PHENOTYPE-SPECIFIC TREATMENT OF HEART FAILURE WITH PRESERVED EJECTION FRACTION: A MULTIORGAN ROADMAP

Abstract

Heart failure (HF) with preserved ejection fraction (EF) (HFpEF) accounts for 50% of HF cases and its prevalence relative to HF with reduced EF (HFrEF) continues to rise. In contrast to HFrEF, large trials testing neurohumoral inhibition in HFpEF failed to reach a positive outcome. This failure was recently attributed to distinct systemic and myocardial signaling in HFpEF and to diversity of HFpEF phenotypes. In this review, a HFpEF treatment strategy is proposed which addresses HFpEF-specific signaling and phenotypic diversity. In HFpEF, extracardiac comorbidities such as metabolic risk, arterial hypertension and renal insufficiency drive left ventricular (LV) remodeling and dysfunction through systemic inflammation and coronary microvascular endothelial dysfunction. The latter affects LV diastolic dysfunction through macrophage infiltration resulting in interstitial fibrosis and through altered paracrine signaling to cardiomyocytes, which become hypertrophied and stiff because of low nitric oxide (NO) and cyclic guanosine monophosphate (cGMP). Systemic inflammation also affects other organs such as lungs, skeletal muscle and kidneys leading respectively to pulmonary hypertension, muscle weakness and sodium retention. Individual steps of these signaling cascades can be targeted by specific interventions: metabolic risk by caloric restriction, systemic inflammation by statins, pulmonary hypertension by phosphodiesterase (PDE) 5 inhibitors, muscle weakness by exercise training, sodium retention by diuretics and monitoring devices, myocardial NO bioavailability by inorganic nitrate-nitrite, myocardial cGMP content by neprilysin or PDE 9 inhibition and myocardial fibrosis by spironolactone. Because of phenotypic diversity in HFpEF, personalized therapeutic strategies are proposed, which are configured in a matrix with HFpEF presentations in the abscissa and HFpEF predispositions in the ordinate.

INTRODUCTION

Heart Failure (HF) with preserved Ejection Fraction (EF) (HFpEF) currently accounts for more than 50% of all heart failure cases and its prevalence relative to HF with reduced Ejection Fraction (HFrEF) continues to rise at an alarming rate of 1% per year¹. In the past three decades, HFrEF evolved to a distinct therapeutic entity partly because large outcome trials demonstrated efficacy of neurohumoral inhibition. No similar evolution has occurred in HFpEF, where large trials testing neurohumoral inhibition consistently failed to reach a positive primary outcome either individually² or on meta-analysis³. In trials testing angiotensin converting enzyme inhibitors (ACEIs), angiotensin II receptor blockers (ARBs) or mineralocorticoid receptor antagonists (MRA) a modest positive trend was sometimes observed but only for secondary outcomes⁴ or retrospectively defined subgroups^5,6. The failure of neurohumoral inhibition in the large HFpEF outcome trials led some investigators to challenge HFpEF as a distinct HF phenotype^7-9. More recent views attributed this failure to different systemic and myocardial signalling in HFpEF and HFrEF¹⁰ or to diverse phenotypes within the HFpEF patient population^11-14. In line with these views, the current HFpEF treatment roadmap first addresses HFpEF specific systemic and myocardial signalling, subsequently configures HFpEF phenotypes in a matrix of predispositions and presentations and finally discusses therapeutic inroads that fit into the phenotypic framework.

SYSTEMIC AND MYOCARDIAL SIGNALLING

Large outcome trials and registries all revealed HFpEF patients to be of advanced age and predominantly women and to suffer from multiple comorbidities such as overweight/obesity (84%)¹⁵, arterial hypertension (60-80%)¹⁶, type 2 diabetes mellitus (20-45%)¹⁶, renal insufficiency and sleep apnea. Aging and the aforementioned comorbidities may initiate chronic systemic inflammation as manifest from biomarker profiles which revealed high plasma levels of soluble interleukin 1 receptor-like 1 (IL1RL1), C-reactive protein (CRP) and growth differentiation factor 15 (GDF15) in HFpEF^17-20. Initial studies revealed plasma levels to be similarly elevated in HFpEF and HFrEF¹⁷ but recent studies observed them to be higher in HFpEF¹⁹ and therefore suggested a larger involvement of systemic inflammation in HFpEF. Systemic inflammation may affect myocardial remodeling and dysfunction in HFpEF through a signalling cascade, which begins with coronary microvascular endothelial dysfunction (Figure 1)^10,21. It subsequently involves myocardial infiltration by activated macrophages, which induce reactive interstitial fibrosis²² and altered paracrine communication between endothelial cells and surrounding cardiomyocytes²¹. The latter deprives cardiomyocytes of nitric oxide (NO) and of cyclic guanosine monophosphate (cGMP), which renders them hypertrophied and stiff²³. High cardiomyocyte stiffness is caused by diminished distensibility of the giant cytoskeletal protein titin, whose elastic properties are dynamically modulated by isoform shifts, phosphorylation and oxidation^24,25. Strong support for an extramyocardial origin of HFpEF came from “parabiosis” experiments in which hearts of young animals acquired HFpEF-like features when exposed to blood from old animals and vice versa, as hearts of old animals reversed HFpEF-like features when exposed to blood of young animals²⁶.

Figure 1. Systemic and myocardial signalling in HFPEF.

