Cardiac Remodelling Patterns and Proteomics: The Keys to Move Beyond Ejection Fraction in Heart Failure?

This article refers to 'Concentric vs. eccentric remodelling in heart failure with reduced ejection fraction: clinical characteristics, pathophysiology and response to treatment' by J.F. Nauta et al., published in this issue on pages 1147–1155. Cardiac remodelling is recognized as a major determinant of disease progression and outcomes in heart failure (HF) and has played a central role in the discovery of disease-modifying treatments for HF with a reduced ejection fraction (HFrEF).1 The sacubitril/valsartan outcome studies in HFrEF2 and in HF with preserved ejection fraction (HFpEF)3 have recently illustrated the limitations of using left ventricular ejection fraction (LVEF) thresholds to examine the extent of treatment efficacy in HF. It is time to move beyond ejection fraction. The early evidence on the importance of cardiac remodelling came from ischaemic models4 and much of the mechanistic evidence is still derived from similar models, although other models (such as rapid pacing) are used. These generally lead to eccentric cardiac remodelling, with eventual reduction in LVEF. HF is now recognized as a multisystem disease involving a vast array of structural, functional, electrophysiological, cellular, and molecular mechanisms.5 The availability of proteomics provides tools to explore the mechanisms involved in HF beyond LVEF and even cardiac remodelling. Aetiologic factors are changing over time, with lower proportions of patients having primarily ischaemic HF. Still, cardiac remodelling phenotypes will remain related to the clinical phenotypes (triggering events, toxic exposures, cardiovascular risk factors, co-morbid conditions, etc.), as will molecular mechanisms underlying the development and progression of HF. Left ventricular (LV) concentric remodelling or hypertrophy are typically seen in the pressure-overload models, remodelling phenotypes frequently observed in HFpEF,6 where one of the dominant co-morbid conditions is hypertension. In a recently published report examining mechanisms involved in HFpEF, the Olink Proseek Multiplex CVD II and III proximity extension assays were used to identify biomarkers associated with coronary flow reserve, a marker of coronary microvascular dysfunction.7 The latter mechanism appears to be common in HFpEF, possibly related to myocardial hypertrophy and fibrosis, insufficient microvascular supply, atherosclerosis, microvascular inflammation and elevated LV filling pressures.7 The authors of the latter study identified pregnancy-associated plasma protein A as an important hub in the network of interest, suggesting greater biological importance of this pathway related to subclinical atherosclerosis. As stated by the authors,7 a selection bias can be introduced in such studies with regard to the proteins 'chosen' in the cardiovascular disease panels. Further, there is a lack of data on how several of these circulating proteins relate to the processes at the cardiac or vascular tissue level. In this issue of the Journal, Nauta and colleagues, using similar technology, present a distinct approach to the mechanisms contributing to the evolution of HFrEF. More attention should be given to phenotyping in HF (beyond LVEF) and LV remodelling pattern is likely relevant to response to treatment in HFrEF. The aims of their study were (i) to investigate the clinical difference in characteristics between the two patterns of LV hypertrophy (LVH) in patients with HFrEF, (ii) to use circulating biomarkers to gain insight into the underlying biological processes most relevant to each type of hypertrophic remodelling, (iii) to explore whether the type of LVH was associated with a differential response to guideline-directed medical therapy (GDMT). The authors performed a retrospective analysis of the BIOSTAT-CHF study8 (A systems BIOlogy Study to TAilored Treatment in Chronic Heart Failure) that recruited patients with worsening HF from 11 European countries, between 2010–2015. Patients were eligible when there was an anticipated need for up-titration of angiotensin-converting enzyme inhibitors (ACEi) or angiotensin receptor blockers (ARBs) and beta-blockers, with the aim of optimizing GDMT. Of the index cohort of 2516 patients with HF, 1819 patients had a LVEF <40%. The echocardiographic variables that define LV geometry (LV mass index and relative wall thickness) were available in 1304 patients (72%). Hypertrophy of either concentric or eccentric type was present in 1015 patients, with 142 (11%) having concentric LVH. In the latter group, patients were on average older and more likely hypertensive compared to those with eccentric LVH. There was no statistical difference between the proportion of women in both groups, although this was likely due to sample size (43 women or 30.3% of the concentric hypertrophy group vs. 220 or 25.2% women in the eccentric hypertrophy group). Sex differences in cardiac remodelling have been observed in multiple HF studies.5 Of note, Nauta and colleagues did not study other patterns of LV remodelling, including concentric remodelling (only 26 patients) and normal geometry patterns (263 patients). The authors employed a paired-antibody, proximity extension assay panel (Olink CVD III; https://www.olink.com/products/cvd-iii-panel/biomarkers/) to measure 92 biomarkers associated with various aspects of cardiovascular pathophysiology, including inflammation, oxidative stress, myocyte injury, endothelial function, atherosclerosis, and angiogenesis in plasma in patients from the index BIOSTAT-CHF study with HFrEF and either eccentric or concentric LVH. This approach yielded semi-quantitative normalized protein expression levels for each biomarker (Nauta