Epigenome-wide association study of heart failure uncovers phenotype-specific DNA methylation

Mykhailo Krolevets (Mainz)1, V. ten Cate (Mainz)2, J. Prochaska (Mainz)3, A. Schulz (Mainz)2, S. Rapp (Mainz)2, E. Yapici (Mainz)2, M. Ingold (Mainz)4, M. A. Andrade-Navarro (Mainz)5, S. Horvath (San Diego)6, C. Niehrs (Mainz)7, P. S. Wild (Mainz)4

1Universitätsmedizin der Johannes Gutenberg-Universität Mainz Centrum für Thrombose und Hämostase Mainz, Deutschland; 2Universitätsmedizin der Johannes Gutenberg-Universität Mainz Preventive Cardiology and Preventive Medicine Mainz, Deutschland; 3Universitätsmedizin der Johannes Gutenberg-Universität Mainz Zentrum für Kardiologie Mainz, Deutschland; 4Universitätsmedizin der Johannes Gutenberg-Universität Mainz Präventive Kardiologie und Medizinische Prävention Mainz, Deutschland; 5Johannes Gutenberg-University Mainz Institute for Immunology, University Medical Center Mainz, Deutschland; 6Altos labs San Diego, USA; 7Institute of Molecular Biology DNA DEMETHYLATION, DNA REPAIR AND REPROGRAMMING Mainz, Deutschland


Introduction: Heart failure (HF) is a disease with severe morbidity and poor prognosis. The role of DNA methylation (DNAm) in the progression and development of phenotypes of heart failure is unclear.

Aim: To investigate the importance of individual CpG site methylation in the development and progression of three HF phenotypes, as well as their role in aging processes.

Methods: Individuals with symptomatic HF (universal definition stage C/D, including HFpEF, HFmrEF and HFrEF) of the MyoVasc cohort (N=3,289; NCT04064450) and individuals without cardiovascular disease (CVD) from a population-based cohort (Gutenberg Health Study, N=18,700) were investigated. DNA was extracted from peripheral blood. DNAm was measured using the Illumina Infinium methylationEPIC v2.0 methylation array (San Diego, USA). Epigenetic age was modeled with GrimAge. Multivariable logistic regression models adjusted for cardiovascular risk factors (CVRFs) were used to assess the relationship between individual CpG sites and each HF phenotype. Ridge regression was used to generate phenotype-specific CpG-scores. Linear regression was used to estimate associations between each HF phenotype CpG score and cardiovascular traits. Cox regression was used to relate CpG scores to clinical outcome. Premature aging was derived using residuals from linear regression of chronological against epigenetic age. Worsening of heart failure was defined as a combination of cardiac-related deaths, hospitalizations resulting from heart failure and progression from stage B to C/D, over 4 years of follow-up, assessed by telephone interviews and by querying medical records.

Results: Data from 1,317 with symptomatic HF and 838 CVD-free individuals were analyzed. After FDR adjustment 73,251 CpG sites were significantly differentially methylated in HFpEF, 42,720 in HFmrEF, 142,639 in HFrEF.  Only 20.6% of CpG sites were shared between 3 phenotypes, while simultaneously 50.9% of CpG-annotated genes were common between 3 phenotypes. C-reactive protein (R² = 0.08), HBA1C (R² = 0.07), obesity (R² = 0.05) and C-peptide (R² = 0.04) had strongest association with methylation in HFpEF, while NT-proBNP (R² = 0.24), Presence of plaques in carotid artery (R² = 0.13) and Troponin I (R² = 0.13) with methylation in HFrEF. Age (ß=-0.22 [0.17- 0.27], p<0.0001) and NT-proBNP (ß=0.21 [0.16- 0.26], p<0.0001) were positively associated with the HFpEF score. The strongest association was observed between the HFrEF score and NT-proBNP (ß =0.24 [0.19- 0.30], p<0.0001). All scores were correlated with premature aging derived from GrimAge (all p<0.0001). HFpEF and HFrEF-scores were positively associated with worsening of heart failure (p=0.013; p=0.0008). All 3 scores were positively associated with all-cause death (p=0.011; p=0.02; p<0.0001)

Conclusion: A large number of CpG sites were identified to be associated with all 3 HF phenotypes vs CVD-free individuals. Phenotype scores generated with these CpGs were strongly associated with multiple CVD-related traits, as well as with clinical outcome. Between-phenotype differences were observed on individual CpG as well as on the score-level.

Diese Seite teilen