Comorbidities induce systemic inflammation, evident from elevated plasma levels of inflammatory biomarkers such as soluble interleukin 1 receptor-like 1 (IL1RL1), C-reactive protein (CRP) and growth differentiation factor 15 (GDF15). Chronic inflammation affects the lungs, myocardium, skeletal muscle and kidneys leading to diverse HFpEF phenotypes with variable involvement of pulmonary hypertension (PH), myocardial remodeling, deficient skeletal muscle oxygen extraction (ΔA-VO₂) during exercise (Ex) and renal Na⁺ retention. Myocardial remodeling and dysfunction begins with coronary endothelial microvascular inflammation manifest from endothelial expression of adhesion molecules such as vascular cell adhesion molecule (VCAM) and E-Selectin. Expression of adhesion molecules attracts infiltrating leukocytes secreting transforming growth factor β (TGF- β), which converts fibroblasts to myofibroblasts with enhanced interstitial collagen deposition. Endothelial inflammation also results in presence of reactive oxygen species (ROS), reduced nitric oxide (NO) bioavailability and production of peroxynitrite (ONOO⁻). This reduces soluble guanylate cyclase (sGC) activity, cyclic guanosine monophosphate (cGMP) content and the favorable effects of protein kinase G (PKG) on cardiomyocyte stiffness and hypertrophy.

The extramyocardial origin of HFpEF differs from the intramyocardial origin of HFrEF, where remodeling is driven by cardiomyocyte cell death because of ischemia, infection or toxicity²⁷. Distinct origins of HFpEF and HFrEF are mirrored by unequal LV structural and ultrastructural remodeling (Table). Biomarker profiles in HFpEF and HFrEF are consistent with the distinct origins of both HF phenotypes as they show lower markers of myocardial injury (high-sensitivity troponin T; hsTNT)) or of myocardial stress (N-terminal pro brain natriuretic peptide; NT-proBNP) in HFpEF^17-20,28-30. Lower hsTNT is explained by less cardiomyocyte damage as a result of limited upregulation in HFpEF myocardium of nicotinamide adenine dinucleotide phosphate oxidase 2 (Nox2) evident in infiltrating macrophages or endothelial cells but not in cardiomyocytes²¹. Lower NT-proBNP is explained by concentric LV remodeling/hypertrophy in HFpEF in contrast to eccentric LV remodeling/hypertrophy in HFrEF³¹ and by visceral distribution of adipose tissue in the mostly overweight or obese HFpEF patients³², which is associated with decreased production and increased clearance of natriuretic peptides (NP).

Table.

Unequal structural, functional and ultrastructural LV characteristics in HFpEF and HFrEF

	HFpEF	HFrEF
LV Structure/Function
End-DiastolicVolume	↔	↑
End-Systolic Volume	↔	↑
Wall Thickness	↑	↔
Mass	↑	↑
Mass/Volume Ratio	↑	↓
Remodeling	Concentric	Eccentric
Ejection Fraction	↔	↓
Stroke Work	↔	↓
End-Systolic Elastance	↔	↓
End-DiastolicStiffness	↑	↓
LV Ultrastructure
Myocyte Diameter	↑	↔
Myocyte Length	↔	↑
Myocyte Remodeling	Concentric	Eccentric
Fibrosis	Interstitial/Reactive	Focal/Replacement

In HFpEF, chronic systemic inflammation affects not only the myocardium but also other organs such as lungs, skeletal muscles and kidneys (Figure 1). Although HFpEF patients may stop exercising because of a rapid and brisk rise in LV filling pressures^33-38, in a substantial subset of patients effort tolerance is limited by inappropriate pulmonary vasoconstriction evident from pulmonary hypertension, or by inadequate peripheral skeletal muscle vasodilation, perfusion and oxygen utilization evident from absent widening of arterio-venous oxygen difference^39-43. Systemic inflammation also affects the renal microcirculation and the ability of the kidneys to excrete a sodium load⁴⁴. Inability to excrete a sodium load contributes to the progressive volume expansion observed during transition from chronic compensated to acute decompensated HFpEF^45-46 and explains the efficacy of diuretics as they restore the pressure-natriuresis relationship.

PHENOTYPIC FRAMEWORK

HFpEF clinically presents as a diverse syndrome initiated by a variety of comorbidities and inflammatory mediators with extracardiac manifestations and cardiac abnormalities ^13,47-49. Despite the diversity of the HFpEF syndrome, the treatment strategy thus far has focused on a “one-size-fits-all” approach that has worked relatively well for chronic HFrEF. However, virtually all clinical syndromes benefit from more tailored, personalized therapy, and this may also be true of HFpEF. Successfully addressing the diversity of HFpEF is an active area of investigation, and solutions for the problem range from simple (e.g., stratifying based on type of clinical presentation⁴⁷) to sophisticated (e.g. machine learning techniques to perform data reduction in order to classify patients based on intrinsic patterns in dense phenotypic data⁴⁹). While the field of machine learning is not new⁵⁰, its application to clinical medicine is still relatively novel and these techniques will require iterative testing and application to clinical trials before they can be applied clinically on a routine basis.

In the absence of compelling outcome data to support individual therapies, we propose a matrix configuration combining predisposition phenotypes with clinical presentation phenotypes as a starting point to guide current clinical care and future prospective research (Figure 2). Rare etiologies such as constrictive pericarditis, valvular heart disease, high output failure or infiltrative cardiomyopathies are presumed to be excluded beforehand. Figure 2 displays a stepwise approach that begins in the left hand upper corner of the matrix with general treatment recommendations, presumed to be beneficial to the vast majority of HFpEF patients as they address the presentation phenotype of lung congestion and the predisposition phenotype of overweight/obesity present in >80% of HFpEF patients¹⁵. Subsequently, supplementary recommendations are suggested for additional predisposition-related phenotypic features when moving downward in the matrix and for additional presentation-related phenotypic features when moving rightward in the matrix. Arterial hypertension, renal dysfunction and coronary artery disease are proposed as additional predisposition phenotypes (Figure 2). Additional clinical presentation phenotypes, in whom specific therapeutic interventions could be meaningful, are chronotropic incompetence, pulmonary hypertension (especially combined precapillary and postcapillary pulmonary hypertension; CpcPH), skeletal muscle weakness and atrial fibrillation. Apart from use of diuretics, caloric restriction diet, exercise training and anticoagulation in the presence of atrial fibrillation, all recommendations need to be confirmed by prospective outcome trials in the respective phenotypic subsets.

