Heart failure with preserved ejection fraction (HFpEF) accounts for half of all heart failure diagnoses and is associated with similar morbidity and mortality to heart failure with reduced ejection fraction (HFrEF) [1]. Until recently, there were no treatments which were shown to have a prognostic benefit in HFpEF ^[2][3][2,3]. It is growing in prevalence, owing largely to increases in risk factors such as obesity, hypertension, and diabetes within an ageing population [1]. Despite this, there is a lack of clear understanding of the pathophysiological processes at play in this condition. Moreover, the possibility of distinct HFpEF phenotypes has emerged, including cardiometabolic, elderly hypertensive, right heart/pulmonary dysfunction, and left atrial myopathy. Broadly speaking, the condition is characterised by impaired cardiac relaxation and diastolic dysfunction, resulting in high filling pressures. The factors leading to this impaired relaxation are complex but are likely to be related to passive factors such as increased myocardial fibrosis and inherent cardiomyocyte stiffness, and active factors involved with ATP-dependent relaxation. This resviearchw focuses predominantly on these active mechanisms.

The heart is a highly energy-demanding organ, consuming more energy than any other organ at around 6 kg of ATP per day [4]. It has been postulated that derangements in this energy production process may play an integral role in the development of heart failure and may even be present before structural and functional sequalae of cardiac failure become apparent ^[4][5][4,5]. This has been most thoroughly investigated in HFrEF, whereby derangements in all three key steps of substrate utilisation, energy production, and energy transport have been shown. Moreover, impairment of myocardial energetics as assessed by phosphorous spectroscopy has been shown to correlate with higher symptom burden and worse prognosis in HFrEF [6].

Despite this clear link between HFrEF and myocardial energetics, the most energetically demanding part of the cardiac cycle is diastole, where ATP is used to break actomyosin cross links and thus allow relaxation [5]. Moreover, the most energetically demanding enzyme in the contractile apparatus, the sarcoplasmic reticular calcium ATPase (SERCA), will be affected early by any derangements in metabolism, preventing calcium lowering in diastole [7].

2. Normal Cardiac Metabolism

The initial steps in ATP generation are the uptake and utilisation of fuels in the myocardium. Under normal conditions, the heart sources the majority of its ATP from fatty acids (70–90%) and to a lesser extent carbohydrates (10–40%), with smaller contributions from ketone bodies and amino acids (Figure 1) [4]. The final end product of these pathways is acetyl coenzyme A which can then be fed into the tricarboxylic acid (TCA) cycle. The heart is metabolically flexible, adapting its substrate usage based on local and systemic conditions, allowing ATP generation to continue in fed, fasted, and high-demand states. The second stage in this process is the generation of ATP via oxidative phosphorylation, which accounts for >95% of cardiac ATP. Following the generation of high-energy electron carriers (NADH and FADH₂) in the TCA cycle, and to a lesser extent glycolysis, electrons are donated to the electron transport chain allowing the establishment of a proton gradient across the inner mitochondrial membrane. ATP synthase then utilises this gradient to phosphorylate adenosine diphosphate (ADP), creating ATP (Figure 1). The final phase is ensuring the highly energetic ATP molecule can be transferred to the myofibrils for utilisation. Given its relatively large size and polarity, this is achieved via a phosphate transfer system. Creatine kinase catalyses the conversion of creatine to phosphocreatine (PCr) at the mitochondria, which can then diffuse freely to the myofibril (Figure 1). Here, the cytosolic isoenzyme catalyses the reverse reaction, allowing phosphorylation of the recently produced ADP, replenishing ATP supplies at the myofibril. Not only does this system function as a spatial energy shuttle, but it also serves an important temporal energy buffer, whereby PCr can replenish ATP supplies by phosphate donation at times of high demand.

