Background: Circulating concentrations of growth differentiation factor 15 (GDF15) are elevated in times of acute cardiac stress. Here we report an epigenome-wide association study (EWAS) of GDF15 in the Multi-Ethnic New Zealand Study of Acute Coronary Syndromes (MENZACS).
Methods: MENZACS enrolled participants admitted with first-time acute coronary syndromes (ACS) from 10 NZ hospitals between 2015 and 2024. Whole-blood DNA methylation (DNAm) was measured using Illumina MethylationEPIC v1 or v2 arrays. The EPIC v1 group was randomly partitioned into training (90%) and test (10%) datasets. In the training set, an EWAS for GDF15 adjusted for age, sex, ethnicity and blood cell composition was performed. Significant CpG sites were used to train a predictor of GDF15 using elastic net regression with 10-fold cross validation. The predictor was projected into the test and EPIC v2 datasets. A previous epigenetic predictor of GDF15[1] was calculated for comparison. Variance explained (R2) was estimated by regressing measured GDF15 on each epigenetic score.
Results: For 1038 participants with GDF15 and DNAm data available, mean age was 57 years. 30% were female, 53% were NZ European, 28% were Māori, 10% were Pacific, and 8% were Indian. Median GDF15 was 1186 (interquartile range, 852, 1730) pg/mL. In the training dataset (n=867, 82%), EWAS identified 459 CpG sites (p<3.6x10-8). The epigenetic score explained 24% of variation in GDF15 levels in the test dataset (n=96, 9%) and 26% in the EPIC v2 group (n=75, 9%). The published epigenetic score explained 12% GDF15 variation in the test group and 17% in the EPIC v2 group. In Māori and Pacific participants of the EPIC v2 group (n=48), the MENZACS derived score explained 33% variation in GDF15.
Discussion: GDF15 levels are associated with DNAm in people with ACS. Tailoring epigenetic scores to disease contexts and increasing diversity within training samples and may improve accuracy.