Gut Microbiota-Immune Interactions in Endometrial Cancer: Causal Mediation and Subtype-Specific Mechanisms

Introduction

Endometrial cancer (EC) is one of the most common gynecological malignancies, and its incidence has increased progressively in high-income regions.¹ According to their histological features, grade, and hormone receptor (ER and PR) expression, EC can generally be classified into types I and II.^1–3

Type I EC is the most frequent subtype, presenting as low-grade, endometrioid, and hormone receptor-positive with a good prognosis (85% 5-year OS rate).^1,2 Conversely, Type II EC is non-endometrioid, high-grade, and hormone receptor-negative, accompanied by high-risked metastasis and poor prognosis (5-year OS rate of ~55%).^2,4 Genetic and environmental factors, including obesity and metabolic and reproductive factors, are the primary risk factors for the occurrence and development of EC.^1,5 Notably, these risks are strongly linked to the gut and vaginal microbiota; for instance, the GM may affect endometrial carcinogenesis through altering the systemic and uterine cavity microenvironment, providing a novel therapy for EC.^2,6,7 In addition, EC, especially type II EC, is regarded as an immunogenic disease, and its infiltrating immune cells affect anti-cancer therapy to facilitate tumor progression, ultimately leading to the development of immune tolerance.^8–10 Malignant and non-malignant immune cells and signaling molecules constitute the tumor microenvironment surrounded by blood vessels, disrupting therapeutic efficacy.^11–13 Immunotherapy is a prospective therapeutic intervention for carcinostasis by activating the immune system.^14,15 Recently, immunotherapy has demonstrated benefits in treating recurrent and advanced EC.^14–17

Microbiota refers to the collection of microorganisms in a particular community, whereas the genome of microorganisms is defined as a microbiome.^18–20 Evidence suggests that crosstalk between microbiota and the gut immune system is crucial.²¹ With further exploration of the gut microbiota (GM), the disrupted balance of the GM may influence various disease states, such as obesity, mental disorders, and autoimmune diseases.^22–24 GM can impact human health by regulating host immunity and metabolism.^25–28 Therefore, regulating host immunity by GM can prominently affect a variety of treatment efficacies and toxicities in cancer.¹⁸

Furthermore, it is recognized that the gut microbiota typically functions within a community rather than as individual bacteria. Increasing evidence suggested that microbial communities as regulated networks, reshape host immunity and tumor progression. Multi-omics analyses indicated that certain microbial communities are associated with immune infiltration in tumors. For instance, an integrated tumor-immune-microbiota network in colorectal cancer performed that, Firmicutes enriched communities are accompanied by enhanced cytotoxic T cell responses, while the Proteobacteria characteristics are related with immunosuppressive myeloid infiltration.²⁹ Moreover, microbiota derived metabolites such as inosine and bile acids, have been proven to regulate anti-tumor immunity through T cells and NKT cells.^25,30 Clinical studies further supported that the enrichment of Ruminococcaceae and Bacteroidetes species are correlated with improved responses to anti-PD-1 treatment in melanoma patients.^31,32 These findings demonstrated that microbiota induced immune regulation represents a conserved mechanism across various tumors. However, it is unclear that whether there are subtype-specific microbiota-immune pathways in EC.

Mendelian randomization (MR), a data analysis tool used in epidemiological studies to evaluate causal inferences, uses genetic variants strongly correlated with exposure factors as instrumental variables to assess the causal relationship between exposure factors and outcomes.^23–35 MR generally presents single nucleotide polymorphisms (SNPs) that serve as genetic variants to estimate the causal effect of exposure on outcomes with less susceptibility to environmental confounders and reverse causality.³³ Single-sample MR data are derived from the same individual, whereas two-sample MR uses large-scale GWAS data from independent study populations.^36,37 Large-scale summary statistics can be used to analyze the relationship between the GM, immune cells, and EC, enhancing the statistical power of the two-sample MR Analysis.

In this study, comprehensive MR Analysis was performed to explore the causal effects of GM, immune cells, and various EC types, including EC and EC with endometrioid and non-endometrioid histology. Next, we discuss whether immune cells serve as mediators from GM to EC.

Materials and Methods

Study Design

Bidirectional two-sample univariable Mendelian randomization (UVMR) was applied to explore the causal association between GM and the three types of EC: EC overall, endometrioid histology, and non-endometrioid histology. Subsequently, two-step MR analysis was performed to determine whether immune cell traits mediated these causal associations (Figure 1).

Figure 1 Overview flow of the two-sample univariable Mendelian randomization (UVMR) and mediation analysis design. GM, gut microbiota; EC, endometrial cancer; SNPs, single nucleotide polymorphisms; IVs, instrumental variables. β0 (total effect): The causal function of GM on EC; β1 (direct effect A): The causal function of GM on immune cell traits; β2 (direct effect B): The causal function of immune cell traits on EC; β (mediating effect) = β1 (Direct effect A) × β2 (Direct effect B); mediated proportion = β (mediating effect) / β0 (total effect). All the bolds present the summary or generic terms of MR analysis. GWAS summary data (in bold): The summary of data sources; Gut microbiota (LifeLines Cohort Study) (in bold): The summary of gut microbiota; Mediator (in bold): The data sources of immune cell traits as mediators; Endometrial cancer (GWAS) (in bold): The summary of 3 types endometrial cancer from GWAS; Exposure variables (in bold): The exposure of MR study; Outcome variables (in bold): The outcomes of MR study; Mediating variables (in bold): The mediators of MR study; Mendelian randomized analysis (in bold): The methods used for MR analysis; Sensitivity analysis (in bold): The methods used for assessing sensitivity.

