Bridging gut microbiota and polycystic ovary syndrome: the mediating role of metabolites and immune pathways via Mendelian randomization and mediation analysis Authors: Yushan Li, Yan Zhang, Wei Liu, Tao Tao Published: 2026 ABSTRACT AbstractBackgroundPolycystic Ovary Syndrome (PCOS) is a common reproductive metabolic disorder impacting women of reproductive age, with its etiology influenced numerous factors. Recent research indicates that the gut microbiota, along with its metabolites and associated immune-inflammatory responses, may contribute to the pathogenesis of PCOS. However, there is finite understanding of the potential intermediate impacts of circulating metabolites, immune cells, and inflammatory proteins on the linkage between intestinal flora and PCOS. This study seeks to explore the interrelationships among the gut microbiome, immune-inflammatory responses, and the metabolome within the context of PCOS.MethodsAggregated statistics pertaining to individual traits were derived from genome-wide association studies accessible to general public. To elucidate mediation mechanisms, a two-step MR mediation framework was applied. In the first step, univariable MR was used to identify causal effects of gut microbial taxa on circulating metabolites, immune cells, and inflammatory proteins. In the second step, multivariable MR and mediation analyses were employed to estimate indirect effects of these mediators on PCOS. All analyses were performed using the inverse variance weighted (IVW) method as the primary model, complemented by MR-Egger, weighted median, and MR-PRESSO sensitivity tests.ResultsMR analysis identified the causal impacts of 11 intestinal flora taxa on PCOS, among which 4 taxas (Class Betaproteobacteria, Genus Eubacterium eligens group, Genus Ruminiclostridium6 and Order Burkholderiales) may exert positive effects on PCOS, 7 taxas (Family Bifidobacteriaceae, Genus Bilophila, Genus Holdemania, Genus Hungatella, Genus Ruminococcaceae UCG004, Genus Veillonella, and Order Bifidobacteriales) may act as protective factors on PCOS. Meanwhile, 124 plasma metabolites, 45 circulating immune cell types, and 5 inflammatory proteins exhibited causal effects on PCOS. Notably, the circulating metabolites predominantly involved in PCOS were correlative of androgenic steroids, leucine, isoleucine and valine metabolism, glutathione metabolism, and secondary bile acid metabolism, and other metabolic pathways. Mediation analysis elucidated 34 metabolites-mediated path-ways linking gut microbiota and PCOS, with 13 immune cells/inflammatory proteins-mediated pathways interconnecting gut flora and PCOS.ConclusionsAccording to MR analysis, circulating metabolites, immune cells, and inflammatory proteins play a role in the correlation between intestinal flora and PCOS. The findings shed new light on the underlying pathophysiological mechanisms of PCOS.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13048-026-02162-2.Keywords: polycystic ovary syndrome, gut microbiome, immune cells, inflammatory proteins, metabolites, mendelian randomization, mediation analysis Introduction Women of reproductive age are highly susceptible to polycystic ovary syndrome (PCOS), a chronic metabolic and reproductive disorder affecting approximately 5% to 13% of women [1]. A diagnosis of PCOS requires at least two of these criteria: hyperandrogenism, oligomenorrhea, and/or polycystic ovarian morphology. Excess androgen is a significant contributor to many of the phenotypic manifestations of PCOS, including follicular dysplasia, ovulatory dysfunction, endometrial pathologies, and metabolic abnormalities [2]. Additionally, insulin resistance and low-grade inflammation in correlation to PCOS contribute to an elevated possibility of developing type 2 diabetes mellitus (T2D) [3], which imposes a heavy socio-economic burden. Numerous research conducted in recent years have suggested that the pathophysiology and development of PCOS may be significantly influenced by the intestinal microbiota [4]. Besides, PCOS is tightly linked to persistent low-grade inflammatory reactions and the immune system [5]. Immune cells influence sex hormone signaling in granulosa cells through the secretion of cytokines, and macrophages with a pro-inflammatory phenotype can induce insulin resistance [6]. Tremellen [7] introduced the DOGMA theory, which posited that dysbiosis of the intestinal microbiota, resulting from unhealthy dietary habits, could alter the gut mucosa’s permeability. Thus, Gram-negative bacteria in the colon may release the pro-inflammatory substance lipopolysaccharide (LPS) into the bloodstream, which would trigger the immune