/group). (C) The immunofluorescence analysis of SPP1 expression in ESCs upon SPP1 knockdown and LPC (16:0) treatment. Scale bars: 100 μm for low-magnification images and 10 μm for high-magnification images. (D) CCK-8 cell viability assay evaluating ESCs proliferative capacity upon SPP1 knockdown and LPC (16:0) treatment. ( n = 5/group). (E) Wound healing assay showing the effect of SPP1 knockdown and LPC (16:0) treatment on ESCs migratory capacity ( n = 3/group). Scale bars, 100 μm. (F) Transwell assay demonstrating the effect of SPP1 knockdown and LPC (16:0) treatment on ESCs migration ( n = 3/group). Scale bars, 100 μm. (G) Western blot confirming the effect of SPP1 knockdown and LPC (16:0) treatment on Vimentin expression in ESCs ( n = 3/group). (H) Western blot confirming the effect of SPP1 knockdown and LPC (16:0) treatment on PI3K-AKT pathway in ESCs ( n = 3/group). (I) The schematic diagram showing the experimental pipeline for generating SPP1-knockdown donor mice via intrauterine AAV injection and subsequent EM model establishment in recipient mice. (J) The effect of SPP1 knockdown and LPC (16:0) treatment on the average weight and diameter of EM lesions in EM mice ( n = 8/group). Data are presented as mean ± standard deviation (SD), analyzed by one-way ANOVA with Fisher’s LSD post hoc test (B and F‑H), Kruskal-Wallis test followed by Dunn’s post hoc test (J), and two-way ANOVA with Fisher’s LSD post hoc test (D and E). ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001. ESCs, endometrial stromal cells; si-NC, negative control group; si-SPP1, SPP1-knockdown group by siRNA; AAV-NC, adeno-associated virus negative control group; AAV-SPP1, SPP1-knockdown group by adeno-associated virus; ig, intragastric. LPC (16:0) induces the proliferation and migration of endometrial stromal cells via SPP1-PI3K-AKT pathway (A) RNA sequencing demonstrating 41 differentially expressed genes (DEGs) in ESCs after LPC (16:0) treatment, comprising 26 upregulated and 15 downregulated transcripts. log 2 FC > 1. p < 0.05. (B) The validation of LPC (16:0) inducing SPP1 expression in ESCs by qPCR ( n = 3/group). (C) The immunofluorescence analysis of SPP1 expression in ESCs upon SPP1 knockdown and LPC (16:0) treatment. Scale bars: 100 μm for low-magnification images and 10 μm for high-magnification images. (D) CCK-8 cell viability assay evaluating ESCs proliferative capacity upon SPP1 knockdown and LPC (16:0) treatment. ( n = 5/group). (E) Wound healing assay showing the effect of SPP1 knockdown and LPC (16:0) treatment on ESCs migratory capacity ( n = 3/group). Scale bars, 100 μm. (F) Transwell assay demonstrating the effect of SPP1 knockdown and LPC (16:0) treatment on ESCs migration ( n = 3/group). Scale bars, 100 μm. (G) Western blot confirming the effect of SPP1 knockdown and LPC (16:0) treatment on Vimentin expression in ESCs ( n = 3/group). (H) Western blot confirming the effect of SPP1 knockdown and LPC (16:0) treatment on PI3K-AKT pathway in ESCs ( n = 3/group). (I) The schematic diagram showing the experimental pipeline for generating SPP1-knockdown donor mice via intrauterine AAV injection and subsequent EM model establishment in recipient mice. (J) The effect of SPP1 knockdown and LPC (16:0) treatment on the average weight and diameter of EM lesions in EM mice ( n = 8/group). Data are presented as mean ± standard deviation (SD), analyzed by one-way ANOVA with Fisher’s LSD post hoc test (B and F‑H), Kruskal-Wallis test followed by Dunn’s post hoc test (J), and two-way ANOVA with Fisher’s LSD post hoc test (D and E). ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001. ESCs, endometrial stromal cells; si-NC, negative control group; si-SPP1, SPP1-knockdown group by siRNA; AAV-NC, adeno-associated virus negative control group; AAV-SPP1, SPP1-knockdown group by adeno-associated virus; ig, intragastric. To further verify whether LPC (16:0) promotes the progression of endometriotic lesions in vivo through SPP1, we employed AAV-mediated SPP1-knockdown mice model ( Figure 5 I; Figures S11 A and S11B; STAR Methods ). As expected, LPC (16:0) gavage increased the average weight, average diameter, and numbers of endometriotic lesions, while SPP1 knockdown by adeno-associated virus (AAV) could significantly suppress the progression of endometriotic lesions induced by LPC (16:0), evidenced by decreased average weight, average diameter, and numbers of endometriotic lesions ( Figure 5 J; Figure S11 C). Histologically, LPC (16:0) gavage thickened the glandular epithelium of endometriotic lesions and promoted the proliferation of ESCs (Vimentin + Ki-67 + ), while AAV-mediated SPP1 knockdown could significantly suppress the progression of endometriotic lesions as well as the proliferation of ESCs induced by LPC (16:0) ( Figure S11 D). These findings support that LPC (16:0) promotes the progression of endometriotic lesions through SPP1. To verify the clinical relevance of O. profusa -LPC (16:0)-SPP1 axis to EM, we detected the expressions of SPP1 pathway in clinical samples of EM patients. The results showed that the expression of SPP1 was significantly higher in EM endometrial tissue compared with non-endometriosis (NEM) endometrial tissue ( Figure 6 A). Meanwhile, PI3K-AKT pathway was activated in endometrial tissue of EM patient, as shown by enhanced phosphorylation of both PI3K and AKT ( Figure 6 B). Furthermore, we confirmed that the levels of SPP1 in serum were significantly higher in EM patients with high abundance of O. profusa and LPC (16:0) compared with EM patients with low abundance of O. profusa and LPC (16:0) ( Figures 6 C and 6D). These results demonstrate the clinical relevance of O. profusa -LPC (16:0)-SPP1 axis to EM. Figure 6 O. profusa and LPC (16:0) serve as potential diagnostic biomarkers for EM (A) The expressions of SPP1 in the endometrium of NEM controls and EM patients ( n = 3/group). Scale bars: 100 μm for low-magnification images and 20 μm for high-magnification images. (B) The expressions of PI3K-AKT pathway in the endometrium of NEM controls and EM patients ( n = 3/group). (C and D) ELISA analysis of serum SPP1 levels in EM patients stratified by O. profusa abundance in feces (C) and LPC (16:0) abundance in serum (D) ( O. profusa -high, n = 35; O. profusa -low, n = 32; LPC (16:0)-high, n = 34; LPC (16:0)-low, n = 33). (E) Odds ratio analysis of different variables for EM risk. (F) ROC curve analysis for the diagnostic performance of O. profusa and LPC (16:0) in EM compared to CA125. (G) ROC curve analysis for the diagnostic performance of O. profusa and LPC (16:0) in the stage I/II of endometriosis (I/II EM). (H) ROC curve analysis for the diagnostic performance of O. profusa and LPC (16:0) in the stage III/IV of endometriosis (III/IV EM). Data are presented as mean ± standard deviation (SD), analyzed by unpaired t test (A‑D). ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001. NEM, non-endometriosis; EM, endometriosis; BMI, body mass index; HE4, human epididymis protein 4; ROMA_pre, risk of ovarian malignancy algorithm-premenopausal; ROMA_post, risk of ovarian malignancy algorithm-postmenopausal; O. profusa, Olsenella profusa ; I/II EM, stage I/II of endometriosis; III/IV EM, stage III/IV of endometriosis. O. profusa and LPC (16:0) serve as potential diagnostic biomarkers for EM (A) The expressions of SPP1 in the endometrium of NEM controls and EM patients ( n = 3/group). Scale bars: 100 μm for low-magnification images and 20 μm for high-magnification images. (B) The expressions of PI3K-AKT pathway in the endometrium of NEM controls and EM patients ( n = 3/group). (C and D) ELISA analysis of serum SPP1 levels in EM patients stratified by O. profusa abundance in feces (C) and LPC (16:0) abundance in serum (D) ( O. profusa -high, n = 35; O. profusa -low, n = 32; LPC (16:0)-high, n = 34; LPC (16:0)-low, n = 33). (E) Odds ratio analysis of different variables for EM risk. (F) ROC curve analysis for the diagnostic performance of O. profusa and LPC (16:0) in EM compared to CA125. (G) ROC curve analysis for the diagnostic performance of O. profusa and LPC (16:0) in the stage I/II of endometriosis (I/II EM). (H) ROC curve analysis for the diagnostic performance of O. profusa and LPC (16:0) in the stage III/IV of endometriosis (III/IV EM). Data are presented as mean ± standard deviation (SD), analyzed by unpaired t test (A‑D). ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001. NEM, non-endometriosis; EM, endometriosis; BMI, body mass index; HE4, human epididymis protein 4; ROMA_pre, risk of ovarian malignancy algorithm-premenopausal; ROMA_post, risk of ovarian malignancy algorithm-postmenopausal; O. profusa, Olsenella profusa ; I/II EM, stage I/II of endometriosis; III/IV EM, stage III/IV of endometriosis. Next, we investigated whether O. profusa and LPC (16:0) can serve as risk indicators and potential diagnostic biomarkers for EM. The multivariate logistic regression revealed that fecal O. profusa , fecal LPC (16:0) and serum LPC (16:0) were risk factors associated with EM, with odds ratio (OR) 2.58 (95% CI: 1.35-4.90, p = 0.004), 1.01 (95% CI: 1.01-1.01, p = 0.002), and 2.17 (95% CI: 1.62-2.92, p < 0.001) respectively ( Figure 6 E). Then, the diagnostic efficacy of O. profusa and LPC (16:0) as potential biomarker was evaluated through receiver operating characteristic (ROC) analysis. The results showed that the AUC of fecal O. profusa , fecal LPC (16:0) and serum LPC (16:0) were 0.9459, 0.7747 and 0.8927 respectively, while the combined AUC of these three indices was 0.9799, outperforming the traditional non-specific biomarker CA125 ( Figure 6 F). Next, we also assessed the diagnostic efficacy of these potential biomarkers in different stages of EM, which consistently showed robust diagnostic performance in both I/II EM ( Figure 6 G) and III/IV EM ( Figure 6 H). These findings demonstrate a causal link between O. profusa -LPC (16:0)-SPP1 axis and EM in clinical setting and suggest that fecal O. profusa , fecal LPC (16:0), and serum LPC (16:0) could serve as valuable diagnostic biomarkers for EM.

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dongxin Zhang ( [email protected] ). All requests for resources and reagents should be directed to and will be fulfilled by the lead contact ( [email protected] ). All materials and reagents may be available on request after completion of a materials transfer agreement. • 16S rRNA gene sequencing data (BioProject: PRJCA047221; GSA: CRA032092), transcriptomic RNA-seq data (BioProject: PRJCA067200; GSA-Human: HRA019199), and metabolomics data (BioProject: PRJCA067082; OMIX: OMIX018303) have been deposited at the National Genomics Data Center (NGDC) database and are publicly accessible as of the date of publication. Accession numbers are listed in the key resources table . • This paper does not report original code. • Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. 16S rRNA gene sequencing data (BioProject: PRJCA047221; GSA: CRA032092), transcriptomic RNA-seq data (BioProject: PRJCA067200; GSA-Human: HRA019199), and metabolomics data (BioProject: PRJCA067082; OMIX: OMIX018303) have been deposited at the National Genomics Data Center (NGDC) database and are publicly accessible as of the date of publication. Accession numbers are listed in the key resources table . This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Our study delineates a complete gut microbe-metabolite-host axis that drives the pathogenesis of EM, offering a perspective for understanding EM. We establish that the gut commensal O. profusa is a potent driver of EM progression, acting through the metabolite LPC (16:0) activating SPP1-PI3K-AKT signaling cascade in ESCs. This work not only identifies O. profusa and LPC (16:0) as previously unrecognized risk factors and potential diagnostic biomarkers but also positions the gut commensal bacteria as a key player in EM pathophysiology, reframing our understanding of this disease to incorporate the host-gut microbiota interaction. The causal influence of gut microbiota on EM was initially suggested by our FMT experiments, wherein gut microbiota from EM patients sufficed to exacerbate EM progression in recipient mice. To pinpoint specific microbial drivers, we employed a clinical cohort with strictly designed groups, which revealed a selective enrichment of O. profusa in EM patients. This clinical observation is further strongly supported by Mendelian randomization studies implicating the Olsenella genus in EM risk. 32 , 33 , 34 , 35 The definitive pathogenic role of O. profusa was demonstrated by our single-bacterium gavage experiment in a pseudo-sterile mouse model, where it potently accelerated EM lesion growth. This finding is notable, as O. profusa has been previously associated with inflammatory pathologies at other body sites, particularly oral infection and intestinal inflammation. 