3. Cardiac Metabolism in HFpEF

3.1. Animal Studies

Much of our knowledge of myocardial metabolism in heart failure comes from ischaemia–reperfusion animal models which reliably progress to HFrEF. However, given the distinct risk factors, pathophysiological mechanisms, and clinical phenotype in HFpEF, any derangement in cardiac metabolism may be different from than observed in HFrEF. Animal models of HFpEF have, however, proven more difficult to develop, partly owing to the heterogeneity in the clinical phenotype of HFpEF patients. Early studies focused on HFpEF risk factors such as obesity, hypertension, diabetes, and ageing, with animal models of these conditions in isolation being used as surrogates for HFpEF. Whilst many of these models showed classic features of HFpEF such as diastolic dysfunction and preserved systolic function, they often lacked key clinical features such as pulmonary congestion and elevated filling pressures [8]. More advanced animal models of HFpEF have sought to combine hypertensive, metabolic, and ageing stressors and have had more success in generating a more comprehensive set of HFpEF features. For instance, a novel ‘three-hit’ HFpEF mouse model has recently been developed, with obesity induced via a high-fat diet and hypertension via an injection of desoxycorticosterone pivalate (DOCP), with ageing as the final ‘hit’. This model accurately replicates key haemodynamic and clinical features of HFpEF including diastolic dysfunction, reduced exercise tolerance, and pulmonary congestion, all whilst maintaining left ventricular ejection fraction [9]. Interestingly, hypertensive and metabolic stresses appear to have competing consequences for substrate utilisation. In single-factor hypertensive models, a shift away from fatty acid oxidation towards glucose metabolism becomes apparent [10], similar to that seen in HFrEF. Whilst traditionally this shift was believed to confer an efficiency benefit in terms of oxygen use, it is becoming more established that this increased glycolytic activity occurs to support aspartate synthesis, a key intermediate in nucleotide synthesis, which is important in driving cardiac hypertrophy [11]. In contrast, metabolic models such as db/db mice ^[12][13][14][12,13,14], ZDF rats ^[15][16][15,16], and HFD/STZ rats ^[17][18][17,18] have increased fatty acid oxidation, presumably due to insulin resistance and decreased access to glucose. In the previously mentioned ‘three-hit’ HFpEF model, the authors found that fatty acid utilisation was slightly increased. It may be the case that the insulin resistance present caused a reciprocal increase in fatty acid oxidation via the Randal cycle, ultimately explaining the slightly increased fatty acid oxidation. It is clear, however, that there are complex pathways governing substrate use and that the flux of these pathways is likely to change with comorbidities and therefore HFpEF phenotype. The increase in fatty acid oxidation does not appear to be sufficient to prevent accumulation of fatty acids and associated lipotoxicity. This intramyocardial lipid accumulation appears to be directly related to the development of diastolic dysfunction ^[19][20][21][19,20,21]. Fatty acids gain entry to the cytoplasm via the fatty acid translocase (CD36) located on the cytoplasmic membrane. Upon entry, they must cross the mitochondrial membrane in order to undergo fatty acid oxidation. Long-chain fatty acids, however, are unable to cross this membrane and require the aid of a carnitine shuttle. The key transporter in this shuttle is carnitine palmitoyltransferase 1 (CPT1), which regulates fatty acid uptake into the mitochondria and thus oxidative phosphorylation. Ketone bodies are an alternative myocardial fuel, and recent work has suggested that they contribute to myocardial metabolism in the healthy human heart ^[22][30]. These are synthesised from acetyl coenzyme A in hepatocytes and provide a carbon source for ATP synthesis in times of fasting, exercise, and exogenous ketone ingestion ^[23][31]. In HFrEF animal models, ketone body utilisation increases in what appears to be an adaptive response ^[24][25][32,33]. Furthermore, mice with cardio-specific knockdown of key ketolytic enzymes experienced worsening hypertrophy and systolic dysfunction ^[26][34], while infusion of ketone bodies appears to be protective ^[26][27][34,35]. Downstream of substrate utilisation, few studies have probed myocardial energetics in animal models of HFpEF [8]. Studies using 31-phosphorous magnetic resonance spectroscopy have shown decreases in PCr/ATP with obesity ^[28][29][36,37] and ageing ^[30][38] in murine models. However, data are lacking for phosphorous spectroscopy studies in two- and three-hit animal models of HFpEF. This is likely to be important in validating these newer models of HFpEF against human HFpEF. Finally, ATP generated at the mitochondria must be transferred to the myofibrils via the creatine kinase (CK) shuttle. Again, this has been studied in conditions predisposing to HFpEF rather than in the condition directly. Forward CK flux ([phosphocreatine] x CK forward rate constant) has been shown to be significantly reduced in canine and porcine models of left ventricular hypertrophy (LVH) ^[31][32][39,40], with greater reductions present upon development of heart failure ^[31][39]. In isolated perfused hearts of obese mice who exhibited diastolic dysfunction, CK flux was maintained due to elevations in CK forward rate constant ^[29][37]. Similarly, in a diabetic rodent model, the CK forward rate constant was increased with CK flux similar to matched controls ^[33][41]. Future studies should seek to investigate CK kinetics in animal models of HFpEF that combine risk factors.

