Deciphering the immunometabolic axis: a mendelian randomization study of a causal cascade network from immune cell phenotypes to metabolites in esophageal cancer

Introduction

Esophageal cancer (EC) is the seventh most prevalent malignancy in China (1), and a leading cause of cancer-related death worldwide (2). Histopathologically, EC comprises two main subtypes: esophageal squamous cell carcinoma, which is predominant in East Asia and linked to tobacco/alcohol exposure, and esophageal adenocarcinoma, which is more common in Western populations, and associated with gastroesophageal reflux. Crucially, immunological mechanisms drive disease progression and treatment response in both subtypes (3). While immune checkpoint inhibitors [e.g., anti-programmed cell death protein 1/programmed death ligand 1 (anti-PD-1/PD-L1)] have transformed EC therapy, response rates remain below 30% due to intrinsic/acquired resistance (4). Thus, understanding the tumor microenvironment (TME) and identifying predictive biomarkers are essential for therapeutic advancement.

In the TME, the degree of immune-cell infiltration and their phenotypic alterations directly govern tumor progression, invasion, and metastasis; indeed, Mendelian randomization (MR) evidence from breast cancer has established that specific immune-cell subpopulations are causally linked to tumor risk (5). Translating this concept to EC, high CD8⁺ tumor-infiltrating lymphocyte (TIL) density predicts prolonged disease-free survival, an enhanced neoadjuvant response, and reduced lymph-node metastasis (6). Conversely, the accumulation of immunosuppressive cell populations—most notably myeloid-derived suppressor cells (MDSCs) and regulatory T cells (Tregs)—correlates strongly with advanced disease stage and inferior survival outcomes (7). Moreover, beyond MDSCs and Tregs, tumor-associated macrophages polarized toward the immunosuppressive M2 phenotype are also associated with shorter overall survival in EC patients (8).

Immunometabolism has emerged as a pivotal research frontier in oncology, revealing the intricate interplay between immune-surveillance networks and metabolic circuitry. Systemic metabolic states can shape tumor risk; recent MR evidence shows that lipid-metabolic traits are causally linked to cancer development (9). At the cellular level, the metabolic milieu directly dictates immune-cell fate and function: activated effector T cells rely on glycolysis and glutamine metabolism, whereas Tregs and memory T cells sustain immune homeostasis through fatty-acid oxidation (10,11). Recent research has shown that the phosphatidylinositol-3-kinase/protein kinase B/mechanistic target of rapamycin (PI3K/AKT/mTOR) and the liver kinase B1–adenosine monophosphate-activated protein kinase (LKB1-AMPK) pathways are key signaling pathways in immunometabolism. Activating or inhibiting these pathways not only affects the metabolic phenotypes of tumor cells but also significantly regulates the function and differentiation of immune cells (12). Consequently, the pharmacological targeting of key immune-metabolic nodal regulators represents a promising therapeutic paradigm for enhancing anti-tumor immunity (13).

MR analysis is an instrumental variable (IV) approach that substantially mitigates the confounding factors and reverse causation inherent in conventional observational designs, thereby providing a robust framework for causal inferencen. This study leveraged multi-omics datasets, comprising immunophenotypes, plasma metabolomic profiles, lipidomic signatures, and EC genome-wide association study (GWAS) summary statistics, to construct a directionally ordered “immune-metabolic-EC pathogenesis” causal cascade using MR frameworks. First, we implemented bidirectional univariable Mendelian randomization (UVMR) for the primary causal evaluation, and then used multivariable Mendelian randomization (MVMR) to account for immunological collinearity. Subsequently, we applied a two-step MR-mediation framework to identify core metabolic mediators, elucidating pathway-specific mechanisms whereby immune phenotypes modulate the risk of EC. This integrative strategy delineated the immune-metabolic causal architecture governing EC, deciphered microenvironmental immune reprogramming dynamics, and provided the mechanistic foundations for biomarker discovery and novel immune-metabolic therapeutic strategies. We present this article in accordance with the STROBE-MR reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-774/rc).

Methods

Study design

This study was conducted using a retrospective MR design with publicly available summary-level data. To this end, we established a multi-phase analytical framework, as schematized in Figure 1. First, bidirectional UVMR was used to assess the reciprocal causal relationships linking immunophenotypic profiles to EC. After excluding reverse causation, significant associations identified in the forward MR were subjected to MVMR analysis to evaluate the independent causal effects of specific immune subsets on oncogenic progression. Subsequently, a tripartite MR-mediation framework was employed as follows: (I) UVMR-based screening of EC-associated metabolic and lipidomic signatures; (II) identification of immunophenotypes linked to candidate metabolites; and (III) effect-direction concordance analysis to ascertain potential mediators. Finally, we employed both UVMR and MVMR to analyze the mediation effects of mediators between exposure and outcome, calculating the effect values and proportions for each qualified mediator. As this MR study only used publicly available summary-level data, preregistration was not required. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Figure 1 Flow chart of the MR analysis. BWMR, Bayesian-weighted Mendelian randomization; MR, Mendelian randomization; MV-IVW, multivariable inverse-variance weighted; MVMR, multivariable MR; SNPs, single nucleotide polymorphisms.