First, the causal relationship between the GM and EC was assessed with UVMR using three filters: SNPs should (1) be strongly associated with EC, (2) affect EC only through the causal effect of the GM, and (3) remove other potential confounders.³⁸ Subsequently, reverse MR analysis was conducted to screen for GMs with reverse functions in the EC. Second, the mediating effects of various immune cells in the corresponding GM and EC were evaluated and quantified using UVMR and Mendelian randomization. The above analysis followed the guidelines for Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STRBOE-MR) (Table S1).³⁹

Data Sources

GM data were summarized from the Dutch Microbiome Project (DMP), a GWAS of 7738 European individuals.⁴⁰ The project applied shotgun metagenomic sequencing of fecal samples, which identified 207 microbial taxa (5 phyla, 10 classes, 13 orders, 26 families, 48 genera, and 105 species).

A total of 731 immune-related genome-wide features with approximately 22 million variants in 3757 individuals of European ancestry were identified using GWAS Summary Statistics.⁴¹ Integrated GWAS summary Statistics, with classified numbers from GCST0001391 to GCST0002121 published in the GWAS directory.

Genetic summary data for the two-sample and mediated MR analyses were primarily obtained from the IEU Open GWAS project and included endometrial cancer (ebi-a-GCST90018838), endometrial cancer with endometrioid histology (ebi-a-GCST006465), and non-endometrioid histology (ebi-a-GCST006466). All data are publicly available as GWAS summary data and received ethical approval (Table 1). This Mendelian randomization study is based solely on publicly available summary statistics from previously published genome-wide association studies. All original studies obtained informed consent and ethical approval from their respective institutional review boards. This study is exempt from approval based on Article 32, Item 1, of the Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects, issued by China on 18 February 2023. As no new human participants were involved and no individual-level data were analyzed in the present study, ethical approval for this specific analysis was not required.

Table 1 Summary of Data Sources

Genetic Instrumental Variable (IVs) Selection

Effective MR analysis relies on three foundations: IVs should (1) be strongly associated with EC, (2) affect EC only through the causal effect of GM, and (3) remove other potential confounders.³⁸ The first step was to select SNPs associated with exposure with a significance threshold of 5e-08, which were further screened via linkage disequilibrium analysis (r² < 0.001; distance within 10,000 kb).⁴² Finally, F values [F=β/r² (square of the standard error)] were calculated to identify the strength of IV, retaining SNPs with an F value of >10.⁴³ The study assumptions were presented in Figure 2, and selected SNPs would view in Table S2.

Figure 2 Assumptions of the two-sample UVMR and mediation analysis. SNPs, single nucleotide polymorphisms. All the bolds present the summary or generic terms of MR analysis. Instrumental variables (in bold): SNPs as the instrumental variables in MR study; Exposure (in bold): Gut microbiota as the exposure of MR study; Outcome (in bold): Endometrial cancer as the outcomes of MR study; Mediator (in bold): Immune cell traits as the mediators of MR study.

Statistical Analysis

All statistical analyses were performed using the R 4.4.0 (https://www.r-project.org). The “VariantAnnotation” package, “ieugwasr” package, and “TwoSampleMR” package were used to perform UVMR. Among the five methods (“MR Egger,” “Weighted median,” “Inverse variance weighted (IVW),” “Simple mode,” and “Weighted mode”), IVW is the primary method of causal estimation; P < 0.05 was identified as a significant causal association. Moreover, heterogeneity and horizontal pleiotropy were based on the Cochran Q statistic and MR-Egger intercept of the IVW and MR-Egger methods. Heterogeneity and pleiotropy were identified by a p-value less than 0.05, which indicated unsustainable causality. The “leave-one-out” test was used to investigate the effects of possible uncorrelated SNPs.

Mediation Analysis

After a two-sample UVMR, GM and immune cells with significant causal relationships to various EC were selected for further mediation analysis. We determined whether the GM had causal effects on immune cells; if so, multiple MR analyses were conducted to explore the mediating effect of immune cells from GM to EC.

Results

Instrumental Variable Selection

Initially, 4031 GM-associated SNPs were identified by P < 1×10⁻⁵ with closely linked SNPs removed (single nucleotide polymorphisms), which served as the IV of 412 GM. R² and F values were calculated for the screened SNPs (r² < 0.001; F-values > 10), which were unlikely to be affected by weak instrument bias (Table S2). The same process was used to select SNPs of immune cell traits (Table S2).