36 , 37 , 38 Concordantly, we found that O. profusa colonization heightened the expression of pro-inflammatory cytokines within EM lesions, thereby linking its presence to the established inflammatory milieu of the disease. Gut bacterial effects are often mediated by metabolites. In fact, emerging evidence suggest the involvement of gut metabolites in the impact of gut microbiota on EM progression. 19 , 20 Our metabolomic profiling of patient cohort revealed a significant enrichment of glycerophospholipid metabolism in EM, redirecting our focus to this metabolism pathway. Consistently, previous studies have showed that glycerophospholipids are potential biomarkers for EM, 39 and glycerophospholipid metabolites are enriched in serum, 40 peritoneal fluid, 41 follicular fluid, 42 , 43 and endometrial tissue 44 of EM patients. Intriguingly, among the altered glycerophospholipid metabolites identified in our cohort, LPC (16:0) emerged as a key candidate, consistently elevated in both feces and serum of EM patients—a finding that aligns with prior identification of LPC (16:0) as a potential biomarker in peritoneal fluid and plasma of EM patients. 45 To test the direct contribution of O. profusa to LPC (16:0) production, we annotated its genome and identified a patatin-like phospholipase family protein (Op-PLA2). Heterologous expression of Op-PLA2 in E. coli enabled conversion of PC to LPC (16:0) in vitro . Moreover, mice colonized with E. coli -PLA2 recapitulated the elevated serum LPC (16:0) levels and the pro-endometriotic effects observed with O. profusa colonization. These findings establish that O. profusa directly generates LPC (16:0) via its PLA2 enzyme, defining a gut O. profusa -LPC (16:0) metabolic axis in EM. Notably, the role of bacterial phospholipase in generating bioactive LPC is not limited to EM. A recent study in Alzheimer’s disease reported that Bacteroides ovatus also produces LPC via its PLA2 and that this metabolite attenuates β-amyloid pathology through GPR119-mediated suppression of ferroptosis. 46 These findings suggest that bacterial phospholipase-LPC axis may represent a common mechanism linking gut microbiota to distinct disease states. The subsequent elucidation of downstream mechanism revealed SPP1 as the crucial molecular target of LPC (16:0). We demonstrated that LPC (16:0) robustly upregulates SPP1 expression in ESCs, thereby enhancing their proliferative and migratory capacities. Our findings resonate with and significantly extend previous reports: SPP1 is known to be elevated in EM patients and has been proposed as a superior diagnostic marker to CA125. 47 , 48 , 49 Furthermore, prior work has linked SPP1 to PI3K pathway activation in ESCs migration. 31 In the study, we provide definitive functional evidence for its indispensable role in this axis. The knockdown of SPP1, either in vitro or in vivo , completely abrogated the pro-proliferative and pro-migratory effects of LPC (16:0) and halted EM progression, establishing SPP1 as a non-redundant molecular switch in this pathway. In summary, the study highlights gut O. profusa as a key pathogenic bacterium contributing to EM, and emphasizes the important role of O. profusa -LPC (16:0) metabolic axis in the pathogenesis of EM. We unraveled a previously unknown pathogenic circuit originating from the gut, wherein O. profusa promotes EM through LPC (16:0) activating SPP1-PI3K-AKT pathway in ESCs. This discovery not only provides a new perspective for understanding EM but also lays a solid theoretical foundation for the development of novel therapies based on gut microbiota. From a translational perspective, targeting O. profusa -LPC (16:0)-SPP1 axis holds immense promise. Our findings contribute to develop potential therapeutic strategies in preventing EM by targeting the axis, and also provide potential biomarkers for early non-invasive diagnosis of this intractable gynecological condition. This study has some limitations. First, while most O. profusa and LPC (16:0)-enriched women presented with EM, some did not develop the condition, indicating incomplete penetrance and EM heterogeneity. We propose that moderator variables, such as host susceptibility, co-microbial communities, maternal immune tone and metabolic status, and timing change of O. profusa and LPC (16:0) relative to onset, may gate whether O. profusa -LPC (16:0) metabolic axis precipitates EM, warranting further studies. Additionally, due to the lack of detailed clinical symptom data, we were unable to assess associations between the identified microbial-metabolic axis and patient-reported symptoms such as pain severity (e.g., dysmenorrhea and dyspareunia) or fertility status; future studies incorporating standardized symptom questionnaires and fertility follow-up are needed to determine whether these microbial markers correlate with clinical presentation and prognosis. Second, sample collection was limited to a single geographical region and failed to cover populations with different dietary habits, climate environment, and geographical backgrounds, which needs to incorporate multi-center designs. This is particularly relevant given that dietary patterns can shape gut microbiota composition and metabolite profiles and that regional differences may influence disease subtype presentation or use of hormonal treatments. Future studies should incorporate multi-center designs across diverse populations to validate our findings and assess the impact of these variables. Third, sampling was performed only at EM onset, preventing continuous tracking of O. profusa and LPC (16:0) abundance throughout follow-up. Future studies should monitor these changes to gain further insights. Fourth, while our mouse model successfully demonstrated the pathogenic role of the O. profusa -LPC (16:0) axis, we recognize that this model, which relies on transplantation of whole uterine tissue, typically generates fluid-filled, cystic lesions that do not fully capture the histopathological complexity of SPE and DIE. 50 Future studies employing more advanced models, such as those utilizing menstrual endometrium or humanized immune systems, may better approximate the full spectrum of human disease pathology. Despite these limitations, our study provided evidence for the contribution of gut O. profusa -LPC (16:0) metabolic axis to EM.