Data sources

EC GWAS summary statistics were obtained from the FinnGen database (https://finngen.gitbook.io/documentation/v/r12; Data ID: finngen_R12_C3_OESOPHAGUS_EXALLC). The dataset comprised 1,277 ICD-10-certified EC cases and 378,749 control cases (14). Data for 731 immune cell traits (Ebi-a-GCST0001391 to Ebi-a-GCST0002121) were retrieved from the IEU Open GWAS project database (https://gwas.mrcieu.ac.uk/) (15). The immunophenotypes under investigation were classified into seven principal groups: B cells; conventional dendritic cells; mature stage T cells; monocytes; myeloid cells; T cells, B cells, and natural killer cells; and Treg panels. Plasma metabolome-wide association study data were sourced from the GWAS Catalog database (https://www.ebi.ac.uk/gwas/) with identifiers ranging from GCST90199621 to GCST90201020. The dataset comprised 1,091 metabolites and 309 metabolite ratios (16). GWAS summary statistics for lipidome were acquired from the GWAS Catalog (https://www.ebi.ac.uk/gwas/) with identifiers ranging from GCST90277238 to GCST90277416. The dataset comprised 7,174 FinnGen participants from the GeneRISK cohort, with lipid species systematically classified into 13 biochemical categories and four super-classes, capturing genetic associations across 179 distinct lipid molecular species (17). It should be noted that the study exclusively included participants with European ancestry. For specific details on all the data, see Table 1.

IV selection criteria

This study employed a MR analysis to determine the causal relationships between immune cells and EC. To be deemed suitable as IVs, the single nucleotide polymorphisms (SNPs) had to meet three essential criteria: (I) each IV had to have a substantial association with the exposure; (II) each IV had to influence the outcome solely through the exposure; and (III) each IV had to be minimally affected by confounding variables to reduce bias fromlinkage disequilibrium (LD).

We implemented rigorous selection criteria for the genetic IVs across exposures. To ensure analytical robustness and result validity, the IVs were selected through three sequential filters: First, genome-wide screening excluded SNPs with weak exposure associations. In genetic studies, the conventional genome-wide significance threshold (P<5×10⁻⁸) often yields insufficient SNPs for robust analysis. Therefore, our forward MR analyses adopted a threshold of P<1×10⁻⁵ (18,19) to ensure adequate IVs while maintaining analytical reliability, which is consistent with empirical evidence demonstrating preserved validity at this threshold. Reverse MR used stricter thresholds (P<5×10⁻⁶) to minimize false-positive associations during the causal directionality assessment (20). Second, LD clumping (r²<0.001, window =10,000 kb) (21) ensured genetic instrument independence. Third, to ensure strong associations between the selected IVs and exposures, the F statistic (screening criterion: F>10) was used to assess the strength of genetic variation as IVs (22,23). The F statistic was calculated using the following formula: F = (N–K–1)*R²/K*(1–R²) (24), where N is the sample size in the exposure database, K is the number of IVs, and R² is the proportion of variance explained by SNP in the exposure database. R² was calculated as follows: R² = B²*EAF*(1–EAF)/SE²*N, where EAF is the effect allele frequency, β is the allele effect value, N is the sample size, and SE is the standard error.

UVMR and MVMR analyses

To investigate the causal effects of immunophenotypes and plasma metabolites/lipids on EC risk, we implemented five MR approaches: inverse-variance weighted (IVW) (25), weighted median (26), MR-Egger regression (27,28), weighted mode (29), and Bayesian-weighted Mendelian randomization (BWMR) (30). The IVW approach, a gold-standard analytical approach in MR, aggregates genetic variant effects through inverse-variance weighting to enhance causal effect estimation precision and statistical power. When IV assumptions are satisfied, the IVW approach provides consistent and asymptotically efficient causal estimates. Conversely, BWMR addresses uncertainties from weak instrument bias and horizontal pleiotropy, while identifying outliers attributable to substantial pleiotropic effects through Bayesian probabilistic modeling. The IVW and BWMR approaches were designated as the primary analytical approaches, with a statistical significance threshold of P<0.05 for causal interpretation. The MVMR analyses adopted the multivariable inverse-variance weighted (MV-IVW) approach as the principal method, applying a P<0.05 threshold for statistical significance (31).

Assessment of heterogeneity and horizontal pleiotropy

The following sensitivity analyses were conducted to ensure the robustness of the MR findings: MR-Egger regression to assess directional pleiotropy, Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR-PRESSO) to identify outliers (32), leave-one-out analysis to evaluate the influence of individual variants, and Cochran’s Q statistic to test for heterogeneity (32). Heterogeneity across the SNPs was quantified using Cochran’s Q statistic (a P<0.05 value indicating significant between-instrument heterogeneity in the causal effect estimates). The MR-Egger intercept analysis was conducted to evaluate horizontal pleiotropy, where non-zero intercepts (P<0.05) indicated systematic pleiotropic effects independent of exposure-outcome pathways. The MR-PRESSO global test (P<0.05) identified pleiotropic outliers through distortion testing, followed by corrected effect estimation after outlier removal. The leave-one-out sensitivity analysis iteratively excluded individual IVs to assess effect estimate stability, with substantial fluctuations indicating reduced robustness.