Causal Effects of GM on Different Types of EC

The IVW method was used to evaluate the potential causal relationship between GM and the invasion of different EC (Figure 3B, D and F). Although other methods, including MRE, WMed, SMod, and WMod, did not show statistical significance, the estimated causal effect demonstrated a tendency similar to that obtained using the IVW method. There was no heterogeneity or horizontal pleiotropy in this MR analysis, as shown by Cochran’s Q statistic, MR-Egger intercept test, and MR-PRESSO test (Table S3). Furthermore, none of the SNP severely interfered with the overall effects of GM on EC.

Figure 3 Various microorganisms of 412 GM presented causal relationships with 3 different types of EC. (A) MR result of GM on overall EC; (B) MR-IVW result of GM on overall EC; (C) MR result of GM on endometrioid EC (type I EC); (D) MR-IVW result of GM on endometrioid EC (type I EC); (E) MR result of GM on non-endometrioid EC (type II EC); (F) MR-IVW result of GM on non-endometrioid EC (type II EC). The symbol “c_/o_/f_/g_/s_” demonstrated the class, order, family, genus, and species. All the bolds present the summary or generic terms of MR analysis. Bacteria (in bold): the microbiota that can impact EC progression; pvalue (in bold): the p value of each bacterium on EC through MR analysis, which less than 0.05; SNPs (in bold): the significant SNPs used for MR analysis; OR (95% CI) (in bold): odds ratio with 95% confidence interval. Abbreviations: MR, mendelian randomization; GM, gut microbiota; EC, endometrial cancer; IVW, inverse variance weighted method; OR, odds ratio; 95% CI, 95% confidence interval.

Overall EC

Genetic predictions of the genus Erysipelotrichaceae_noname (odds ratio (OR) = 1.157, 95% Confidence Interval (CI) [1.021, 1.316], P = 0.021), genus Dialister (OR = 1.192, 95% CI [1.017, 1.398], P = 0.030), species Dialister_invisus (OR = 1.236, 95% CI [1.055, 1.450], P = 0.009) and species Ruminococcus_torques (OR = 1.238, 95% CI [1.034, 1.482], P = 0.020) were associated with an increased risk of overall EC. Species Bacteroides faecis (OR = 0.931, 95% CI [0.868, 0.997], P = 0.042), species Bacteroides massiliensis (OR = 0.769, 95% CI [0.642, 0.922], P = 0.004), and species Ruminococcus obeum (OR = 0.830, 95% CI [0.702, 0.981], P = 0.029) were associated with a reduced risk of EC (Figure 3A and B).

Endometrioid EC (Type I EC)

For endometrioid EC, class Bacilli (OR = 0.880, 95% CI [0.801, 0.968], P = 0.008), family Bacteroidaceae (OR = 0.863, 95% CI [0.757, 0.985], P = 0.028), order Lactobacillales (OR = 0.905, 95% CI [0.820, 0.998], P = 0.046), species Eggerthella_unclassified (OR = 0.855, 95% CI [0.734, 0.996], P = 0.044), and species Alistipes_sp_AP11 (OR = 0.900, 95% CI [0.814, 0.996], P = 0.042) were protective factors for endometrioid EC. Species Aspergillus senegalensis (OR = 1.154, 95% CI [1.015, 1.311], P = 0.029) and species Holdemania_unclassified (OR = 1.109, 95% CI [1.003, 1.225], P = 0.044) were risk factors for endometrioid EC (Figure 3C and D).

Non-Endometrioid EC (Type II EC)

Focusing on non-endometrioid EC, species Bacteroides_stercoris (OR = 1.450, 95% CI [1.010, 2.082], P = 0.044) and species Lachnospiraceae_bacterium_5_1_63FAA (OR = 1.251, 95% CI [1.027, 1.525], P = 0.026) were associated with an increased risk of developing this type of EC, whereas species Ruminococcus_bromii (OR = 0.633, 95% CI [0.446, 0.900], P = 0.011) was associated with a decreased risk of EC (Figure 3E and F).

Reverse MR Analysis of GM on Different Types of EC

Next, a reverse MR analysis was conducted on the three EC subtypes and their causal GM. The results showed no obvious causal effects (p > 0.05, Table 2), except for species Lachnospiraceae_bacterium_5_1_63FAA, between the three EC types and the relevant GM using the IVW method of reverse analysis (Table 2). Therefore, species Lachnospiraceae_bacterium_5_1_63FAA should not be further analyzed in non-endometrioid EC.

Table 2 Reverse MR_IVW Analysis of Gut Microbiota on 3 Types of EC

Effect of Immune Cell Traits on Different Types of EC

The IVW method, as the primary assessment method, was used to explore the effect of immune cells on different types of EC, with causal effects similar to those of MRE, WMed, SMod, and WMod. In addition, the MR-Egger intercept test, MR-PRESSO test, and leave-one-out sensitivity analysis showed no heterogeneity or horizontal pleiotropy, and the removal of any single SNP did not significantly affect the overall influence of immune cells on EC (Table S3).

The analysis also revealed that protecting 6 immune cell traits and 14 genetically predicted immune cell traits enhanced the overall risk of EC (Figure 4A). Thirteen immune cell traits were associated with increased risk, whereas 14 traits were associated with reduced risk of endometrioid EC (Figure 4B). In addition, MR analysis showed that non-endometrioid EC were amplified by 19 immune cell traits and suppressed by 19 other immune cell traits (Figure 4C).

