The causal role of plasma metabolites in breast cancer susceptibility mediated by immune cells: A two-sample Mendelian randomization analysis Authors: Zhe Sun, Fei Ge, Wenjing Duan, Jinxian Qian, Liang Chen Published: 2026 ABSTRACT Breast carcinoma (BC) is a significant oncological issue, categorized based on estrogen receptor (ER) status into ER-positive and ER-negative types. The ER-positive variant relies on hormonal signaling, whereas the ER-negative form is characterized by increased aggressiveness. There exists a potential interplay between plasma metabolites and immune effectors that may influence BC susceptibility, although the underlying mechanisms remain largely undefined. This study employs Mendelian randomization (MR) to elucidate the causal pathways through which plasma metabolites affect both ER-positive and ER-negative BC susceptibility via immune mediators. In the context of ER-positive BC, sphingolipid metabolites (e.g., Sphingomyelin [d18:1/17:0] concentrations, odds ratio [OR] = 1.299, 95% confidence interval [CI]: 1.140–1.480) exhibited positive associations, whereas bile acid metabolites (e.g., Glycochenodeoxycholate 3-sulfate concentrations, OR = 0.876, 95% CI: 0.818–0.938) displayed negative correlations; immune phenotypes such as CD8br % leukocyte (OR = 1.165, 95% CI: 1.005–1.351) were positively linked. For ER-negative BC, the phosphate-to-tryptophan ratio (OR = 1.371, 95% CI: 1.047–1.796) showed a positive correlation, while indolepropionate (OR = 0.762, 95% CI: 0.647–0.896) had an inverse relationship; CD28 on CD39 + CD8br (OR = 1.052, 95% CI: 1.016–1.090) demonstrated a positive association. Mediation analyses revealed pathways: In ER-positive BC, 2R,3R-dihydroxybutyrate was mediated through EM DN (CD4−CD8−) %T cell (intermediation fraction 23.4%), and Sphingomyelin via CD38 on transitional cells (7.41%); in ER-negative BC, Dihomo-linoleoylcarnitine was mediated through human leukocyte antigen-DR on CD14+ CD16+ monocytes (9.98%), and homocitrulline through human leukocyte antigen-DR on plasmacytoid dendritic cells (12.7%). Plasma metabolites significantly influence susceptibility to different BC variants by modifying immune effector functions. This study provides genetic evidence supporting interventions targeting the metabolic-immune interface to improve prevention and management strategies. Utilizing publicly available genome-wide association study datasets, we examined 1440 plasma metabolites (identifiers GCST90199621 to GCST90201020), 731 immune phenotypes (ebi-a-GCST90001391 to ebi-a-GCST90002121), and the FinnGen R12 BC cohorts. Instrumental variables were selected based on criteria of P < 5 × 10−8/10−7, r2 < 0.001, and F-statistic > 10. We employed bidirectional two-sample MR, primarily using inverse-variance weighted methods, supplemented by MR-Egger and other techniques, to assess causal relationships and intermediary roles. 1. Introduction Breast carcinoma (BC) is a prevalent neoplastic condition characterized by significant heterogeneity. It is classified into estrogen receptor (ER)-positive and ER-negative groups based on the presence of ERs. The ER-positive variant typically progresses at a slower rate and is reliant on hormonal signals, while the ER-negative variant tends to advance more rapidly and is associated with increased invasiveness. According to worldwide data, breast cancer accounted for around 2.3 million incident cases and 670,000 fatalities in 2022, accompanied by significant health complications such as impaired movement after surgery and ongoing discomfort.[1] Forecasts point to a rise exceeding 3.1 million fresh detections and 1.1 million fatalities by the year 2050. In addition to personal well-being effects, this malignancy generates profound community and financial strains, driving up medical expenditures, lowering overall living standards, and imposing severe economic hardship on as many as 48% of affected individuals globally. Work productivity is notably impaired owing to disease-linked deaths and illnesses, leading to an anticipated worldwide fiscal impact nearing 2 trillion international dollars (at 2017 values) over the period from 2020 to 2050, akin to imposing a yearly average levy of 0.043% on the planet’s GDP.[2] Our investigation distinctively applies a two-sample Mendelian randomization (MR) technique to explore the interplay between metabolites and immune factors in breast cancer development, bridging an essential void in grasping causal links to enhance strategies for avoidance and management. Current therapeutic strategies encompass excisional methods (such as lumpectomy or mastectomy), radiation therapy, cytotoxic drugs (including taxanes and anthracyclines), hormonal therapies (like tamoxifen), and precision medicine (for instance, trastuzumab). These treatment modalities facilitate targeted management of both localized and metastatic diseases; however, they are limited by complications associated with surgery, cardiac effects from radiation, and hematologic suppression induced by chemotherapy. Despite achieving a 5-year survival rate exceeding 80%, challenges such as therapeutic resistance and disease relapse persist. The underlying mechanisms contributing to breast cancer remain incompletely understood, resulting in a variety of clinical outcomes. Consequently, a comprehensive investigation into susceptibility factors and intermediary processes is essential for the development of preventive strategies and personalized treatment approaches. Plasma metabolites exhibit extensive interactions with immune effectors, which modulate their activation and functions through alterations in cellular bioenergetics. For example, short-chain fatty acids can stimulate G-protein-coupled receptors or induce epigenetic modifications, thereby regulating inflammatory pathways and the differentiation of immune cells – such as promoting anti-inflammatory macrophage transformations or inhibiting antigen processing by dendritic cells (DCs).