All statistical analyses were performed using R software (version 4.4.0; R Foundation for Statistical Computing). The following R packages were utilized for MR analyses: TwoSampleMR (version 0.5.7) [Hemani et al., 2018] (33) for data harmonization and primary MR analyses; MR-PRESSO (version 1.0) [Verbanck et al., 2018] (32) for outlier detection and correction; and MendelianRandomization (version 0.9.0) [Yavorska and Burgess, 2017] (34) for additional sensitivity analyses, including MVMR.

Mediation analysis

We employed a four-stage analytical framework to identify and validate immune cell-esophageal tumor pathway mediators. First, UVMR was used to establish the initial causal associations between potential mediators and EC. Mediators demonstrating statistically significant causal effects (P<0.05) were selected, and their effect sizes (β₁) were quantified. Next, the causal relationships were re-evaluated through UVMR by reversing the analytical direction (i.e., the identified immune cells served as exposures, and the previously selected mediators acted as outcomes). This reciprocal analysis identified immune cells exerting significant causal effects on mediators (P<0.05), with corresponding effect sizes (β₂). Following this, the mediators underwent directional consistency validation relative to the overall exposure-outcome relationship. Specifically, when the total exposure-outcome effect (β) was positive, both β₁ and β₂ had to exhibit concordant directionality (both positive or both negative). Conversely, negative total β necessitated opposing directionalities between β₁ and β₂. Finally, the validated mediators underwent a MVMR analysis to evaluate their independent causal effects on the outcome. The mediators with adjusted MV-IVW P values <0.05 were considered robust mediators. The final mediation analysis employed the product-of-coefficients method to calculate the mediation effect (β₁×β₂) and the proportion mediated [(β₁×β₂)/β] (35).

Results

IV screening results

Following rigorous quality control, qualified SNPs were identified as IVs. For the forward MR analysis (exposures on EC), these included 16,289 for immune cell phenotypes, 33,281 for plasma metabolomes, and 4,415 for plasma lipidome traits. The reverse MR analysis (EC on traits) used 151, 454, and 41 IVs, respectively. All IVs were strong (F-statistics >10), ruling out weak instrument bias (Tables available at https://cdn.amegroups.cn/static/public/jgo-2025-774-1.xlsx).

Two-sample MR analysis between immune cells and EC

The IVW and BWMR analyses revealed 22 immunophenotypes exhibiting causal associations with EC (Figure 2A). Among these, 11 immunophenotypes demonstrated positive causal effects, while the remaining 11 showed negative causal effects (Figure 2B). Notably, seven immunophenotypes demonstrated robust multi-method validity (P<0.05) (Figure 2C), of which, the CD19 on IgD⁺CD38⁻ naive B cells exhibited the most significant correlation [IVW: odds ratio (OR) =0.911, 95% confidence interval (CI): 0.867–0.957, P=0.0001; BWMR: OR =0.863, 95% CI: 0.788–0.945, P=0.002]. The other three analytical methods also yielded results consistent with those of the IVW and BWMR analyses (P<0.05).

Figure 2 MR analysis of the causal effects of immunophenotypes on EC. (A) Hot plot of univariable Mendelian randomization results showing that 22 immune cell phenotypes had significant causal effects on esophageal cancer. (B) Forest plot showing the seven most robustly and significantly associated immune phenotypes. CI, confidence interval; EC, esophageal cancer; MR, Mendelian randomization; OR, odds ratio; SNP, single nucleotide polymorphism.

We performed multiple sensitivity analyses on the initial MR results to ensure their robustness. The MR-Egger regression intercept test revealed no evidence of bias due to genetic pleiotropy (P>0.05), and the MR-PRESSO analysis corroborated the absence of horizontal pleiotropy across the MR estimates (P>0.05). Cochran’s Q test indicated no significant heterogeneity in most results (P>0.05), except for CD19 on Ig D⁺ CD24^–B cells (P=0.04) (Table S1). The reverse MR analyses revealed no statistically significant causal effects of genetically predicted EC on the 22 immune cell phenotypes (IVW_P>0.05), indicating an absence of reverse causation (Figure 3).

Figure 3 Forest plots of the reverse MR analysis results showing non-significant causal associations between EC and the 22 immune cell phenotypes. CI, confidence interval; EC, esophageal cancer; IVW, inverse-variance weighted; OR, odds ratio; SNP, single nucleotide polymorphism.

MVMR-adjusted causal effects of immunophenotypes on EC

Given the potential inter-phenotype correlations among the immune cells, we conducted MVMR analyses to delineate the independent causal effects of individual immunophenotypes. After simultaneously adjusting for 22 candidate immunophenotypes, three retained statistically significant independent associations with the risk of EC (P<0.05), while 19 phenotypes showed attenuated effects in the multivariable model. The multivariable-adjusted ORs were: CD38⁺CD20⁻ B cells (OR =1.084, 95% CI: 1.005–1.169, P=0.04), CD33⁺CD14⁺ monocytes (OR =1.129, 95% CI: 1.003–1.271, P=0.044), and CD45⁺CD33⁺HLA-DR⁺CD14dim myeloid cells (OR =1.045, 95% CI: 1.007–1.085, P=0.007) (Figure 4).

Figure 4 The MVMR assessment of the independent causal effects of 22 immune phenotypes showed that the phenotypes still maintained significant independent associations with the risk of esophageal cancer. CI, confidence interval; IVW, inverse-variance weighted; MV-IVW, multivariable inverse-variance weighted; MVMR, multivariable Mendelian randomization; OR, odds ratio; SNP, single nucleotide polymorphism.

