Decoding Immune Influence: A Mendelian Randomisation Study on Dilated Cardiomyopathy

Dilated cardiomyopathy (DCM) is a condition characterised by left ventricular or biventricular dilation and systolic dysfunction in the absence of abnormal loading conditions, such as primary valve disease or coronary artery disease, that can account for the ventricular remodelling.¹ The aetiology of DCM is complex and multifactorial, involving both genetic and acquired factors, such as infections, toxins and immune-mediated diseases.¹ However, the exact mechanisms underlying the initiation and progression of DCM remain elusive and current treatments are largely symptomatic and palliative.²

Immune cells are essential components of both the innate and adaptive immune system, playing vital roles in pathogen defence, maintaining homeostasis and modulating inflammation.³ Immune cells can be categorised based on their origin, phenotype, function and interaction with other cells.⁴ Previous studies have demonstrated that immune cells are intricately involved in various cardiovascular diseases, such as atherosclerosis, myocarditis, ischaemia-reperfusion injury and heart failure.^5–8 Recent immunophenotyping research has further underscored the heterogeneity of immune responses in cardiovascular conditions, highlighting differences in cell-mediated inflammation that may influence disease outcomes.

Mendelian randomisation (MR) is a statistical method that leverages genetic variants as instrumental variables to infer the causal effect of an exposure on an outcome, exploiting the random allocation of alleles at conception.⁹ MR can overcome key limitations of conventional observational studies, including confounding, reverse causation and measurement error. Previous MR studies have investigated the causal association between immune cell traits and various diseases, such as cancers and psychiatric disorders.^10,11 Recent advances in MR have enabled researchers to infer causal relationships between specific immune cell traits and cardiovascular pathologies. For example, our recent study on aortic dissection demonstrated that several immunophenotypes - including the median fluorescence intensity of CD19 on transitional B-cells and activation markers on CD39⁺ regulatory T-cells - had significant causal effects on disease risk, thereby offering novel insights into risk stratification and potential therapeutic targets.¹² Similarly, another MR investigation provided compelling evidence that immune cell traits, such as CD20 expression on immunoglobulin D (IgD)⁺CD38⁻ B-cells and CD127 levels on CD28⁺CD4⁺ T cells are causally linked with the development of abdominal and thoracic aortic aneurysms, even after accounting for confounding factors, such as lipids (e.g. cholesterol and triglyceride levels) and smoking.¹³

Despite these advances, the role of immunophenotypes in the pathogenesis of DCM remains poorly defined. In this study, we aimed to fill this critical gap and performed a comprehensive two-sample MR analysis using genetic variants from large-scale genome-wide association studies (GWAS) to test the causal effect of 731 distinct immune cell traits on DCM risk. Notably, we categorised these immune traits into four groups - median fluorescence intensity (MFI), relative counts (RC), absolute counts (AC) and morphological parameters (MP). This classification was selected because it captures different dimensions of immune regulation, providing insights into cell surface antigen expression, circulating cell proportions, overall cell numbers and cell morphology, all of which may differentially contribute to DCM pathogenesis. Our goal was to identify the immune cell types that have a causal association with DCM, thereby providing new insights into the underlying mechanisms and potential therapeutic targets for this devastating disease.

Methods

Study Design

Our MR study aligns with the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomisation (STROBE-MR) guidelines.¹⁴ We applied a two-sample Mendelian randomisation approach to investigate the causal relationship between 731 immune cell features and DCM. We also contrasted the MR results of DCM with another outcome, chronic heart failure (CHF), to identify the immune cell types that have a specific causal relationship with DCM. Mendelian randomisation uses genetic variants as proxies for risk factors, and thus, valid instrumental variables for causal inference must satisfy three key criteria: genetic variants are strongly associated with the exposure; genetic variants are independent of potential confounders between the exposure and the outcome; and genetic variants affect the outcome only through the exposure.⁹

Data Sources

GWAS data sources for CHF and DCM data were derived from a cross-population atlas of genetic associations for 220 human phenotypes.¹⁵ CHF data were obtained from large-scale GWAS data from the UK Biobank, which used a fast and efficient sparse matrix-based algorithm – fastGWA generalised linear mixed model (GLMM) – for GWAS applied to 456,348 individuals.¹⁶ This UK Biobank and FinnGen meta-analysis (total=628,000) identified about 5,000 new loci, increasing the resolution of the human trait genomic atlas.