Figure 4 Various immune cell traits of 713 traits presented causal relationships with 3 different types of endometrioid cancer. (A) Results of MR with IVW method on causality between immune cells and overall EC. (B) Results of MR with IVW method on causality between immune cells and endometrioid EC. (C) Results of MR with IVW method on causality between immune cells and non-endometrioid EC. All the bolds present the summary or generic terms of MR analysis. Immune cells (in bold): the immune cell traits that can impact EC progression; pvalue (in bold): the p value of each bacterium on EC through MR analysis, which less than 0.05; SNPs (in bold): the significant SNPs used for MR analysis; OR (95% CI) (in bold): odds ratio with 95% confidence interval. Abbreviations: MR, mendelian randomization; EC, endometrioid cancer; IVW, inverse variance weighted method; AC, absolute cell counts; Treg, regulatory T cell; NKT, natural killer T cells; OR, odds ratio; 95% CI, 95% confidence interval.

Effect of GM on Immune Cell Traits

We have demonstrated the role of GM and immune cell traits in various EC subtypes, and the causation of GM on immune cells in the three EC subtypes was explored through MR Analysis (Table S4).

Overall EC

Overall, 7 GM and 20 immune cell traits were analyzed, of which species Dialister_invisus was a risk factor for CD25hi CD45RA-CD4 but not for Treg AC (OR = 1.222, 95% CI [1.026, 1.456], P = 0.089). Species Bacteroides_massiliensis was protected against CD62L-HLA Dr + monocyte AC (OR = 0.761, 95% CI [0.617, 0.939], P = 0.011). In addition, species Ruminococcus_obeum was protective against CD86+ myeloid DC AC (OR = 0.817; 95% CI [0.687, 0.971], P = 0.222). However, it is also dangerous for CD27 on IgD+ CD38-unsw mem (OR = 1.356, 95% CI [1.081, 1.701], P = 0.008) and CD27 on IgD-CD38BR (OR = 1.222, 95% CI [1.036, 1.443], P = 0.018).

Endometrioid EC (Type I EC)

MR analysis revealed seven GM and 27 immune cell traits. The results showed that class Bacilli was harmful to IgD+ CD38br % lymphocytes (OR = 1.172, 95% CI [1.034, 1.327], P = 0.013) and Transitional AC (OR = 1.175, 95% CI [1.038, 1.329], P = 0.011), whereas species Holdemania_unclassified was also dangerous relative to Transitional AC (OR = 1.201, 95% CI [1.055, 1.369], P = 0.006), and order Lactobacillales was a risk factor for CD39+ activated Treg %CD4 Treg (OR = 1.121, 95% CI [1.007, 1.247], P = 0.036).

Non-Endometrioid EC (Type II EC)

For non-endometrioid EC with three GM and 38 immune cell traits, species Ruminococcus_bromii for IgD-CD38dim % lymphocytes (OR = 1.217, 95% CI [1.008, 1.468], P = 0.041), CD8dim % leukocytes (OR = 1.243, 95% CI [1.025, 1.508], P = 0.027), and CD8br NKT % lymphocytes (OR = 1.223, 95% CI [1.010, P = 0.027) 1.482], P = 0.039) were detrimental; however, they were protective against CD25hi AC (OR = 0.772, 95% CI [0.632, 0.943], P = 0.011).

Mediation Analysis

The proportion of mediating effects was quantified by determining the ratio of indirect to direct effects after confirming the cause-effect relationship between GM, immune cell traits, and various EC types (Table 3).

Table 3 Mediation Effect of Various GM on 3 Types of EC via Immune Cell Traits

Overall EC

Species Ruminococcus_obeum affected EC through three different mediators: CD86+ myeloid DC AC (mediated effect β = −0.017, mediated proportion = 9.17%), CD27 on IgD+ CD38- unsw mem (β = 0.017, mediated proportion = −16.10%), and CD27 on IgD- CD38br (β = 0.031, mediated proportion = −16.80%) (Figure 5A). Notably, the mediated proportion of the mediator CD27 on IgD+ CD38- unsw mem and CD27 on IgD- CD38br was negative, indicating that exposure had the opposite effect on the outcome, as expected, through the mediator. This may be owing to 1) the reverse effect of the mediating variables. The mediating variable may have a reverse effect, and its regulation may lead to a reverse change in the outcome owing to various factors such as biological mechanisms, environmental factors, or individual differences. 2) Other unconsidered variables: Other unidentified variables may influence the mediating effect. In addition, statistical errors during calculations may have contributed to this result.

