Skip to content

Genetic Architecture of Heart Failure Subtypes

Summary

Two large genome-wide association studies show that clinically-defined heart failure (HF) subtypes have genuinely distinct genetic architectures: HF with reduced ejection fraction (HFrEF) is substantially more heritable and gene-rich than HF with preserved ejection fraction (HFpEF), and stripping the ischemic component of HF removes the atherosclerosis-gene signal. This page documents that evidence and frames it as the phenotype-first paradigm of genetic subtyping — define the subtype by clinical or mechanistic criteria first, then dissect it with genetics — in deliberate contrast to the genetics-first PRS-clustering approaches surveyed under Polygenic-Score-Based Disease Subtyping.

Why This Is the Phenotype-First Contrast Case

Lab interpretation: These HF studies do the opposite of mechanism-first PRS clustering. Rather than clustering variants by multi-trait association and reading a subtype off the result, they take a subtype that is already defined externally — by ejection fraction, and by presence/absence of ischemic and structural antecedents — and ask what its genetics look like. The subtypes are grounded in clinical reality and, for HFpEF vs HFrEF, in a differential treatment-response history (large HFpEF pharmacologic trials have been largely unsuccessful, whereas HFrEF therapies succeeded).[1] When the subtype is defined this cleanly, genetics behaves mechanistically — which is exactly the property the genetics-first clustering methods struggle to demonstrate.

HFpEF vs HFrEF in a Single Cohort: Joseph et al. 2022

Joseph, Liu, Hui et al. used the Million Veteran Program — a single, uniformly-phenotyped cohort — to curate HFrEF and HFpEF by current consensus clinical definitions and run separate GWAS.[1] The primary study population was 258,943 controls with unclassified HF (n=43,344), HFpEF (n=19,589), and HFrEF (n=19,495) from MVP, plus 8,227 HF cases and 379,788 controls from UK Biobank, all European ancestry.[1]

  • Unclassified HF meta-analysis identified 20 genome-wide-significant (GWS) loci (10 novel), replicating all 12 GWS SNPs from a prior HF GWAS.[1]
  • Despite comparable sample sizes, HFrEF yielded 13 GWS loci while HFpEF yielded only one (FTO).[1]
  • Nine of the 13 HFrEF loci had significantly different associations with HFrEF versus HFpEF, indicating divergent pathobiology rather than a shared HF signal.[1]

The authors' central conclusion is that clinically-defined HFpEF likely represents an amalgamation of several distinct pathobiological entities, and that consensus sub-phenotyping of HFpEF is needed before its genetics can be dissected.[1] This is an important caution for the whole field: even a clinically-defined subtype can be too crude to be genetically coherent.

HF and Its Subtypes at Scale: HERMES / Lumbers et al. 2024

The HERMES consortium meta-analysis spanned 42 studies and 1,946,349 individuals, including 153,174 HF cases (HFall).[2] To capture genetic factors acting directly on HF susceptibility rather than via upstream ischemic disease, cases were stratified into non-ischemic HF (ni-HF, n=44,012; excluding antecedent myocardial infarction, coronary revascularization, and marked structural/valvular disease) and, where left-ventricular data allowed, ni-HFrEF (LVEF<50%, n=5,406) and ni-HFpEF (LVEF≥50%, n=3,841).[2] The meta-analysis was cross-ancestry (91% European, 6.2% East Asian, 2.2% African, 0.5% South Asian, 0.1% admixed American).[2]

Key findings:[2]

  • 66 independent loci across the HF phenotypes (37 not previously reported); 76/87 (87%) prior HF variants replicated.
  • SNP-heritability differs sharply by subtype: HFall 5.4±0.2%, ni-HF 6.1±0.5%, ni-HFrEF 11.8±2.6%, ni-HFpEF 1.8±1.3% — the ni-HFpEF estimate is both far lower and highly uncertain, echoing Joseph et al.'s "amalgamation" conclusion. Genetic correlation between ni-HFrEF and ni-HFpEF was only 0.42.
  • Mechanistic behaviour of a clean subtype: genes with established atherosclerosis roles (LDLR, LPL, ABCG5, LPA) were notable among loci associated with HFall but not ni-HF — i.e. removing the ischemic component removed the atherosclerosis signal. Prioritized ni-HF effector genes were instead cardiomyopathy/cardiac-development genes (ACTN2, BAG3, FLNC, HSPB7, NKX2-5), hypertrophy genes (CAMK2D, CAND2), and arrhythmia genes (PITX2, KLF12, ATP1B1).
  • Tissue/cell etiology tracked the subtype: cardiac tissue showed the highest heritability enrichment for HFall, ni-HF, and ni-HFrEF, whereas kidney and pancreas were the top-enriched tissues for ni-HFpEF; cardiomyocytes were the major effector cell type for ni-HF. IGFBP7, a senescence-associated secretory-phenotype gene, was prioritized at a ni-HFpEF locus.
  • The LPA sentinel variant was the single locus showing significant ancestry heterogeneity of effect — one of several places in the vault where lipoprotein(a) surfaces as mechanistically special (see Polygenic Subtyping of Cardiovascular Disease).
  • A polygenic score from HFall (OR 1.37 per SD in UK Biobank; top-decile OR 1.70 vs the fifth decile) predicts overall HF but, consistent with the low ni-HFpEF heritability, is expected to transfer poorly to the HFpEF subtype.

Relationship to Polygenic Subtyping of CVD

This page is the phenotype-first exemplar referenced by Polygenic Subtyping of Cardiovascular Disease. The contrast it draws is not that one paradigm is always right — genetics-first clustering is a legitimate hypothesis-generating tool — but that the clinical grounding of a subtype comes from mechanism and therapy, and genetics is most informative when it dissects such a subtype rather than being asked to invent one.

See Also

Citations

[1] Joseph, J., Liu, C., Hui, Q., Aragam, K., Wang, Z., Charest, B., Huffman, J. E., Keaton, J. M., Edwards, T. L., Demissie, S., Djoussé, L., Casas, J. P., Gaziano, J. M., Cho, K., Wilson, P. W. F., Phillips, L. S., VA Million Veteran Program, O'Donnell, C. J., & Sun, Y. V. (2022). Genetic architecture of heart failure with preserved versus reduced ejection fraction. Nature Communications, 13, 7753. DOI: 10.1038/s41467-022-35323-0. Source: s41467-022-35323-0.pdf. Supports: MVP/UKB cohort and sample sizes, 20 unclassified-HF loci, 13-HFrEF-vs-1-HFpEF (FTO) locus contrast, 9/13 differential loci, and the HFpEF-amalgamation conclusion above. Location: Full text — Abstract and Results (GWAS of unclassified HF; GWAS of HFrEF and HFpEF).

[2] Lumbers, R. T., et al. (HERMES Consortium) (2024/2025). Genome-wide association study meta-analysis provides insights into the etiology of heart failure and its subtypes. Nature Genetics, 57, 815–828. DOI: 10.1038/s41588-024-02064-3. Source: s41588-024-02064-3.pdf. Supports: 1.9M-individual meta-analysis, ni-HF/ni-HFrEF/ni-HFpEF definitions and sample sizes, 66 loci, subtype SNP-heritabilities (11.8% vs 1.8%) and rg=0.42, atherosclerosis-genes-in-HFall-not-ni-HF, tissue/cell enrichment, LPA ancestry heterogeneity, and PGS_HF performance above. Location: Full text — Abstract and Results (GWAS meta-analysis; Genetic architecture and heritability; Prioritization of effector genes; Identifying organs, tissues and cells).