Integrating metabolomics and proteomics to identify novel drug targets for heart failure and atrial fibrillation

Abstract

Background: Altered metabolism plays a role in the pathophysiology of cardiac diseases, such as atrial fibrillation (AF) and heart failure (HF). We aimed to identify novel plasma metabolites and proteins associating with cardiac disease.

Methods: Mendelian randomisation (MR) was used to assess the association of 174 metabolites measured in up to 86,507 participants with AF, HF, dilated cardiomyopathy (DCM), and non-ischemic cardiomyopathy (NICM). Subsequently, we sourced data on 1567 plasma proteins and performed cis MR to identify proteins affecting the identified metabolites as well as the cardiac diseases. Proteins were prioritised on cardiac expression and druggability, and mapped to biological pathways.

Results: We identified 35 metabolites associating with cardiac disease. AF was affected by seventeen metabolites, HF by nineteen, DCM by four, and NCIM by taurine. HF was particularly enriched for phosphatidylcholines (p = 0.029) and DCM for acylcarnitines (p = 0.001). Metabolite involvement with AF was more uniform, spanning for example phosphatidylcholines, amino acids, and acylcarnitines. We identified 38 druggable proteins expressed in cardiac tissue, with a directionally concordant effect on metabolites and cardiac disease. We recapitulated known associations, for example between the drug target of digoxin (AT1B2), taurine and NICM risk. Additionally, we identified numerous novel findings, such as higher RET values associating with phosphatidylcholines and decreasing AF and HF. RET is targeted by drugs such as regorafenib which has known cardiotoxic side-effects. Pathway analysis implicated involvement of GDF15 signalling through RET, and ghrelin regulation of energy homeostasis in cardiac pathogenesis.

Conclusions: This study identified 35 plasma metabolites involved with cardiac diseases and linked these to 38 druggable proteins, providing actionable leads for drug development.

Keywords: Atrial fibrillation; Cardiomyopathy; Drug development; Heart failure; Mendelian randomisation; Metabolomics; Proteomics.

Fig. 1

Study design to identify plasma proteins and metabolites affecting cardiac disease. A Flowchart of the analysis steps. B A triangulation diagram, illustrating how a robust set of directionally concordant proteins was identified affecting plasma metabolite levels as well as cardiac disease. The red plus indicates a risk-increasing effect, while the blue minus indicates a risk-decreasing effect. C Annotated network of prioritised metabolites, proteins, and outcomes for which the metabolites have at least 20% associated common proteins and belong to the metabolite class acylcarnitines. Prioritised directionally concordant proteins are represented by circles, metabolites by diamonds, outcomes by triangles. Circle colours represent protein druggability. Increasing effects are displayed by a red arrow, decreasing effects by a blue arrow. See Supplementary Note for more details. Abbreviations: AF = atrial fibrillation, DCM = dilated cardiomyopathy, HF = heart failure, MR = mendelian randomisation, NICM = non-ischemic cardiomyopathy

Fig. 2

Plasma metabolites associating with at least one cardiac outcome. N.B. For visualisation purposes, the p-values were truncated to a −log(10) of 10. Multiplicity corrected significant associations are depicted by a star, non-significant associations by a dot. The second row indicates the number of druggable proteins reflecting information retrieved from BNF and ChEMBL. Genetic associations with the cardiac outcomes were obtained from Nielsen et al. (60,620 AF cases) [33], Shah et al. (47,309 HF cases) [34], Garnier et al. (2719 DCM cases) [35], and Aragam et al. (2038 NICM cases) [36]. See the “Methods” section additional details, and Appendix Table S2 for the underlying numerical data. Abbreviations: a = acyl residue, aa = diacyl residue, ae = acyl-alkyl residue, AF = atrial fibrillation, DCM = dilated cardiomyopathy, HF = heart failure, LPC = lysophosphatidylcholines, NICM = non-ischemic cardiomyopathy, PC = phosphatidylcholines, SM = sphingomyelins

Fig. 3

Percentage of mRNA expression in cardiac tissue of the subset of directionally concordant proteins affecting cardiac disease. N.B. The percentage mRNA expression in cardiac tissue was calculated by dividing cardiac expression by total expression in all organs. The horizontal line indicates 1% cardiac expression. Proteins are referred to using their Uniprot label and proteins in bold font are overexpressed in cardiac tissue relative to non-cardiac tissue. Data were sourced from the human protein atlas; see “Methods” section

Fig. 4

Forest plots of prioritised drugged or druggable proteins effect on cardiac outcome. N.B. Metabolite class of the metabolites affected by the protein are indicated on the right y-axis. Genetic associations with the cardiac outcomes were obtained from Nielsen et al. (60,620 AF cases) [33], Shah et al. (47,309 HF cases) [34], Garnier et al. (2719 DCM cases) [35], and Aragam et al. (2038 NICM cases) [36]. See the “Methods” section for a more detailed description and Appendix Table S7 for the full numerical results. Abbreviations: AF = atrial fibrillation, CI = confidence interval, DCM = dilated cardiomyopathy, HF = heart failure, NICM = non-ischemic cardiomyopathy, OR = odds ratio

Fig. 5

Forest plot of prioritised proteins associating with more than one cardiac outcome. N.B. Metabolite class of the metabolites affected by the protein are indicated on the right y-axis. Genetic associations with the cardiac outcomes were obtained from Nielsen et al. (60,620 AF cases) [33], Shah et al. (47,309 HF cases) [34], Garnier et al. (2719 DCM cases) [35], and Aragam et al.(2038 NICM cases) [36]. See the “Methods” section for a more detailed description and Appendix Table S7 for the full numerical results. Abbreviations: AF = atrial fibrillation, CI = confidence interval, DCM = dilated cardiomyopathy, HF = heart failure, NICM = non-ischemic cardiomyopathy, OR = odds ratio