Immunity-wide GWAS data sources GWAS summary statistics data for 731 immune phenotypes were publicly available from the GWAS Catalog (ID from GCST0001391 to GCST0002121).⁴ Immune cells included absolute cell counts AC (n=118), median MFI reflecting surface antigen levels (n=389), MP (n=32) and relative cell (RC) counts (n=192). MFI, AC and RC traits included B cells, classic dendritic cells, T-cell maturation stages, monocytes, myeloid cells, T- and B-lymphocyte and natural killer cell profiles and Treg panels. MP traits included classic dendritic cells and T- and B-lymphocyte and natural killer panels. The immune trait GWAS was based on 3,757 Europeans; covariates included sex, age and approximately 22 million single nucleotide polymorphisms (SNPs) genotyped using high-density arrays imputed using a reference panel based on European sequences.

Quality Control of the Single Nucleotide Polymorphisms

To ensure the validity of the analysis, we performed quality control on SNP, including: removing all non-biallelic SNPs; removing all SNPs with strand-ambiguous alleles (SNPs with A/T, C/G allele); removing SNPs without rs IDs, duplicated rs IDs or base-pair (bp) position; removing SNPs not in 1,000 Genomes Project Phase 3 (1,000GP) (or any other reference panel); removing SNPs whose bp positions or alleles did not match those in 1,000GP Phase 3 (or any other reference panel); removing SNPs with imputation INFO score less than 0.9; removing all SNPs on chromosome X, Y and mitochondrion; removing SNPs with sample size 5 standard deviations away from the mean.¹⁷ This was to avoid scenarios where some SNPs were genotyped on a specialised genotyping array and had substantially more samples than the rest.

Selection of Instrumental Variables

For each immune cell phenotype, we selected SNPs that met a significance threshold of p<1×10⁻⁵.¹⁸ This relatively lenient criterion was adopted to maximise the number of potential instruments given the modest sample size for immune cell traits - contrasting with outcomes such as DCM/CHF, for which a more stringent threshold (p<5×10⁻⁸) was applied.^12,19 Subsequently, we implemented the clumping procedure using PLINK software to remove SNPs in linkage disequilibrium, applying an r² cut-off of 0.001 based on the European 1,000GP reference panel. In cases where a specific instrumental variable was absent from the outcome dataset, we substituted it with a proxy variant that exhibited an r²>0.8.

To quantify the strength and relevance of each instrument, we calculated both the proportion of variance explained (R²) and the corresponding F-statistic, thereby determining how effectively each genetic variant captured the variability in the exposure. We used scatter plots, funnel plots and leave-one-out cross-validation (LOOCV) to visualise the associations of genetic variants with exposure and outcome, and to evaluate the validity and consistency of genetic variants as instrumental variables. Supplementary Table 1 summarises the details of the SNPs selected as genetic instruments.

Mendelian Randomisation Study

In our study, we assessed the overall impact of risk factors on the likelihood of outcome through a two-sample MR analysis, using the fixed/random-effects inverse-variance weighting (IVW) method as our primary analytical tool. The association between each exposure and outcome was represented as an OR for each unit increase in the genetically predicted exposure attribute. We applied a Bonferroni-adjusted significance threshold of 0.005 to account for multiple comparisons in our main analysis. To ensure the robustness of our findings, we conducted sensitivity analyses using methods such as the weighted median and Mendelian Randomisation-Pleiotropy RESidual Sum and Outlier (MR-PRESSO) to evaluate the impact of horizontal pleiotropy. The MR-PRESSO framework identifies horizontal pleiotropy by regressing the effects of genetic variants on outcomes against their effects on exposure, removing outliers and reapplying the IVW method without them. The weighted median approach provides a robust causal estimate, assuming that the majority of the weight comes from valid genetic variants. MR-Egger regression evaluates the associations between variants and outcomes against variant-exposure correlations, weighted by the precision of the estimates, to identify horizontal pleiotropy under the instrument strength independent of direct effect (InSIDE) assumption. We also employed debiased inverse variance weighted (D-IVW) and MR-robust adjusted profile score, D-IVW (MR-RAPS). D-IVW, a modification of IVW, accounts for directional pleiotropy, providing more accurate causal estimates. MR-RAPS enhances robustness by adjusting for weak instruments and residual pleiotropy. These methods were implemented using various R packages for MR analysis, ensuring a comprehensive evaluation of causal relationships. Multivariable MR analysis addressed potential horizontal pleiotropy across traits, and reverse MR analysis alleviated concerns about reverse causation.