Endometriosis (EM), a prevalent chronic inflammatory gynecological condition manifesting as pelvic pain, menstrual irregularities, and infertility, affects around 10% of women of reproductive age globally. 1 Histologically, it involves the ectopic growth of endometrial glands and stroma beyond the uterine cavity, predominantly in the pelvic and abdominal regions, including the ovary, bladder, abdominal wall, and gastrointestinal tract. 2 The EM lesion forms a complex microenvironment composed of glandular epithelium, stroma, vascular compartment (including endothelial cells and perivascular cells), diverse immune cell infiltrates, and sensory nerve fibers. 3 These components interact dynamically to drive key pathophysiological processes, including chronic inflammation, tissue fibrosis, and neuroangiogenesis, which collectively promote lesion survival, disease progression, and pain symptoms. 4 , 5 , 6 Clinically, EM presents as three pelvic subtypes based on lesion location and infiltration depth: superficial peritoneal endometriosis (SPE), ovarian endometrioma (OMA), and deep infiltrating endometriosis (DIE). 1 For surgical staging, the revised American Society for Reproductive Medicine (rASRM) classification remains the most widely used system, categorizing the disease as stage I/II (minimal to mild) and stage III/IV (moderate to severe), 7 which provides a standardized framework for assessing disease burden. Despite extensive research, the etiology of EM remains elusive, with hormones, 8 inflammation, 4 nerves, 6 and immunity 9 implicated in its pathogenesis. Early diagnostic challenges persist due to the reliance on invasive procedures like laparoscopy, leading to diagnostic delays averaging 7–10 years. 10 , 11 Current treatments are limited in efficacy, often associated with hormonal side effects, surgical complications, and high recurrence rates, seriously impacting patients’ health and life quality. 12 , 13 Therefore, it is urgent to explore the key pathogenesis of EM, identify effective non-invasive biomarkers for early diagnosis, as well as develop promising targeted therapies. Emerging evidence from our team and others has shown an association between gut microbiota dysbiosis and EM, suggesting a pathogenic role of the gut microbiota in EM. 14 , 15 , 16 , 17 , 18 Yu et al. observed reduced alpha diversity and an elevated ratio of Firmicutes to Bacteroidetes in the gut of patients with stage III/IV EM compared with healthy controls, 14 and significant differences in the abundance of several taxa (e.g., Actinomyces , Mollicutes , Blautia , Bifidobacterium , Dorea , and Streptococcus ) were noted between the two groups. 14 Moreover, Urman et al. identified the predominance of Shigella / Escherichia in the gut of individuals with stage III/IV EM. 15 In addition, the animal study revealed alterations in the gut microbiota of EM mice. 16 Intriguingly, depleting intestinal microbiota with antibiotics could alleviate EM progression in mice. 17 In 2021, we published data from a cross-sectional cohort study and found that EM was associated with gut dysbiosis (depletion of short-chain fatty acid [SCFA]-producing Lachnospiraceae/ Ruminococcus ), as well as enrichment of potential pathogens (e.g., Pseudomonas ) in peritoneal fluid, while cervical microbiota showed minimal alterations. A random forest classifier based on gut microbial profiles robustly discriminated EM cases from controls (area under the curve [AUC] = 0.840), significantly outperforming cervical microbiota-based models (AUC = 0.672), highlighting the gut microbiome as a superior non-invasive diagnostic biomarker. 18 Together, these findings underscore the important roles of gut microbiota in the pathogenesis of EM. Gut microbiota-derived metabolites are required for intricate host-gut bacteria interactions, which profoundly influence host physiological regulation and disease development. 19 , 20 , 21 In a previous study, quinic acid, a metabolite of enterobacteria, had been confirmed to promote the survival of EM epithelial cells in vitro and lesion growth in vivo , 19 whereas Talwar et al. discovered that 4-hydroxyindole, a metabolite of gut microbiota, could prevent the formation and progression of EM lesions and alleviate associated pain. 20 Moreover, Chadchan et al. observed lower levels of SCFAs, such as n-butyrate, in the feces of mice with EM compared with control mice, and demonstrated that n-butyrate treatment reduced EM lesion growth. 21 Collectively, these investigations highlight a robust link between gut-derived metabolites and EM. However, the composition and metabolic output of the gut microbiota is profoundly influenced by host factors, particularly dietary habits. 22 , 23 Variations in dietary patterns may partly explain the discrepancies observed across published microbiome studies in EM, as diet can directly modulate the abundance of metabolite-producing taxa. 24 Furthermore, emerging evidence suggests that interventions using prebiotics or probiotics could potentially alleviate EM symptoms or restore gut dysbiosis, although data from robust clinical trials remain limited. 25 , 26 In the present study, while we did not collect detailed dietary records, we minimized the confounding effects of these factors at the enrollment stage by excluding participants who had used antibiotics, prebiotics, or probiotics within the six months prior to sample collection. This strict exclusion criterion helps reduce the potential impact of these modulators on the gut microbiome, allowing for a more focused investigation of EM-associated microbial and metabolic signatures. Although increasing studies have implicated gut microbiota in EM pathogenesis, it remains unclear whether specific gut bacteria directly influence EM development. To investigate the exact causal mechanism of specific gut bacteria and their metabolites in EM, we screened for characteristic gut bacteria and metabolites in EM patients through 16S rRNA gene sequencing and metabolomics, respectively. Here, we discovered that Olsenella profusa (O. profusa) and lysophosphatidylcholine (LPC (16:0)) were specifically enriched in EM patients, with a significant positive correlation between them. O. profusa promotes EM progression through phospholipase-mediated production of LPC (16:0) that induces endometrial stromal cells (ESCs) proliferation and migration via the secreted phosphoprotein 1 (SPP1) pathway. These findings emphasize the important role of the gut O. profusa -LPC (16:0) metabolic axis in EM, providing promising diagnostic and therapeutic targets for the disease.

The authors D.Z., Y.M., H.Z., Y. Chen, Z.Z., Y. Cao, and L.P. have a patent application related to this work.

