Clinical-transcriptional prioritization of the circulating proteome in human heart failure

Abstract

Given expanding studies in epidemiology and disease-oriented human studies offering hundreds of associations between the human "ome" and disease, prioritizing molecules relevant to disease mechanisms among this growing breadth is important. Here, we link the circulating proteome to human heart failure (HF) propensity (via echocardiographic phenotyping and clinical outcomes) across the lifespan, demonstrating key pathways of fibrosis, inflammation, metabolism, and hypertrophy. We observe a broad array of genes encoding proteins linked to HF phenotypes and outcomes in clinical populations dynamically expressed at a transcriptional level in human myocardium during HF and cardiac recovery (several in a cell-specific fashion). Many identified targets do not have wide precedent in large-scale genomic discovery or human studies, highlighting the complementary roles for proteomic and tissue transcriptomic discovery to focus epidemiological targets to those relevant in human myocardium for further interrogation.

Keywords: heart failure; multiomics; proteomics; snRNA-seq; transcriptomics.

Graphical abstract

Figure 1

Proteomic architecture of cardiac structure and function in CARDIA (A) Protein associations that are directionally consistent in both CARDIA derivation and validation samples and are significant at a 5% FDR in both after adjusting for age, gender, and race. “Pearson” reports the Pearson correlation between beta coefficients from models in CARDIA derivation and validation samples for each respective phenotype. Displayed effect sizes (beta) are from regression models performed in the derivation sample. (B) An UpSet plot with the number of proteins related to each echocardiographic domain, demonstrating broadly shared proteins implicated across cardiac phenotypes. (C) Beta coefficients from linear models of echocardiographic measures as a function of circulating proteins from the derivation (x axis) and validation (y axis) CARDIA samples (with Pearson r).

Figure 2

Proteomics of incident HF in two independent studies (the FHS and UK Biobank) (A) Volcano plots in FHS (adjusted for age and gender) and UK Biobank (adjusted for age, gender, race). Blue color indicates <5% FDR. Note that in the FHS where SomaScan proteomics was used, there can be multiple aptamers for the same protein (e.g., F9). (B) Pathway analysis of the 134 targets (KEGG, Reactome) prioritized by echocardiographic associations with LV systolic function, LV diastolic function, and LV morphology, which were associated with incident HF in FHS or UK Biobank. The assayed proteome from CARDIA was used as background. p.adjust indicates FDR-corrected p values.

Figure 3

Four-chamber deconvolution of proteomic associations with echocardiographic measures of HF Made in BioRender. Genes of significant proteins (FDR <0.05 in both CARDIA derivation and validation sets) associated with echocardiographic measures of HF were filtered to those significantly differentially expressed in the left atrium and LV as published.

Figure 4

Single-nucleus myocardial prioritization of clinical targets in human HF (A) The mean count expression (pseudobulked) between the full list of genes expressed versus those 134 prioritized by our approach, demonstrating a higher myocardial expression for those prioritized genes. (B) The number of differentially expressed genes by cell type in single-nucleus RNA-seq in HF (dilated cardiomyopathy) versus non-failing (donor) myocardium. “Unique” represents genes uniquely differentially expressed in that cell type; “Shared_Once” and “Shared_Multi” represent genes differentially expressed in one other cell type or multiple, respectively. The number of differentially expressed genes in the pseudobulk population is shown in black as a reference and not considered in the shared or unique count for the cell type differentially expressed genes. Fibroblasts, pericytes, and cardiomyocytes are among the top 3 cell types by number of differentially expressed genes. (C) A heatmap of the number of shared differentially expressed genes across cell types. Fibroblast and pericytes had the most frequent shared differentially expressed genes from our prioritized set. (D) A heatmap of the log₂ fold change for genes differentially expressed in the three prioritized cell types (now visualized across cell types). Genes with blanks signify the absence of reporting of differential expression in the parent data. “lfc” signifies log fold change, and “mean” signifies mean expression. (E–G) Individual genes across failing and non-failing myocardial samples in fibroblasts, pericytes, and cardiomyocytes by single-nucleus RNA-seq for responders and non-responders across the three cell types. The y axis represents the fold change in individuals who experienced cardiac recovery (more positive = greater expression with hemodynamic unloading with assist device). The asterisks reflect an FDR < 10% (executed only across the displayed transcripts).