The ongoing enigma of NT-proBNP in HFpEF: Insights from Proteomics

B-type natriuretic peptides (BNPs) are widely utilized as circulating biomarkers for diagnosing, risk stratifying and monitoring heart failure (HF) regardless of left ventricular ejection fraction (LVEF). Importantly, BNPs are also central in the pathophysiology of HF, produced by cardiomyocytes as a counter-regulatory hormone in response to myocardial stress from stretch, injury and maladaptive neurohormonal responses. Expression of BNPs is intricately and tightly regulated from initial translation of messenger RNA in the cardiomyocyte nucleus to the mediation of biological effects on target cells. Higher circulating levels of BNPs are indicative of more advanced disease and a poorer prognosis in HF, while increasing BNP levels often serve as an early indicator of worsening HF. However, the precise biological processes and pathophysiological pathways that stimulate BNP expression are not fully understood. In response to myocardial stress, proBNP is cleaved to an inactive N-terminal fragment (NT-proBNP) and the biologically active the C-terminal peptide, which exerts its effect on target cells. Circulating concentrations of NT-proBNP demonstrate significant variability and are influenced by several factors beyond cardiomyocyte stretch including adiposity, renal function, sex, and genetic ancestry.¹ NT-proBNP concentrations are below the diagnostic thresholds in a subset of patients with well-treated and compensated HF, particularly those a preserved LVEF (HFpEF). The biological mechanisms underlying this variability is also incompletely understood.

Large-scale proteomics has emerged as a powerful tool to interrogate the complex mechanisms underlying HF. Assays quantifying up to one-third of an individual’s circulating proteome have been leveraged to identify novel circulating proteins associated with risk of developing HF and with risk of adverse outcomes among individuals with prevalent HF. Despite recognized limitations with target assay specificity, these high through-put assays therefore hold promise for discovery of novel HF mechanisms and biomarkers. Importantly, despite the quantification of thousands of proteins, NT-proBNP has remained the strongest and most consistent predictor of risk of incident HF² and outcomes in prevalent HF³ – highlighting the enduring significance of NT-proBNP and importance of understanding the biologic mechanisms influencing NT-proBNP levels.

In this issue of the Circulation Heart Failure, Azzo et al. ⁴ leveraged large-scale proteomics to investigate circulating protein correlates of NT-proBNP levels among patients with chronic HFpEF. They used the SomaScan aptamer-based assay to measure levels of approximately five thousand circulating proteins in two independent HFpEF patient cohorts: 253 patients in the Penn HF Study (PHFS) and 218 patients enrolled in the Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist Trial (TOPCAT). They identified 278 proteins significantly associated with aptamer-based NT-proBNP levels in both cohorts in adjusted models and after multiple testing correction (393 in PHFS, 546 in TOPCAT). Pathway enrichment analysis suggested that these NT-proBNP-related proteins enriched for pathways involved in fibrosis, inflammation, cellular regulation, and angiogenesis. Among the individual proteins that stood out as the strongest correlates of higher NT-proBNP were pentraxin domain containing-1 (SVEP-1) and angiopoietin-2 (ANGPT2). SVEP-1 is a large extracellular matrix protein that promote inflammatory responses and fibrosis, while ANGPT2 is a key regulator of vascular stability and is expressed from stressed endothelium where it stimulates new vessel growth. These findings confirm previous studies demonstrating their prognostic importance in HF and for ANGPT2 with incident HF, suggesting they may play key roles in the pathophysiology HF. ^3,5.More studies are needed to further understand the role of these proteins in the disease and their potential as a therapeutic target.

The authors are to be commended on this novel approach to interrogating the biologic correlates of circulating NT-proBNP. Parallel analyses in two independent HFpEF cohorts provide confidence in the generalizability of the 278 identified proteins and has proven essential in large-scale proteomic studies. As highlighted by the authors, several of the proteins most strongly associated with higher or lower NT-proBNP levels have known biologic functions related to fibrosis, inflammation, and endothelial dysfunction. One remarkable observation is the sheer number of proteins correlated with NT-proBNP. This large number may reflect the complexity of NT-proBNP regulation, but is likely also related to the heterogeneity of the HFpEF syndrome.⁶ For example, a HFpEF phenotype characterized by obesity and metabolic disease is associated with lower levels of NT-proBNP,⁷ while a phenotype linked to older age and chronic renal disease typically exhibits higher NT-proBNP levels. The biological mechanisms regulating NT-proBNP release and clearance may differ in these different contexts. Further mechanistic heterogeneity may exist related to sex and ancestry.¹ The analysis of Azzo et al also highlights some of the challenges in deriving mechanistic insight using these broad proteomic technologies. While proBNP is produced by the cardiomyocytes, circulating levels are primarily determined by upstream pathways influencing cardiac remodeling and hemodynamics. The extent to which proteins associated with NT-proBNP are specific to NT-proBNP itself or simply epiphenomena or downstream consequences of these primary upstream pathways is unknown. Furthermore, the correlation structure of the proteomic measures makes interpretation of associations complicated – is the protein identified biologically relevant or just a correlate of a biologically relevant protein? These are challenges common to all large-scale proteomic studies in HF. Understanding the relationships of each NT-proBNP-associated protein with risk of adverse HF outcomes (HF hospitalization, cardiovascular death) would have helped refine the interpretation of the study findings. Furthermore, one unique strength of proteomics is the large number of genetic determinants. Genetic causal inference analysis with Mendelian randomization (MR) – and reverse MR – may have been another approach to help clarify which of the 278 proteins had causal effects of NT-proBNP levels, and vice versa.

As noted by the authors, NT-proBNP is also increasingly used as an eligibility criteria for HF clinical trials, especially for HFpEF trials. This serves the dual purpose of increasing confidence that enrolled patients truly have HF, and of selecting patients with higher baseline risk to improve study power. However, this approach may introduce bias in patient selection as several clinical characteristics influence NT-proBNP levels beyond cardiac structure and function as noted above. Novel biomarkers that either enhances the diagnostic precision for HF irrespective of demographic status and comorbidities or biomarkers that improves risk prediction in HF beyond that of NT-proBNP are therefore highly warranted. Although the current study does not address this topic, proteomic studies hold the potential to identify a biomarker or combination of biomarkers that may surpass the prognostic value of NT-proBNP.

The authors adopted a novel approach in aptamer-based proteomics by utilizing one of the aptamers, the one binding to NT-proBNP, as the outcome. A key concern with this strategy is the comparability of the aptamer assay with other proteomic platforms and - most importantly - immunoassays used in the clinic. Significant variations in quantitative measures are observed for many proteins in comparative analyses between aptamer-based (SomaScan), proximity-extension assay (Olink Proseek), and targeted immunoassays.^{8 9} Reassuringly, NT-proBNP demonstrates strong correlation between the SomaScan and Olink platforms (r=0.91), although correlations with clinically used NT-proBNP assays were not provided. ⁸. Another consideration associated with the use of NT-proBNP as the outcome measure in this study is the potential for confirmation bias, particularly in patients from the TOPCAT cohort where elevated BNP or NT-proBNP was one of the inclusion criteria. This may have led to artificially high NT-proBNP levels (median ~1000 ng/L in individuals with available immunoassay measurements) that may not accurately represent HFpEF patients with lower NT-proBNP levels. Further investigations are therefore necessary to explore the pathophysiology reflected by NT-proBNP in other HF phenotypes and individuals at risk of developing HF. Despite these limitations, Azzo et al.’s research deserves commendation for bringing us closer to unraveling this perplexing biomarker that continues to occupy a central role in our evaluation of HF and HF therapies.