Association between human blood metabolome and risk of myocarditis: a mendelian randomization study Authors: Ziyi Wang, Haonan Tian, Jun Wang Published: 2024 ABSTRACT Myocarditis is a common disease of the cardiovascular and immune systems, but the relationship between relevant blood metabolites and the risk of myocarditis has not been well-established. To identify potential biometabolic markers associated with myocarditis, we conducted a two-sample Mendelian randomization (MR) study. We performed preliminary MR analysis using the inverse variance weighted (IVW) method, supplemented by MR-Egger, weighted median, and weighted mode methods to adjust for false discovery rate (FDR). Confounders were screened using the GWAS Catalog website. Sensitivity analyses included Cochrane Q-test, Egger regression, Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR-PRESSO), scatterplots, funnel plots, and forest plots. For genetic and directional analysis, we employed co-localization analysis and the Steiger test. MR analysis was performed using the FinnGen database and meta-analysis was performed using the IEU database. MR analysis identified significant correlations for five metabolic biomarkers after FDR correction. These included four known metabolites: kynurenine, 1-stearoyl-GPE (18:0), deoxycarnitine, and 5-acetylamino-6-formylamino-3-methyluracil, as well as one unknown metabolite, X-25,422. Among these, kynurenine (OR = 1.441, 95%CI = 1.089–1.906, p-value = 0.018) and 1-stearoyl-GPE (18:0) (OR = 1.263, 95%CI = 1.029–1.550, p-value = 0.029) were identified as risk factors for myocarditis, while deoxycarnitine (OR = 0.813, 95%CI = 0.676–0.979, p-value = 0.029), 5-acetylamino-6-formylamino-3-methyluracil (OR = 0.864, 95% CI = 0.775–0.962, p-value = 0.018), and X-25,422 (OR = 0.721, 95%CI = 0.587–0.886, p-value = 0.009) were found to be protective factors. No evidence of heterogeneity, horizontal pleiotropy, or sensitivity issues was observed, and no shared genetic factors between exposure and outcome were detected. The causality was in the correct direction. Meta-analysis further confirmed the causal relationship between the five metabolites and myocarditis. This study identifies a causal relationship between five circulating metabolites and myocarditis. Kynurenine, 1-stearoyl-GPE (18:0), deoxycarnitine, X-25,422, and 5-acetylamino-6-formylamino-3-methyluracil may serve as potential drug targets for myocarditis, providing a theoretical basis for the prevention, diagnosis, and treatment of the condition.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-024-78359-6. Introduction Myocarditis is an inflammatory condition of the heart that can result from infections, toxic substance exposure, or immune system activation. It is frequently underdiagnosed, contributing to acute heart failure, sudden death, and chronic dilated cardiomyopathy1,2. Clinically, myocarditis may present as acute heart failure, ventricular arrhythmias, or cardiogenic shock, which can lead to significant complications and mortality; thus, early diagnosis and treatment are crucial. Biomarkers play a vital role in drug development and are essential for the accurate diagnosis and prognosis of myocarditis, potentially paving the way for new treatment options3. For instance, several studies noted that troponin is a common and significant biomarker in assessing cardiac conditions related to myocarditis4,5. Other immune serum biomarkers, including the erythrocyte sedimentation rate, complement levels, cytokines, and anti-cardiac antibodies, have yet to be prospectively validated for accurately screening biopsy-confirmed myocarditis4. Recent studies have identified various genomic, myocardial injury, cytoarchitectural, and immune cell-related markers of myocarditis3. Known and novel pathways of metabolic dysfunction, previously relevant in older populations or animal models, such as transcriptional regulation, brain-derived neurotrophic factor (BDNF), nitric oxide, and renin-angiotensin pathways, have been recognized in young adults at risk for coronary issues6. In contrast, Tzoulaki et al. demonstrated that metabolites linked to atherosclerosis show disturbances in lipid and carbohydrate metabolism, branched-chain and aromatic amino acid metabolism, oxidative stress, and inflammatory pathways. Their analysis of cardiovascular events revealed negative correlations with creatine, creatinine, and phenylalanine, while direct correlations were found with mannose, acetaminophen-glucuronide, lactate, and ApoB7. Ruiz et al. emphasized