system [8, 9]. This activation may hinder insulin receptor function, elevate serum insulin concentrations, stimulate the ovaries to produce more androgens, and further impair normal follicle development [7]. Qiao et al. [10] demonstrated notable augmentation in the prevalence of Bacteroides vulgatus (B. vulgatus) within the intestines of PCOS individuals. In addition, glycodeoxycholic acid (GDCA), inversely related to B. vulgatus, was found to induce the secretion of interleukin-22 (IL-22) from intestinal group 3 innate lymphoid cell (ILC3) via GATA binding protein 3, hence mitigating local ovarian inflammation and contributing to the amelioration of the PCOS phenotype [10]. Metabolomics represents an emerging platform for the comprehensive analysis of metabolites within biological systems, and it has been widely utilized in the examination of diverse clinical states and the detection of diagnostic biomarkers [11]. This approach leverages qualitative and quantitative analyses of blood and other bodily fluids samples to strengthen researchers’ understanding of particular metabolites, metabolic pathways, and their relationships with diseases [12]. Metabolomics, being the final output of the genome and transcriptome, is regarded as more of the actual phenotype and capable of reflecting the microenvironment in the body [13]. Multiple investigations have established a strong correlation between metabolites, metabolic pathways, and PCOS [14–18]. Moreover, the gut flora produces several bioactive metabolites, incorporating bile acids, short-chain fatty acids, and polyamines, which are essential for sustaining metabolic homeostasis [8]. Integrating metabolomics with microbiome analysis holds promise for further elucidating the pathogenesis of PCOS. Specifically, the arginine derivative agmatine, synthesized by B. vulgatus via arginine decarboxylase activity, inhibits glucagon-like peptide-1 (GLP-1) secretion in intestinal L cells, hence worsening the PCOS phenotype [19]. In summary, intestinal flora may influence the initiation and advancement of clinical phenotypes related to PCOS, including chronic low-grade inflammatory response, metabolic disorders, elevated androgen levels, and follicular dysplasia. Despite previous studies examining the correlations among the gut microbiota, immunological inflammation, metabolome, and PCOS, the exact causal linkages and the degree of their mediation remain ambiguous. Mendelian randomization (MR) analysis provides a dependable approach for deducing the causal impacts of exposures on outcomes through the utilization of genetic [20]. Modern high-throughput experimental technologies, coupled with comprehensive genetic information, provide extensive resources for investigating the causal effects of various diseases. These high-throughput experimental biomarkers encompass metabolites, proteins, intestinal microbes, cell phenotypes, and imaging data, thereby providing an optimal foundation for MR studies [21]. Consequently, our objective is to clarify the causative association and determine if the connection between gut microbiota and PCOS is mediated by circulating metabolites, immune cells, and inflammatory proteins by a two-sample MR analysis and mediation analysis. Materials and methods Research design This study sought to carry out a preliminary two-sample MR analysis to examine the causal connection between intestinal flora and PCOS. The second objective was to evaluate the causal association among circulating immune cells, inflammatory proteins, metabolites, and PCOS. Furthermore, mediation analysis was employed to clarify the mediating roles of circulating immune cells, inflammatory proteins, and metabolites in the relationship between intestinal microbiota and PCOS. The datasets included in Mendelian randomization research were obtained from publicly accessible, legitimate, and traceable databases. The literature on database research routinely documented ethical approval and related certificates. The literature on database studies consistently included documentation of ethical approval and associated certificates [22]. Data origin For the investigation, we extracted Genome-Wide Association Study (GWAS) data from the Finngen-public-data-R10 version of the Finnish adult female participant database [23]. This database integrates the national biobank, structured national healthcare data, and specific population cohorts, combining biobank genotype data with digital health records from the health registry, which is ethnically homogeneous. The GWAS dataset for PCOS (finngen_R10_E4_PCOS.gz) comprises 1,639 patients diagnosed with PCOS and 218,970 control subjects. The