Figure 2. Phenotype-specific HFpEF treatment strategy using a matrix of predisposition phenotypes and clinical presentation phenotypes.

A stepwise approach is proposed that begins in the left hand upper corner of the matrix with general treatment recommendations, presumed to be beneficial to the vast majority of HFpEF patients as they address the presentation phenotype of lung congestion and the predisposition phenotype of overweight/obesity present in >80% of HFpEF patients. Subsequently, supplementary (+) recommendations are suggested for additional predisposition-related phenotypic features when moving downward in the matrix and for additional presentation-related phenotypic features when moving rightward in the matrix. Arterial hypertension, renal dysfunction and coronary artery disease are proposed as additional predisposition phenotypes. Additional clinical presentation phenotypes, in whom specific therapeutic interventions could be meaningful, include chronotropic incompetence, pulmonary hypertension (especially combined precapillary and postcapillary pulmonary hypertension; CpcPH), skeletal muscle weakness and atrial fibrillation. Only therapeutic measures indicated in bold are currently established. All other therapeutic measures require further testing in specific phenotypes.

PHENOTYPIC TREATMENT STRATEGY

Numerous steps of the HFpEF signalling cascade, which ranges from systemic inflammation to myocardial titin elasticity, are valid treatment targets either for the vast majority of the HFpEF population (i.e. the lung congestion/metabolic risk phenotype in the upper left hand corner of Figure 2) or for specific presentation/predisposition HFpEF phenotypes (Figure 2).

Lung congestion/metabolic risk phenotype

The lung congestion/metabolic risk phenotype is considered the “garden variety” of HFpEF because by definition, HF patients have evidence of lung congestion at rest or during exercise and because overweight/obesity (BMI > 25 kg/m²) is highly prevalent in HFpEF (>80%)¹⁵ and increasingly recognized to drive HFpEF development. The latter was evident from recent longitudinal non-invasive studies, which revealed close correlations over a 4 year time interval between diastolic LV stiffness and body mass index (BMI) and concluded that central adiposity predisposed to HFpEF^51,52. Similar evidence was already provided by the ALLHAT trial, which enrolled patients with arterial hypertension and one additional cardiovascular risk factor, and observed a high BMI at enrolment to be the strongest predictor of HFpEF development⁵³.

Diuretics

Lowering of LV filling pressures with diuretics is of paramount importance for HFpEF patients to achieve symptomatic benefit, to reduce pulmonary artery pressures and to improve RV loading⁵⁴. Their efficacy relates to a restored pressure-natriuresis relationship in the presence of renal microvascular inflammation⁵⁵. Administration of diuretics can be guided by use of implantable hemodynamic monitors that either directly and continuously measure diastolic LV pressures or provide surrogates of pressure⁴⁵. Studies evaluating hemodynamic monitoring have demonstrated that even in HFpEF patients considered to be “compensated” by expert HF clinicians, diastolic LV pressures are elevated and these elevations have important prognostic implications⁵⁶. When transition to decompensated HF occurs, diastolic LV pressures progressively increase over weeks. During this time interval hemodynamic monitoring allows for early uptitration of diuretics, which improves outcome as demonstrated in the CHAMPION trial. In this study, treatment guided by implantable hemodynamic monitoring significantly decreased cardiovascular death and HF hospitalizations in HFPEF patients^57,58.

Caloric restriction

As increased body adiposity promotes inflammation and impairs cardiac, arterial, renal and skeletal muscle function, weight loss should be considered in a treatment strategy for the vast majority of HFpEF patients. Kitzman et al. recently reported that a 20-week caloric restriction diet was feasible and appeared safe in older, obese HFpEF patients, and significantly improved their symptoms, peak oxygen consumption (VO₂), and quality of life (QOL) scores (Figure 3)⁵⁹. The QOL improvement was significantly greater with diet than exercise. The combination of diet with endurance exercise training was additive and produced a large (2.5 mL/kg/min) increase in peak VO₂ (Figure 3), similar or larger than what most drug or other treatments produced in HFrEF patients. The validity of the increase in peak VO₂ was supported by significant increases in 5 other measures of physical performance that are independent of body mass: VO₂ reserve, exercise time to exhaustion, workload, six-minute walk distance, and leg power. The increase in peak VO₂ was strongly correlated with reduced body fat mass, increased percent lean body mass, higher thigh muscle/intermuscular fat ratio and lower biomarkers of inflammation supporting the hypothesis that overweight/obesity contributes to exercise intolerance in HFpEF through systemic inflammation⁵⁹.

Figure 3. Effects of a 20 week caloric restriction diet on exercise capacity and quality of life in HFpEF.

The graph displays percent changes ± standard errors at the 20-week follow-up relative to baseline by randomized group for: peak VO2 (ml/kg/min, panel A), and KCCQ (Kansas City Cardiomyopathy Questionnaire) overall score (= Quality of Life Score) (panel B). AT=aerobic exercise training; CR=caloric restriction diet. P-values represent effects for AT and CR.