the scarcity of longitudinal studies exploring associations between comprehensive metabolomic profiles and cardiovascular disease risk, suggesting that standardizing metabolomics techniques and combining novel methodologies could enhance research in this area8. While studies on metabolites and cardiovascular diseases are prevalent, those specifically targeting myocarditis are rare. Previous research has predominantly focused on more common cardiovascular conditions such as atherosclerosis, hypertension, myocardial infarction, and ischemic stroke8–11. Randomized controlled trials (RCTs) are recognized as the gold standard for inferring causal relationships, but they are often costly and impractical. In contrast, Mendelian randomization (MR) studies offer several advantages; they are generally quicker and less expensive, utilizing existing large-scale genome-wide association study (GWAS) data. MR studies can elucidate potential causal relationships between modifiable risk factors and rare diseases, addressing the need for large sample sizes and extended follow-up inherent in RCTs. Moreover, MR allows the exploration of exposures deemed unethical to test in trials, thereby serving as a valuable research design that mitigates some limitations of traditional observational studies and RCTs12. In this paper, we employed a two-sample MR study design to systematically evaluate the causal relationship between 1,400 blood metabolites and the risk of developing myocarditis. MR serves as a standardized analytical tool to investigate the association between human blood metabolites and myocarditis, aiming to identify key metabolic markers as potential drug targets and to establish a theoretical foundation for the diagnosis and treatment of myocarditis. Methods Study design Three major assumptions that need to be fulfilled by Mendelian randomization are the assumption of association: there is a strong correlation between Single Nucleotide Polymorphism (SNP) and exposure factors; the assumption of independence: SNPs are independent of confounders; and the assumption of exclusivity: SNPs can only contribute to outcomes through exposure factors and confounding factors; and the exclusivity assumption: single nucleotide polymorphisms can only affect outcome through exposure factors. All analyses were performed using the TwoSampleMR (version 0.6.2), MendelianRandomization (version 0.10.0), MRPRESSO package (version 1.0) and coloc (version 5.2.3) in R Software 4.3.1 (https://www.R-project.org). Genetic information on the blood metabolome and myocarditis was based on two large GWAS databases from European populations. This MR study strictly followed the guidelines of STROBE-MR (Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization)13. (Table S1). Blood metabolome data sources A summary of single nucleotide polymorphisms (SNPs) associated with the human blood metabolome comes from a study by Chen et al.14. The study focused on a European population consisting of 8,299 males and females, with plasma metabolic markers selected from the Canadian Longitudinal Study on Aging (CLSA) cohort, encompassing 15,400,000 SNPs. A total of 1,091 metabolites and 309 metabolite ratios were examined through genome-wide association studies, which identified novel gene-metabolite associations that may serve as potential targets for various traits and diseases (https://www.ebi.ac.uk/gwas/publications/36635386). Source of data on myocarditis Genetic information on myocarditis was obtained from the FinnGen database (https://storage.googleapis.com/finngen-public-data-r10/summary_stats/finngen_R10_I9_MYOCARD.gz)15. The population was a European population of males and females containing a sample size of 212,306, SNPs of 19,338,525, 1,654 cases, and 210,652 noncases. To further strengthen the robustness of the results, we selected another myocarditis GWAS from the IEU database for meta-analysis, which included 633 cases and 427,278 noncases from a European population, resulting in a total sample size of 427,911 and comprising 24,180,570 SNPs (https://gwas.mrcieu.ac.uk/datasets/ebi-a-GCST90018882/). Combining the same phenotypes from the two databases for data analysis. Screening of blood metabolomic data First, association analysis was performed to select SNPs that were strongly associated with exposure factors, with a filtering condition of p-value < 5e-8. And to remove SNPs in linkage disequilibrium, which refers to the fact that genetic variants with similar genomic locations are more likely to be inherited together, resulting