GWAS data meta-analysis from 2021, published by the MiBioGen consortium (https://mibiogen.gcc.rug.nl) [22], seeks to examine the correlation between genetic variation and intestinal flora. This database includes 16 S rRNA sequencing data and genotype information, encompassing 211 taxa, 9 phyla, 16 classes, 20 orders, 35 families, 131 genera [24]. The immune cell phenotypic summary data comprised 3,757 individuals of European descent in a distinct cohort, containing 731 immune traits: 32 morphological parameters, 389 median fluorescence intensities (MFIs) of surface antigens, 118 absolute cell counts, and 192 relative counts (ratios between cell levels) [25]. Summarized statistics for plasma metabonomics are available in the GWAS directory (https://www.ebi.ac.uk/gwas/) under accession numbers GCST90199621 to GCST90201020, accessible without login. The dataset includes 1,091 plasma metabolites and 309 metabolite ratios derived from 8,299 European people [26]. The summary data for plasma inflammatory proteins was obtained from a genome-wide protein quantitative trait locus (pQTL) study involving 11 cohorts and 14,824 persons of European descent [27, 28]. This database incorporates measurements of 91 circulating inflammatory proteins acquired with the Olink Target platform. Selection of instrumental variables In Mendelian randomization analysis to investigate causal relationships between exposure factors and outcome variables, three fundamental assumptions must be met: relevance (the instrumental variable must correlate with the exposure factor X), independence (the instrumental variable must be independent of confounders), and exclusion restriction (the instrumental variable influences the outcome solely through the exposure factor) [29]. Quality control techniques were instituted to identify single nucleotide polymorphisms (SNPs) as instrumental variables linked to the intestinal flora, circulating metabolites, and immune cells, employing a significance threshold of P < 1 × 10− 5 [30]. For plasma proteins, the screening criterion was set at P < 5 × 10− 6 [31]. Linkage disequilibrium (LD) analysis, which assesses the non-random co-occurrence of alleles [32], was conducted. The effect threshold was established at r2 = 0.001, with a clumping window size of 10,000 kb to guarantee the mutual independence of SNPs [28]. To prevent alleles from confounding the analysis, palindromic sequences were excluded to ensure homology. To exclude weak instrumental variables, those with F-value statistics exceeding 10 were chosen, utilizing the formula = R²(n-2)/(1-R²), in which F > 10 signifies a robust connection between SNPs and exposure variables [33]. Statistical analysis Initially, Univariable Mendelian Randomization (UVMR) was conducted. The Inverse Variance Weighted (IVW) method was designated as the principal statistical approach, complemented by additional robust statistical techniques including the Weighted Mode, Weighted Median (WM), MR Egger and Simple Mode. The horizontal pleiotropy of instrumental variables is further evaluated using the P-value of the intercept obtained via MR Egger regression. To identify potential outliers and detect gene pleiotropy, as well as to verify robustness, the MR Pleiotropy Residual Sum and Outlier (MR-PRESSO) test is employed. A combined evaluation of horizontal pleiotropy using both MR Egger and MR-PRESSO suggests the absence of significant horizontal pleiotropy if the P-value exceeds 0.05 [32]. among the instrumental variables is assessed using Cochran’s Q statistic, with a P-value exceeding 0.05 indicating the absence of significant heterogeneity. Two-step mediation analysis based on Mendelian randomization To explore whether circulating metabolites, immune cells, or inflammatory proteins mediate the causal relationship between gut microbiota and PCOS, we adopted a two-step MR mediation. In the first step, UVMR was applied to estimate the genetic associations between gut microbiota and potential mediators, including metabolites, immune cell subtypes, and inflammatory proteins. The causal effect of each microbial taxon on these mediators was quantified as BETA₁ using the IVW method as the primary approach. Metabolites or immune/inflammatory traits showing significant associations (P < 0.05) were selected for the second-step analysis. In the second step, multivariable MR (MVMR) was performed to estimate the direct and indirect causal effects of gut microbiota on PCOS while adjusting for the mediating variables identified in step one. The MVMR model provided the effect estimate BETA₂ for the mediator–outcome (PCOS) relationship. The total effect of gut microbiota on PCOS