Figure 5 Mediation effect of various GM on EC with traditional definition via immune cell traits. (A) Mediation analysis of Dialister_invisus on EC via CD25hi CD45RA- CD4 not Treg AC. (B) Mediation analysis of Bacteroides_massiliensis on EC via CD62L- HLA DR++ monocyte AC. (C) Mediation analysis of Ruminococcus_obeum on EC via CD86+ myeloid DC AC. The symbol “c_/o_/f_/g_/s_” demonstrated the class, order, family, genus, and species. All the bolds present the summary or generic terms of MR analysis. Exposure (in bold): The exposures impact the outcomes in MR study; outcome (in bold): The outcomes impacted by exposures in MR study; nsnp (in bold): the significant SNPs used for MR analysis; method (in bold): The methods used for MR analysis; pval (in bold): the p value of each exposure on outcomes through 5 different methods, values less than 0.05 will be marked in bold; OR (95% CI) (in bold): odds ratio with 95% confidence interval. Abbreviations: EC, endometrioid cancer; AC, absolute cell counts; Treg, regulatory T cell; OR, odds ratio; 95% CI, 95% confidence interval.

The results showed that species Bacteroides_massiliensis affected EC via CD62L- HLA DR++ monocyte AC (β = −0.033, mediated proportion = 12.60%) with a negative mediating role (Figure 5B). CD25hi CD45RA- CD4, not Treg AC, positively mediated species Dialister_invisus acting on EC (β = 0.025, mediated proportion = 11.80%) (Figure 5C).

Endometrioid EC (Type I EC)

The Class Bacilli protected endometrioid EC through IgD+ CD38br % lymphocytes (β = −0.010, mediated proportion = 7.49%) and transitional AC (β = −0.011, mediated proportion = 8.48%) with negative mediations (Figure 6A and B). Transitional AC also mediated the influence of species Holdemania_unclassified influence on endometrioid EC (β = −0.008, mediated proportion = −7.43%), which was illogical. In addition, the order Lactobacillales was protective against endometrioid EC by CD39+ activated Treg %CD4 Treg (β = −0.010, mediated proportion = 10.20%) (Figure 6C).

Figure 6 Mediation effect of various GM on EC with endometrioid histology via immune cell traits. (A) Mediation analysis of Bacilli on EC via IgD+ CD38br %lymphocyte. (B) Mediation analysis of Bacilli on EC via Transitional AC. (C) Mediation analysis of Lactobacillales on EC via CD39+ activated Treg %CD4 Treg. The symbol “c_/o_/f_/g_/s_” demonstrated the class, order, family, genus, and species. All the bolds present the summary or generic terms of MR analysis. Exposure (in bold): The exposures impact the outcomes in MR study; outcome (in bold): The outcomes impacted by exposures in MR study; nsnp (in bold): the significant SNPs used for MR analysis; method (in bold): The methods used for MR analysis; pval (in bold): the p value of each exposure on outcomes through 5 different methods, values less than 0.05 will be marked in bold; OR (95% CI) (in bold): odds ratio with 95% confidence interval. Abbreviations: EC, endometrioid cancer; AC, absolute cell counts; Treg, regulatory T cell; OR, odds ratio; 95% CI, 95% confidence interval.

Non-Endometrioid EC (Type II EC)

Species Ruminococcus_bromii reduced the risk of non-endometrioid EC by IgD- CD38dim %lymphocyte (β = 0.042, mediated proportion = −9.25%), CD25hi AC (β = 0.074, mediated proportion = −16.10%), CD8dim % leukocytes (β = 0.055, mediated proportion = −12.10%), and CD8br NKT % lymphocytes (β = −0.036, mediated proportion = 7.87%), although only CD8br NKT % lymphocytes were logical (Figure 7).

Figure 7 Mediation effect of Ruminococcus_bromii on EC with non-endometrioid histology via CD8br NKT %lymphocyte. The symbol “c_/o_/f_/g_/s_” demonstrated the class, order, family, genus, and species. All the bolds present the summary or generic terms of MR analysis. Exposure (in bold): The exposures impact the outcomes in MR study; outcome (in bold): The outcomes impacted by exposures in MR study; nsnp (in bold): the significant SNPs used for MR analysis; method (in bold): The methods used for MR analysis; pval (in bold): the p value of each exposure on outcomes through 5 different methods, values less than 0.05 will be marked in bold; OR (95% CI) (in bold): odds ratio with 95% confidence interval. Abbreviations: EC, endometrioid cancer; AC, absolute cell counts; NKT, natural killer T cells; OR, odds ratio; 95% CI, 95% confidence interval.

Discussion

Intestinal microbiota and its metabolites affect several pathophysiological processes covering host metabolism and immune response, known as the “second endocrine organ”.⁴⁴ Previous studies have verified that the regulation of host immunity is closely interrelated with the microbiota, and a disrupted balance would motivate the occurrence and deterioration of diseases;^45,46 hence, it is crucial to equilibrate the complex communions between them. In this study, we conducted a comprehensive large-scale two-sample MR analysis based on the LifeLine Biobank and GWAS databases, followed by reverse MR analysis for each of the three types of EC. There were 17 causal relationships, including 7 GM species and overall EC, 7 species and endometrioid EC, and 3 species and non-endometrioid EC, which were evaluated by predicting 158 SNP loci. Therefore, the three EC subtypes were primarily causally related to the phyla Firmicutes and Bacteroidetes, and type I EC presented a causal correlation with the Phylum Actinobacteria. This indicates that there are differences in the microbiome-associated profiles among different EC subtypes. Type I (estrogen-related, highly differentiated endometrial adenocarcinoma) may be influenced by specific Actinobacteria, whereas type II (non-endometrial, high-grade cancer) does not show such a relationship.