[3] Recent studies have linked plasma metabolites to breast cancer susceptibility; notably, a prospective metabolomic study identified alterations in nucleosides and various metabolites as potential biomarkers for early detection. These findings suggest that plasma metabolites may play a role in the initiation and progression of breast cancer by reconfiguring bioenergetic processes.[4] Furthermore, plasma metabolites might indirectly influence breast cancer susceptibility by modulating the capabilities of immune effectors. Immune effectors can exert dual effects on breast cancer development, potentially limiting tumor growth or facilitating immune evasion. Mechanistic pathways involve metabolites that modify the cytotoxic abilities of immune cells, leading to immunological changes in breast cancer. For instance, macrophages can secrete cytokines that remodel the tumor microenvironment, enhancing endocrine signaling in ER-positive tumors or increasing inflammation in ER-negative tumors, thereby promoting tumor growth and dissemination.[4] Metabolites such as cortisol have been associated with increased breast cancer risk, possibly due to the suppression of immune effectors.[5] This evidence highlights the continuum between plasma metabolites and immune effectors in breast cancer, warranting further investigation. In summary, existing research has initially revealed potential pathways through which plasma metabolites influence breast cancer via immune effectors; however, the intricate mechanisms and specific variations require further clarification. This study aims to examine the effects of plasma metabolites on the susceptibility to both ER-positive and negative breast cancer, mediated by immune effectors. The hypothesis posits that plasma metabolites regulate immune effector functions, thereby indirectly increasing or decreasing the risk of developing either subtype of breast cancer. Employing MR, this approach offers enhanced causal inference compared to traditional epidemiological methods by mitigating issues related to reverse causation and confounding, thereby providing stronger evidence and translational potential. 2. Materials and methods 2.1. Study design The data for this investigation were exclusively sourced from publicly accessible genome-wide association study (GWAS) databases, encompassing 1440 plasma metabolites, 731 immune effector profiles, and summaries pertaining to hormone receptor-positive (ER-positive) and hormone receptor-negative (ER-negative) variants. Adherence to the STROBE-MR guidelines for MR was strictly maintained, utilizing genetic polymorphisms as instrumental variables (IVs). The study adheres to 3 essential principles: a strong association between plasma metabolites and osteoarthritis; independence from confounding variables; and an exclusively indirect influence on outcomes through the exposure, without any direct effects. To enhance credibility, the genomic data for exposure and outcomes were sourced from separate GWASs. A two-sample bidirectional MR analysis was conducted to investigate the causal relationships between plasma metabolites and osteoarthritis, while mediation MR was utilized to evaluate the mediating role of immune effectors in the effects of plasma metabolites on breast cancer. The methodological framework is illustrated in Figure 1. Two-sample bidirectional MR and mediation analysis framework. Schematic illustrating the two-sample bidirectional Mendelian randomization (MR) and mediation analysis framework using 1400 serum metabolites as exposure,[6] 731 immune cell phenotypes as potential mediators,[7] and estrogen receptor-negative/positive breast cancer (BREAST_ERNEG and BREAST_ERPLUS) as outcomes from the FinnGen R12 cohort. SNP = single nucleotide polymorphism. 2.2. Data sources Data acquisition involved the following: Exposure variables: GWAS data for plasma metabolites obtained from the Catalog GWAS repository, with codes ranging from GCST90199621 to GCST90201020. Exposure variables: GWAS data for plasma metabolites obtained from the Catalog GWAS repository, with codes ranging from GCST90199621 to GCST90201020. Mediators: Immune effector data sourced from the IEU-OPEN GWAS repository, identified by codes ebi-a-GCST90001391 to ebi-a-GCST90002121. Mediators: Immune effector data sourced from the IEU-OPEN GWAS repository, identified by codes ebi-a-GCST90001391 to ebi-a-GCST90002121. Outcomes: Data on hormone receptor-positive(ER-positive) and hormone receptor-negative (ER-negative) variants from the FinnGen R12 project (https://www.finngen.fi/en). Detailed datasets are presented in Attachment (File S1, Supplemental Digital Content). Outcomes: Data on hormone receptor-positive(ER-positive) and hormone receptor-negative (ER-negative) variants from the FinnGen R12 project (https://www.finngen.fi/en). Detailed datasets are presented in Attachment (File S1, Supplemental Digital Content). 2.3. Selection of IVs The criteria for selecting IVs included single nucleotide polymorphisms (SNPs) related to plasma metabolites and immune effectors with a significance threshold of P < 5 × 10−8, and for breast cancer, a threshold of P < 5 × 10−7. Linkage disequilibrium was addressed by excluding SNPs with r2 < 0.001 within a 10 Mb distance to ensure independence. IVs were finalized by selecting SNPs exhibiting F-statistics above 10 (with average values ranging from 25 to 35 across instruments, varying by exposure data source) and minor allele frequencies >0.01. This ensured strong and reliable proxies, effectively reducing the risk of bias from weak instruments, in line with standard protocols for MR studies. 