Mediation analysis

Causal prioritization of metabolites and lipids in EC

We conducted two-sample MR analyses of 1,400 plasma metabolome traits and 179 plasma lipidome traits using selection criteria identical to that used in the immunophenotype analysis. Among the 1,400 metabolic features, 43 showed consistent causal associations with the risk of EC (Figure 5A). Protective effects were observed for 22 features (18 absolute metabolite levels, 4 ratios), while 21 showed risk effects (20 levels, 1 ratio) (Figure 5B). Six metabolites demonstrated robust significance (IVW and BWMR P<0.05) with concordant directionality across the three supplementary methods (Figure 5C). For instance, 3-formylindole levels showed a significant negative correlation with EC (IVW: OR =0.808, 95% CI: 0.701–0.932, P=0.003; BWMR: OR =0.806, 95% CI: 0.697–0.932, P=0.004). In the MR screening of 179 plasma lipidomes, five lipids demonstrated significant associations with EC (Figure 6). For example, Phosphatidylcholine (18:0_20:4) levels showed a significant negative correlation with EC (IVW: OR =0.884, 95% CI: 0.801–0.975, P=0.01; BWMR: OR =0.883, 95% CI: 0.799–0.975, P=0.01). Conversely, phosphatidylethanolamine (18:0_18:2) demonstrated a significant positive correlation (IVW: OR =1.155, 95% CI: 1.004–1.327, P=0.043; BWMR: OR =1.184, 95% CI: 1.020–1.375, P=0.03).

Figure 5 MR analysis of plasma metabolites and EC risk. (A) Hot plot showing the UVMR identified significant causal effects of 43 plasma metabolomes on esophageal cancer. (B) Forest plot showing the six most robustly and significantly associated immune phenotypes. CI, confidence interval; EC, esophageal cancer; MR, Mendelian randomization; OR, odds ratio; SNP, single nucleotide polymorphism; UVMR, univariable Mendelian randomization.

Figure 6 Results from the bidirectional univariable Mendelian randomization examining the causal effects of lipid species on esophageal cancer. Color lines represent different MR methods: red lines indicate results from the IVW method, blue lines indicate results from the BWMR method. BWMR, Bayesian-weighted Mendelian randomization; CI, confidence interval; EC, esophageal cancer; IVW, inverse-variance weighted; OR, odds ratio; SNP, single nucleotide polymorphism.

The results of the heterogeneity and pleiotropy analyses were not significant, confirming the robustness of our results (Table S2). The reverse MR analysis showed that EC had no causal effects on these lipids and metabolic features (P>0.05), excluding reverse causation (Figure 7).

Figure 7 In the reverse MR analyses, the forest plots showed no significant causal associations between esophageal carcinoma and 43 plasma metabolites or five lipid species. CI, confidence interval; EC, esophageal cancer; IVW, inverse-variance weighted; MR, Mendelian randomization; OR, odds ratio; SNP, single nucleotide polymorphism.

Causal relationships between immune phenotypes and metabolites/lipidome

Using IVW as the primary method, we assessed the causal effects of three immune phenotypes (CD38⁺CD20⁻ B cells, CD33⁺CD14⁺ monocytes, and CD45hiCD33⁺CD14dim cells) on 43 plasma metabolites. The MR analysis identified causal relationships between three immune phenotypes and four metabolites (Figure 8A). These associations remained robust in the sensitivity analyses, showing no significant heterogeneity or horizontal pleiotropy (Table S3). The subsequent IVW analysis of the three phenotype effects on five lipids revealed two causal associations: CD20-CD38-B cells reduced PC (18:0_20:4) levels (IVW: OR =0.9405, 95% CI: 0.8915–0.9921, P=0.02), and CD45hiCD33⁺HLA^–DR⁺CD14dim cells decreased both PC (16:0_20:5) (IVW: OR =0.9513, 95% CI: 0.9159–0.9881, P=0.009) and PC (18:0_20:4) levels (IVW: OR =0.9558, 95% CI: 0.9141–0.999, P=0.047). BWMR confirmed these findings (P<0.05) (Figure 8B). The causal relationships between the two immune phenotypes and two lipid species remained robust in the sensitivity analyses (Table S3).

Figure 8 MR analysis of candidate immunophenotypes and potential mediators. (A) Results from the bidirectional univariable Mendelian randomization examining the causal effects of candidate immune cells on candidate plasma metabolomes. (B) Results of the bidirectional univariable Mendelian randomization examining the interactions between candidate immune cells and candidate liposomes. Color lines represent different MR methods: red lines indicate results from the IVW method, blue lines indicate results from the BWMR method. BWMR, Bayesian-weighted Mendelian randomization; CI, confidence interval; IVW, inverse-variance weighted; MR, Mendelian randomization; OR, odds ratio; SNP, single nucleotide polymorphism.