was expressed as BETA, and the mediated (indirect) effect was calculated as BETA₁ × BETA₂, while the direct effect was BETA – (BETA₁ × BETA₂). The proportion of mediation was calculated as (BETA₁ × BETA₂) / BETA. The aforementioned analyses were conducted using R program (version 4.3.2). Particular software tools, such as TwoSampleMR, TSMRhelper, MRPRESSO, and Plinkbinr, were utilized to analyze different facets of the data. Packages such as ggplot2, forestploter, circlize, and ComplexHeatmap were employed for graphical representation and visualization. Results The causative impacts of intestinal flora on PCOS The IVW analysis identified 11 intestinal microbial taxa potentially involved in the development of PCOS (P < 0.05) (Fig. 1A and B and Supplementary Table S1). Among these, four taxa may increase the risk of PCOS: Class Betaproteobacteria (OR = 1.554, P = 0.040), Genus Eubacterium eligens group (OR = 1.751, P = 0.023), Genus Ruminiclostridium6(OR = 1.385, P = 0.035) and Order burkholderiales (OR = 1.959, P = 0.002). Seven taxa: Family Bifidobacteriaceae (OR = 0.678, P = 0.047), Genus Bilophila (OR = 0.675, P = 0.035), Genus Holdemania (OR = 0.707, P = 0.008), Genus Hungatella (OR = 0.708, P = 0.028), Genus Ruminococcaceae UCG004 (OR = 0.616, P = 0.013), Genus Veillonella (OR = 0.639, P = 0.036), and Order bifidobacteriales (OR = 0.678, P = 0.047) acted as protective factors. Q statistics indicated no evidence of heterogeneity. SNP scatter plots and leave-one analysis plots were provided in supplementary figure (Figure S1 and Figure S2). The MR-PRESSO analysis and MR-Egger regression demonstrated the lack of horizontal pleiotropy (Supplementary Table S2). Outcomes of two sample MR analyses regarding the causal relationship from gut microbiota to PCOS. A, forest plot of two sample MR analysis outcome. B, circular heatmap of MR outcomes from gut microbiota to PCOS. MR-Egger, Mendelian randomization of Egger regression; IVW, inverse variance weighted; OR, odds ratio; CI, confidence interval. WM, weighted median; MRPRESSO, MR Pleiotropy Residual Sum and Outlier; Q, Cochran’s Q test The causative impacts of circulating metabolites, immune cells, and inflammatory proteins on PCOS The IVW analysis identified 124 causal association from plasma metabolomics to PCOS (P < 0.05) (Figs. 2A and 3). These relationships encompassed lipids (29), amino acids (27), peptides (2), carbohydrates (2), nucleotides (8), Xenobiotics (9), partially characterized molecules (4), cofactors and vitamins (2), unknown substances (19), and metabolite ratios (23). Several metabolic pathways are involved: circulating metabolites prominently included Androgenic Steroids, Leucine, Isoleucine and Valine Metabolism, Glutathione Metabolism, and Secondary Bile Acid Metabolism, et al. (Fig. 2D). Additionally, the IVW method uncovered 45 significant associations (P < 0.05) involving circulating immune cell phenotypes (Figs. 2B and 4). The study identified 5 absolute counts, 16 relative counts, and 24 MFIs. Regarding cell types, the analysis included 12 B cells, 4 maturation stages of T cells, 8 regulatory T cells (Tregs), 7 TBNK cells, 5 myeloid cells, 7 monocytes, and 2 conventional dendritic cells (cDCs). MR revealed five associations between inflammatory proteins and PCOS (Figs. 2C and 4). The inflammatory proteins recognized as possible risk factors for PCOS include C-C motif chemokine 25, interferon-gamma, and neurotrophin-3. Conversely, interleukin-6 and matrix metalloproteinase-10 were identified as potential protective inflammatory proteins. Causal relationship estimates of plasma metabolites, circulating immune cells, and inflammatory proteins on PCOS and involved metabolic sub-pathways. A - C, volcano plots of the IVW method for the associations from plasma metabolites, circulating immune cells, and inflammatory proteins to PCOS. D, count of metabolic sub-pathways involved in significant outcomes of plasma metabolites (n≥2) Forest plot of two sample MR analysis from plasma metabolites to PCOS. OR, odds ratio; CI, confidence interval. A, amino acid, carbohydrate and peptide. B lipid, nucleotide and cofactors and Vitamins. C, partially characterized molecules, xenobiotics and partially characterized molecules Forest plot of two sample MR analysis from circulating immune cells and inflammatory proteins to PCOS. OR, odds ratio; CI, confidence interval Mediation analysis of the causal pathways linking circulating metabolites, immune cells, and inflammatory proteins from the intestinal flora to the pathogenesis of PCOS To explore the factors contributing to the onset of PCOS, we adopted the two-sample MR analysis previously