in a higher chance of alleles belonging to more than two genetic loci to co-occur on a single chromosome than randomly, and therefore we need to remove SNPs with this characteristic16. The kb refers to the genomic range of linkage disequilibrium, while r2 is a metric used to assess the degree of linkage disequilibrium between two SNPs. It measures the correlation of allele frequencies between the two SNPs, ranging from 0 to 1. An r2=1 indicates perfect linkage disequilibrium between the SNPs, whereas an r2=0 signifies complete linkage equilibrium. The SNPs with a strong correlation with the outcome were filtered by setting the range of the region of linkage disequilibrium, kb, to 10,000 and r2 to 0.00117. Next, weak instrumental variables were removed, which implies that the instrumental variables do not have a strong correlation with the SNPs, i.e., SNPs with F < 10 were removed. MR analysis First, perform a Steiger test to exclude SNPs exhibiting reverse causation. The Steiger test showed that the test p-value and direction p-value were both less than 0.05, indicating no reverse causation. Consequently, SNP data with evidence of reverse causation were filtered out. The causal relationship between human blood metabolites and myocarditis was assessed primarily based on inverse-variance weighted (IVW) results. In the “MR analysis” section, we used genetic variant k on outcome Y to calculate the ratio Y/X. Here, X represents the exposure variable, and σ denotes the standard deviation of the estimate of Y. The standard error of the ratio estimate can be approximated using the delta method, where the leading term is σ/X. In the Inverse Variance Weighted (IVW) meta-analysis under a fixed-effects model, the combined ratio estimates from each variant are utilized16,18. In addition, MR-Egger, weighted median, and weighted mode were used as supplementary analyses. The weighted median and weighted mode have been shown to exhibit superior Type 1 error rates in finite sample settings and complement the recently proposed MR-Egger method. Therefore, integrating these methods for analysis is warranted19–21. Subsequently, FDR multiple correction was performed, which aimed at controlling the false-positive error rate in multiple comparisons. The four methods were selected for results with mid-beta values in the same direction. Meanwhile, power calculations were performed using the online tool (https://sb452.shinyapps.io/power/). Finally after clarifying that the phenotype was the relevant phenotype for the outcome, SNPs with confounding factors were screened and removed using the GWAS Catalog website. Sensitivity analysis Sensitivity analyses include heterogeneity analysis, horizontal pleiotropy analysis, outlier detection, and visualization analysis. Heterogeneity analysis includes heterogeneity Q-test18, which requires a p-value > 0.05 to represent the absence of heterogeneity (mainly looking at IVW). The horizontal pleiotropy analysis, such as the MR-Egger test, assesses whether an instrumental variable influences the outcome through factors other than the exposure. Egger regression is a tool used to detect small study biases in meta-analyses and can be applied to pleiotropy bias tests, with the slope coefficients providing estimates of causal effects. Under the assumption that the association of each genetic variant with exposure is independent of the variant’s pleiotropic effect (rather than through exposure), the Egger test gives a valid test of the null hypothesis of causality and a consistent estimate of the causal effect, even if all the genetic variants are null instrumental variables, requiring a p-value > 0.05 to represent no horizontal pleiotropy19; outlier detection Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR-PRESSO), which effectively removes outlier SNPs22. Visualization analysis includes leave-one-out sensitivity analysis, where the effect value after removing a specific SNP should be close to the overall effect value, indicating that the removal did not significantly impact the MR analysis. If the results after removing a single SNP are consistently on one side of the black dotted line, it indicates a minimal effect on the MR analysis. The scatterplot illustrates the effect of SNPs on exposure along the X-axis and on the outcome along the Y-axis, with dots representing SNPs and crosses indicating the fluctuation range of effect values. Horizontal and vertical lines depict the fluctuation range of SNPs on exposure and outcome, respectively, with proximity of the lines