2.4. Statistical analysis strategy To enhance reliability and counteract possible distortions, multiple MR techniques were implemented. The inverse-variance weighted method served as the core analytical tool, offering optimal precision when all variants meet validity criteria. Complementary approaches included MR-Egger for identifying and adjusting directional pleiotropic effects; weighted mode for diminishing outlier effects via emphasis on the predominant estimate; weighted median for yielding stable results despite invalidity in up to half of the variants; MR-PRESSO for spotting and rectifying horizontal pleiotropy by eliminating anomalous data points; and robust adjusted profile score to bolster resilience to pleiotropic influences and variability in models with excess identification. For solitary SNP evaluations, the Wald ratio determined causal associations from individual proxies; weighted median was favored in scenarios involving suspect invalid variants owing to its robustness against such flaws; and weighted mode was utilized to attenuate distortions from anomalies in asymmetric data sets. These techniques were chosen for their synergistic capabilities in managing breaches of MR prerequisites, adhering to the STROBE-MR framework for clear documentation in MR investigations. Variability among estimates was examined via Cochran Q test, where a P-value below .05 denoted notable inconsistency. Pleiotropic tendencies were probed using the MR-Egger intercept, with a P-value under .05 signaling potential directional bias. Result credibility hinged on lacking clear signs of inconsistency or pleiotropy, supported by additional checks to verify alignment between approaches. A schematic diagram in Figure 1 depicts the sequence for obtaining and handling data, encompassing GWAS retrieval from repositories (such as the IEU OpenGWAS platform), variant identification and alignment, proxy confirmation (including F-statistic thresholds), core MR procedures, and validation assessments. Mediation was analyzed through a two-phase MR approach: Direct MR: Assessing the relationship between plasma metabolites/immune effectors and breast cancer. Direct MR: Assessing the relationship between plasma metabolites/immune effectors and breast cancer. Inverse MR: Evaluating the influence of breast cancer on plasma metabolites/immune effectors. Inverse MR: Evaluating the influence of breast cancer on plasma metabolites/immune effectors. Exclusion of reverse-significant items from direct outcomes. Exclusion of reverse-significant items from direct outcomes. Calculation of the mediation fraction using βm = (βem × βmo)/βeo, where βm represents the fraction, βem denotes the effect of plasma metabolites on immune effectors, βmo indicates the effect of immune effectors on breast cancer, and βeo signifies the overall effect of plasma metabolites on breast cancer. The delta method was utilized to derive confidence intervals (CIs), with P < .05 indicating statistical significance. Analyses were performed using R version 4.5.1 (R Foundation) with the TwoSampleMR and related packages. Calculation of the mediation fraction using βm = (βem × βmo)/βeo, where βm represents the fraction, βem denotes the effect of plasma metabolites on immune effectors, βmo indicates the effect of immune effectors on breast cancer, and βeo signifies the overall effect of plasma metabolites on breast cancer. The delta method was utilized to derive confidence intervals (CIs), with P < .05 indicating statistical significance. Analyses were performed using R version 4.5.1 (R Foundation) with the TwoSampleMR and related packages. 3. Results Utilizing two-sample bidirectional MR, this investigation explored causal relationships with plasma metabolites as initiators, immune effectors as intermediaries, and ER-positive breast cancer as the outcome. By leveraging GWAS data, the interactions between metabolites and immune effectors were examined, along with their individual contributions to cancer susceptibility. The objective was to identify intermediary pathways, providing genetic evidence for novel preventive strategies and highlighting the interconnected roles of immune modulation and bioenergetic reconfiguration in tumorigenesis. 3.1. Two-sample bidirectional MR results 3.1.1. Breast cancer (ER-positive) results 3.1.1.1. Causal influences of plasma metabolites on ER-positive BC An investigation into the causal relationships between plasma metabolites and ER-positive BC has revealed significant associations. Notably, the concentrations of various sphingomyelin species (d18:1/17:0, d17:1/18:0, and d19:1/16:0) were found to have an odds ratio (OR) of 1.299 (95% CI: 1.140–1.480). Additionally, the ratio of N-palmitoyl-sphingosine (d18:1 to 16:0) to N-palmitoyl-sphinganine (d18:0 to 16:0) exhibited an OR of 1.330 (95% CI: 1.134–1.559). Further, sphingomyelin concentrations (d18:1/22:1, d18:2/22:0, and d16:1/24:1) presented an OR of 1.236 (95% CI: 1.097–1.392), while N1-methyladenosine concentrations had an OR of 1.305 (95% CI: 1.048–1.625). Moreover, sphingomyelin concentrations (d17:2/16:0 and d18:2/15:0) yielded an OR of 1.272 (95% CI: 1.101–1.469; P < .05), indicating notable positive correlations (Fig. 2, details in Files S2–S3, Supplemental Digital Content, pleiotropy in File S2, Supplemental Digital Content, and heterogeneity in File S2, Supplemental Digital Content). These findings suggest that sphingolipid metabolites may facilitate tumor growth and invasion by enhancing membrane integrity and signaling pathways, thus underscoring the potential for targeting lipid metabolic pathways. Conversely, inverse associations were identified, including Glycochenodeoxycholate 3-sulfate concentrations (OR = 0.876, 95% CI: 0.818–0.938), glycohyocholate concentrations (OR = 0.781, 95% CI: 0.689–0.886), Caffeic acid sulfate concentrations (OR = 0.727, 95% CI: 0.582–0.907), N,N-dimethylalanine concentrations (OR = 0.875, 95% CI: 0.769–0.997), and 5alpha-pregnan-3beta,20alpha-diol disulfate concentrations (OR = 0.915, 95% CI: 0.841–0.996; P < .05), all demonstrating significant negative correlations (Fig. 2, details in Files S2–S3, Supplemental Digital Content, pleiotropy in File S2, Supplemental Digital Content, and heterogeneity in File S2, Supplemental Digital Content). These results indicate that bile acids and steroid derivatives may impede tumor progression through pathway modulation, thereby reducing cancer susceptibility. Collectively, these findings highlight the bidirectional influence of plasma metabolomic profiles in the etiology of ER-positive BC. Causal associations of serum metabolites with ER-positive breast cancer. Forest plot showing significant causal associations between serum metabolites and estrogen receptor-positive breast cancer (BREAST_ERPLUS) identified by inverse-variance weighted (IVW) two-sample Mendelian randomization analysis (nominal P < .05). CI = confidence interval, ER = estrogen receptor, OR = odds ratio. 