REAGENT or RESOURCE SOURCE IDENTIFIER Antibodies β-Actin Antibody Abmart Cat# T40104 ; RRID: AB_2936320 Osteopontin Recombinant Rabbit Monoclonal Antibody [PSH09-24] HUABIO Cat# HA723082 Osteopontin Recombinant Rabbit Monoclonal Antibody [PSH09-23] HUABIO Cat# HA723081; RRID: AB_3674154 Goat Anti-Rabbit Mouse IgG-HRP Abmart Cat# M21003S Anti-Ki67 antibody Abcam Cat# ab16667; RRID: AB_302459 Vimentin Recombinant Rabbit Monoclonal Antibody [SC60-05] HUABIO Cat# ET1610-39; RRID: AB_3069923 Ki67 Recombinant Mouse Monoclonal Antibody [PD00-10] HUABIO Cat# HA601053 ; RRID: AB_3071776 Phospho-Akt1 (Ser473) (D7F10) XP Rabbit mAb (Akt1 Specific) Cell Signaling Technology Cat# 9018; RRID: AB_2629283 Akt1 (C73H10) Rabbit mAb Cell Signaling Technology Cat# 2938; RRID: AB_915788 PI3 Kinase p85 Antibody Cell Signaling Technology Cat# 4292; RRID: AB_329869 Phospho-PI3K p85 (Y467) + PI3K p55 (Y199) Recombinant Rabbit Monoclonal Antibody [PSH01-38] HUABIO Cat# HA721672; RRID: AB_3072785 Bacterial and virus strains Escherichia coli K-12 MG1655 ATCC ATCC700926 Olsenella profusa DSMZ DSMZ13989 Escherichia coli BL21 (DE3) Guangzhou IGE Biotechnology Co., Ltd. China N/A Biological samples Human endometriosis tissues Obstetrics and Gynecology Medical Center, Zhujiang Hospital, Southern Medical University, Guangzhou, China N/A Human endometrium tissues Obstetrics and Gynecology Medical Center, Zhujiang Hospital, Southern Medical University, Guangzhou, China N/A Endometriosis patients’feces/serum Obstetrics and Gynecology Medical Center, Zhujiang Hospital, Southern Medical University, Guangzhou, China N/A Healthy controls’ feces/serum Obstetrics and Gynecology Medical Center, Zhujiang Hospital, Southern Medical University, Guangzhou, China N/A Benign gynecological diseases controls’ feces/serum Obstetrics and Gynecology Medical Center, Zhujiang Hospital, Southern Medical University, Guangzhou, China N/A Chemicals, peptides, and recombinant proteins 1- Palmitoyl- sn -glycero-3-phosphocholine (LPC (16:0)) Macklin Cat# P863742; CAS# 17364-16-8 1, 2-dipalmitoyl- sn -glycero-3-phosphocholine (PC) Macklin Cat# D863688; CAS# 63-89-8 β-Estradiol MedChemExpress (MCE) Cat# HY-B0141; CAS# 50-28-2 TRIzol Invitrogen Cat# 15596026 ExtractRNA reagent Vazyme Cat# R401-01 DMEM/F-12 culture medium Gibco Cat# C11330500BT Type I collagenase Yeasen Biotechnology Cat# 40121ES76 Cell strainers Falcon Cat# 352340 RNAfit reagent Hanheng Biotechnology Cat# HB-RF-1000 RIPA lysis buffer Beyotime Cat# P0013B SDS-PAGE Yamei Cat# PG112 0.45 μm polyvinylidene fluoride membranes Millipore Cat# IPVH00010 PMSF Beyotime Cat# ST506 Phosphatase inhibitor Sigma Cat# P5726 chemiluminescence (ECL) substrate Millipore Cat# WBKLS0100 Columbia Blood Agar Plate HKM Cat# CP0160 Modified PYG Medium BNCC Cat# BNCC 341095 LB Agar Medium HKM Cat# HKM 02-001 Agar HKM Cat# 9002-18-0 Isopropyl-b-D-thiogalactopyranoside (IPTG) MedChemExpress (MCE) Cat# HY-15921 Ampicillin (AMP) MedChemExpress(MCE) Cat# 69-53-4 Vancomycin hydrochloride Yuanye Bio-Technology Cat# S17018-25g Ampicillin Yuanye Bio-Technology Cat# S17059-5g Chloramphenicol Yuanye Bio-Technology Cat# S17079-25g Neomycin sulfate Yuanye Bio-Technology Cat# S17028-25g Goat-serum Biologic Industries Cat# AR0009 DAPI Sigma-Aldrich Cat# C0065-10ML Triton X-100 Amresco Cat# 9,002,931 Critical commercial assays Human Osteopontin (OPN) ELISA Kit J&L Biological Cat# JL10368-96T QIAamp Fast DNA Stool Mini Kit (50) QIAGEN Cat# 51604 Evo M-MLV RT Premix for qPCR kit Accurate Biology Cat# AG11706 TB Green Premix Ex Taq II (Tli RNase H Plus) Takara Bio Cat# RR820A BCA Protein Assay Kit SolelyBio Cat# CW0014S Cell Counting Kit-8 MedChemExpress (MCE) Cat# HY-K0301 Deposited data RNA transcriptome sequence of primary endometrial stromal cells This paper BioProject ID: PRJCA067200 GSA-Human ID: HRA019199 16S rRNA sequence of fecal bacteria in the cohort This paper BioProject ID: PRJCA047221 GSA ID: CRA032092 Untargeted metabolomics of feces in the cohort This paper BioProject ID: PRJCA067082 OMIX ID: OMIX018303 Experimental models: Cell lines Human primary endometrial stromal cells (Non-endometriosis) Endometrial tissues obtained from non-endometriosis patients N/A Experimental models: Organisms/strains C57BL/6J Mice Zhuhai BesTest Bio-Tech Co, .Ltd, China N/A Oligonucleotides Primers for qRT-PCR See Table S2 N/A Control siRNA (Sense: 5′-UUCUCCGAACGUGUCACGUTT-3'; Antisense: 5′-ACGUGACACGUUCGGAGAATT-3′) This paper N/A SPP1 siRNA#1 (Sense: 5′-CCUUUCCAAAGUAAGUCCAATT-3'; Antisense: 5′-UUGGACUUACUUGAAGGGTT-3′) This paper N/A SPP1 siRNA#2 (Sense: 5′-CCGUGGGAAGGACAGUUAUTT-3'; Antisense: 5′-AUAAACUGUCCUUCCCACGGTT-3′) This paper N/A SPP1 siRNA#3 (Sense: 5′-CCUGUAAUUCCACAGCCAATT-3'; Antisense: 5′-AUGGCUGUGGAAUUCCAGGGTT-3′) This paper N/A Recombinant DNA AAV9-Negative control (shRNA target sequence: TTCTCCGAACGTGTCACGT) This paper N/A AAV9-Spp1-RNAi(P25E1446) (shRNA target sequence: GGTGATAGCTTGGCTTATGGA) This paper N/A AAV9-Spp1-RNAi(P25E1447) (shRNA target sequence: TGAGTCAAGTCAGCTGGATGA) This paper N/A AAV9-Spp1-RNAi(P25E1448) (shRNA target sequence: TCGGATGTGATCGATAGTCAA) This paper N/A Software and algorithms ImageJ v2.0.0 National Institutes of Health https://imagej.nih.gov/ij/ GraphPad Prism9.0 GraphPad Software https://www.graphpad.com/ Adobe Illustrator Adobe Inc. https://www.adobe.com/products/illustrator.html/ R4.2.2 CRAN https://www.r-project.org Other Liquid chromatography-tandem mass spectrometry (LC-MS/MS) (AB SCIEX 4500) AB SCIEX Model 4500 SCIEX OS Software (v2.1.5) SCIEX Version 2.1.5 ViiA™ 7 Real-Time PCR System Applied Biosystems ViiA™ 7 The women of reproductive age (18–48 years) were recruited at Zhujiang Hospital of Southern Medical University, Guangzhou, China, from June 2021 to November 2024. The healthy control group (HC, n = 50) was composed of individuals who underwent routine physical examinations; the benign gynecological diseases control (BGC, n = 50) comprised patients undergoing surgery for other benign gynecological conditions, including ovarian cysts (e.g., simple cysts, mature teratomas), uterine fibroids, and tubal pathologies, who had been confirmed free of endometriosis through laparoscopic exploration; the EM group (EM, n = 114) consisted of those patients whose EM had been pathologically confirmed by laparoscopic surgery. Based on the staging and scoring system formulated