Mediation analysis of potential mediators

Through the two preceding steps of UVMR and targeted screening based on β, β1, and β2, we preliminarily identified two immune cell phenotype-lipid-EC pathways and two immune cell phenotype-plasma metabolite-EC pathways (Figure 9), involving two plasma metabolome and two plasma lipids. Subsequently, the MVMR analysis confirmed the independent causal effect of the aspartate to citrate (Asp/Cit) ratio on EC (MVMR-IVW P<0.05) (Table 2). The mediation analysis using the product-of-coefficients method revealed that the Asp/Cit ratio mediated 10.6% of the causal relationship between CD33⁺CD14⁺ monocytes and EC, and persisted after MVMR adjustment for confounding factors (Table 3).

Figure 9 Potential pathways linking immunophenotypes to esophageal cancer through specific metabolites. (A) The figure shows the mediation mode of “immune cells-lipidomes-EC”. (B) The figure shows the mediation mode of “immune cells-plasma metabolomes-EC”. EC, esophageal cancer.

Discussion

This multi-omics study integrated immunophenotypic, plasma metabolomic, and lipidomic data, applying two-sample MR to infer causal relationships in the EC immune microenvironment. Genetic evidence confirmed the causal pro-malignant effects of immunosuppressive cellular subpopulations. Specifically, our analyses revealed significant positive causal associations between CD38⁺CD20⁻ B cells, CD33⁺CD14⁺ monocytes, and CD45⁺CD33⁺HLA-DR⁺CD14dim myeloid cell subsets and EC risk. Through MVMR and mediation analyses, the Asp/Cit ratio was identified as a key metabolic mediator, mediating the pro-tumorigenic effects of CD33⁺ monocytes.

The TME is the internal environment in which tumor cells survive, and comprises malignant cells alongside stromal components, immune populations, inflammatory mediators, and extracellular matrix constituents. Among these, immune cells have significant prognostic value in EC. Their plasticity in the TME leads to diverse phenotypic expression that can either promote or inhibit tumor progression (36). Our MR analyses confirmed previous observations; for example, the anti-tumor components (e.g., CD8⁺ T cells) showed inverse causal associations with oncogenesis, while immunosuppressive subsets (e.g., MDSCs and Tregs) exerted positive causal effects (37). Notably, these causal genetic insights corroborate previous observational studies. For example, a high abundance of CD8⁺ TILs predicts prolonged disease-free survival and enhanced neoadjuvant response rates in EC (38). Conversely, the accumulation of MDSCs in peripheral blood or tumor tissue has been shown to be independently associated with advanced tumor-node-metastasis (TNM) staging and poor prognosis (39). Our MR analysis results provide causal support for these associations.

This study delineates a novel immunometabolic pathway linking MDSCs to metabolic reprogramming in EC. While previous research has emphasized MDSC-derived immunosuppressive factors such as ARG1, iNOS and IDO (40), our MR analysis reveals an underappreciated role of MDSCs in systemic metabolic perturbation, particularly through modulation of the aspartate-to-citrate (Asp/Cit) ratio. Aspartate is an amino acid essential for DNA and ATP synthesis, often serving as a rate-limiting substrate for tumor cell proliferation (41). Citrate, a core intermediate of the tricarboxylic acid cycle and a substrate for fatty acid synthesis, is often taken up by highly invasive tumor cells to support their rapid growth (42). Our Mendelian-randomisation analysis establishes that an elevated Asp/Cit is an independent risk factor for EC (OR =1.44). Mechanistically, this metabolic imbalance reflects a reprogramming process wherein increased aspartate availability promotes nucleotide synthesis while citrate depletion enhances compensatory uptake by tumor and stromal cells (43). Given the tumor-promoting effects of this metabolic imbalance, intervention strategies targeting this pathway show therapeutic potential. Previous studies have demonstrated that restricting aspartate availability or supplementing citrate can produce anti-tumor effects, establishing this metabolic axis as a valuable therapeutic target. For instance, MYC-driven tumors display marked aspartate dependency, and disrupting mitochondrial aspartate transport can effectively inhibit tumor progression (44). Conversely, exogenous citrate administration remodels tumor metabolism and context-dependently suppresses malignant progression (45). The identified Asp/Cit pathway suggests a potential metabolic target for further investigation. Future studies could explore whether modulating the Asp/Cit ratio might complement existing therapeutic approaches, though the modest mediation proportion indicates this represents just one component of a complex network of mechanisms.

B cell subpopulations exert dualistic effects in EC pathogenesis. Univariable MR analysis identified naïve B-cell phenotypes (e.g., CD19 on IgD⁺CD38⁻ B cells) as protective factors, aligning with evidence that tumor-infiltrating B cells contribute to antitumor immunity via antigen presentation and lymphoid structure formation (46). A previous MR analysis of 190 B cell phenotypes identified five with causal relationships to EC risk, mostly in a protective direction, which were further validated in an independent population (47). Notably, “CD19⁺IgD⁺CD38⁻ naive B cells” were identified as protective factors in both that study and ours. This suggests that B cell-mediated humoral immunity may inhibit the development of EC, with mechanisms potentially involving the production of anti-tumor antibodies, T cell assistance, and the formation of immunological memory. In contrast, CD38⁺CD20⁻ plasmablast-like B cells were associated with elevated EC risk. While some studies correlate plasmablast infiltration with favorable prognosis, possibly reflecting antitumor antibody responses (48), others report that these cells can produce IL-10, inhibit T-cell function, and foster an immunosuppressive microenvironment (49). Our genetic evidence supports the latter paradigm, suggesting excessive plasma blast responses may drive immune evasion. Notably, CD38 is not only a marker of cell differentiation but also an enzyme (NAD⁺ hydrolase) that can deplete NAD⁺ and generate immunosuppressive molecules such as cyclic ADP-ribose and adenosine (50). Whether CD38⁺ plasmablasts impair T cell function via this enzymatic activity warrants further investigation. These results highlight a functional dichotomy in B cell-mediated antitumor immunity: naïve B subsets exert protective effects, whereas terminally differentiated plasmablasts may adopt an immunosuppressive role under specific microenvironmental conditions. Thus, the heterogeneity of B cells needs to be considered when developing immunotherapies.