Pathway and Tissue Enrichment Analysis

We identified significant cis-acting Expression Quantitative Trait Loci (cis-eQTLs) using the Genotype-Tissue Expression Program (GTeX) Consortium database, revealing variations in gene expression. We mapped risk loci genes linked to causative immune cells at a significance threshold of p<0.10. Subsequent tissue-specific expression and pathway analyses were conducted on these key genes using Functional Mapping and Annotation of GWAS (FUMAl; https://fuma.ctglab.nl/) for tissue-specific expression and enrichment analyses. Additionally, we employed Metascape tools and the clusterProfiler package in R software to integrate Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), WikiPathways and cell-specific gene set databases for comprehensive pathway analysis.

Statistical Analysis

Using R software (version 4.2.1), we employed various MR techniques to explore the causal relationships between 731 immunophenotypes and outcomes. Instrumental variable heterogeneity was assessed with Cochran’s Q statistic. If the null hypothesis was rejected, we used a random effects IVW model. Horizontal pleiotropy was examined with the MR-Egger method, while MR-PRESSO was used to identify and correct outlier-induced biases. We used the false discovery rate (FDR) method to control for false positives in multiple tests, with FDR<0.15 and p<0.05 considered significant. Figure 1 provides a schematic overview of the study methodology, including the criteria used for selecting instrumental variables.

Ethics Statement

This research adhered to the World Medical Association’s Code of Ethics (Declaration of Helsinki). The datasets from the UK Biobank database and the GWAS Catalog were obtained from publicly accessible databases and had received prior approval from their respective Ethics Committees. Written informed consent was collected from all participants in the original studies for the UK Bank Consortium and GWAS Catalog datasets used in this study.

Results

Alignment of Genetic Variants with Causal Effect Estimates

Scatter plots demonstrate that all genetic variants align with the overall causal effect estimate and cluster closely around the regression line (Supplementary Figure 1A). The funnel plot shows a symmetrical distribution, with horizontal lines for each SNP on both sides of the vertical line representing the overall causal effect estimate (Supplementary Figure 1B). The LOOCV bar chart displays a unimodal distribution, with the peak value near the overall p-value (Supplementary Figure 1C). These plots support the assumptions of MR analysis, indicating that our SNPs are valid instrumental variables and our results are robust and reliable.

Causal Effect of Immunophenotypes on the Onset of Dilated Cardiomyopathy

We applied a two-sample MR analysis with the IVW method as the primary analysis to examine the causal effects of immunophenotypes on DCM and CHF. Figure 2 shows the immune cell types that are significantly associated with either of these outcomes. The figure reveals that the immune cell type profiles related to DCM and CHF are distinct and that none of the immune cell types are significantly associated with the broad definition of CHF, which encompasses both ischaemic and non-ischaemic causes (Supplementary Table 2). We considered immune cells with FDR <0.15 as significant causal associations. After adjusting for multiple testing using the FDR method, we detected a positive causal effect of one immunophenotype on DCM: the relative count (%) of CD4⁺ Treg T cells (Figure 3). The IVW method estimated the OR of the relative count (%) of CD4⁺ Treg T cells for DCM risk to be 1.144 (95% CI [1.069–1.223]; p=9.36E-05, FDR=0.0684) (Supplementary Table 2).

Sensitivity Analysis

We obtained consistent results using other MR methods: weighted mode (OR 1.114; 95% CI [1.025–1.211]; p=0.011); weighted median (OR 1.102; 95% CI [0.995–1.220]; p=0.063); DIVW (OR 1.151; 95% CI [1.070–1.238]; p<0.001), MR-RAPS (OR 1.149; 95% CI [1.067–1.238]; p<0.001), MR-Egger (OR 1.100; 95% CI [1.007–1.202]; p=0.035) and MR-PRESSO (OR 1.144; 95% CI [1.084–1.206]; p<0.001) (Figure 2, Supplementary Table 3). No heterogeneity was detected by the heterogeneity test (p=0.948). Moreover, the MR-Egger intercept (p=0.186, Supplementary Table 3) and the MR-PRESSO global test (p=0.96, Supplementary Table 4) ruled out horizontal pleiotropy for the association. Since we identified only one immune cell trait with a causal relationship with DCM, we did not perform a multivariate MR analysis.

