Large-scale metabolomic profiling identifies novel biomarkers for incident coronary heart disease

Abstract

Analyses of circulating metabolites in large prospective epidemiological studies could lead to improved prediction and better biological understanding of coronary heart disease (CHD). We performed a mass spectrometry-based non-targeted metabolomics study for association with incident CHD events in 1,028 individuals (131 events; 10 y. median follow-up) with validation in 1,670 individuals (282 events; 3.9 y. median follow-up). Four metabolites were replicated and independent of main cardiovascular risk factors [lysophosphatidylcholine 18∶1 (hazard ratio [HR] per standard deviation [SD] increment = 0.77, P-value<0.001), lysophosphatidylcholine 18∶2 (HR = 0.81, P-value<0.001), monoglyceride 18∶2 (MG 18∶2; HR = 1.18, P-value = 0.011) and sphingomyelin 28∶1 (HR = 0.85, P-value = 0.015)]. Together they contributed to moderate improvements in discrimination and re-classification in addition to traditional risk factors (C-statistic: 0.76 vs. 0.75; NRI: 9.2%). MG 18∶2 was associated with CHD independently of triglycerides. Lysophosphatidylcholines were negatively associated with body mass index, C-reactive protein and with less evidence of subclinical cardiovascular disease in additional 970 participants; a reverse pattern was observed for MG 18∶2. MG 18∶2 showed an enrichment (P-value = 0.002) of significant associations with CHD-associated SNPs (P-value = 1.2×10-7 for association with rs964184 in the ZNF259/APOA5 region) and a weak, but positive causal effect (odds ratio = 1.05 per SD increment in MG 18∶2, P-value = 0.05) on CHD, as suggested by Mendelian randomization analysis. In conclusion, we identified four lipid-related metabolites with evidence for clinical utility, as well as a causal role in CHD development.

Figure 1. Study flow chart.

Overview of the study design and analyses performed.

Figure 2. Association between four metabolites and cardiovascular traits and genotypes.

Panel A: Association with main cardiovascular risk factors in three population-based studies. Panel B: Minus log₁₀(P-value) for association with markers of inflammation, oxidative stress and subclinical CVD in PIVUS. Sex-adjusted analysis (upper panel) and adjusted by sex, systolic blood pressure, body mass index, current smoker, antihypertensive treatment, LDL-C, HDL-C, log-triglycerides and diabetes at baseline (lower panel). * indicates the alpha threshold after multiple-testing correction. Panel C: Minus log₁₀(P-value) for association with 51 SNPs previously reported for association with CHD (44 SNPs) or selected from candidate pathways (7 SNPs).

Figure 3. Mendelian randomization analysis.

A significant deviation from zero of the estimate of causal effect using all SNPs (solid red line) suggests a causal relationship between the metabolite and CHD.