Next, a two-sample MR analysis was conducted on immune cell traits and different types of EC, and causal relationships based on various SNPs are discussed. Twenty immune traits were associated with typical EC, 27 were causally linked to endometrioid EC, and non-endometrioid EC showed a causal correlation with 38 immune cells, further demonstrating the heterogeneity of the immune microenvironment among the different EC subtypes.

Finally, through two-step MR mediation analysis, we identified seven significant “GM-immune-EC” causal chains and quantified the proportion of the immune mediation effect in the total effect. These results support the hypothesis from a genetic perspective that GM may affect EC occurrence by regulating the immune system.

Notably, types I and II may have unique pathways in their pathogenic mechanisms. Owing to the differences in hormone dependence and molecular characteristics between type I and type II EC, it is impossible to present a common “ GM-immune-EC” pathway for both subtypes. The partially common associations observed in this study were more likely the result of indirect immune regulation. Therefore, caution should be exercised when discussing these results, emphasizing that this study is a hypothesis generating exploration. The causal inferences from this study require further biological validation and in-depth mechanistic exploration. Moreover, targeted research and clinical studies should be conducted on different EC subtypes with different pathologies to clarify their GM and immune mechanisms.²

Overall EC

Bacteroides massiliensis can reduce the risk of EC by repressing the immune cell trait “CD62L HLA DR++ monocyte AC”. Bacteroides regulate the immune system to maintain homeostasis, and their metabolites maintain immune system stability.⁴⁷ Bacteroides are the major producers of short-chain fatty acids (SCFAs) in the gut,⁴⁸ which are important for maintaining microecological equilibrium, mainly in the form of acetic and propionic acids. Acetate and propionate are potent anti-inflammatory mediators that inhibit the release of pro-inflammatory cytokines from neutrophils and macrophages.⁴⁹ The anticancer effects of propionic acid on apoptosis have been described in human colon cancer cells.⁵⁰ The anti-tumor effects of cytotoxic T lymphocyte antigen 4 (CTLA-4) blockers depend on different Bacteroides species, with specific T cell responses to B. thetaiotaomicron or B. Fragilis affecting the efficacy of CTLA-4 blockers in mice and patients.⁵¹ In addition, the HLA-DR phenotype is closely related to anti-tumor immunotherapy, with CD62L as a marker of monocyte activation,^52,53 and CD62L HLA DR++ monocyte were analyzed to facilitate EC progression. Nevertheless, the immunosuppressive cytokines interleukin (IL)-10 and pro-inflammatory factors IL-6 and TNFα are negatively correlated with HLA-DR in B-cell non-Hodgkin lymphoma.^54,55 Mengos et al found that monocytes with reduced or no HLA-DR expression are crucial mediators of tumor-induced immunosuppression and negatively affect programmed death-1 (PD-1) and CTLA-4 checkpoint inhibition,^56–62 chimeric antigen receptor T-cell (CAR-T) immunotherapy,^53–65 cancer vaccines,^66–70 and hematopoietic stem cell transplantation.^71–74 Previous studies have presented conflicting views on our analysis; therefore, further exploration is necessary to investigate the effect of CD62L-HLA DR++ monocytes on EC.

Ruminococcus obeum, which belongs to the genus Blautia, protects EC via the trait “CD86+ myeloid DC AC”. GM analysis of immune checkpoint inhibitor (ICI)-treated patients showed that the high diversity and presence of immunogenic bacteria, such as Ruminococcus, resulted in more significant CD8+ T cells and CD4+ Th1-dependent anti-tumor responses, leading to better outcomes.^31,32,75,76 However, Xu et al found that Blautia obeum was enriched in patients who were nonreactive to anti-PD-1 and ICI treatments and may have antibiotic properties along with antibiotic resistance.^77,78 Therefore, the distinct mechanism of action of Ruminococcus obeum on EC warrants further investigation. In addition, CD86 expression is considered a poor prognostic indicator and is down-regulated by transendocytosis of CTLA-4.^78,79 Blocking CD28: CD80/CD86 in vivo re-sensitizes multiple myeloma cells to chemotherapy and significantly reduces the tumor load.⁸⁰ Combined with MR analysis, Ruminococcus obeum may inhibit CD86+ on myeloid DC to enhance immunotherapy, thereby restraining EC.

Dialister invisus is a risk factor for EC functioning through the immune cell trait “CD25hi CD45RA- CD4, not Treg AC.” Enriching Dialister invisus, isolated from the human oral cavity, indicates a high risk and rapid progression of tumors in colorectal cancer and increases the risk of HPV-infected cervical cancer in female reproductive system tumors.^81–85 However, Byrd et al found that dialister status was associated with higher survival in MSI-H colorectal tumors.⁸⁶ CD4+ CD25hi T cells maintain immune tolerance to autoantigens in various cancers, including ovarian and cervical.^87–89 Typically, T cells can be divided into CD45RA+ initial and CD45RA- memory subtypes. Activated CD45RA- T cells inhibit the anti-tumor function of CD8+ T cells via IL-10 secretion and intercellular contacts. This facilitates immunosuppression and gastric cancer progression.^90–92 Tassi et al confirmed that CD45RA-T lymphocytes were increased in the epithelial ovarian tumor microenvironment.⁹³ Therefore, together with our analysis, Dialister invisus down-regulates the anti-tumor function of CD8+ T cells by stimulating the immunosuppressive trait “CD25hi CD45RA-CD4 not Treg,” following accelerated tumor progression of EC.

