Proteomic pathways across the ejection fraction spectrum in patients with heart failure and diabetes mellitus: an EXSCEL trial substudy

Ejection fraction (EF) is a key component of heart failure (HF) classification. However, the biologic basis of HF with mildly reduced EF (HFmrEF) as a distinct biologic entity distinct from HF with preserved EF (HFpEF) and reduced EF (HFrEF) has not been well characterized. The EXSCEL trial randomized participants with type 2 diabetes (T2DM) to a once-weekly glucagon-like peptide receptor agonist (GLP-1 RA) exenatide (EQW) vs. placebo. For this study, profiling of ~ 5000 proteins using the SomaLogic SomaScan platform was performed in baseline and 12-month serum samples from N = 1199 participants with prevalent HF at baseline. Unsupervised principal component analysis (PCA) and ANOVA (FDR p < 0.1) were used to identify proteins that were significantly between three EF groups (EF > 55% [HFpEF], EF 40-55% [HFmrEF], EF < 40% [HFrEF], categories as previously curated in the parent trial). Cox proportional hazards was used to assess association between baseline levels of proteins significantly different between groups, and changes in protein level between baseline and 12-month, with time-to-HF hospitalization. Mixed models were used to assess whether significant proteins changed differentially with exenatide vs. placebo therapy. Of N = 1199 EXSCEL participants with prevalent HF, 284 (24%), 704 (59%) and 211 (18%) had HFpEF, HFmrEF and HFrEF, respectively. Eight principal components analysis (PCA) protein factors differed significantly across the three EF groups, of which 270 individual proteins within those factors were significant. The majority of proteins (75%) demonstrated similar levels in HFmrEF and HFpEF with higher levels in HFrEF. Biologic pathways of epithelial-mesenchymal transition, ECM receptor interaction (tenascin C [TNC], COL28A1), complement and coagulation cascades, and epithelial apical surface and junctions demonstrated enrichment among proteins with this dominant pattern. A minority of proteins (1%) demonstrated similar levels between HFmrEF and HFrEF with lower levels in HFpEF, including MMP-9 (p < 0.0001). Baseline levels of the majority of the 270 proteins (92%) were also associated with time-to-incident HF hospitalization including domains of extracellular matrix (COL28A1, TNC), angiogenesis (VEGFa, VEGFd), myocyte stretch (NT-proBNP), and renal function (cystatin-C). Change in levels of 8% of these from baseline to 12 months (including increase in TNC) predicted incident HF hospitalization (p < 0.05). Levels of 41% of the 270 significant proteins (including TNC and NT-proBNP; p < 0.0001) were reduced differentially by EQW compared with placebo. In conclusion, we found that serum levels of the vast majority of proteins across multiple biologic domains were similar between HFmrEF and HFpEF suggesting that HFmrEF may be more biologically similar to HFpEF than HFrEF. We also identified specific biomarkers showing this pattern, most notably TNC, for which baseline and change levels predicted incident HF hospitalization and were beneficially modified by the GLP-1 RA EQW. These results may offer unique data on prognosis and pharmacotherapy modification with variability by EF.