suggesting consistency across the four methods. Funnel plots assess symmetry to indicate the absence of heterogeneity, while forest plots reveal that an effect value greater than 0 signifies a risk factor, and a value less than 0 indicates a protective factor. With the above methods, we screened blood metabolic markers with the following characteristics: (1) the adjusted p-values after False Discovery Rate (FDR) under the IVW method; (2) consistent directionality among the four analytical methods; (3) confounders of exposure-related phenotypic non-endpoints; (4) absence of heterogeneity, horizontal pleiotropy, and outliers; and (5) little effect of individual SNPs on the results of the MR analyses. Co-localization analysis The purpose of co-localization is to assess whether two input phenotypes share the same causal variant within a given region23. Venkateswaran et al. conducted the co-localization analysis using the “coloc package” to determine whether causal variants driving two distinct traits are shared or distinct. Coloc employs a Bayesian framework to generate posterior probabilities for five mutually exclusive hypotheses regarding the sharing of causal variants between the two traits: H0 (neither trait has a causal variant); H1 or H2 (causal variant affects one trait only); H3 (two distinct causal variants, one per trait); and H4 (a single causal variant shared between both traits)24. Therefore, a positive co-localization of two GWAS phenotypes indicates shared genetic factors between them, suggesting that genetic variants may jointly influence the development of both diseases or that shared pathological mechanisms exist between the two phenotypes. These shared genetic variants may participate in common biological pathways contributing to the occurrence of both diseases. Conversely, a negative co-localization result between two GWAS phenotypes suggests the absence of shared genetic factors. In such cases, a positive Mendelian randomization result would indicate that any observed relationship between the traits is entirely due to the effects of exposure on the outcomes. Replicative and meta-analysis We selected myocarditis GWAS from the FinnGen and IEU databases for Meta-analysis. The heterogeneity of included data was assessed by calculating the I2 statistic. If I2 ≤ 50% and p-value ≥ 0.1, it was considered that there was no significant heterogeneity among studies, and a fixed-effects model was used. If I2 > 50% and p-value < 0.1, significant heterogeneity among studies was deemed present, and a random-effects model was employed, with a calculation of the 95% confidence interval (CI)25. Statistical significance was defined as p-value < 0.05. Fig. 1Flow chart of data processing. Flow chart of data processing. Results Preliminary MR analysis The Steiger test was used to detect reverse causality between exposure and outcome, and the results showed that the individual and overall SNP tests were in the right direction (steiger_dir = TRUE), suggesting that there is no reverse causality between the blood metabolites and myocarditis. We set the threshold parameter at 5e-8 and, after performing linkage disequilibrium analysis, extracted exposure data that met the criteria, resulting in a total of 2,695 SNPs. Subsequently, we extracted the corresponding outcome data based on the exposure data, filtering out instrumental variables (IVs) that were associated with the outcome. This process yielded a total of 1,494 outcome SNPs. Following data harmonization, we arrived at a final set of 2,586 instrumental variables (1211 blood metabolites) for analysis (Fig. 1, Table S2). Based on these blood metabolites, we performed IVW analysis (after FDR adjustment) and power calculation. Subsequently, we screened for metabolites that showed consistent directional results across four methods: IVW, MR-Egger, weighted median, and weighted mode, with the IVW p-value < 0.05. This process identified five metabolic markers significantly associated with the outcome, including four known metabolites: kynurenine, 1-stearoyl-GPE (18:0), deoxycarnitine, and 5-acetylamino-6-formylamino-3-methyluracil, as well as one unknown metabolite: X-25422 (Fig. 1, Table S3). Kynurenine (OR = 1.441, 95%CI = 1.089–1.906, p-value = 0.018, power = 81.6%) and 1-stearoyl-GPE (18:0) OR = 1.263, 95%CI = 1.029–1.550, p-value = 0.029, power = 100%) were risk factors for myocarditis, Deoxycarnitine (OR = 0.813, 95%CI = 0.676–0.979, p-value = 0.029, power = 64%), 5-acetylamino-6-formylamino-3-methyluracil (OR = 0.864, 95%CI = 0.775–0.962, p-value = 0.018, power = 88%) and X-25422 (OR = 0.721, 95%CI = 0.587–0.886, p-value = 0.009, power = 73.7%) were protective factors against myocarditis (Table 1). Heterogeneity Q-test analysis showed no heterogeneity (p-value > 0.05) (Fig. 2). Egger intercept, MR-PRESSO analysis was used to detect horizontal pleiotropy, and the results showed no horizontal pleiotropy (p-value > 0.05), and consistency among the four methods (Fig. 3). The results of sensitivity analysis indicated that removing a single SNP did not overly affect the results, i.e., there was no sensitivity (p-value < 0.05) (Fig. 4). Table 1Results of MR analysis.OutcomeExposurenSNPIVWHeterogeneityPleiotropySteiger testPowerNumbersPval (FDR)OROR_ci95QpvalpvaldirectionpvalMyocarditisKynurenine50.0181.4411.089–1.9064.6970.3200.652TRUE3.35E − 7081.6%Myocarditis1-stearoyl-GPE (18:0)70.0291.2631.029–1.5505.9300.4310.796TRUE3.80E − 111100%MyocarditisDeoxycarnitine50.0290.8130.676–0.9792.4460.6540.522TRUE3.70E − 16964%MyocarditisX-25,42260.0180.7210.587–0.8860.7630.9790.868TRUE2.91E − 11288%Myocarditis5-acetylamino-6-formylamino-3-methyluracil40.0090.8640.775–0.9622.5130.4730.906TRUE073.7% Table 1Results of MR analysis.OutcomeExposurenSNPIVWHeterogeneityPleiotropySteiger testPowerNumbersPval (FDR)OROR_ci95QpvalpvaldirectionpvalMyocarditisKynurenine50.0181.4411.089–1.9064.6970.3200.652TRUE3.35E − 7081.6%Myocarditis1-stearoyl-GPE (18:0)70.0291.2631.029–1.5505.9300.4310.796TRUE3.80E − 111100%MyocarditisDeoxycarnitine50.0290.8130.676–0.9792.4460.6540.522TRUE3.70E − 16964%MyocarditisX-25,42260.0180.7210.587–0.8860.7630.9790.868TRUE2.91E − 11288%Myocarditis5-acetylamino-6-formylamino-3-methyluracil40.0090.8640.775–0.9622.5130.4730.906TRUE073.7% Results of MR analysis. Fig. 2Funnel plot. (A) Kynurenine; (B) 1-stearoyl-GPE (18:0); (C) Deoxycarnitine; (D) X-25,422; (E) 5-acetylamino-6-formylamino-3-methyluracil. The main point to look at is that the straight line of IVW shows symmetry at both ends i.e. there is no heterogeneity exists. Funnel plot. (A) Kynurenine; (B) 1-stearoyl-GPE (18:0); (C) Deoxycarnitine; (D) X-25,422; (E) 5-acetylamino-6-formylamino-3-methyluracil. The main point to look at is that the straight line of IVW shows symmetry at both ends i.e. there is no heterogeneity exists. Fig. 3Scatter plot. (A) Kynurenine; (B) 1-stearoyl-GPE (18:0); (C) Deoxycarnitine; (D) X-25,422; (E) 5-acetylamino-6-formylamino-3-methyluracil. The results of the four methods show directional consistency. Scatter plot. (A) Kynurenine; (B) 1-stearoyl-GPE (18:0); (C) Deoxycarnitine; (D) X-25,422; (E) 5-acetylamino-6-formylamino-3-methyluracil. The results of the four methods show directional consistency. Fig. 4Forest plot. (A) Kynurenine; (B) 1-stearoyl-GPE (18:0); (C) Deoxycarnitine; (D) X-25,422; (E) 5-acetylamino-6-formylamino-3-methyluracil. Removal of individual SNPs did not have a disproportionate effect on the results, i.e., no sensitivity existed. Forest plot. (A) Kynurenine; (B) 1-stearoyl-GPE (18:0); (C) Deoxycarnitine; (D) X-25,422; (E) 5-acetylamino-6-formylamino-3-methyluracil. Removal of individual SNPs did not have a disproportionate effect on the results, i.e., no sensitivity existed. Results of genetic and directional analyses Co-localization analysis Co-localization analyses showed no shared genetic factors between kynurenine (PP.H4 = 12%), 1-stearoyl-GPE (18:0) (PP.H4 = 10%), Deoxycarnitine (PP.H4 = 20%), X-25,422 (PP.H4 = 13%), 5-acetylamino-6-formylamino-3-methyluracil (PP.H4 = 8%) and myocarditis phenotypes all did not share genetic factors (Fig. 5). Fig. 5Results of co-localization analysis. Results of co-localization analysis. Replicative and meta-analysis results We extracted another GWAS related to myocarditis from the IEU database for Meta-analysis, and we found that the results about IEU database were not significant due to the difference in sample size (Kynurenine: p-value = 0.221, 1-stearoyl-GPE(18:0): p-value = 0.349, Deoxycarnitine: p-value = 0.464, X-25422: p-value = 0.600, 5-acetylamino-6-formylamino-3-methyluracil: p-value = 0.214). But the directions showed consistency and low heterogeneity (I2 < 50%), which further confirmed the causal relationship between the five metabolites and myocarditis. Among them, Kynurenine (OR = 1.383, 95%CI = 1.102–1.738, p-value = 0.005), 1-stearoyl-GPE(18:0) (OR = 1.231, 95%CI = 1.037–1.460, p-value = 0.017) were risk factors; Deoxycarnitine (OR = 0.840, 95%CI = 0.721–0.980, p-value = 0.026) were risk factors; X-25,422 (OR = 0.781, 95%CI = 0.627–0.974, p-value = 0.028), 