Endometrioid EC (Type I EC)

Class Bacilli play a protective role in endometrioid EC by colonizing the gastrointestinal tract.⁹⁴ The regulates tumor pathophysiological processes, and their secretions exhibit bacteriological and anti-tumor properties. Typically, SCFAs from Bacilli are beneficial for gut peristalsis and secretion and inhibit tumor proliferation by inducing apoptosis and controlling epigenetic modification.^94–98 Bacteriocins and other secretions derived from bacilli can suppress phospholipase A2, downregulate pro-inflammatory cytokines, and upregulate anti-inflammatory cytokines.^99,100 Extracellular vesicles (EVs) of bacilli induce apoptosis in HepG2 cells by increasing the expression ratio of bax/bcl-2.¹⁰¹ Bacilli stimulate the production of insulin-like growth factor 1 (IGF-1), participates in the regulation of blood glucose and lipids, and is closely associated with the pathogenesis of type I EC.⁹⁷ In a clinical study on colorectal cancer, the abundance of bacilli in the metastatic group was significantly decreased compared with that in tumor patients,¹⁰² indicating that it may restrain tumor invasion. CD38, a type II transmembrane glycoprotein, is a crucial metabolic enzyme located on the cell surface and its products are vital during immune regulation.^103,104 CD38 is an anti-tumor therapeutic target involved in adenosine formation and exerts remarkable immunosuppressive effects on the solid tumor microenvironment.^103,105 Thavaneswaran et al reported that patients with EC with high CD38 + expression in peripheral tertiary lymphoid structures (TLSs) had favorable survival outcomes during chemotherapy.^106,107 Our MR analysis showed that lymphocytes with lgD+ and bright CD38 are protective against endometrioid EC. Bacilli play a protective role by enhancing lymphocyte function. However, their specific mechanisms require further investigation.

In particular, the order Lactobacillales belongs to the Class Bacilli; hence, its metabolites and secretions, such as SCFAs, exhibit similar antitumor mechanisms. Bactericin mainly disrupts the membrane intimal potential, leading to uncontrolled ion leakage and cell death.^108,109 Moreover, these bacteria can reduce the concentration of soluble bile salts in faeces, thereby neutralizing the cancer-promoting effects of bile acids, including DNA damage and apoptosis.^109–112 Importantly, the Lactobacillales is the main component of a healthy vaginal microecosystem, and its disturbance promotes inflammation and cancer progression.^113–115 In contrast, CD39+-activated Treg %CD4 Treg mainly play an anti-inflammatory role. CD39, an extranuclear triphosphate diphosphate hydrolase-1, can hydrolyze ATP into AMP, obstructing inflammatory signal transduction of extracellular ATP.^116–118 Maddaloni et al showed that lactic acid bacteria can deliver IL-35 in collagen-induced arthritis in mice, thus stimulating CD39+ CD4+ Tregs to release increased IL-10, exerting anti-inflammatory and immune effects.¹¹⁸ Furthermore, lactic acid bacteria increase the proliferation of CD39+ Tregs in mouse allergic asthma models, following the regulation of the inflammatory response induced by immune disorders.¹¹⁹ Therefore, we speculate that Lactobacillales activates and promotes the proliferation of CD39+-activated Treg %CD4 Tregs by delivering special inflammatory mediators, such as IL-35, in endometrioid EC, indicating their protective role.

Non-Endometrioid EC (Type II EC)

Ruminococcus bromii is beneficial for patients with non-endometrioid EC. A colon cancer study proposed the microbiota characteristics of R. bromii as a reliable indicator, combined with the immune rejection constant, to create a score for evaluating survival prognosis, which contributed to the discovery of personalized therapy.²⁹ Notably, colon cancer has a genetic phenotype similar to that of non-endometrioid EC: DNA mismatch repair (MMR) defects or microsatellite instability (MSI).¹²⁰ The increased abundance of R. bromii following symbiotic treatment (MS-20) combined with anti-PD-1 treatment positively correlated with reduced tumor load and CD8+ T cell infiltration in xenograft mouse models.¹²¹ Additionally, rumen cocci can bind to castalagin and promote anti-cancer responses; castalagin rich in R. bromii influences the efficacy of anti-PD-L1 immunotherapy and increases the ratio of CD8 + / FOXP3 + CD4 + in the tumor microenvironment.¹²² NKT cells specifically assemble T-cell receptor (TCR) and natural killer (NK) cell receptor on the surface, which produce numerous cytokines and play a similar cytotoxic role as NK cells.^123–128 This enhances the immune response and mediates the inhibition of tumor growth in various cancers, including liver, ovarian, and colon cancers.^30,129–132 In mismatch repair-deficient EC, a large infiltration of CD8+ and NKT cells is a marker for better survival prognosis, indicating a favorable immune microenvironment.^132,133 Therefore, combined with the MR analysis, R. bromii may improve the tumor immune response and the efficacy of immunotherapy by increasing CD8+ NKT cells, thus playing a protective role. Recently, universal CAR-engineered NKT (UCAR-NKT) cells were developed and demonstrated powerful anti-tumor efficacy against blood and solid tumors in vitro and in vivo, with multiple tumor-targeting mechanisms.^134,135 These cells regulate the tumor microenvironment by selectively depleting immunosuppressive macrophages. Therefore, based on this analysis and previous research, targeting R. bromii and the immune microenvironment may help improve the prognosis of non-endometrioid EC.