3.1.1.2. Causal influences of immune effector profiles on ER-positive BC The examination of immune effector profiles in relation to ER-positive BC demonstrated significant positive associations: CD8br and CD8dim percentages of leukocytes (OR = 1.165, 95% CI: 1.005–1.351), CD11c expression on granulocytes (OR = 1.071, 95% CI: 1.009–1.136), CD4 expression on effector memory CD4 + T cells (OR = 1.061, 95% CI: 1.001–1.125), and CD40 expression on monocytes (OR = 1.033, 95% CI: 1.003–1.064; P < .05), all reflecting notable positive correlations (Fig. 3, details in Files S4–S5, Supplemental Digital Content, pleiotropy in File S4, Supplemental Digital Content, and heterogeneity in File S4, Supplemental Digital Content). These outcomes suggest that specific T cell and monocyte subsets may contribute to the remodeling of the tumor microenvironment through enhanced inflammatory signaling, thereby emphasizing the pivotal role of immune activation in cancer susceptibility. Conversely, inverse associations were observed, including lymphocyte activation (AC; OR = 0.818, 95% CI: 0.734–0.913), T cell activation (OR = 0.818, 95% CI: 0.720–0.930), CD8dim percentage of T cells (OR = 0.866, 95% CI: 0.766–0.979), CD8dim percentage of leukocytes (OR = 0.858, 95% CI: 0.754–0.977), and effector memory double-negative T cells (CD4−CD8−; OR = 0.903, 95% CI: 0.820–0.994; P < .05), with significant negative correlations (Fig. 3, details in Files S4–S5, Supplemental Digital Content, pleiotropy in File S4, Supplemental Digital Content, and heterogeneity in File S4, Supplemental Digital Content). These findings imply that reduced lymphocyte counts may compromise the antitumor immune response, thereby enhancing susceptibility to neoplastic development. Causal associations of immune cell phenotypes with ER-positive breast cancer. Forest plot showing significant causal associations between immune cell phenotypes and estrogen receptor-positive breast cancer (BREAST_ERPLUS) identified by inverse-variance weighted (IVW) two-sample Mendelian randomization analysis (nominal P < .05). CI = confidence interval, ER = estrogen receptor, OR = odds ratio. MR analysis identified positive causal links of plasma metabolites (e.g., sphingomyelins and hydroxybutyrates) to granulocyte/T cell profiles (OR > 1), and inverse associations (e.g., pregnane sulfates) with B cell markers (OR < 1) in ER + breast cancer (details, pleiotropy, and heterogeneity in Files S6–S8, Supplemental Digital Content). These results suggest that pregnane sulfates may inhibit the expression of B cell activation markers, thereby diminishing adaptive immune responses and highlighting their potential role in immune evasion in neoplastic contexts. Collectively, these findings support the notion that plasma metabolites may serve as critical mediators of immune effector profiles, laying the foundation for integrative therapeutic approaches in the context of ER-positive BC. 3.1.2. Breast cancer (ER-Negative) results 3.1.2.1. Causal influences of plasma metabolites on ER-negative BC The examination of causal relationships involving plasma metabolites in ER-negative breast cancer has revealed significant associations. Specifically, the phosphate to tryptophan ratio (OR = 1.371, 95% CI: 1.047–1.796), concentrations of 4-hydroxyphenylacetylglutamine (OR = 1.308, 95% CI: 1.044–1.639), levels of 2-hydroxydecanoate (OR = 1.285, 95% CI: 1.076–1.534), the choline to taurocholate ratio (OR = 1.212, 95% CI: 1.021–1.440), and the aspartate to asparagine ratio (OR = 1.144, 95% CI: 1.021–1.282) were all statistically significant (P < .05; Fig. 4, details in Files S9–S10, Supplemental Digital Content, pleiotropy in File S9, Supplemental Digital Content, and heterogeneity in File S9, Supplemental Digital Content). These findings suggest that imbalances in phosphate and amino acids may facilitate tumor invasion and survival by disrupting bioenergetics and the equilibrium of amino acids, highlighting the potential therapeutic implications of modulating these metabolic pathways. Conversely, inverse associations were identified, including indolepropionate concentrations (OR = 0.762, 95% CI: 0.647–0.896), glycohyocholate concentrations (OR = 0.807, 95% CI: 0.694–0.939), dopamine 3-o-sulfate concentrations (OR = 0.739, 95% CI: 0.573–0.952), the arginine to ornithine ratio (OR = 0.832, 95% CI: 0.738–0.938), and plasma free asparagine concentrations (OR = 0.925, 95% CI: 0.867–0.988; P < .05; Fig. 4, details in Files S9–S10, Supplemental Digital Content, pleiotropy in File S9, Supplemental Digital Content, and heterogeneity in File S9, Supplemental Digital Content). These results imply that indole compounds and bile acids may reduce cancer incidence by inhibiting inflammatory signaling and amino acid utilization, thereby underscoring their potential preventive roles. Collectively, these findings underscore the importance of monitoring plasma metabolic dynamics in the context of ER-negative breast cancer pathology. Bidirectional MR results for ER-negative breast cancer. Forest plot displaying significant bidirectional Mendelian randomization associations (IVW method) involving serum metabolites, immune cell phenotypes, and estrogen receptor-negative breast cancer (BREAST_ERNEG; nominal P < .05). CI = confidence interval, ER = estrogen receptor, MR = Mendelian randomization, OR = odds ratio. 