5-acetylamino-6-formylamino-3-methyluracil (OR = 0.871, 95%CI = 0.793–0.955, p-value = 0.003) as protective factors (Fig. 6). Fig. 6Meta-analysis results. Meta-analysis results. Discussion To the best of our knowledge, this study is the first to leverage MR to investigate the relationships between blood metabolites and myocarditis. After screening, we identified five metabolites—Kynurenine, 1-stearoyl-GPE(18:0), Deoxycarnitine, X-25,422, and 5-acetylamino-6-formylamino-3-methyluracil—that are significantly associated with myocarditis. This study utilized data from two distinct databases, enhancing the robustness of our findings compared to relying on a single cohort. The integration of GWAS with metabolomic data revealed significant findings. Furthermore, conducting sensitivity analysis, colocalization analysis, and meta-analysis further confirmed these relationships. This study addresses critical gaps in understanding the potential causal links between these metabolites and myocarditis, exploring their roles within a genomic framework. The five blood metabolites identified in this study encompass various types of metabolite levels, including amino acid metabolites, lipid metabolites, nucleotide metabolites, and one unidentified metabolite. Robustness of MR analysis results In this study, we conducted Mendelian randomization (MR) analysis to explore the causal relationship between specific blood metabolites and myocarditis. The robustness of MR analysis is crucial, and our findings provide valuable insights into this area, particularly regarding the negative results from co-localization and some power calculations. Firstly, our MR analysis identified significant causal associations between various blood metabolites and myocarditis. Kynurenine and 1-stearoyl-GPE (18:0) were identified as risk factors, while deoxycarnitine, 5-acetylamino-6-formylamino-3-methyluracil, and X-25,422 acted as protective factors. These positive results demonstrated consistency across multiple analytical methods (IVW, MR-Egger, weighted median, and weighted mode), indicating that our identified associations are strong and reliable, thereby enhancing our understanding of the potential role of these metabolites in the mechanisms of myocarditis. Additionally, co-localization analysis revealed no strong evidence of shared genetic factors between the metabolites and myocarditis phenotypes, as indicated by the coloc PP.H4 < 0.8. This suggests that the same genetic variants are unlikely to influence both the exposure (metabolites) and the outcome (myocarditis), indicating a potential risk of confounding due to genetic pleiotropy. Although none of the significant MR results showed PP.H4 > 0.8, this does not undermine our findings but rather highlights that these metabolites may influence myocarditis through independent or indirect pathways, rather than through direct genetic overlap. Therefore, while the absence of significant co-localization results suggests caution regarding pleiotropy, our findings remain robust and highlight the complexity of the biological mechanisms involved23. However, we must address the low power calculations for some metabolites. For example, deoxycarnitine and 5-acetylamino-6-formylamino-3-methyluracil had power values of 64% and 73.7%, respectively. Although these values indicate potential limitations in detecting true associations, they do not undermine the overall robustness of our analysis. The power of MR studies is influenced by various factors, including the effect size of the exposure on the outcome, the prevalence of the exposure, and the genetic architecture of the traits involved. The consistent directional effects across different methods further support the reliability of our findings. Burgess et al. noted that although large sample sizes are typically required for Mendelian randomization studies, it is not always necessary to measure risk factors for all participants in the study, and thus, sample size is not the sole criterion for evaluating MR analysis26. In summary, despite the negative co-localization results and some low power values, our MR analysis remains reliable and robust. The assessment of the previously positive results, along with the consistency across different analytical methods and the absence of confounding genetic factors, enhances the validity of the identified causal relationships. Future research should focus on larger sample sizes and potentially explore additional metabolic pathways to deepen our understanding of the complex relationships between metabolites and