et al., online supplementary Table S2). The results were validated in an independent validation cohort, consisting of 1738 patients from six centres in Scotland (730 patients with a LVEF <40%). Data on LV geometry were available for 328 patients in the validation cohort (45%). Pairwise correlations from the discovery dataset of all 92 biomarkers were performed separately in patients with concentric and eccentric LVH. Patients with concentric and eccentric hypertrophy showed, respectively, 37 and 28 significant correlations that could be successfully validated. Within the concentric subgroup, 15 of the 37 correlations were unique to concentric LVH whereas within the eccentric subgroup 8 of the 28 correlations were unique to eccentric LVH. Protein–protein correlations that are unique to either concentric or eccentric LVH are presented in Figure 1. Biomarkers that correlated with a network of two or more others were designated as central hub wherein the number of protein–protein correlations signifies the importance of this biomarker in the network, presumably reflecting the most important mechanisms at play for each type of remodelling studied. The main hubs identified in concentric LVH included tumour necrosis factor receptor 1 (TNF-R1), urokinase plasminogen activator surface receptor (U-PAR), paraoxonase (PON3), TNF-R2, tartrate-resistant acid phosphatase type 5 (TR-AP), insulin-like growth factor-binding protein 2 (IGFBP-2), trefoil factor 3 (TFF3) and P-selectin (SELP). Whereas U-PAR and TNF-R1 serve as hubs correlating with four other biomarkers, the remaining hubs correlate with two other proteins. The hubs in eccentric LVH were N-terminal pro-B-type natriuretic peptide (NT-proBNP), junctional adhesion molecule A (JAM-A) and matrix metalloproteinase-2 (MMP-2): each hub correlating with two other proteins. Network analysis indicated the importance of inflammation (TNF-R1, U-PAR, SELP) and oxidative stress (PON3) in concentric hypertrophy whereas LV wall stress (NT-proBNP), endothelial dysfunction (JAM-A) are important factors in eccentric hypertrophy (Figure 1). Hence, network analysis of circulating biomarkers further highlights the differences in myocardial remodelling between concentric and eccentric LVH. We agree with Nauta et al. summarizing the above findings by the statement 'HFrEF with concentric hypertrophy is characterized by markers of oxidative stress and inflammation'. However, fibrosis is also an important mechanism underlying a concentric hypertrophy pattern in HFrEF, similar to HFpEF. Biomarkers that reflect collagen homeostasis and fibrosis have been correlated with the presence and severity of disease in HFpEF.9 The subclinical atherosclerosis hypothesis (via the PAPP-A pathway) proposed earlier for HFpEF,7 using a comparable analysis, is also in agreement with the hubs identified by Nauta and colleagues. Furthermore, a similar network analysis from BIOSTAT-CHF showed that biomarker profiles observed in HFpEF were related to inflammation and extracellular matrix reorganization.10 With the objective to identify predictors of cardiovascular outcomes in a population setting, data from 6814 participants in the Multi-Ethnic Study of Atherosclerosis (MESA), aged 45 to 84 years, from four ethnicities, and six centres across the United States were analysed.11 When using the random survival forests technique (including 735 variables), LV structure and function and cardiac troponin T were among the top predictors for incident HF. Such findings demonstrate that LV structure, at the centre of the study by Nauta et al., is critical in the development of HF. However, considering a significant imbalance of study group sizes and the cohort study design of the original BIOSTAT-CHF, propensity matching should ideally have been performed to address the study question and eliminate as much confounding as possible. Considering the impact of HF treatment based on the type of LV remodelling or hypertrophy is clearly a concept worth further exploring. For beta-blockers, the data presented by Nauta et al. would suggest a mortality benefit of up-titration higher in HFrEF patients with eccentric hypertrophy compared to those with concentric hypertrophy; whereas for ACEi/ARBs, there was no statistically significant interaction between the LVH type and the effect of up-titration on mortality. However, with 142 patients having concentric hypertrophy in the index study, these observations must be viewed as hypothesis-generating. Furthermore, the study is not current in terms of GDMT.12 Part of precision medicine,13 the proteomic approach to biomarkers relevant to HF has been used increasingly over the past few years and can provide a less biased approach to identify important mechanisms of disease. It is a promising tool to eventually develop and test new treatments, and could help in tailoring HF therapies. However, the findings of such studies must be replicated and tested prospectively, ideally including advanced cardiac imaging and/or other forms of in vivo proofs of concepts. One of the pitfalls of using 'pre-made' cardiovascular panels could be missing important mechanisms related to the specific cause of HF that could develop through unique or unknown pathways (e.g. infiltrative cardiomyopathy, genetic mutations, toxic agents). The concepts proposed by Nauta et al. could be tested in a prospective study including a full spectrum of LVEFs and LV remodelling patterns, with current GDMT and in a more homogeneous population setting. Conflict of interest: E.O.M. or her institution have received financial support for clinical trials from Novartis, Bayer, AstraZeneca, Merck and Amgen. She has served as a consultant or speaker for AstraZeneca, Novartis, Servier and Amgen. B.G.A. declares no conflicts of interest.