3.1.2.2. Causal influences of immune effector profiles on ER-negative BC An analysis of immune effector profiles in ER-negative BC revealed significant associations: the presence of CD28 on CD39 + CD8bright T cells (OR = 1.052, 95% CI: 1.016–1.090), the percentage of CD39 + CD8bright T cells (OR = 1.067, 95% CI: 1.008–1.129), the proportion of CD39 + CD8bright T cells among total CD8bright T cells (OR = 1.066, 95% CI: 1.005–1.130), and the expression of CD40 on CD14− CD16 + monocytes (OR = 1.038, 95% CI: 1.005–1.073), as well as CD39 + CD8bright T cell activation (OR = 1.061, 95% CI: 1.002–1.123; P < .05), which were all associated with notable positive correlations (Fig. 5, details in Files S11–S12, Supplemental Digital Content, pleiotropy in File S11, Supplemental Digital Content, and heterogeneity in File S11, Supplemental Digital Content). These findings imply that the activation of CD8 T cells and monocytes could enhance tumor evasion mechanisms through increased immunological tolerance and inflammatory pathways, thereby opening new avenues for effector-targeted immunotherapy. Conversely, negative associations were observed: human leukocyte antigen-DR (HLA-DR) expression on plasmacytoid DCs (OR = 0.950, 95% CI: 0.923–0.976), CD20 expression on B cells (OR = 0.871, 95% CI: 0.799–0.948), CD20 on IgD + CD38dim B cells (OR = 0.897, 95% CI: 0.842–0.957), HLA-DR on DCs (OR = 0.951, 95% CI: 0.922–0.981), and CD20 on IgD + CD38bright B cells (OR = 0.897, 95% CI: 0.832–0.967; P < .05), which also demonstrated significant inverse correlations (Fig. 5, details in Files S11–S12, Supplemental Digital Content, pleiotropy in File S11, Supplemental Digital Content, and heterogeneity in File S11, Supplemental Digital Content). These results indicate that diminished markers on dendritic and B cells may impair antigen presentation and humoral immunity, increasing susceptibility to cancer and highlighting the critical need for restoring immunological vigilance in therapeutic strategies. Collectively, these observations underscore the pivotal role of immune effector variability in the pathogenesis of ER-negative BC. Causal associations of immune cell phenotypes with ER-negative breast cancer. Forest plot showing significant causal associations between immune cell phenotypes and estrogen receptor-negative breast cancer (BREAST_ERNEG) identified by inverse-variance weighted (IVW) two-sample Mendelian randomization analysis (nominal P < .05). CI = confidence interval, ER = estrogen receptor, OR = odds ratio. MR analysis showed positive causal effects of metabolites (e.g., homocitrulline and androgen glucuronides) on HLA-DR in DCs/B cells (OR > 1), and negative influences (e.g., acylcarnitines and asparagine) on monocyte/B cell activation (OR < 1) in ER-breast cancer (details, pleiotropy, and heterogeneity in Files S13–S15, Supplemental Digital Content). In summary, these results underscore the critical role of immune effector profile variability in the pathogenesis of BC, highlighting the intricate interplay between plasma metabolites and immune responses in this malignancy. 3.2. Mediation MR results Mediation analyses revealed various causal pathways linking metabolites to breast cancer, mediated by immune effectors, with notable variability across different breast cancer subtypes (Table 1; File S16, Supplemental Digital Content). In the case of BREAST_ERNEG, Dihomo-linoleoylcarnitine (C20:2) levels demonstrated significant mediation through HLA-DR in CD14 + CD16 + monocytes, while homocitrulline levels influenced susceptibility via HLA-DR in plasmacytoid DCs, suggesting that immunological processes linking lipid and amino acid bioenergetics are critical in this variant. For BREAST_ERPLUS, 3 notable pathways emerged, including positive mediation by Sphingomyelin (d18:1/17:0, d17:1/18:0, and d19:1/16:0) levels through CD38 in transitional B cells, and inverse mediation by X-21470 levels and 2R,3R-dihydroxybutyrate levels via the percentage of EM DN (CD4−CD8−) T cells and SSC-A in granulocytes, respectively, emphasizing the significant influence of lipid pathways in ER-positive tumors. Table 1Mediation analysis for breast cancer.MetaboliteImmune cellOutcomeEffectProportionBetaDirectDihomo-linoleoylcarnitine (C20:2) levelsHLA-DR on CD14 + CD16 + monocyteBREAST_ERNEG0.00989.98%0.0886Homocitrulline levelsHLA-DR on plasmacytoid DCBREAST_ERNEG0.011912.70%0.08172R,3R-dihydroxybutyrate levelsEM DN (CD4−CD8−) %T cellBREAST_ERPLUS−0.019323.40%−0.0634Sphingomyelin (d18:1/17:0, d17:1/18:0, and d19:1/16:0) levelsCD38 on transitionalBREAST_ERPLUS0.01947.41%0.2421X-21470 levelsSSC-A on granulocyteBREAST_ERPLUS−0.016014.90%−0.0910Mediation analysis results showing the indirect effect, proportion mediated (%), and direct effect (BetaDirect) for significant metabolite–immune cell–breast cancer triads in estrogen receptor-negative (BREAST_ERNEG) and estrogen receptor-positive (BREAST_ERPLUS) subtypes. Mediation analysis for breast cancer. Mediation analysis results showing the indirect effect, proportion mediated (%), and direct effect (BetaDirect) for significant metabolite–immune cell–breast cancer triads in estrogen receptor-negative (BREAST_ERNEG) and estrogen receptor-positive (BREAST_ERPLUS) subtypes. Mediation analyses systematically uncovered indirect pathways from metabolites to the onset of breast cancer via immune effectors and their relative contributions. In ER-positive breast cancer, 2R,3R-dihydroxybutyrate significantly enhanced protective effects by increasing the percentage of EM DN (CD4−CD8−) T cells, accounting for 23.4% of the bridging effect (β indirect = −0.0193), thereby reducing the risk of disease development. Sphingomyelin (d18:1/17:0, d17:1/18:0, and d19:1/16:0) exhibited positive mediation by decreasing CD38 levels in transitional B cells, contributing 7.41% to the overall impact (β indirect = 0.0194), indicating its role in transitional B cell functions that increase susceptibility. The unidentified metabolite X-21470 contributed 14.9% of the inverse impact through SSC-A elevation in granulocytes, providing notable protective effects. In ER-negative breast cancer, the lipid metabolite Dihomo-linoleoylcarnitine (C20:2) accounted for 9.98% of the impact (β indirect = 0.00983) by reducing HLA-DR expression in CD14 + CD16 + monocytes, suggesting that long-chain acylcarnitines may promote disease progression by impairing antigen presentation in atypical monocytes. Notably, homocitrulline