Reverse Mendelian Randomisation Causal Effect Analysis

We performed a two-sample MR analysis with the IVW method as the primary analysis to examine the causal effects of DCM on the immune trait. No causal effect was identified at a significance level of 0.05 after adjusting for multiple testing using the FDR method (Supplementary Table 5). The DIVW and MR-RAPS methods confirmed the robustness of the causal association observed.

Pathway Analysis of Risk Single Nucleotide Polymorphisms-related-eQTLs

We performed genetic mapping of risk SNPs for the causal immune cell trait relative count of CD4⁺ Treg T-cells (Supplementary Table 6) and conducted tissue-specific and pathway analysis of the mapped genes. These genes were broadly expressed in various solid tissue organs, with the cardiovascular system being one of the main tissue types enriched with specificity (Figure 4). Tissue enrichment analysis results showed that these differentially expressed genes were predominantly distributed in lung, aorta, omental adipose, spleen, thyroid, EB virus-transformed lymphocytes, left atrium and left ventricle (Figure 5). The cardiovascular tissue types included coronary artery, aorta, left atrium and left ventricle, and the associated cell types included fibroblasts and EB virus-transformed lymphocytes. The most enriched term for biological process was ‘positive regulation of leukocyte cell-cell adhesion’ (Figure 6A, Supplementary Table 7). Comprehensive pathway analysis conducted by Metascape tool showed that ‘regulation of inflammatory response’ was the most important pathway (Figure 6B).

Discussion

We applied a two-sample MR analysis to examine the causal associations between 731 immune cell traits and DCM. To our knowledge, this is the first MR analysis to explore the causal relationship between multiple immunophenotypes and DCM. Among four types of immune traits (MFI, RC, AC and MP), we found that only one immunophenotype of relative count of CD4⁺ Treg T-cell had a significant causal effect on DCM (FDR<0.15). The IVW method of MR analysis showed that a one-unit increase in the relative count (%) of CD4⁺ Treg T cells was associated with a 14.4% increase in the odds of DCM. This finding is inconsistent with previous studies that have reported the protective role of Treg cells in DCM by suppressing the inflammatory response and attenuating the cardiac fibrosis.^20,21 A potential explanation for this apparent discrepancy lies in the distinction between relative and absolute immune cell metrics. The relative count of CD4⁺ Treg cells may not accurately reflect their absolute abundance, functional phenotype or suppressive capacity. We hypothesise that while Treg cells may exert a protective role in DCM under physiological conditions, an elevated relative proportion could result from a reduction in other immune cell subsets, rather than from an actual expansion of functional Tregs. This shift in immune cell composition may reflect an underlying dysregulation that contributes to disease progression rather than protection. Moreover, Treg cell function is highly context-dependent and influenced by a range of factors, including cytokine milieu, metabolic state, surface marker expression and transcriptional regulation.^22,23 For instance, while Foxp3 expression is a canonical marker of Treg identity, functional stability and suppressive activity can be compromised under inflammatory or metabolically stressed conditions, leading to phenotypic plasticity or even conversion into pro-inflammatory subsets.²⁴

It is therefore plausible that a higher relative count of CD4⁺ Treg cells may coexist with impaired regulatory function, insufficient to counteract the inflammatory milieu characteristic of DCM. To test this hypothesis, future studies should incorporate quantitative assessments of absolute Treg cell counts, phenotypic markers, such as CD25, CD127 and Foxp3, and functional assays, such as suppression of effector T-cell proliferation or cytokine secretion. Flow cytometry and ex vivo suppression assays remain gold-standard tools for such evaluations.²⁵ These additional data would help clarify whether the observed causal association reflects a compensatory immune response, a dysfunctional regulatory phenotype or a broader shift in immune homeostasis. Ultimately, integrating MR findings with immunophenotypic and functional validation could provide a more comprehensive understanding of Treg cell dynamics in DCM pathogenesis.