Clinical Implication

However, it is necessary to consider whether the alteration of the gut microbiota actively drive the EC progression, or instead present as a consequence of the EC presence.¹³⁶ If specific microbial dysbiosis is the cause of the EC development, the microbiota can serve as a therapeutic target. For instance, interventions including probiotics and diet could be attempted to prevent the progression of EC.^136,137 On the other hand, if the tumors induce the observed microbiota changes, the gut microbiota may tend to serve as a diagnostic or prognostic marker for the disease state, rather than a treatment therapy. Our bidirectional MR results support the former point, we found non-significant evidence of reverse causality from EC to microbiota, while significant causal effects were shown from gut microbiota to EC. It indicated that gut microbiota perform a directional impact on the development of EC, rather than a resulting phenomenon. And other recent studies revealed that altering the unfavorable microbiota can affect tumor outcomes.^122,138 Therefore, the gut microbiota is not only a biomarker, also a potential therapeutic target for EC, which need experimental supports.

Besides, our analysis emphasized the gut microbiota-immune axis as a promising new therapy or biomarker for improving the diagnosis and risk forecast of EC. Firstly, from the perspective of prevention strategies, many high-risk factors for EC are closely related to intestinal microbiota, such as obesity and metabolic disorders. Hence, the gut microbiota regulation may become a new therapy to reducing the EC risk. For example, correcting the microbiota imbalance through dietary intervention and probiotics, is expected to improve the pro-tumor inflammation and metabolism, thereby preventing especially type I EC.^139–141 Secondly, in terms of immunotherapy, the immunogenicity of EC especially type II suggests that patients may benefit from immune checkpoint inhibitors (ICB), and our results further supported that the gut microbiota may affect the efficacy of immunotherapy.¹³⁹ Studies have shown that patients with different compositions of the gut microbiota respond significantly differently to immunotherapy. Therefore, it is prospective to incorporate microecological regulation into EC treatment. And it is effective to enhance the efficacy of immunotherapy in patients with advanced or recurrent EC, combining with probiotic preparations or fecal microbiota transplantation (FMT) to improve the microbiota composition.¹³⁹ At the same time, the gut microbiota and its metabolites have the potential to serve as biomarkers for predicting the efficacy of immunotherapy in EC patients.^140,142,143 In summary, the gut microbiota-immune axis provides a new perspective for the prevention and treatment of EC, microecological interventions can both regulate the EC progression and improve the immune status to enhance the therapeutic effect.¹⁴³

Strengths of This Study

The advantage of this study lies in the first application of a comprehensive causal mediation analysis of the EC subtypes. It systematically integrated bidirectional two-sample MR and mediation analyses to explore the potential causality between GM and EC mediated by immune cells. We performed comparative analyses for overall, type I, and type II EC, revealing similarities and differences in the risks of GM and immune-mediated patterns among the different subtypes. Moreover, our study utilized large-scale GWAS data and a strict two-sample MR design to identify multiple robust causal associations between microbiota, immunity, and EC, and excluded major confounding factors through multiple sensitivity tests. These findings provide new insights into the biological basis of EC heterogeneity and support evidence for the development of treatments targeting the GM and the immune microenvironment.

Limitations of This Study

This study has some limitations. First, the data were derived mainly from large-scale European population databases. The homogeneity of race and region may have limited the results to other populations. Second, EC occurrence and development are the outcomes of the combined effects of multiple factors. Our genetic instrumental variable analysis cannot consider the influence of non-genetic factors such as the environment and lifestyle. In addition, causal inference in the MR analysis relies on this hypothesis. Although we controlled for horizontal pleiotropy, unobserved biases may still be present. Finally, our inference lacks in vivo and in vitro verification and the mechanistic explanation remains speculative. Further research should combine clinical cohorts and functional experiments to confirm and deepen our understanding of microbiota-immune interactions, especially for different EC subtypes.

Conclusion

Through a two-step MR and mediation analysis, our findings provide robust evidence that immune traits act as significant mediators in connecting the causal pathway of GM to EC development. Importantly, this study highlights distinct causal relationships and immune-mediated mechanisms across the three major EC subtypes (overall, endometrioid, and non-endometrioid). These subtype-specific insights into the gut-immune-cancer axis provide novel perspectives for developing therapeutic strategies targeting the GM and immune microenvironment in different forms of EC.