Statins

The presence of systemic inflammation supports the use of statins in HFpEF. Statins improve endothelial redox balance and restore NO bioavailability, independently of low-density lipoprotein lowering^60-61. Analysis of endomyocardial biopsy material revealed statin-treated HFpEF patients to have less myocardial nitrotyrosine, higher myocardial PKG activity, less cardiomyocyte hypertrophy and lower cardiomyocyte resting tension¹⁰. In an observational study, statin-treated HFpEF patients were also less prone to develop atrial fibrillation⁶². These findings support the positive outcome of small phase 2 trials and HF registries that showed statin use to improve outcome of HFpEF patients^63-65. It remains to be explored whether other novel approaches to treat systemic inflammation might be effective in HFpEF⁶⁶.

Inorganic nitrite/nitrate

In the HFpEF signaling cascade cardiomyocytes are deprived of NO and cGMP because of altered paracrine communication between inflamed microvascular endothelial cells and cardiomyocytes (Figure 1). Organic NO donors were therefore suggested to be potentially useful in HFpEF as they could restore myocardial NO content and concomitantly correct the elevated arterial load. Recently however, Redfield et al. demonstrated in patients with HFpEF that the organic nitrate isosorbide mononitrate tended to reduce chronic activity levels measured by accelerometry, with no improvement in submaximal exercise capacity⁶⁷. This result might be interpreted as disproving the NO hypothesis in HFpEF but there are some important caveats to consider. Organic nitrates may produce greater than expected hypotensive effects in people with HFpEF or potentially impair cardiac output owing to excessive preload reduction⁶⁸. Organic nitrates tonically increase local NO levels and require bioactivation in the tissues. The latter can cause pharmacologic tolerance, while the former can chronically lower renal perfusion pressure, which, as alluded to before, is countered by renal sodium retention. This may override any beneficial reduction in filling pressures, a phenomenon known as pseudo-tolerance. Perhaps more importantly, organic nitrates such as isosorbide mononitrate have also been shown to cause endothelial dysfunction^69,70 which plays a central role in the HFpEF signaling cascade.

In contrast to organic nitrates, the inorganic nitrate-nitrite pathway represents an important alternative route to restore NO signaling in HFpEF⁷¹. Formerly considered as an inert byproduct of NO metabolism, nitrite is now known to function as an important in vivo NO reservoir. Importantly, nitrite is preferentially reduced to NO in the presence of hypoxia and acidosis, which occurs during physical exercise, thus delivering NO at the time and locations (i.e. skeletal and cardiac muscles) of greatest need. Nitrate-nitrite preparations have been shown to improve conduit artery stiffness in healthy volunteers and improve systemic vasodilation during exercise in patients with HFpEF^72,73. More recently, acute infusion of sodium nitrite was shown in a placebo-controlled trial of patients with HFpEF to preferentially reduce diastolic LV pressures and pulmonary artery pressures during exercise while restoring cardiac output reserve toward normal (Figure 4)⁷⁴. Part of this benefit was mediated by vasodilation, but evidence for a direct myocardial benefit such as increased stroke work was also observed. Another recent study found that inorganic nitrate, delivered as one week of once-daily beetroot juice consumption, improved submaximal exercise endurance⁷⁵.

Figure 4. Effects of acute infusion of inorganic nitrite on exercise hemodynamics in HFpEF.

ΔPCWP_EX : change in exercise induced pulmonary capillary wedge pressure; ΔExerciseCO : increase in exercise induced cardiac output; mean PAP: mean pulmonary artery pressure; ΔCO/ ΔVO₂: ratio of exercise induced increase in cardiac output over exercise induced increase in oxygen consumption.

Sacubitril and other PKG stimulating drugs

A substantial number of HFpEF patients have pathological ventricular hypertrophy, with interstitial fibrosis and diastolic chamber stiffening. This has encouraged efforts to block key activators and to stimulate intrinsic suppressors of these changes. Among the attractive pathways representing the latter approach are those coupled to cyclic guanosine monophosphate (cGMP) and its cognate kinase – protein kinase G (PKG). PKG stimulation has potent anti-fibrotic and anti-hypertrophic effects in cultured myocytes and fibroblasts^76-79, and has been protective in a wide array of experimental cardiac disease models including pressure-overload hypertrophy^80-82. Moreover, there are multiple therapeutic approaches to stimulate PKG already in clinical use or under active investigation, which increases the potential translational relevance of this pathway.

Stimulation of PKG requires cGMP which is either synthesized by soluble guanylate cyclase (sGC) activated by NO or by receptor guanylate cyclase (rGC) linked to the natriuretic peptide (NP) receptor^83-85. This is in turn counterbalanced by hydrolysis of cGMP back to GMP by select members of the phosphodiesterase (PDE) superfamily, and their inhibition, which leads to increased cGMP, can also increase PKG activity (Figure 5). cGMP also controls cyclic adenosine monophosphate (cAMP) levels by feedback modulation of PDE2 and PDE3. At low levels of cGMP, pro-inotropic effects via cAMP have been observed while at higher levels and with cAMP co-stimulation, cGMP induces an anti-adrenergic effect. Four members of the PDE superfamily (PDE1, PDE2, PDE5, and PDE9) regulate cGMP in the heart of which both PDE5 and PDE9 are selective for cGMP. They are not redundant, but target different intracellular pools, with PDE5 largely impacting NO-sGC derived cGMP, while PDE9 regulates NP-rGC derived pools⁸⁰. These local pools impact different intracellular compartments of PKG as detected by differences in net phosphokinomes and effects on transcriptional regulation⁸⁰.

Figure 5. Myocardial cGMP signaling and pharmacological interventions in HFpEF.

Nitric Oxide (NO) produced by nitric oxide synthases (NOS) stimulates soluble guanylate cyclase (sGC) to produce cyclic guanosine monophosphate (cGMP), which activates protein kinase G (PKG). Inorganic nitrate/nitrite, sGC stimulators and phosphodiesterase (PDE) 5 inhibitors target this pathway. The natriuretic peptides ANP and BNP attach to the natriuretic peptide receptors A/B (NPRA/NPRB). This stimulates receptor guanylate cyclase (rGC) to produce cGMP, which again activates PKG. Neprilysin inhibitors such as sacubitril and PDE9 inhibitors act through this pathway.