by the American Society for Reproductive Medicine (ASRM), EM group was further subdivided into the I/II EM group (I/II EM, n = 50) and the III/IV EM group (III/IV EM, n = 64). The gathered samples included paired fecal samples and serum samples. Endometrial tissue and ectopic endometrial tissue were collected during laparoscopic surgery. Fresh endometrial tissues were immediately frozen at −80°C or fixed in 4% paraformaldehyde for paraffin embedding, and subsequently stored in Biobank. The clinical characteristics of enrolled volunteers were presented in Table S1 . All studies were approved by the Ethics Committee of Zhujiang Hospital, Southern Medical University (2021-KY-03), and registered under Clinical trial ( NCT05086484 ). All procedures relevant to the study were carried out in strict compliance with pertinent guidelines and regulations. The informed consent was obtained from all participants. C57BL/6J female mice (6–8 weeks old) were purchased from Zhuhai BesTest Bio-Tech Co., Ltd (Guangdong, China). Animals were housed individually in plastic cages with corncob fragments as bedding under pathogen-free conditions, in a controlled environment of temperature at 20°C–25°C and 12 h cycles of light and dark. Mice were fed a standard laboratory diet and water ad libitum . All animal experiments were conducted according to the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All studies involving mice were approved by the Ethical Committee on Animal Experimentation of Zhujiang Hospital, Southern Medical University (LAEC-2024-027). For the induction of EM in mice, we employed a well-established EM model, in which the uterine fragments of donor mice in estrus were intraperitoneally injected into recipient mice. 51 Briefly, all mice were injected with estrogen at the back of neck before modeling. Meanwhile, the recipient mice were gavaged with antibiotics combination (vancomycin, 100 mg/kg; neomycin sulfate, metronidazole and ampicillin, 200 mg/kg) for five consecutive days to deplete gut microbiota. On the day of modeling, the donor mice were euthanized with CO 2 . Uterus was excised and immersed in sterile PBS for cleaning to remove surface red blood cells, then minced into 1 mm 3 tissue fragments. The abdomen of the recipient mice was disinfected with iodophor, and the uterine fragments were injected into the abdominal cavity. The uterus of one donor mouse was evenly injected into two recipient mice. 21 For O. profusa -mediated EM model in mice, after modeling EM, the recipient mice were gavaged daily with 10 8 colony-forming units (CFU) of O. profusa for 21 days. PBS and E. coli MG1655 were used as blank control and bacterial negative control respectively. The EM lesions of mice were dissected for further analyses and experiments. For FMT model in EM mice, the feces collected from healthy donors ( n = 5) and EM patients ( n = 5) were pooled respectively, and then resuspended in PBS with 20% glycerol (at a ratio of 10 g/50 mL). Following filtration through a 70 μM sieve to eliminate solid components, 200 μL of the resultant fecal filtrate was gavaged daily to gut microbiota-depleted pseudo-sterile EM mice for 21 days. PBS with 20% glycerol was used as blank control. The EM lesions of mice were dissected for further analyses and experiments. For LPC (16:0)-mediated EM model in mice, after modeling EM, the recipient mice were gavaged daily with 200μL of LPC (16:0) for 21 days. PBS was used as blank control. The EM lesions of mice were dissected for further analyses and experiments. For the knockdown of SPP1 in EM mice, the SPP1-knockdown adeno-associated virus (AAV-SPP1) and negative control (AAV-NC) were constructed by GeneChem (Shanghai, China). The sequences of SPP1 knockdown were shown in Table S3 . According to the manufacturer’s instructions, 20 μL of AAV at the same titer was injected into the uterine cavity of donor mice, with the outcome of knockdown validated after three weeks. After modeling EM, the recipient mice were gavaged daily with 200μL of LPC (16:0) for 21 days. PBS was used as blank control. The EM lesions of mice were dissected for further analyses and experiments. For the heterologous expression validation of Op-PLA2 in EM mice, the EM model was induced in recipient mice as described above. After EM establishment, mice were randomly assigned to three groups: Negative control, receiving daily gavage of 200 μL sterile PBS; E. coli -Vector control, receiving daily gavage of 200 μL E. coli carrying the empty pET-21b(+) vector (1 × 10 8 CFU/mL); and E. coli -PLA2, receiving daily gavage of 200 μL E. coli expressing Op-PLA2 (1 × 10 8 CFU/mL). Bacterial suspensions were freshly prepared each day and administered via oral gavage daily for 21 consecutive days. The EM lesions of mice were dissected for further analyses and experiments. Primary endometrial stromal cells (ESCs) were isolated from female donors aged between 25 and 40 who were confirmed to have no EM through laparoscopic examination. After dissecting the endometrial tissue, it was immediately immersed in pre-cooled PBS at 4°C and transferred to the laboratory to extract ESCs. The extraction method was referenced to Masuda et al. 52 All cells of 3–6 passages were used in the study, and were cultured in DMEM/F12 medium (Gibco) supplemented with 10% FBS at 37°C in humidified atmosphere with 5% CO 2 . The identity of primary ESCs was confirmed by their spindle-shaped morphology and immunocytochemical staining (vimentin >95%) ( Figure S8 B). As primary cells, STR authentication was not performed. Mycoplasma testing was not performed because no signs of contamination were observed in the cell cultures. Olsenella profusa ( O. profusa , DSM13989) was from Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ, Braunschweig, Lower Saxony). E. coli strain MG1655 was from American Tissue Culture Collection (ATCC, Manassas, VA). O. profusa was cultured anaerobically at 37°C in PYG Medium (modified) using Whitley A35 Anaerobic Workstation (Don Whitley Scientific Limited, UK). E. coli strain MG1655 was cultured at 37°C in LB broth medium using bacteriological incubator (Thermo Fisher Scientific, USA). Fecal samples were subjected to genomic DNA extraction using the CTAB method. The purity and concentration of extracted DNA was assessed by spectrophotometer (NanoDrop One, Thermo Fisher Scientific, USA). PCR