Additionally, the CD45⁺CD33⁺HLA-DR⁺CD14dim subset was identified as a risk factor for EC. This population displays phenotypic features of MDSCs, which can deplete local arginine via the arginase-1 pathway in the TME to inhibit T-cell function (51). Our study further revealed significant associations between this subset and two phosphatidylcholine metabolites, suggesting a potential role of phospholipid metabolism in immune regulation. However, lipids such as PC(16:0_20:5) did not demonstrate statistically significant mediation effects in our analyses. This lack of significant mediation at the individual lipid level may reflect the complex nature of lipid metabolic networks, where coordinated changes across multiple lipid species collectively influence biological processes rather than single molecules acting in isolation (52). Future investigations employing composite lipidomic indices or enzyme-centric MR approaches targeting key lipid-metabolizing enzymes may help elucidate the system-level relationships between lipid metabolism and immunosuppressive activity in the TME.

In summary, our genetic epidemiology study provides evidence supporting causal effects of specific immune cell subsets on EC and identifies metabolite-mediated pathways. These findings highlight the potential of considering both immunological and metabolic pathways for future therapeutic strategies. While current immunotherapies focus predominantly on T-cell checkpoints, our results suggest myeloid cells and their metabolic crosstalk represent complementary therapeutic targets. Approaches targeting MDSCs or modulating aspartate metabolism warrant further investigation as potential adjuvants to existing regimens. However, as with all MR findings, these results require validation through experimental studies and clinical trials before any translational applications can be considered.

This study had several limitations. First, the EC GWAS data were sourced exclusively from the FinnGen R12 release. This cohort includes a limited number of cases (N=1,277) and is predominantly composed of individuals of Finnish and Nordic ancestry. This not only constrains the statistical power to detect modest genetic associations and increases the risk of false negatives but may also limit the generalizability of the findings across diverse populations and EC subtypes. Future studies should aim to validate these findings in larger, multi-ancestry datasets, such as the UK Biobank and the International Cancer Genome Consortium. Furthermore, the summary-level GWAS data did not allow for the assessment of potential confounding by comorbidities (e.g., hematological cancers) that might influence both immune cell phenotypes and EC risk. Future studies with access to individual-level data are needed to adjust for these important clinical factors and validate the independence of the identified genetic effects. Finally, this study was constrained by the lack of histological subtyping in the FinnGen dataset, which combined both EAC and ESCC cases. This precludes subtype-specific inferences, and future research with stratified GWAS data is warranted to explore the potentially distinct immunometabolic mechanisms in each subtype.

Conclusions

Through two-sample MR, we found that genetic predispositions to increased levels of CD38 on CD20⁻ B cells, CD33 on CD14⁺ monocytes, and CD45 on CD33⁺HLA^–DR⁺CD14dim cells are significantly positively correlated with the risk of EC, suggesting that these immune cell subsets may promote the development of EC. Additionally, we demonstrated the mediating role of plasma metabolomes and the plasma lipidomes in the impact of immune cells on EC. Among them, the Asp/Cit ratio was identified as a key metabolic node mediating the pro-carcinogenic effect of monocytes, accounting for approximately 10.6% of the effect. These findings offer novel insights into the causal relationships between immune cells, metabolites, and EC progression, providing new perspectives and potential therapeutic targets for understanding and treating EC.

Acknowledgments

We would like to thank the investigators and participants from the genome-wide association study (GWAS) Catalog, University of Bristol, IEU Open GWAS initiative, and FinnGen consortium for providing open-access data resources critical to this study.

Footnote

Reporting Checklist: The authors have completed the STROBE-MR reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-774/rc