applied to ascertain relationships between intestinal flora and PCOS-related plasma metabolites, immune cells, and inflammatory proteins. We conducted mediation analysis to clarify the causative relationships between circulating metabolites, immune cells, and inflammatory proteins from the intestinal flora to the pathogenesis of PCOS. Our analysis identified 70 associations with metabolites, 19 associations with immune cells, and 3 associations with proteins, thereby offering insights into the causal linkages among gut microbiota, metabolic profiles, immunological responses, and inflammation within the framework of PCOS. At the metabolite level, our mediation analysis identified 34 intermediary relationships involving 9 types of intestinal flora and 25 types of metabolites (Fig. 5A and B; Table 1). Seven microbial types exhibited a variety of metabolite mediators that regulate the occurrence of PCOS, while eight metabolites were found to mediate the association between different microbial types and PCOS. At the level of PCOS risk factors, the class Betaproteobacteria was positively associated with PCOS through 9 metabolites: Homocitrulline (1.81%), Tauro-beta-muricholate (5.90%), Glutarylcarnitine (c5-dc) (7.26%), Cysteinylglycine disulfide (6.58%), N-lactoyl valine (7.71%), 5-oxoproline (5.44%), Dihydroorotate (5.22%), X-23,659 (6.58%), and Uridine to pseudouridine ratio (9.53%). What’s more, the order Burkholderiales was found to increase the risk of PCOS through 11 metabolites: Homocitrulline (1.34%), Glutarylcarnitine (c5-dc) (6.39%), Cysteinylglycine disulfide (5.06%), N-lactoyl valine (4.76%), 5-oxoproline (4.76%), Pseudouridine (9.37%), Dihydroorotate (4.61%), X-18,886 (15.32%), X-23,659(6.10%), X-25,371(4.16%) and Uridine to pseudouridine ratio (7.29%). The favorable correlation between the Eubacterium eligens group and PCOS was facilitated by Epiandrosterone sulfate and Glutarylcarnitine (C5-DC), with mediating ratios of 3.75% and 6.60%, respectively. Additionally, the positive relationship between genus Ruminiclostridium6 and PCOS was mediated by N2, N5-diacetylornithine (5.83%), Phenylacetylcarnitine (7.06%), Sulfate of piperine metabolite C16H19NO3 (3) (15.34%), and N-palmitoyl-sphingosine (d18:1 to 16:0) to N-palmitoylsphinganine (d18:0 to 16:0) ratio (6.44%). Mediation effects of plasma metabolites, circulating immune cells, inflammatory pro-teins in the associations between gut microbiome and PCOS. A, sankey plot revealing plasma metabolites pathways linking gut microbiome to PCOS. B, the proportion of association degree between the gut microbiome and PCOS mediated by plasma metabolites. C, sankey plot re-vealing circulating immune cells and inflammatory proteins pathways linking gut microbiome to PCOS. D, the proportion of association degree between the gut microbiome and PCOS mediated by circulating immune cells and inflammatory proteins Table 1Mediation effect of gut microbiota on PCOS via plasma metabolitesExposureMediatorOutcomeTotal effectDirect effectMediation effect(95%CI)Mediation Proportionclass BetaproteobacteriaHomocitrullinePCOS0.44090.43290.008(-0.044-0.067)1.81%Tauro-beta-muricholatePCOS0.44090.41490.026(-0.021-0.096)5.90%Glutarylcarnitine (c5-dc)PCOS0.44090.40890.032(-0.001-0.083)7.26%Cysteinylglycine disulfidePCOS0.44090.41190.029(-0.012-0.093)6.58%N-lactoyl valinePCOS0.44090.40690.034(-0.014-0.107)7.71%5-oxoprolinePCOS0.44090.41690.024(-0.002-0.065)5.44%DihydroorotatePCOS0.44090.41790.023(-0.003-0.064)5.22%X-23,659PCOS0.44090.41190.029(-0.006-0.084)6.58%Uridine to pseudouridine ratioPCOS0.44090.39890.042(-0.002-0.113)9.53%genus Bilophila7-methylguaninePCOS-0.3926-0.3526-0.04(-0.128-0.022)10.19%genus Eubacterium eligens groupEpiandrosterone sulfatePCOS0.56040.53940.021(-0.055-0.109)3.75%Glutarylcarnitine (c5-dc)PCOS0.56040.52340.037(-0.003-0.102)6.60%genus Holdemania1-palmitoyl-2-linoleoyl-GPI (16:0/18:2)PCOS-0.3473-0.3153-0.032(-0.085-0.002)9.21%QuinolinatePCOS-0.3473-0.3263-0.021(-0.064-0.006)6.05%genus Hungatella1-stearoyl-2-docosahexaenoyl-GPE (18:0/22:6)PCOS-0.3453-0.3313-0.014(-0.051-0.012)4.05%1-palmitoyl-2-oleoyl-GPE (16:0/18:1)PCOS-0.3453-0.3143-0.031(-0.075–0.002)8.98%genus Ruminiclostridium6N2, N5-diacetylornithinePCOS0.32600.30700.019(-0.039-0.092)5.83%PhenylacetylcarnitinePCOS0.32600.30300.023(-0.006-0.068)7.06%Sulfate of piperine metabolite C16H19NO3 (3)PCOS0.32600.27600.050(0.002–0.122)15.34%N-palmitoyl-sphingosine (d18:1 to 16:0) to N-palmitoylsphinganine(d18:0 to 16:0) ratioPCOS0.32600.30500.021(-0.019-0.079)6.44%genus Ruminococcaceae UCG0045alpha-androstan-3alpha,17alpha-diol monosulfatePCOS-0.4848-0.4808-0.004(-0.038-0.025)0.83%genus