Recent studies have defined multiple targets relevant to the lusitropic, anti-hypertrophic and anti-fibrotic impact of PKG. HFpEF cardiomyocytes display greater passive diastolic stiffness that has been linked to changes in titin phosphorylation at PEVK-region residues modulated by PKA and PKG^23,86-88. The latter is a particularly potent regulator of titin stiffness, which in turn impacts cardiac muscle stiffness²³. Anti-hypertrophic and anti-fibrotic mechanisms include PKG-suppression of transforming growth factor-β (TGF-β) signaling by phosphorylation of Smad proteins that blocks their nuclear translocation and signaling⁸⁹.

Data regarding myocardial cGMP/PKG signaling in HFpEF remain fairly limited, but several studies have revealed likely critical features in this disease that could ultimately dictate how a successful therapy would need to work. Most pertinently, human LV biopsy analysis from HFpEF has reported very low levels of cGMP and associated PKG activity, particularly when compared to HFrEF and aortic stenosis patients²³. This may help explain reduced titin phosphorylation and muscle stiffening, as well as contributory signaling to hypertrophy and fibrosis – i.e. the brake has been removed. It also raises questions regarding the initiating mechanism. Myocardial oxidative stress coupled with a pro-inflammatory microvascular environment has been proposed¹⁰ and was recently supported by comparative analysis of HFpEF, HFrEF and aortic stenosis samples, which revealed in HFpEF higher microvascular expression of adhesion molecules and Nox2 with higher hydrogen peroxide and lower nitrite/nitrate content²¹.

Administration of sGC activators or stimulators could provide downstream correction for the low myocardial NO bioavailability in HFpEF. Use of the sGC activator cinaciguat in HFrEF was hampered by hypotension⁹⁰. The oral sGC stimulator riociguat improved exercise tolerance or quality of life in pulmonary arterial hypertension (PATENT)⁹¹, in chronic thromboembolic pulmonary hypertension (CHEST)⁹² and in pulmonary hypertension due to HFrEF (LEPHT)⁹³. In these three studies, arterial blood pressure also fell by up to 9 mmHg and this is especially worrisome for HFpEF patients because of their limited ability to raise LV stroke volume⁶⁸. The use of vericiguat, another sGC stimulator, was well tolerated in HFrEF but failed to lower NP except at the highest dose⁹⁴. It is currently being tested in HFpEF in the SOCRATES-PRESERVED trial.

Because of concentric LV remodeling, NP stimulation is less marked in HFpEF than HFrEF, a finding that may limit counter stimulation via this pathway. NPs are degraded by circulating neprilysin. Inhibition of this peptidase could augment deficient NP-rGC signaling and therefore be beneficial in HFpEF as suggested by the fall in NP following administration of valsartan/sacubitril in the phase 2 PARAMOUNT study⁹⁵. Use of valsartan/sacubitril is currently being tested in the multicenter PARAGON-HF trial.

Another approach is to block PDEs to increase cGMP levels and hence PKG activity. PDE5 upregulation in HFrEF was reported by multiple^96,97 but not all⁹⁸ laboratories. Data in HFpEF did not support a similar elevation²³, and two PDE5-inhibitor trials in HFpEF yielded a neutral outcome^99,100. An alternative may therefore be inhibiting PDE9. A recent study found marked upregulation of PDE9 protein in human left ventricular biopsies from HFpEF patients as well as HFrEF and aortic stenosis patients⁸⁰. This suggests the low cGMP levels might be related to enhanced expression of PDE9, and if so, inhibiting this PDE should have beneficial effects. In mice subjected to sustained pressure-overload, blocking PDE9 by gene deletion or selective pharmacological inhibition suppressed hypertrophy, fibrosis, and chamber dysfunction⁸⁰. PDE9 inhibition has been previously examined clinically for its potential to alter cognition¹⁰¹ but these new data may trigger interest in HFpEF and other forms of heart failure.

Spironolactone and E-matrix modification

The extracellular matrix is composed of fibrillary proteins (such as collagen, elastin), non-fibrillary proteins (such as aminoglycans, fibronectin, laminin) and bioactive proteins such as TGF-β, matrix metalloproteinases (MMPs), tissue inhibitors of matrix metalloproteinases (TIMPs) and matricellular proteins. The homeostatic control of collagen is especially important for abnormal diastolic function in HF¹⁰². Important differences in geometry, composition and homeostatic mechanisms are seen in HFpEF vs. HFrEF. HFpEF is more often associated with interstitial, reactive fibrosis and HFrEF with focal, replacement fibrosis (Table). The extent of collagen cross-linking tends to be higher in HFpEF¹⁰³ and homeostasis in HFpEF is profibrotic while fibrinolytic in HFrEF. The latter is reflected in the respective plasma biomarker profiles¹⁰⁴.

Resident myocardial fibroblasts control collagen homeostasis in normal hearts. Whether resident fibroblasts remain responsible for increased collagen production or whether recruitment of fibroblasts occurs from a different source such as bone-marrow or microvascular endothelium remains uncertain. In murine HFpEF models, resident fibroblasts and not bone-marrow derived cells or endothelial-mesenchymal transition were primarily responsible for myocardial collagen production following transverse aortic constriction^105,106. Recruited cells could however still be involved through secretion of cytokines or matricellular proteins. Myofibroblasts have also been implicated in collagen deposition in HFpEF as they are closely associated with fibrotic collagen deposition and scar contracture.In HFpEF, fibroblasts are presumed to convert to myofibroblasts because of exposure to TGF-β as a result of monocyte/macrophage myocardial infiltration²².