concentrations also exhibited significant associations. Accounted for 12.7% of the impact (β indirect = 0.0119) through HLA-DR on the reduction of plasmacytoid DCs, suggesting that disruptions in amino acid bioenergetics may promote the onset of the disorder by impairing effector functions. These findings distinctly demonstrate the essential bridging roles of lipid compounds across various immune effector subgroups in both types of BC, providing novel insights into the variant-specific characteristics of metabolic-immunologic interactions in the etiology of BC. 4. Discussion BC represents a prevalent neoplastic disorder primarily arising from the hyperproliferation of mammary epithelial cells, and it is classified into hormone receptor-positive (ER-positive) and hormone receptor-negative (ER-negative) subtypes. Typically, ER-positive variants respond favorably to hormonal treatments, whereas ER-negative variants exhibit increased aggressiveness. This study employed bidirectional MR and bridging analyses, utilizing data from the UBK cohort that includes 1440 plasma metabolites and 731 immune effector profiles, alongside FinnGen R12 endpoint data, to elucidate the causal pathways linking plasma metabolites to different BC subtypes via immune effectors. This investigation holds potential for advancing precision oncology by identifying biomarkers and developing strategies that target the metabolic-immunologic continuum. The bidirectional MR analysis revealed strong associations between plasma metabolites and neoplasms, including BC, which govern cellular proliferation and inflammatory processes that drive tumorigenesis and progression. Among the metabolites analyzed, carnitines were specifically examined for their role in fatty acid transport and bioenergetics, with disturbances in these processes facilitating bioenergetic reconfiguration in BC.[8] In ER-negative BC, elevated levels of Ximenoylcarnitine were associated with an increased risk, attributed to enhanced long-chain fatty acid bioenergetics that support tumor invasion.[9] Specifically through increased mitochondrial β-oxidation rates that fuel cancer cell energy demands under hypoxic conditions. Additionally, dihomolinoleoylcarnitine and succinylcarnitine exhibited positive correlations, corroborated by multi-omic studies,[10] where dysfunction in the carnitine cycle alters intracellular oxidative and inflammatory status to promote tumorigenesis. Conversely, lignoceroylcarnitine demonstrated an inverse relationship, potentially limiting the accumulation of very-long-chain fatty acids to impede tumor progression.[11] In ER-positive BC, adipoylcarnitine showed a positive association while glutarylcarnitine was inversely correlated, highlighting the distinct transport disparities specific to each variant. Notably, OCTN2 transporter overexpression in ER + tumors, driven by estrogen signaling, enhances carnitine uptake and fatty acid metabolism, contributing to cell proliferation.[12] Furthermore, causal relationships were identified between amino acids and their derivatives with BC: Amino acid bioenergetics provides polypeptide precursors and regulates mTOR pathways, thereby promoting the growth of BC. In ER-negative BC, increased levels of homocitrulline were positively associated, resulting from alterations in the urea cycle that enhance ammonia clearance and support tumor cell survival under metabolic stress via mTORC1 activation for anabolic processes like protein synthesis.[13] Gamma-glutamylglutamine and 4-hydroxyphenylacetylglutamine also showed positive associations, reinforcing antioxidant defenses.[14] Dimethylarginine was positively associated, inhibiting nitric oxide synthase to promote vascular complications.[15] In contrast, plasma free asparagine was inversely correlated, with its depletion adversely affecting tumor viability due to impaired amino acid uptake and mTOR signaling adaptations that sustain tumor development.[16] In ER-positive BC, N,N,N-trimethyl-5-aminovalerate and 3-amino-2-piperidone were inversely associated, limiting polyamine pathways. N-acetylglucosaminylasparagine was inversely correlated, while aspartate showed a positive association, supporting nucleoside synthesis.[17] The bidirectional MR analysis also identified associations between sphingolipids and BC: Sphingolipids play crucial roles in apoptosis regulation, with anomalies enhancing tumor survival. In ER-positive BC, various sphingolipids, such as sphingomyelin, were positively associated, resulting from reduced enzymatic activity stabilizing lipid domains to facilitate signal transduction, specifically through decreased sphingomyelinase activity that stabilizes lipid rafts, enhancing PI3K/AKT and MAPK pathways for cell survival and proliferation. Overexpression of sphingosine kinase 1 in ER + cells induces proliferation via estrogen-dependent mechanisms linked to extracellular signal-regulated kinase activation.[18] Additionally, steroid hormones and sulfates were found to have close relationships with BC, as steroid hormones regulate endocrine signaling, influencing hormone-dependent neoplasms.[19] In ER-negative BC, 11beta-hydroxyandrosterone was notably implicated. The presence of glucuronide was positively correlated, while dopamine 3-o-sulfate exhibited an inverse relationship, in the context of neoplastic infiltration. In ER-positive breast cancer, both androsterone sulfate and epiandrosterone sulfate showed a positive association, facilitating estrogen conversion. DHEAS and 16α-hydroxy DHEA 3-sulfate demonstrated varying connections. Androstenediol (3β,17β) monosulfate and 5α-androstan-3α,17α-diol monosulfate, among others, mediate sulfation via sulfotransferases (SULTs), which regulate the availability of metabolites. A reduction in SULT activity leads to an accumulation of active hormones, which exacerbates endocrine disruption and highlights the critical role of sulfotransferases in promoting the progression of breast cancer.