amplification, construction of high-throughput sequencing libraries, and 16S rRNA gene sequencing were performed by LC-Bio Technology Co., Ltd. (Hangzhou, China). The V3-V4 hypervariable regions were amplified with PCR using the primer pair 341F (5′-CCTACGGGNGGCWGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′), and sequenced on an Illumina NovaSeq platform. Raw sequencing data were processed using DADA2 (v1.28) to infer amplicon sequence variants (ASVs). Taxonomic assignment was performed against the SILVA v138.1 database using a naive Bayesian classifier. Species-level annotations were assigned only when ≥99% sequence identity was achieved. For data analysis, FastQC and MultiQC were employed to evaluate key sequencing metrics, including the number of raw reads, total base pairs, sequencing error rate, Q20 score, and GC content. The raw sequencing data have been deposited in the National Genomics Data Center (NGDC) database (ID: PRJCA047221). DNA was extracted from fecal samples using QIAamp Fast DNA Stool Mini Kit (QIAGEN, Germany) according to the manufacturer’s protocols. Quantitative PCR was performed to detect the abundance of O. profusa in feces. In all PCR amplifications, ultrapure water was used as negative control to show that the amplification system was free of nucleic acid contamination. The primers used were listed in Table S2 . The relative level of O. profusa was calculated relative to the 16s rRNA gene in the fecal samples. Total RNA was extracted from tissues and cells with ExtractRNA reagent (R401-01, Vazyme, China) according to the manufacturer’s protocols. cDNA synthesis was performed using the Evo M-MLV RT Premix for qPCR kit (AG11706, Accurate Biology, China) according to the manufacturer’s instructions. Quantitative PCR analysis was carried out on ViiATM 7 Real-Time PCR System (Applied Biosystems). The specific primers were shown in Table S2 . The samples were homogenized with 400μL of ice-cold methanol: water (7:3, v/v) containing internal standard, vortexed (3min), sonicated (10min, ice bath), and incubated at −20°C for 30min. After centrifugation (12000×g, 10min, 4°C), the supernatant was collected and re-centrifuged (12000×g, 3min, 4°C). A 200μL aliquot of the clarified supernatant was analyzed by LC-MS. The samples were analyzed via two LC/MS methods. For positive ion mode, aliquots were separated on a Waters ACQUITY Premier HSS T3 Column (1.8 μm, 2.1 × 100 mm) with 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) using the following gradient: 5% B (0-2min), 5–20% B (2-5min), 20–60% B (5-8min), 60–99% B (8-9min), 99% B (9-10.5min), 99-5% B (10.5-10.6min), and 5% B (10.6-13min). Column temperature: 40°C; flow rate: 0.4 mL/min; injection volume: 4μL. Negative ion mode analysis used an identical gradient profile. Data acquisition (IDA mode, Analyst TF 1.7.1) parameters were as follows: ISVF±5000/-4000 V, DP±60/-60 V, GAS1/GAS2 50 psi, CUR 25 psi, 550°C. TOF-MS: m/z 50–1000, 200 ms, BG subtraction. MS/MS: m/z 25–1000, 40 ms, CE±30/-30 ± 15 eV, unit resolution, >100cps, 4 Da isotope exclusion, 50ppm tolerance, 18 ions/cycle. Unsupervised PCA was performed using the statistical function prcomp in R ( www.r-project.org ), and differential metabolites were determined by VIP(VIP>1) and p - value ( p - value<0.05, ANOVA). Identified metabolites were annotated with the KEGG Compound database ( http://www.kegg.jp/kegg/compound/ ), and the annotated metabolites were then mapped to KEGG Pathway database ( http://www.kegg.jp/kegg/pathway.html ). The samples were subjected to targeted metabolomic analysis using liquid chromatography-tandem mass spectrometry (LC-MS/MS) (AB SCIEX 4500, USA). Briefly, 50 μL of samples was mixed with 200 μL of ice-cold acetonitrile containing 0.1% formic acid for protein precipitation. After vortexing (5min) and centrifugation (16000×g, 10min, 4°C), 100 μl of supernatant was taken and transferred to another new tube for testing. Mass spectrometric detection was conducted on an AB SCIEX 4500 triple quadrupole system equipped with electrospray ionization (ESI) in positive mode. Data acquisition and quantification were performed using SCIEX OS Software (v2.1.5). Calibration curves (1-1,000 ng/mL, R 2 > 0.998) were established with quality control samples (QC; intra-day RSD <5%, inter-day RSD <8%). The coding sequence of the patatin-like phospholipase family protein (Op-PLA2) was amplified by PCR from synthetic DNA using primers with homologous arms to the pET-21b(+) vector (forward: 5′- AGCATGACTGGTGGACAGCAAATGGGTCGCGGATCCATGGATACC-3′; reverse: 5′- AGCCGGATCTCAGTGGTGGTGGTGGTGGTGCTCGAGTGGTTCCAG-3′). The pET-21b(+) vector was double-digested with BamHI and XhoI, and the linearized vector was then combined with the PCR product using a homologous recombination kit. The resulting recombinant plasmid pET-21b(+)-Op-PLA2 was transformed into E. coli BL21(DE3) by heat shock transformation. Transformants were selected on LB agar plates containing ampicillin, and positive clones were designated E. coli -PLA2. The E. coli carrying the empty pET-21b(+) vector served as the negative control ( E. coli -Vector). For protein expression, fresh LB broth containing ampicillin was inoculated with 1% (v/v) overnight culture of E. coli -PLA2. When the cell density reached an OD600 of 0.6, IPTG was added to a final concentration of 0.5 mM to induce expression of the fusion protein for 6 h at 37°C. Bacterial cells were harvested by centrifugation at 5000 rpm for 5 min and resuspended in sterile PBS for subsequent experiments. The proliferation of ESCs was determined by a colorimetric method with Cell Counting Kit-8 (CCK8) according to manufacturer’s instructions. The absorbency was measured with microplate reader (Synergy HTX, BioTek, USA) at a wavelength of 450 nm. The ESCs were inoculated and cultured until reaching 80–90% confluence in 6-well plates. Then, scratches by a sterile 200μL pipette tip were made while drugs were added for treatment. Photos were snapped at 0, 12 and 24 h of co-culture by inverted fluorescence microscope. The proportion of migration distance was calculated for assessing the effect of different treatments on ESCs migratory ability. The migration ESCs was assessed using Transwell (Corning, NY, USA) with 8 μm pore-sized filters. First, 200 μL of ESCs suspension (1.5×10 5 cells/mL) was plated into the