Collagen metabolism requires sequential, highly orchestrated and regulated steps: 1) procollagen synthesis and secretion; 2) procollagen post-synthetic processing; 3) collagen post-translational modification and 4) collagen degradation. Each of these steps is altered in HFpEF, contributes either individually or in aggregate to LV diastolic dysfunction, is mirrored in plasma biomarkers and serves as a unique treatment target (Figure 6). Procollagen I and III are synthesized in myocardial fibroblasts and secreted as a soluble molecule with NH2- (N) and COOH- (C) terminal propeptides attached. These are removed to create insoluble collagen¹⁰⁷ and appear in plasma as procollagen I C-terminal peptide (PICP), procollagen III C-terminal peptide (PIIICP), procollagen I N-terminal peptide (PINP) and procollagen III N-terminal peptide (PIIINP), all of which reflect rate of collagen synthesis. Subsequent formation of insoluble collagen requires enzymatic formation of crosslinks by lysyl- or hydroxylysyloxidase. Non-enzymatic crosslinks can also be formed by advanced glycation endproducts (AGEs), which can activate profibrotic pathways through binding with the receptor for AGE (RAGE). Insoluble collagen formation is promoted by matricellular proteins (SPARC, thrombospondin, osteopontin). To maintain collagen homeostasis, insoluble collagen is continuously degraded by matrix metalloproteinases (MMPs), which are in turn regulated by tissue inhibitors of MMPs (TIMPs) ¹⁰⁸. Collagen degradation results in formation of collagen telopeptides (CITP, CIIITP). MMPs, CITP and CIIITP can be measured in plasma and in combination with PICP, PINP, PIIICP and PIIINP allow for an integrated multi-biomarker assessment of collagen homeostasis^103,104. In HFpEF such an assessment revealed collagen synthesis to be increased and collagen degradation to be decreased resulting in a net increase in collagen content¹⁰⁹. Additional biomarkers that are useful estimates of myocardial collagen content are galectin-3 (Gal-3) and solubleST2 (sST2). The former is secreted by infiltrating macrophages and stimulates fibroblasts while the latter is a member of the interleukin 1 receptor family which is also profibrotic as it acts as a decoy for interleukin-33 (IL-33), which inhibits profibrotic signaling.

Figure 6. Sequential steps of collagen metabolism.

Collagen metabolism involves sequential steps consisting of procollagen synthesis, procollagen processing to collagen fibrils, post-translational modification of collagen fibrils and collagen degradation.

To date, three pharmaceutical agents that affect the extracellular matrix have been tested in HFpEF: spironolactone in TOPCAT, valsartan/sacubitril in PARAMOUNT and torasemide. In TOPCAT, spironolactone (a mineralocorticoid receptor antagonist, MRA) failed to reduce the composite primary endpoint in the overall trial population¹¹⁰ but not in patients with elevated BNP, which was a marker of enrollment in the Americas (p=0.003)¹¹¹. The neutral outcome in the overall population may have been related to aberrant patient enrolment in Russia/Republic of Georgia rather than to inefficacy of spironolactone. In PARAMOUNT, salutary effects of valsartan/sacubitril were observed in HFpEF patients consisting of a significant decrease in NT-proBNP and left atrial volume⁹⁵. These effects support fibrosis-specific therapy for HFpEF patients with advanced extracellular matrix modification¹¹². The loop diuretic torasemide affects collagen crosslinking and its use has been shown to improve diastolic LV dysfunction in patients with hypertensive heart disease¹¹³. Finally, use of mesenchymal stem cells has been examined in Dahl salt-sensitive rats with promising results. In this model, a single intracoronary dose of allogeneic mesenchymal stem cells reduced myocardial collagen volume fraction and normalized diastolic LV function without effect on cardiomyocyte hypertrophy¹¹⁴.

Arterial Hypertension

Arterial hypertension is found in 80% or more of HFpEF patients. Treatment of arterial hypertension in older people without HF reduces incident HF⁵³. In acutely decompensated HFpEF patients with elevated blood pressure, symptoms may improve markedly with blood pressure lowering alone even before diuresis is achieved. However, in chronic stable HFpEF patients there is uncertainty whether adding blood pressure lowering medications provides additional benefit. A discordance was indeed present between substantial blood pressure lowering and outcome in large trials testing neurohumoral inhibition in HFpEF². This was even more surprising, since along with blood pressure lowering, there were numerous other mechanisms whereby neurohumoral inhibition was expected to benefit HFpEF, including improvements in myocardial hypertrophy, myocardial fibrosis and vascular stiffness. However, treating arterial hypertension for non-HF related macrovascular indications (e.g. stroke, myocardial infarction) remains an important goal also in HFpEF patients. In this regard, it is worth noting that large outcome trials confirmed ACEIs and ARBs to be safe and well-tolerated as antihypertensives^2,4. Diuretics, spironolactone and ACEIs/ARBs are therefore reasonable first choices to control blood pressure based upon the currently available data. While it is true that previously tested ACEIs and ARBs did not reduce mortality, increasing the quality of life may be the better strategy in HFpEF patients because they are often elderly and debilitated and because some of the previously completed trials of ACEIs/ARBs had relevant symptomatic benefits such as reduced HF hospitalization^2,4.

Arterial hypertension can affect myocardial remodeling and dysfunction in HFpEF through myocardial overload¹¹⁵ or systemic inflammation¹¹⁶. The importance of overload is unclear because in a concentrically remodeled left ventricle with normal EF, a favorable late-systolic Laplace relation protects LV myocardium from loading increments provoked by large reflected arterial pressure waves. However, in the presence of a minor LV shortening deficit, hypertensive HFpEF patients may develop late-peaking systolic LV wall stress. This may explain the favorable effects in HFpEF patients of nitrate-rich beetroot juice, which reduces magnitude of reflected arterial pressure waves⁷⁵, or of the sodium-restricted DASH (Dietary Approaches to Stop Hypertension) diet, which improves ventricular-arterial coupling¹¹⁷.