[20] Particularly via the sulfatase pathway that hydrolyzes inactive steroid sulfates to generate biologically active estrogens, stimulating epithelial proliferation in ER + tumors. In ER + cases, estrogen/ER signaling interacts with stromal cells to modulate the tumor microenvironment, enhancing fibroglandular tissue and promoting cancer growth. For ER-negative tumors, androgen excess acts as a common stimulatory factor, potentially via cross-talk with EGFR/Her-2 pathways.[21] Additionally, bile acids and their derivatives are implicated in breast cancer. For instance, chenodeoxycholic acid activates farnesoid X receptor (FXR), which inhibits tumor development, whereas dysregulation can trigger inflammatory tumor transformation. FXR activation suppresses NF-κB signaling, thereby reducing inflammation (for example, through the IL-6/STAT3 pathway); however, a deficiency in FXR can induce ferroptosis, facilitating the growth and spread of breast cancer cells.[22] The bidirectional molecular responses also revealed significant interactions between immune effectors and tumors such as breast cancer, categorized into 5 principal groups: B cells, T cells, myeloid cells (including monocytes and DCs), granulocytes, and natural killer (NK) cells. In the case of B cells, mature B cells, memory B cells, and CD38-positive naive B cells perform dual roles in breast cancer. These cells enhance antitumor immunity through the production of immunoglobulins and the activation of the complement system; however, in ER-positive breast cancer, their excessive activation may promote tumor infiltration and suppress local defenses, leading to immune evasion.[23] Conversely, in ER-negative breast cancer, decreased expression of B cell-related markers may correlate with increased invasiveness, thereby impacting prognosis.[24] Such effects may arise from disruptions in endocrine signaling, which influences B cell development and migration. The T cell populations include naive T cells, effector memory T cells, terminally differentiated effector memory T cells, exhausted T cells, and regulatory T cells, all of which modulate cytotoxicity and auxiliary functions within the tumor microenvironment.[25] In ER-positive breast cancer, co-stimulatory signals (for example, those associated with CD28) may be diminished, resulting in T cell depletion and promoting tumor growth. In contrast, stimulation of CD8 + T cell subsets in ER-negative breast cancer may enhance tumor elimination, although this effect is counterbalanced by increases in regulatory T cells.[26] Myeloid cell-associated immune effectors have also been observed to participate in inflammatory responses and antigen presentation during the progression of breast cancer. In ER-positive cases, the expression of HLA-DR on these cells may increase, fostering immunosuppressive environments.[27] Conversely, in ER-negative breast cancer, signaling through CCR2 may facilitate the recruitment of myeloid cells to tumor sites, enhancing angiogenesis and metastasis.[28] These mechanisms may involve the regulation of chemokine receptors, influencing tumor heterogeneity. Granulocytes, including those characterized by CD33 on CD33dim HLA-DR + CD11b− and CD66b++ myeloid cells, along with specific markers on granulocytes, primarily regulate neutrophil and eosinophil functions in breast cancer. In ER-positive breast cancer, morphological changes in these cells may reflect heightened inflammatory states, contributing to therapeutic resistance.[29] In ER-negative breast cancer, CD33-related expression may direct granulocyte orientation towards pro-tumor activities.[30] These dynamics may be linked to oxidative stress and alterations in the extracellular matrix, affecting tumor infiltration. NK cells, closely associated with tumors, such as those expressing CD16–CD56, play a crucial role in innate immune surveillance in breast cancer. The manifestation of immune responses may diminish, thereby reducing direct cytotoxic effects against neoplastic cells.[31] In contrast, stimulation of NK cells in ER-negative BC may be enhanced but simultaneously limited by the suppressive environment created by neoplastic niches.[32] These interactions may arise from the influence of cytokines on the evasion of immune defense mechanisms. Overall, the relationships between immune effector populations and the characteristics of ER-positive versus ER-negative BC exhibit specificity related to the variants. ER-positive cases tend to experience immunological suppression, whereas ER-negative cases are associated with vigorous inflammatory responses. These findings provide valuable insights for the development of precision immunotherapies; however, it is crucial to consider individual variability and contextual factors. Future research should integrate multi-omic data to elucidate the precise etiological roles of these factors in BC. Mediation MR analyses have identified several metabolites, including Dihomo-linoleoylcarnitine, Homocitrulline, and X-21470, which demonstrate causal associations with ER-negative BC variants, while 2R,3R-dihydroxybutyrate and Sphingomyelin are linked to ER-positive variants. The relationships with bridging immune effectors primarily manifest as metabolites that regulate immune effector profiles and functions, influencing the activation states of monocytes and DCs. This aligns with bridging outcomes such as HLA-DR expression on CD14 + CD16 + monocytes, HLA-DR on plasmacytoid DCs, and CD38 on transitional cells, where metabolites either promote immunological suppression or inflammatory responses. Immune effectors, acting as bridges, play essential roles in regulating inflammation and immune evasion within neoplastic environments, ultimately affecting the outcomes associated with different BC variants.