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-774/prf

Funding: This work was supported by the Shenzhen’s San Ming Project (No. SZSM202211036), the Shenzhen Science and Technology Innovation Commission Project (No. JCYJ20230807095102004), the Huilan Charity – Special Research Fund for Precision Medicine in Lung Cancer (No. HL-HS2020-93), the Medical and Health Guiding Project of Xiamen (No. 3502Z20214ZD1011), and the Natural Science Foundation of Fujian Province (No. 2022J011380).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2025-774/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Han B, Zheng R, Zeng H, et al. Cancer incidence and mortality in China, 2022. J Natl Cancer Cent 2024;4:47-53. [Crossref] [PubMed]
2. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
3. Sheikh M, Roshandel G, McCormack V, et al. Current Status and Future Prospects for Esophageal Cancer. Cancers (Basel) 2023;15:765. [Crossref] [PubMed]
4. Shoji Y, Koyanagi K, Kanamori K, et al. Immunotherapy for esophageal cancer: Where are we now and where can we go. World J Gastroenterol 2024;30:2496-501. [Crossref] [PubMed]
5. Xu W, Zhang T, Zhu Z, et al. The association between immune cells and breast cancer: insights from Mendelian randomization and meta-analysis. Int J Surg 2025;111:230-41. [Crossref] [PubMed]
6. Kiyozumi Y, Baba Y, Okadome K, et al. IDO1 Expression Is Associated With Immune Tolerance and Poor Prognosis in Patients With Surgically Resected Esophageal Cancer. Ann Surg 2019;269:1101-8. [Crossref] [PubMed]
7. Yagi T, Baba Y, Okadome K, et al. Tumour-associated macrophages are associated with poor prognosis and programmed death ligand 1 expression in oesophageal cancer. Eur J Cancer 2019;111:38-49. [Crossref] [PubMed]
8. Li J, Xie Y, Wang X, et al. Prognostic impact of tumor-associated macrophage infiltration in esophageal cancer: a meta-analysis. Future Oncol 2019;15:2303-17. [Crossref] [PubMed]
9. Guo C, Wu L, Xu L. Mendelian randomization analysis reveals potential causal relationships between serum lipid metabolites and prostate cancer risk. Discov Oncol 2025;16:602. [Crossref] [PubMed]
10. Leone RD, Zhao L, Englert JM, et al. Glutamine blockade induces divergent metabolic programs to overcome tumor immune evasion. Science 2019;366:1013-21. [Crossref] [PubMed]
11. O'Sullivan D, Sanin DE, Pearce EJ, et al. Metabolic interventions in the immune response to cancer. Nat Rev Immunol 2019;19:324-35. [Crossref] [PubMed]
12. Liu Y, Yan W, Tohme S, et al. Hypoxia induced HMGB1 and mitochondrial DNA interactions mediate tumor growth in hepatocellular carcinoma through Toll-like receptor 9. J Hepatol 2015;63:114-21. [Crossref] [PubMed]
13. Poznanski SM, Singh K, Ritchie TM, et al. Metabolic flexibility determines human NK cell functional fate in the tumor microenvironment. Cell Metab 2021;33:1205-1220.e5. [Crossref] [PubMed]
14. Kurki MI, Karjalainen J, Palta P, et al. FinnGen provides genetic insights from a well-phenotyped isolated population. Nature 2023;613:508-18. [Crossref] [PubMed]
15. Orrù V, Steri M, Sidore C, et al. Complex genetic signatures in immune cells underlie autoimmunity and inform therapy. Nat Genet 2020;52:1036-45. [Crossref] [PubMed]
16. Chen Y, Lu T, Pettersson-Kymmer U, et al. Genomic atlas of the plasma metabolome prioritizes metabolites implicated in human diseases. Nat Genet 2023;55:44-53. [Crossref] [PubMed]
17. Ottensmann L, Tabassum R, Ruotsalainen SE, et al. Genome-wide association analysis of plasma lipidome identifies 495 genetic associations. Nat Commun 2023;14:6934. [Crossref] [PubMed]
18. Rühlemann MC, Hermes BM, Bang C, et al. Genome-wide association study in 8,956 German individuals identifies influence of ABO histo-blood groups on gut microbiome. Nat Genet 2021;53:147-55. [Crossref] [PubMed]
19. Qin Y, Havulinna AS, Liu Y, et al. Combined effects of host genetics and diet on human gut microbiota and incident disease in a single population cohort. Nat Genet 2022;54:134-42. [Crossref] [PubMed]
20. Wang Q, Shi Q, Wang Z, et al. Integrating plasma proteomes with genome-wide association data for causal protein identification in multiple myeloma. BMC Med 2023;21:377. [Crossref] [PubMed]
21. 1000 Genomes Project Consortium; Auton A, Brooks LD, et al. A global reference for human genetic variation. Nature 2015;526:68-74.
22. Cai J, He L, Wang H, et al. Genetic liability for prescription opioid use and risk of cardiovascular diseases: a multivariable Mendelian randomization study. Addiction 2022;117:1382-91. [Crossref] [PubMed]
23. Lv X, Hu Z, Liang F, et al. Causal relationship between ischemic stroke and its subtypes and frozen shoulder: a two-sample Mendelian randomization analysis. Front Neurol 2023;14:1178051. [Crossref] [PubMed]
24. Shi X, Wang T, Teng D, et al. A mendelian randomization study investigates the causal relationship between immune cell phenotypes and cerebral aneurysm. Front Genet 2024;15:1333855. [Crossref] [PubMed]
25. Burgess S, Scott RA, Timpson NJ, et al. Using published data in Mendelian randomization: a blueprint for efficient identification of causal risk factors. Eur J Epidemiol 2015;30:543-52. [Crossref] [PubMed]