VeillonellaIndolebutyratePCOS-0.4474-0.4134-0.034(-0.097-0.006)7.60%Bilirubin degradation product, C17H20N2O5 (1)PCOS-0.4474-0.4434-0.004(-0.036-0.024)0.89%order BurkholderialesHomocitrullinePCOS0.67250.66350.009(-0.043-0.068)1.34%Glutarylcarnitine (c5-dc)PCOS0.67250.62950.043(0.000-0.103)6.39%Cysteinylglycine disulfidePCOS0.67250.63850.034(-0.014-0.105)5.06%N-lactoyl valinePCOS0.67250.64050.032(-0.016-0.106)4.76%5-oxoprolinePCOS0.67250.64050.032(-0.001-0.079)4.76%PseudouridinePCOS0.67250.60950.063(-0.004-0.155)9.37%DihydroorotatePCOS0.67250.64150.031(-0.004-0.081)4.61%X-18,886PCOS0.67250.56950.103(0.016–0.227)15.32%X-23,659PCOS0.67250.63150.041(-0.003-0.105)6.10%X-25,371PCOS0.67250.64450.028(-0.034-0.101)4.16%Uridine to pseudouridine ratioPCOS0.67250.62350.049(0.000-0.125)7.29%CI, confidence interval Mediation effect of gut microbiota on PCOS via plasma metabolites 1-stearoyl-2-docosahexaenoyl-GPE (18:0/22:6) Sulfate of piperine metabolite C16H19NO3 (3) N-palmitoyl-sphingosine (d18:1 to 16:0) to N-palmitoylsphinganine (d18:0 to 16:0) ratio CI, confidence interval In the context of the protective factor pathway of PCOS, the Genus Bilophila exhibited a negative association with PCOS mediated through 7-methylguanine, with a mediator ratio of 10.19%. The protective effect of the Genus Holdemania on PCOS is facilitated by 1-palmitoyl-2-linoleoyl-GPI (16:0/18:2) and Quinolinate, with mediation ratios of 9.21% and 6.05%, respectively. The Genus Hungatella demonstrated a negative association with PCOS via 1-stearoyl-2-docosahexaenoyl-GPE (18:0/22:6) and 1-palmitoyl-2-oleoyl-GPE (16:0/18:1), with mediation ratios of 4.05% and 8.98%, respectively. Furthermore, the negative association of the Genus Veillonella with PCOS is mediated by Indolebutyrate (7.60%) and Bilirubin degradation product, C17H20N2O5(1) (0.89%). In addition, 5alpha-androstan-3alpha and 17alpha-diol monosulfate mediated the protective effect of Genus Ruminococcaceae UCG004 on PCOS at a mediating ratio of 0.83%. At the immune cell level, we identified 12 intermediary relationships through mediation analysis, involving 7 types of gut microbes and 9 types of immune cell phenotypes (Fig. 5C and D; Table 2). Specifically, the Class betaproteobacteria, identified by the HLA DR++ monocyte percentage of leukocytes (intermediate ratio: 12.25%), was found to promote the occurrence of PCOS. Moreover, the Genus Eubacterium eligens group exhibited a positive correlation with PCOS through the presence of CD20 on CD20−CD38−B cells. Furthermore, the absolute count of CD4+ CD8dim T cells facilitated the positive causal relationship between the Genus Ruminiclostridium6 and PCOS, with an intermediate ratio of 1.23%. Central Memory CD8+ T cells mediated the positive effect of the Order burkholderiales on PCOS, with a mediation ratio of 2.38%. Table 2Mediation effect of gut microbiota on PCOS via circulating immune cells and inflammation proteinExposureMediatorOutcomeTotal effectDirect effectMediation effectMediation Proportionclass BetaproteobacteriaHLA DR + + monocyte %leukocytePCOS0.44090.38690.054(-0.022-0.172)12.25%genus BilophilaCD20 on CD20- CD38- B cellPCOS-0.3926-0.3596-0.033(-0.101-0.014)8.41%CD45 on lymphocytePCOS-0.3926-0.3566-0.036(-0.103-0.005)9.17%HLA DR on CD14- CD16 + monocytePCOS-0.3926-0.3796-0.013(-0.051-0.013)3.31%genus Eubacterium eligens groupCD20 on CD20- CD38- B cellPCOS0.56040.52940.031(-0.028-0.115)5.53%genus HoldemaniaCD45 on lymphocytePCOS-0.3473-0.3333-0.014(-0.056-0.019)4.03%genus Ruminiclostridium6CD4 + CD8dim T cell Absolute CountPCOS0.32600.32200.004(-0.033-0.044)1.23%genus Ruminococcaceae UCG004Memory B cell %lymphocytePCOS-0.4848-0.4428-0.042(-0.118-0.006)8.66%CD4 + CD8dim T cell Absolute CountPCOS-0.4848-0.4798-0.005(-0.051-0.037)1.03%CD25 on CD24 + CD27+ B cellPCOS-0.4848-0.4788-0.006(-0.051-0.034)1.24%HLA DR on CD33- HLA DR+PCOS-0.4848-0.4818-0.003(-0.064-0.057)0.62%order BurkholderialesCentral Memory CD8 + T cell %T cellPCOS0.67250.65650.016(-0.034-0.080)2.38%CI, confidence interval Mediation effect of gut microbiota on PCOS via circulating immune cells and inflammation protein CI, confidence interval The Genus Bilophila was negatively associated with PCOS through three mediators: CD20 on CD20− CD38− B cells, CD45 on lymphocytes, and HLA-DR on CD14− CD16+ monocytes, with mediation ratios of 8.41%, 9.17%, and 3.31%, respectively. Additionally, the Genus Bilophila was negatively associated with PCOS through Neurotrophin-3 levels (8.14%). The Genus Ruminococcaceae UCG004 exerted protective effects on PCOS through Memory B cell %lymphocyte (8.66%), the absolute count of CD4+ CD8dim T cells (1.03%), CD25 on CD24+ CD27+ B cells (1.24%) and