3.2. Human Studies

In contrast to the animal studies mentioned above, the bulk of work probing myocardial metabolism in human subjects with HFpEF has been directed at downstream ATP production. Most of this work has utilised phosphorous spectroscopy to non-invasively assess PCr/ATP ratio. Decreases in PCr/ATP in human subjects with HFrEF are well recognised ^{[6][34][35][36]}[6,42,43,44], and PCr/ATP has been shown to be a better predictor of mortality than left ventricular ejection fraction ^[34][42]. Similarly, in HFpEF patients, a 20–27% reduction in PCr/ATP has been shown ^[37][38][39][45,46,47]. Moreover, PCr/ATP has been shown to be reduced in the hearts of patients with risk factors for HFpEF such as obesity ^[40][41][48,49], diabetes ^[37][42][45,50], hypertension ^[43][51], and ageing ^[44][52]. Interestingly, recent work has linked this energetic deficit in HFpEF to impaired left ventricular diastolic reserve, left atrial dilation, and pulmonary congestion during exercise, supporting the importance of energetic derangement as a key mechanism in HFpEF ^[37][45]. Once ATP is produced, the creatine kinase shuttle is crucial in the transportation of high-energy phosphates from mitochondria to myofibrils for utilisation. In patients with non-ischaemic cardiomyopathy, myocardial CK flux has been shown to predict heart failure outcomes independent of left ventricular ejection fraction and NYHA class ^[45][53]. Whilst limited data are available for CK flux in HFpEF, it has been studied in left ventricular hypertrophy (LVH) ^[46][54] and in obesity ^[40][48]. In hypertension-related LVH, the forward rate constant of myocardial CK reaction was normal, with CK flux reduced by 30% compared with healthy controls. However, in those with LVH and associated systolic dysfunction, the forward rate constant was halved and CK flux was reduced by almost two-thirds ^[46][54]. In obesity, the forward rate constant of the myocardial CK reaction at rest was found to be increased, yielding no overall difference in resting ATP delivery (CK flux) ^[40][48]. However, upon exercise, this flux was unable to be augmented, resulting in cardiopulmonary exercise intolerance ^[40][48]. Given exercise intolerance is a cardinal feature of HFpEF, it will be interesting to examine CK flux at rest and stress in individuals living with HFpEF. Upstream substrate utilisation has not been fully assessed in patients living with HFpEF. Studies in patients living with obesity have shown correlations between insulin resistance and increased myocardial fatty acid uptake, utilisation, and oxidation ^[47][55], while studies in patients with diabetic cardiomyopathy have shown increased myocardial fatty acid uptake and oxidation and decreased glucose uptake ^[48][49][56,57]. A mismatch between the uptake and oxidation of fatty acid species leads to their accumulation and resultant lipotoxicity. Proton (¹H) spectroscopy can be used to non-invasively assess myocardial triglyceride content and has been used to show that HFpEF patients have significantly more intramyocardial fat than either HFrEF or control subjects ^[50][61]. Furthermore, intramyocardial fat correlated with the severity of diastolic dysfunction independently of risk factors. The extent to which myocardial lipotoxicity drives HFpEF is unknown, but it represents an attractive treatment target.

4. Therapeutic Strategies

Given the metabolic derangements in HFpEF discussed above, therapeutic modulation of metabolic pathways represents an exciting new treatment strategy for HFpEF. This may involve manipulating fatty acid, glucose, or ketone oxidation (Table 1) to yield improved ‘fuel efficiency’, improving contractile function by increased ATP availability.