myocarditis. These insights provide increasing evidence for the potential of metabolites as therapeutic targets in cardiovascular diseases. The potential link between metabolites and myocarditis Kynurenine, produced primarily in the liver by the enzymes tryptophan dioxygenase (TDO) and indoleamine 2,3-dioxygenase (IDO), plays a significant role in the kynurenine pathway (KP) of tryptophan metabolism. This endogenous system possesses immunosuppressive features that help control inflammation and induce long-term immune tolerance across various organs, making it relevant to inflammatory diseases27–29. Activation of the kynurenine pathway is typically mediated by IDO1 and appears crucial in linking innate and adaptive immune processes. During systemic inflammation, concentrations of kynurenine in the central nervous system can increase through IDO-independent mechanisms, such as enhanced transport into the brain28,30,31. Kynurenine has also been implicated in several cardiovascular diseases. For instance, it has been shown to lower blood pressure in a dose-dependent manner in spontaneously hypertensive rats and mediate coronary vasodilation independently of the endothelium. Additionally, it reduces pulmonary arterial pressure by activating nitric oxide (NO) and cyclic GMP pathways. Evidence suggests that endothelial IDO may serve as a protective mechanism against pulmonary arterial hypertension (PAH) and remodeling32,33. In addition, the relationship between kynurenine and myocarditis has been somewhat validated in animal models. Inhibition of indoleamine 2,3-dioxygenase (IDO), which catalyzes the degradation of tryptophan (TRP) to kynurenine (KYN). The kynurenine pathway (KP) ameliorates EMCV-induced myocarditis34. In contrast, Kubo et al.35 showed that the knockdown of kynurenine 3-monooxygenase (KMO) in KP led to an increase in serum levels of KP metabolites, thereby reducing mortality in mice with acute viral myocarditis. Not surprisingly, Kynurenine has been implicated in the fields of inflammation, immunity, and cardiovascular disease, whereas studies targeting myocarditis-related studies exist only in animal models, so our study builds on this foundation by reinforcing the causal relationship between Kynurenine and myocarditis in human blood. In mammals, carnitine is synthesized from the protein trimethyllysine in the liver, brain, and (in humans) kidney. Other tissues lack the hydroxylase enzyme needed for the final step of converting deoxycarnitine to carnitine, making them reliant on carnitine uptake from the bloodstream36. The carnitine-related drug Mildronate (mildronate; 3-(2,2,2-trimethylhydrazine) propionate; THP; MET-88) is a clinically used cardioprotective drug effective in the treatment of common cardiovascular diseases such as myocardial infarction, heart failure, arrhythmia, and atherosclerosis, with a mechanism of action based on the regulation of energy metabolism pathways through the lowering of the action of levocarnitine. And we explored L-carnitine Deoxycarnitine as a protective factor against myocarditis, which is in line with the findings of the above existing studies. Deoxycarnitine, a member of the trimethylamine group, serves as a precursor metabolite of trimethylamine N-oxide (TMAO). Elevated levels of TMAO-related metabolites have been associated with atherosclerosis, poor prognoses for atherosclerotic cardiovascular disease (ASCVD), and increased rates of major adverse cardiovascular events (MACE)37–40. This suggests a potential correlation between deoxycarnitine and cardiovascular disease, supporting our study’s conclusions. 5-acetylamino-6-formylamino-3-methyluracil, a major metabolite of caffeine, is destabilized in the presence of dilute bases and/or methanol, leading to the formation of a deformylated compound41. In two MR studies, causal associations between 5-acetylamino-6-formylamino-3-methyluracil and cardiovascular diseases such as myocardial infarction and ischemic stroke were identified. Both studies demonstrated a positive correlation, indicating that the risk of these conditions increases with higher metabolite levels42,43. In contrast, our study suggests that 5-acetylamino-6-formylamino-3-methyluracil acts as a protective factor against myocarditis, as higher levels of this metabolite were associated with a decreased prevalence of the condition. Therefore, we had to revisit the association between caffeine intake cardiovascular disease, and myocarditis. Turnbull et al.44 evaluated the effect of caffeine intake on potential cardiovascular disease