26. Bowden J, Davey Smith G, Haycock PC, et al. Consistent Estimation in Mendelian Randomization with Some Invalid Instruments Using a Weighted Median Estimator. Genet Epidemiol 2016;40:304-14. [Crossref] [PubMed]
27. Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol 2015;44:512-25. [Crossref] [PubMed]
28. Hartwig FP, Borges MC, Horta BL, et al. Inflammatory Biomarkers and Risk of Schizophrenia: A 2-Sample Mendelian Randomization Study. JAMA Psychiatry 2017;74:1226-33. [Crossref] [PubMed]
29. Hartwig FP, Davey Smith G, Bowden J. Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. Int J Epidemiol 2017;46:1985-98. [Crossref] [PubMed]
30. Zhao J, Ming J, Hu X, et al. Bayesian weighted Mendelian randomization for causal inference based on summary statistics. Bioinformatics 2020;36:1501-8. [Crossref] [PubMed]
31. Burgess S, Thompson SG. Multivariable Mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects. Am J Epidemiol 2015;181:251-60. [Crossref] [PubMed]
32. Verbanck M, Chen CY, Neale B, et al. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 2018;50:693-8. [Crossref] [PubMed]
33. Hemani G, Zheng J, Elsworth B, et al. The MR-Base platform supports systematic causal inference across the human phenome. Elife 2018;7:e34408. [Crossref] [PubMed]
34. Yavorska OO, Burgess S. MendelianRandomization: an R package for performing Mendelian randomization analyses using summarized data. Int J Epidemiol 2017;46:1734-9.
35. Carter AR, Sanderson E, Hammerton G, et al. Mendelian randomisation for mediation analysis: current methods and challenges for implementation. Eur J Epidemiol 2021;36:465-78. [Crossref] [PubMed]
36. O'Donovan C, Davern M, Donlon NE, et al. Chemokine-targeted therapies: An opportunity to remodel immune profiles in gastro-oesophageal tumours. Cancer Lett 2021;521:224-36. [Crossref] [PubMed]
37. Wu P, Wu D, Ni C, et al. γδT17 cells promote the accumulation and expansion of myeloid-derived suppressor cells in human colorectal cancer. Immunity 2014;40:785-800. [Crossref] [PubMed]
38. Bindea G, Mlecnik B, Tosolini M, et al. Spatiotemporal dynamics of intratumoral immune cells reveal the immune landscape in human cancer. Immunity 2013;39:782-95. [Crossref] [PubMed]
39. Koinis F, Vetsika EK, Aggouraki D, et al. Effect of First-Line Treatment on Myeloid-Derived Suppressor Cells' Subpopulations in the Peripheral Blood of Patients with Non-Small Cell Lung Cancer. J Thorac Oncol 2016;11:1263-72. [Crossref] [PubMed]
40. Zhang Z, Huang W, Hu D, et al. E-twenty-six-specific sequence variant 5 (ETV5) facilitates hepatocellular carcinoma progression and metastasis through enhancing polymorphonuclear myeloid-derived suppressor cell (PMN-MDSC)-mediated immunosuppression. Gut 2025;74:1137-49. [Crossref] [PubMed]
41. Garcia-Bermudez J, Baudrier L, La K, et al. Aspartate is a limiting metabolite for cancer cell proliferation under hypoxia and in tumours. Nat Cell Biol 2018;20:775-81. [Crossref] [PubMed]
42. Drexler K, Schmidt KM, Jordan K, et al. Cancer-associated cells release citrate to support tumour metastatic progression. Life Sci Alliance 2021;4:e202000903. [Crossref] [PubMed]
43. Icard P, Prieto M, Coquerel A, et al. Why and how citrate may sensitize malignant tumors to immunotherapy. Drug Resist Updat 2025;78:101177. [Crossref] [PubMed]
44. Wang J, Hong M, Cheng Y, et al. Targeting c-Myc transactivation by LMNA inhibits tRNA processing essential for malate-aspartate shuttle and tumour progression. Clin Transl Med 2024;14:e1680. [Crossref] [PubMed]
45. Miglionico R, Matera I, Ventola GM, et al. Gene Expression Reprogramming by Citrate Supplementation Reduces HepG2 Cell Migration and Invasion. Int J Mol Sci 2024;25:6509. [Crossref] [PubMed]
46. Ma J, Wu Y, Ma L, et al. A blueprint for tumor-infiltrating B cells across human cancers. Science 2024;384:eadj4857. [Crossref] [PubMed]
47. Guo J, Si G, Song X, et al. Association of B cells and the risk of Esophageal cancer: a bidirectional two-sample mendelian randomization study. BMC Cancer 2024;24:1416. [Crossref] [PubMed]
48. Schlößer HA, Thelen M, Lechner A, et al. B cells in esophago-gastric adenocarcinoma are highly differentiated, organize in tertiary lymphoid structures and produce tumor-specific antibodies. Oncoimmunology 2019;8:e1512458. [Crossref] [PubMed]
49. Bao J, Betzler AC, Hess J, et al. Exploring the dual role of B cells in solid tumors: implications for head and neck squamous cell carcinoma. Front Immunol 2023;14:1233085. [Crossref] [PubMed]
50. Lin L, Roccuzzo G, Yakymiv Y, et al. The CD39/CD73/Adenosine and NAD/CD38/CD203a/CD73 Axis in Cutaneous T-Cell Lymphomas. Cells 2025;14:309. [Crossref] [PubMed]
51. Pan J, Lin Y, Liu X, et al. Harnessing amino acid pathways to influence myeloid cell function in tumor immunity. Mol Med 2025;31:44. [Crossref] [PubMed]
52. van Meer G, Voelker DR, Feigenson GW. Membrane lipids: where they are and how they behave. Nat Rev Mol Cell Biol 2008;9:112-24. [Crossref] [PubMed]