HLA DR on CD33- HLA DR+ (0.62%). The Genus Holdemania exhibited a negative association with PCOS via CD45 on lymphocytes (4.03%). Discussion This study utilized two-sample MR and mediation analysis to clarify the causal links among gut microbiota, circulating metabolites, immune cells, inflammatory proteins, and PCOS. The results suggested possible causal relationships among 11 gut microbial taxa, 124 circulating metabolites, 45 immune cell phenotypes, and 5 inflammatory proteins concerning PCOS. Notably, the subpathways associated with circulating metabolites prominently included Androgenic Steroids, Leucine, Isoleucine and Valine Metabolism, Glutathione Metabolism and Secondary Bile Acid Metabolism, et al. In our prior investigations, we observed that patients with PCOS demonstrated markedly increased concentrations of secondary unconjugated bile acids relative to the control group [34], substantiating the results of the current investigation. Furthermore, the mediation analysis results confirmed the mediating influence of particular factors on intestinal microbiota in the pathogenic processes of PCOS. Our investigation revealed a greater prevalence of the Class betaproteobacteria, Genus Eubacterium eligens group, Genus Ruminiclostridium6, and Order burkholderiales, which may correlate with an elevated risk of developing PCOS. In accordance with the literature, an unidentified strain 0319_6G20 from the phylum Proteobacteria has been determined to be crucial for the identification of PCOS [12]. LEfSe analysis of the non-obese PCOS group and the control group revealed an enrichment of Proteobacteria in the PCOS group [35]. Bilophila, generally not categorized as a probiotic and regarded as a small constituent of the gut microbiota with limited quantity and distribution, possesses a contentious mechanism. Although not commonly acknowledged as a probiotic, studies suggest that Bilophila may play a preventive effect in PCOS [36]. Intake of animal products is a major source of trimethylamine (TMA), and the subsequent production of trimethylamine N-oxide (TMAO) is closely linked to an elevated risk of diabetes and cardiovascular disease [37]. Annunziata et al. proposed a potential enhancement of the DOGMA theory, indicating that TMAO may be linked to an elevated cardiovascular risk in women with PCOS [38]. Furthermore, there is an expansion of the genetic code within the genome of Bilophila, where a stop codon in the TMA methyltransferase gene can encode pyrrolysine. This modification enables Bilophila to metabolize TMA without producing TMAO, thereby potentially mitigating the cardiovascular disease risk associated with an animal-based diet by circumventing TMAO production [39]. Metabolite mediation analysis demonstrated that 7-methylguanine mediated the inverse correlation between Bilophila and PCOS. The genus Holdemania includes several strains, some of which have been identified in the human gastrointestinal tract. Notwithstanding this, the present comprehension of Holdemania is nonetheless constrained. It is claimed that Holdemania facilitates the breakdown and fermentation of carbohydrates, yielding advantageous metabolic byproducts such short-chain fatty acids (SCFAs). Moreover, a negative connection exists between Holdemania and poor lipid metabolism [36]. Furthermore, randomized controlled trials have demonstrated that the abundance of Hungatella in obese patients with PCOS significantly increased following treatment with a combination of exenatide and metformin [40]. We found that both Bifidobacteriales and Bifidobacteriaceae were protective factors for PCOS. Bifidobacterium pseudocatenulatum, as one of the beneficial Guild producing SCFA enriched by dietary fiber, rapidly responded to high dietary fiber signal and significantly improved body weight, body fat and blood glucose indicators in mice [41]. A functional guild dominated by SCFA producing bacteria is named as “cornerstone functional group” because these constituent members acidify the intestinal environment by producing SCFA, thereby inhibiting the overgrowth of opportunistic pathogens, which is similar to the stabilizing effect of trees as “cornerstone species” on forest ecosystems [42]. In addition to compositional alterations in gut microbiota, recent evidence suggests that functional enzymatic activities of intestinal bacteria may play a pivotal role in PCOS pathogenesis. Patel et al. [43] demonstrated that women with PCOS exhibit significantly elevated fecal β-glucuronidase and β-glucosidase activities, which were positively correlated with serum testosterone and estradiol concentrations, implicating these microbial