upper chamber of the Transwell. In the lower compartment, 600μL serum-free medium and serum-free medium supplemented with 4μM LPC (16:0) were added separately. After 24–48 h of incubation at 37°C, the non-migrated cells in the upper chamber were wiped away with cotton swabs, and cells that migrated into the lower surface of the filters were fixed and stained with crystal violet, and photographed with microscope. Total RNA was isolated from the samples and purified using standard protocols. Sequencing libraries were constructed and subjected to paired-end (PE) sequencing on the Illumina platform, generating raw data in FASTQ format. Gene expression quantification was performed using HTSeq (v0.9.1) to obtain raw read counts. Expression levels were normalized using FPKM (Fragments Per Kilobase per Million fragments) and TPM (Transcripts Per Million). Differential gene expression analysis between comparative groups was conducted with DESeq2 (v1.38.3). Significantly enriched Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways among differentially expressed genes were identified using the clusterProfiler package (v4.6.0). Gene Set Enrichment Analysis (GSEA) was further performed using clusterProfiler (v4.6.0) to explore biologically relevant pathways and functional categories. The siRNA transfection was conducted according to protocols provided by Shanghai HANBIO (Shanghai, China). Briefly, ESCs were cultured to 50%–70% confluence, and then transfected with SPP1 siRNA or control siRNA in serum-free, antibiotic-free DMEM/F12 medium containing RNAFit reagent (HB-RF-1000, Hanheng Biotechnology, China). The sequences of siRNA were shown in Table S3 . After transfection for 48 h, the knockdown of SPP1 gene was validated by PCR and western blot. Total protein of cells or tissues was extracted, and protein concentration was quantified by BCA protein detection kit (CW0014S, SolelyBio, China). Proteins were separated on 10% SDS-PAGE, and then transferred onto PVDF membranes. The membranes were blocked with 5% fat-free milk for 1 h at room temperature, and then incubated with primary antibodies (listed in Table S3 ) overnight at 4°C. Next, membranes were incubated with HRP-labeled secondary antibodies at room temperature for 1 h. The signals were detected using ECL kit (WBKLS0100, Merck Millipore, Amecrian). Bands intensities analysis were performed by Tanon 5200 imaging system (Tanon, China). The expression of target protein was quantified by image analysis software ImageJ (National Institutes of Health, USA), and normalized to GAPDH. The concentration of SPP1 in serum was measured using a commercial Human Osteopontin (OPN) ELISA Kit (JL10368-96T, J&L Biological, China). All procedures, including sample addition, incubation, washing, and color development, were performed in strict accordance with the manufacturer’s instructions. The optical density (OD) was measured at a wavelength of 450 nm using a microplate reader, and the SPP1 levels were calculated according to standard curve. Tissues were fixed in 4% paraformaldehyde, and then embedded in paraffin. Paraffin-embedded tissues were cut into 4 μm sections. For HE staining of mice EM lesions, the sections were stained with HE reagent. Pictures were processed using Panoramic MIDI by panoramic scanner (3D HISTECH, Hungary). Data were analyzed by Caseviewer software. Tissues were fixed in 4% paraformaldehyde. The paraffin-embedded sections were deparaffinized through xylene and rehydrated by graded ethanol. Antigen retrieval was performed using EDTA antigen retrieval solution (pH 8.0) and microwave heating (15 min). Then, 3% hydrogen peroxide was used to quench endogenous peroxidase activity, and the slides were blocked with PBS containing 2.5% goat serum (AR0009, Biological Industries) for 1 h, and then incubated with primary antibodies overnight (listed in Table S3 ). After washing with PBS, sections were incubated with secondary antibodies (M21003S, abmart, China) for 1 h at room temperature, and cell nucleus was stained with DAPI. Tissues were fixed in 4% paraformaldehyde. The paraffin-embedded sections were deparaffinized through xylene and rehydrated by graded ethanol. Antigen retrieval was performed using EDTA antigen retrieval solution (pH 8.0) and microwave heating (15 min). Then, 3% hydrogen peroxide was used to quench endogenous peroxidase activity, and the slides were blocked with 5% BSA. The slices were incubated overnight at 4°C with primary antibody against SPP1 (HA723081, HUABIO, China). Next, the sections were incubated with an HRP-conjugated secondary antibody for 30 min at room temperature, and developed using DAB. Finally, the sections were counterstained with hematoxylin for 30 s and mounted with neutral resin. Statistical analyses were conducted with GraphPad Prism 9.0 and SPSS. The results are expressed as means ± SD of multiple independent experiments. For parametric data, the results of different groups were analyzed using two-tailed unpaired t test or one-way or two-way ANOVA with Fisher’s LSD post-hoc test. For nonparametric data, the results were analyzed using Mann-Whitney test or Kruskal-Wallis test with Dunn’s post-hoc test. Correlations were examined using Spearman’s rank correlation analysis. A value of p < 0.05 was considered statistically significant. All statistical details, including the specific tests used and exact p values, are reported in the figure and/or figure legends as appropriate.

This work is supported by the 10.13039/501100001809 National Natural Science Foundation of China (grant nos. 82572635 , 81925026 , 82341218 , 82130068 , and 82402710 ), 10.13039/501100012166 National Key R&D Program of China (grant no. 2022YFA0806400 ), 10.13039/501100021171 Guangdong Basic and Applied Basic Research Foundation (grant nos. 2026A1515012331 and 2023A1515011688 ), Science and Technology Program of Guangzhou (grant no. 2025A04J3781 ), 10.13039/501100002858 China Postdoctoral Science Foundation (grant no. 2024M751324 ), President Foundation of Zhujiang Hospital , 10.13039/501100010096 Southern Medical University (grant no. yzjj2022ms18 ), and Guangdong Provincial Clinical Research Center for Laboratory Medicine (grant no. 2023B110008 ). We thank all the research assistants from the Microbiome Medicine Center of Zhujiang Hospital who kindly provided invaluable support and help for this study.