Renal Dysfunction

HFpEF and renal dysfunction are mutually promoting (Figure 1)¹¹⁸. HFpEF promotes renal dysfunction by 1) an elevated central venous pressure, which results from pulmonary hypertension and right ventricular dysfunction; 2) inability to raise cardiac output following arterial vasodilation because of chronotropic incompetence and fixed LV stroke volume¹¹⁹; 3) systemic inflammation, endothelial dysfunction and low NO bioavailability, which reduces renal blood flow ^55,120 and sodium excretion⁴⁴. Renal dysfunction promotes HFpEF by worsening systemic inflammation, endothelial dysfunction and NO bioavailability partially because of renal-specific mediators such as high levels of fibroblast growth factor 23, phosphorus, parathyroid hormone or uremic toxins and low levels of vitamin D or erythropoietin¹¹⁸. Limited tolerability of systemic vasodilation and impaired sodium excretion are of therapeutic importance⁶⁸. Impaired sodium excretion implies the arterial pressure-natriuresis relationship to be shifted to the right. Under these conditions, a fall in arterial pressure because of systemic vasodilation without cardiac output increase is especially deleterious as it leads to additional sodium retention and extracellular volume expansion, which wipes out any direct beneficial effect of vasodilation on LV filling pressures⁶⁸. This mechanism could partially account for the neutral outcome of the RELAX trial⁹⁹, where sildenafil lowered arterial pressure, raised plasma creatinine and urea levels and failed to improve exercise tolerance.

HFpEF in the presence of renal dysfunction recently emerged as a distinct phenotype with more LV hypertrophy, a larger LV systolic functional deficit, impaired left atrial mechanics, right ventricular dysfunction and poor prognosis^121,122. The latter relates to exaggerated reactive pulmonary hypertension and right ventricular dysfunction. Because of right ventricular dysfunction, renal venous congestion importantly contributes to renal dysfunction in HFpEF. Vigorous diuresis (and ultrafiltration if necessary) are therefore important in HFpEF patients with renal dysfunction.

Coronary Artery Disease

The presence of coronary artery disease (CAD) also identifies a distinct HFpEF phenotype with a larger LV systolic functional deficit, poor prognosis^123,124 and a high incidence of sudden death¹²⁵. Use of ACEIs is recommended for prevention of new cardiovascular events. In HFpEF patients with CAD, observational data suggest that complete revascularization is associated with better preservation of LV systolic function and an improved prognosis, though prospective trial data are still lacking¹²³.

Chronotropic Incompetence

Many patients with HFpEF display marked impairments in cardiac output reserve during exercise, despite normal resting values¹²⁶. Impaired cardiac output reserve in HFpEF is related not only to decreased stroke volume augmentation but also to chronotropic incompetence^127-129. One study actually indicated that chronotropic incompetence was the major contributor to reduced cardiac output reserve in HFpEF⁴². The importance of chronotropic incompetence is further supported by the worsened exercise capacity when heart rate was slowed by the If blocker ivabradine¹³⁰. Chronotropic incompetence was previously shown to be related to endothelial dysfunction and systemic inflammation¹³¹ and therefore fits well into the multiorgan signaling cascade, which appears to drive HFpEF development. Because there is a direct relationship between heart rate response to activity and aerobic capacity¹²⁸, a clinical trial is currently testing whether rate adaptive atrial pacing can improve exercise capacity in patients with HFpEF (NCT02145351).

Pulmonary Hypertension

Recent evidence stressed the importance in HFpEF of pathophysiologic targets beyond the heart (Figure 1)⁵⁴. Pulmonary hypertension is frequently present at rest¹³² and patients can also develop an exaggerated pulmonary hypertensive response to exercise^36,37,39. In HFpEF, pulmonary pressures can be augmented by increased left atrial pressure and/or by pulmonary vasoconstriction. When both mechanisms prevail, combined precapillary and postcapillary pulmonary hypertension (CpcPH) is present.

Because of pulmonary hypertension and shared predisposing mechanisms, right ventricular dysfunction is common in HFpEF and is associated with increased morbidity and mortality^133-135. The right ventricle in HFpEF displays heightened afterload-sensitivity, suggesting favorable potential for benefit from reduction in pulmonary pressures (Figure 7)¹³³. An early single-center trial recruiting mainly CpcPH patients indeed reported salutary effects on hemodynamics and right ventricular function following treatment with the phosphodiesterase 5 inhibitor (PDE5I) sildenafil¹³⁶. However, two subsequent larger trials in patients with IpcPH (isolated postcapillary pulmonary hypertension) and CpcPH failed to corroborate this finding^99,100. A recent trial reported significant improvement in pulmonary vascular function in response to dobutamine in HFpEF patients, greatly exceeding the pulmonary vasodilatory response seen in non-HF controls¹³⁷. Improved right ventricular-pulmonary artery coupling in this study was achieved predominantly through reduction in afterload rather than enhanced right ventricular function, highlighting the importance of management of pulmonary hypertension in HFpEF. A number of trials have or are currently testing the effects of pulmonary vasodilators targeting cGMP¹³⁸, endothelin¹³⁹ and NO (NCT02713126; NCT02262078) in subjects with HFpEF.

Figure 7. Heightened afterload sensitivity of the right ventricle in HFpEF.

The relation between echocardiographic right ventricular (RV) fractional area change (FAC) and mean pulmonary artery (PA) pressure is flat in controls but steep in HFpEF patients.