To further investigate the biological role of immune cells in DCM development, we mapped the risk SNPs of the causal immune cell and identified 22 key genes. Tissue enrichment analysis results showed that these differentially expressed genes were predominantly distributed in lung, aorta, omental adipose, spleen, thyroid, lymphocytes, left atrium and left ventricle, with the cardiovascular system being one of the main tissue types enriched with specificity. The most enriched term for biological process, molecular function and functional pathway was ‘positive regulation of leukocyte cell-cell adhesion’, ‘transcription coregulation activity’, and ‘human T-cell leukaemia virus 1 infection’, respectively. Comprehensive pathway analysis conducted by the Metascape tool showed that ‘regulation of inflammatory response’ was the most important pathway. Our findings are consistent with previous studies that have suggested the involvement of immune cells and inflammatory response in the pathogenesis of DCM. For example, Wang et al. reviewed the molecular role of various immune cells, such as macrophages, T-cells, B-cells and mast cells in DCM and highlighted the importance of cytokine signalling and immune regulation in cardiac remodelling and fibrosis of Treg cells, a subset of CD4⁺ T cells with immunosuppressive function with DCM.³ Wei et al. found that the frequency of Treg cells was significantly decreased in DCM patients compared with healthy controls, and that the Treg cells from DCM patients had impaired suppressive function and reduced expression of Foxp3, a key transcription factor for Treg cells.²¹ It was also demonstrated that the Treg expansion could attenuate cardiac dysfunction and fibrosis in a mouse model of DCM.²⁶ Similarly, Martin et al. also found in a mouse model of myocarditis that a significant increase in the number of Treg cells and the expression levels of anti-inflammatory cytokines can reduce cell infiltration, fibrosis and malnutrition calcification, which is a new beneficial tool for intervening in inflammatory cardiomyopathy.²⁷

Given this immunological context, our identification of CD4⁺ Treg cells as a potential causal factor in DCM raises the possibility of their use as a biomarker for early diagnosis or risk stratification. Integrating Treg-related markers - such as CD25⁺CD127^{^low} expression or Foxp3 levels - into existing diagnostic frameworks, for example, alongside echocardiographic or biomarker-based assessments, such as N-terminal pro-B-type natriuretic peptide, could enhance the sensitivity of detecting subclinical immune dysregulation in DCM patients. Moreover, longitudinal monitoring of Treg cell dynamics may offer prognostic value or guide immunomodulatory interventions.

In addition, our genetic mapping and pathway enrichment analyses revealed that loci associated with CD4⁺ Treg traits were enriched in immune-regulatory and cardiac-relevant pathways, including cytokine signalling, T-cell receptor signalling, and extracellular matrix remodelling. These findings align with recent multi-omics studies that implicate immune-metabolic and redox pathways in DCM pathogenesis.²⁸ By linking immune cell traits to specific genomic regions and biological processes, our study contributes to a more integrated understanding of how immune dysregulation may drive myocardial remodelling and dysfunction in DCM.

Limitations

This study has several limitations, categorised into methodological, sample-specific and data-related aspects:

Methodological limitations

As with all MR studies, our analysis depends on key assumptions - relevance, independence and exclusion restriction. While we conducted comprehensive sensitivity analyses (MR-Egger, weighted median, leave-one-out) to test robustness, residual pleiotropy cannot be completely excluded. In addition, immune traits are inherently dynamic and may fluctuate over time. Although the original dataset used standardised protocols to reduce technical variation, its cross-sectional nature limits insights into long-term immune behaviour - an acknowledged constraint in MR studies involving immunophenotypes.

Sample-specific limitations

Our immune cell data were derived from peripheral blood, which may not fully reflect immune activity in cardiac tissue. Validation in myocardial biopsies or experimental models is needed to confirm tissue relevance. Moreover, DCM cases were drawn from the UK Biobank, a largely European population, which may limit generalisability. Replication in diverse, clinically enriched cohorts is essential.

Data-related limitations

Risk gene identification was based on positional mapping, which may overlook regulatory elements or context-specific expression. To enable broader discovery in exploratory analyses, we applied an FDR threshold of <0.15, accepting a moderate false-positive risk to facilitate downstream pathway exploration. This balance between sensitivity and specificity is acknowledged. Future studies integrating multi-omics approaches are warranted to refine gene-pathway relationships.

Conclusion

In this study, we identified a significant causal association between the relative abundance of CD4⁺ Treg cells and the risk of DCM, using a robust MR framework. We further mapped 22 candidate genes and several immune-related pathways potentially involved in DCM pathogenesis. These findings offer novel insights into the immunogenetic underpinnings of DCM and suggest that Treg cell dynamics may serve as a promising biomarker or therapeutic target. Future research should focus on validating these associations in cardiac tissue, elucidating the mechanistic role of Treg cells in myocardial remodelling, and exploring the translational potential of immunomodulatory strategies in the prevention and treatment of DCM.

Click here to view Supplementary Material.