enzymes in the reactivation of conjugated sex hormones and the maintenance of systemic hyperandrogenism. Complementarily, Łagowska et al. [44]reported that β-glucuronidase activity was strongly associated with carbohydrate metabolism rather than androgen status in overweight and obese women with PCOS. This suggests that bacterial enzymatic functions contribute to both endocrine and metabolic regulation, providing a mechanistic bridge between microbial activity and metabolic dysfunction. Fecal metabolomics analysis in a rat model of PCOS has demonstrated that dehydroepiandrosterone (DHEA) induces alterations in the biosynthesis of arginine, which exerts protective effects on metabolic syndrome parameters [45]. Additional research indicates that arginine levels are reduced in PCOS patients, and supplementation with arginine ameliorates hormonal disturbances in PCOS rats [46]. In accordance with these findings, our study determined that arginine levels serve as a protective factor for PCOS (OR = 0.81), accompanied by modifications in the urea cycle pathway, although no significant mediating effects were detected. Research demonstrated that the quantity and activity of B lymphocytes in the peripheral blood of women with PCOS was elevated [47]. Moreover, androgen activation in the elevated-androgen milieu typical of PCOS directly influences alterations in B cell proportions, with B cell activation facilitating the emergence of PCOS-like metabolic abnormalities in murine models [48]. Our study found that CD20 on CD20−CD38−B cells can concurrently mediate the effects of the genus Bilophila and the genus Eubacterium eligens group on PCOS. Additionally, the molecular function of CD20 is linked to the signaling pathways of B cell receptors. CD20 has been shown to engage with many surface proteins on B cells [49]. An imbalance in T cell subsets and aberrant cytokine levels in the blood and ovaries of women with PCOS [50], alterations in the Treg to Th17 ratio [51], and the upregulation of Th1 and Th17 cells alongside a reduction in Th2 and Treg cells may play a role in the etiology of PCOS [52], potentially leading to a synergistic exacerbation of inflammation [53]. A connection was discovered between HLA-DR expression on CD14−CD16+ monocytes and an increased incidence of PCOS. Additionally, Bilophila was identified to reduce this risk by suppressing HLA−DR expression on CD14−CD16+ monocytes. Compared to conventional monocytes (CD14+CD16−), the CD16+ subset demonstrates an increased ability to secrete pro-inflammatory factors, which are enhanced in persons with PCOS [54]. This work represents the inaugural application of two-sample MR and mediation analysis to thoroughly examine causal links among intestinal microbiota, circulating metabolites, immune cells, inflammatory proteins, and PCOS, and reveal the underlying pathways. Nevertheless, the study is subject to certain limitations. The restricted demographic information precludes the possibility of conducting subgroup analyses, and the study’s focus is confined to a European population. Consequently, further investigation into the universality across multiple racial groups is warranted. While MR analysis provides a compelling basis for establishing causality, additional clinical and fundamental research is necessary to substantiate these findings in the future. Supplementary Information Supplementary Material 1. Acknowledgements We acknowledge the researchers and participators of the data base extracted in our article. Authors’ contributions All authors contributed to the study conception and design. Data collection and analysis were performed by Yushan Li. The first draft of the manuscript was written by Yushan Li and Tao Tao. All authors commented on previous versions of the manuscript. Wei Liu, Yan Zhang and Tao Tao supervised research. All writers reviewed and endorsed the final manuscript. Funding This work was supported by [the National Natural Science Foundation of China, (Grant numbers [82370797] and [82170807]), [Bethune Charitable Foundation] (Grant numbers [J202103E006]), and [National Nature Promotion Project, Renji Hospital, Shanghai Jiao Tong University School of Medicine] (Grant numbers [RJTJ23-ZD-005]). Author Tao Tao has received research support from Foundations mentioned. Data availability The original data presented in our research are available in online database. The name and stash number of the consortium can be found in our article. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The original data presented in our research are available in online database. The name and stash number of the consortium can be found in our article.