outcomes and showed that typical moderate caffeine intake was not associated with an increased risk of overall cardiovascular disease. Additionally, light to moderate coffee and caffeine intake (2–3 cups per day) may be beneficial for metabolic syndrome, which includes conditions like hypertension and diabetes. Coffee consumption has been associated with reduced risks of coronary heart disease, heart failure, arrhythmias, stroke, cardiovascular disease, and all-cause mortality45,46. Mechanistically, Lebeau et al. showed that caffeine increases calcium levels in the hepatic endoplasmic reticulum (ER), inhibiting the activation of sterol regulatory element-binding protein 2 (SREBP2), a key regulator of PCSK9. This mechanism enhances the expression of LDL receptors (LDLR) and improves the clearance of low-density lipoprotein cholesterol (LDLc). However, higher intakes of coffee, tea, and caffeine may elevate the risk of all-cause mortality and cardiovascular death, particularly in patients with pre-existing cardiovascular conditions47,48. In summary, the role of caffeine intake as either a protective or risk factor for cardiovascular health appears to depend on the quantity consumed. While previous research has focused on caffeine’s protective mechanisms, our findings regarding 5-acetylamino-6-formylamino-3-methyluracil suggest a potential breakthrough in understanding its impact on myocarditis. Overall, there is ongoing controversy regarding the relationship between caffeine intake and cardiovascular diseases. The findings of this study suggest that the metabolite 5-acetylamino-6-formylamino-3-methyluracil, a primary metabolite of caffeine, serves as a protective factor against myocarditis. This indicates a more positive perspective on the effects of caffeine consumption on cardiovascular health. Future research should focus on the potential value of this metabolite, utilizing larger sample sizes to further investigate the relationship between caffeine intake and the risk of cardiovascular diseases. 1-stearoyl-GPE(18:0) (1-stearoyl-glycerophosphoethanolamine), where “18:0” indicates the structure of the fatty acid portion. Lebkuchen et al. performed metabolomic and lipidomic analyses of patients suffering from the signs of sleep apnea (OSA) and found glycerophosphoethanolamines to be potential markers of OSA in the early stages of the disease49. OSA is associated with higher cardiovascular morbidity and mortality; in conjunction with myocarditis, it may serve as an early indicator of cardiovascular disease onset. Regarding X-25,422, an unidentified metabolite, current literature lacks specific information. However, advancements in artificial intelligence techniques may enable its identification through machine learning approaches. This highlights the potential for novel discoveries in understanding metabolic profiles related to cardiovascular conditions50. In summary, while there is research linking the screened metabolites to cardiovascular, inflammatory, and immune disorders, studies specifically addressing myocarditis are limited. Therefore, our study effectively addresses this gap. Given the limitations of observational studies, such as small sample sizes and potential reverse causality, the findings from our Mendelian randomization analysis provide stronger evidence for causal inference. Limitation This study identifies potential causal relationships between five human blood metabolites and myocarditis; however, specific metabolic pathways and mechanisms remain unexplored. Although the metabolomics data are derived from the most comprehensive and recent research, the relatively small sample size limits the generalizability of our findings. Future research should focus on investigating these pathways and mechanisms in greater depth and consider conducting randomized controlled trials to validate these findings and uncover effective therapeutic targets for myocarditis. Conclusion This MR study explored the relationship between human blood metabolites and the risk of developing myocarditis, screening for five metabolites that were causally associated with myocarditis. These were risk factors: kynurenine, 1-stearoyl-GPE (18:0); and protective factors: deoxycarnitine, X-25,422, and 5-acetylamino-6-formylamino-3-methyluracil. The discovery of these serum metabolites offers opportunities for early screening for myocarditis, prevention, and treatment, as well as the design of future clinical studies, which provide valuable guidance. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1