In contrast to the pulmonary vasculature, changes in the lung parenchyma are less characterized in HFpEF but likely also play an important role. Impairments in pulmonary function predict incident development of HFpEF independent of cardiac function¹⁴⁰. Patients with HFpEF display gas exchange abnormalities manifest by reduced alveolar capillary membrane conductance^141,142. These impairments become more dramatic during exercise owing to high LV filling pressures during stress¹⁴¹. HFpEF patients with increased interstitial pulmonary edema display greater pulmonary vascular abnormalities and RV dysfunction, supporting aggressive therapies to reduce left heart filling pressures chronically in patients with HFpEF¹⁴².

Skeletal Muscle Weakness

Exercise intolerance can be objectively measured as peak VO₂. By the Fick equation, VO₂ is the product of cardiac output and arteriovenous oxygen difference (ΔA-VO₂). Multiple studies indicate that peak exercise ΔA-VO₂ is significantly reduced in HFpEF and accounts for 50% or more of their severely reduced peak VO₂^40,143. What are the causes of the reduced peak ΔA-VO₂ in HFpEF patients? HFpEF patients have abnormalities in skeletal muscle mass, composition, capillary density, and oxidative metabolism. Haykowsky et al. showed that compared to age-matched healthy controls, older HFpEF patients have significantly reduced percent total lean body mass and percent leg lean mass⁴³. When peak VO₂ was indexed to total lean body mass or leg lean mass, it remained significantly reduced. Thus, HFpEF patients have abnormal O₂ utilization that is independent of and in addition to their reduced muscle mass. HFpEF patients also have abnormal skeletal muscle composition with infiltration of adipose tissue, which is directly related to their reduced peak VO₂¹⁴⁴. Increased intramuscular fat reduces capillary density, thereby increasing the distance O₂ must traverse from the capillaries to the muscle fibers. In HFpEF patients the reduced thigh muscle capillary density is associated with their reduced peak VO₂ (Figure 8)¹⁴⁵. Multiple studies indicate that HFpEF patients also have impaired skeletal muscle oxidative metabolism. Kitzman et al. showed that compared to healthy age-matched controls, HFpEF patients have a shift in skeletal muscle fiber type distribution from oxidative, slow type 1 fibers to glycolytic, fast type 2 fibers, which results in a lower type 1/type 2 fiber ratio. Similarly to capillary density, these alterations are also associated with their severely reduced peak exercise VO₂ (Figure 8)¹⁴⁵. A consequence of this fiber type shift is reduced skeletal muscle oxidative metabolism during exercise, which was evident after cessation of exercise from the delayed regeneration of quadriceps muscle phosphocreatine stores using phosphorus magnetic resonance spectroscopy¹⁴⁶.

Figure 8. Peak VO₂ and skeletal muscle histology in HFpEF.

Relationship of capillary to fiber ratio (Panel A) and percent type 1 muscle fibers (Panel B) with peak VO₂ in older HFpEF patients (squares) and age-matched healthy controls (triangles).

What are the implications of these extensive skeletal muscle abnormalities in HFpEF? First, they confirm that HFpEF is a systemic disorder involving not only the heart but also other organ systems and that skeletal muscle and cardiac abnormalities are incited by common, circulating factors such as proinflammatory cytokines originating from multiple comorbidities¹⁴⁷. Second, they suggest opportunities for novel interventions. Unlike the myocardium, which is terminally differentiated and has minimal capacity for regeneration, skeletal muscle has robust capacity for rapid repair, regeneration, and growth, which can be exploited by participation in an exercise training program¹⁴⁸. Exercise training, shown in multiple studies to significantly improve peak VO₂ in HFpEF^148-151, achieves this primarily by improving skeletal muscle mitochondrial mass or function¹⁴⁸. Most studies to date have used endurance training; high-intensity and strength training might produce even larger improvements but have not been examined systematically¹⁴⁸.

Atrial Fibrillation

Prevalent atrial fibrillation in HFpEF goes along with a more advanced stage of cardiac remodeling evident from a larger left atrium and uniformly carries a worse prognosis^62,152,153. Incident atrial fibrillation in HFpEF also accompanies worse LV diastolic dysfunction and was inversely related to statin use⁶². Prevalent atrial fibrillation was shown to be associated with incident HFpEF and prevalent HFpEF with incident atrial fibrillation¹⁵⁴. These interactions suggest atrial fibrillation to beget HFpEF and vice versa and suggest efforts to restore sinus rhythm could be included in a HFpEF treatment strategy. Similar to HFpEF, atrial fibrillation reacts favorably to exercise training and weight loss¹⁵⁵. To restore sinus rhythm, only cardioversion is recommended as catheter ablation of atrial fibrillation had limited long term success in HFpEF with single- and multiple-procedure drug-free success rates of respectively 27 and 45%¹⁵⁶. If cardioversion is unsuccessful, rate control and permanent anticoagulation become mandatory.

CONCLUSIONS

HFpEF is the most common form of HF, is increasing out of proportion to HFrEF, and is associated with significant morbidity and mortality. Medication trials to date have been largely neutral on their primary outcomes, and so far only exercise training and weight loss appear to improve exercise intolerance and quality of life. Recent insights provide an understanding of the fundamental basis of left ventricular dysfunction in HFpEF, which involves systemic inflammation, coronary microcirculatory disturbances, cardiomyocyte stiffening and myocardial fibrosis. These insights also provide a more expanded view on HFpEF that includes involvement of the pulmonary circulation, right ventricular failure, skeletal muscle weakness and renal dysfunction. These new perspectives on HFpEF open an array of novel therapeutic targets either in the “garden variety” phenotype of lung congestion/metabolic risk or in specific phenotypes that may propel future advances in treatment and prevention of this important disorder.