[33] Key findings from bridging analyses revealed various metabolites that influence susceptibility to BC through distinct immune effectors, with higher bridging coefficients observed in ER-positive variants (e.g., 0.234). This elevation is attributed to the ability of these metabolites to disrupt the progression and functions of immune effectors, thereby accelerating neoplastic development. Dihomo-linoleoylcarnitine (C20:2) reduces HLA-DR expression on CD14 + CD16 + monocytes; studies suggest that metabolites modulate lipid signaling pathways to diminish inflammatory responses in monocytes, with ex vivo experiments indicating that DGLA decreases the release of pro-inflammatory mediators. In the context of ER-negative variants, this bridging effect enhances immune evasion, as empirical evidence reveals shifts in monocyte profiles in oncological patients experiencing disease progression.[34] Homocitrulline reduces the quantity of HLA-DR on plasmacytoid DCs, and related studies indicate that this metabolite affects the antigen presentation capabilities of DCs, potentially due to disruptions in amino acid bioenergetics that impair cellular signaling, supported by evidence of reduced DC activity in neoplastic conditions.[35,36] For ER-negative variants, this bridging facilitates immune tolerance. 2R,3R-dihydroxybutyrate increases the percentage of EM DN (CD4−CD8−) T cells[37]; investigations suggest that this metabolite promotes T cell proliferation through glycolytic pathways, leading to alterations in T cell profiles. In ER-positive variants, it enhances the dampening effects on T cells, with variability in T cell populations linked to resilience against BC.[38] Sphingomyelin levels decrease CD38 expression on transitional cells, with mechanisms involving sphingolipid regulation of B cell development, resulting in operational anomalies in transitional B cells.[39] In ER-positive variants, this influences immunoglobulin production, with shifts in B cell subgroups within neoplastic niches creating barriers to progression.[40] In ER-positive cases, granulocytes exhibit dual roles in oncology, as evidenced by the concentrations of 5alpha-pregnan-3beta,20alpha-iol disulfate, which reduce SSC-A expression on granulocytes, with potential mechanisms involving hormone sulfate signaling pathways.[41] The investigation elucidates the significant role of the metabolic-immunologic continuum in breast cancer by providing insights for personalized treatment approaches. However, the study is not without limitations, including cohort bias and incomplete mechanistic understanding, which direct future research towards the integration of multi-omic data for therapeutic validation. Future research should prioritize multi-omic integration (e.g., metabolomics, proteomics, and single-cell RNA sequencing) to validate these causal pathways and elucidate cell-type-specific mechanisms. Functional studies, such as in vitro/in vivo models with metabolite supplementation or knockdown, are needed to confirm biological roles. Longitudinal cohort studies in diverse populations and clinical trials targeting key metabolites (e.g., carnitines or sphingolipids) could translate these findings into precision oncology strategies. 5. Limitations This study highlights the causal relationships between plasma metabolites and various breast cancer subtypes through the mediation of immune effectors. Nonetheless, it encounters several general limitations, such as potential pleiotropy and confounding influences from genetic IVs that may introduce bias. Dataset limitations arise from the fact that the UBK metabolite and immune effector data are sourced from European populations, potentially limiting generalizability to other ethnic groups where BC subtypes and metabolic profiles differ significantly. This ancestry bias could overestimate or underestimate causal effects due to population-specific genetic variants or environmental factors. Methodological limitations are evident as the assumptions of MR rely on the strength of IVs, where weak instruments can exacerbate bias, and temporal variations have not been adequately addressed. Furthermore, the framework of this inquiry does not consider prospective confounders such as contextual factors, and the endpoint data (FinnGen R12) may be subject to selection bias. The absence of functional validation studies necessitates ex vivo and in vivo confirmation to substantiate etiological claims. Given these limitations, future research endeavors could facilitate multicenter collaborations for validation, incorporate single-cell profiling to explore underlying mechanisms, and develop pharmacological interventions targeting the identified bridging pathways. When comparing our findings to existing scientific literature on similar topics: Our results excel in identifying additional bridging pathways (e.g., 6 specific bridges), enhancing the understanding of ER variant disparities and addressing gaps in the influence of metabolites on granulocytes. Our methodology employs bridging MR analysis, which surpasses previous unidirectional MR approaches by quantifying bridging proportions and improving causal inference accuracy. The selection of the UBK dataset (comprising 1440 metabolites and 731 immune effectors) and FinnGen R12 provides a broader scope and genetic diversity advantages over earlier studies with smaller cohorts, thereby increasing analytical power. 6. Conclusion This study utilizes bidirectional MR and bridging analyses to uncover causal relationships between plasma metabolites and breast cancer variants via immune effectors, elucidating the causal mechanisms linking plasma metabolites to both ER-positive and ER-negative breast cancer, with immune effectors serving as crucial mediators. It enhances our understanding of the interconnected metabolic-immunologic frameworks that underpin breast cancer progression. These findings offer new perspectives for the timely detection and management of breast cancer, presenting innovative avenues for therapeutic interventions. Despite the identified limitations, the outlook remains optimistic and may guide future advancements in precision oncology. Acknowledgments We thank our colleagues for their assistance with this study. Author contributions Conceptualization: Zhe Sun. Data curation: Jinxian Qian. Formal analysis: Wenjing Duan. Methodology: Zhe Sun. Supervision: Zhe Sun. Writing – original draft: Fei Ge, Liang Chen. Writing – review & editing: Zhe Sun.