Landscape of Glycolytic Metabolites and their Regulating Proteins in Myocardium from Human Heart Failure with Preserved Ejection Fraction

Abstract

Aims:

Heart failure with preserved ejection fraction (HFpEF) reflects half of all clinical HF yet has few therapies. Obesity and diabetes are now common co-morbidities which has focused attention towards underlying myocardial metabolic defects. The profile of a major metabolic pathway, glycolytic intermediates and their regulating enzymes and ancillary pathways, remains unknown.

Methods and Results:

Endomyocardial biopsies from HFpEF (n=37) and non-failing controls (n=21) were assayed by non-targeted or targeted metabolomics and immunoblot to determine glycolytic and ancillary pathway metabolites and protein expression of their regulating enzymes. Glucose and GLUT1 expression were higher in HFpEF, but prominent glycolytic metabolites: glucose-6-phosphate, fructose-1,6-biphosphate, and 3-phosphoglycerate were reduced by −79%, −91%, and −73% respectively, versus controls. Expression of their corresponding synthesizing enzymes hexokinase, phospho-fructokinase, and phosphoglycerate kinase were also significantly lower (all P<0.0005). Pentose phosphate and hexosamine bio-synthetic pathway metabolites were reduced while glycogen content increased. Despite proximal reduction in key glycolytic intermediates, pyruvate increased but mitochondrial pyruvate transporter (MPC1) expression was reduced. Pyruvate dehydrogenase converting pyruvate to acetyl-CoA was more activated but some Krebs cycle intermediates were reduced. This HFpEF glycolytic profile persisted after adjusting for body mass index (BMI), diabetes, age, and sex, or in subgroup analysis with controls and HFpEF matched for BMI and diabetes/insulin history. In HFpEF, BMI but not hemoglobin-A1c negatively correlated with F1,6bP (P=7e-5, r=−0.61) and phosphoenolpyruvate (P=0.006, r=−.46).

Conclusions:

Human HFpEF myocardium exhibits reduced glycolytic and ancillary pathway intermediates and expression of their synthesizing proteins. This combines features reported in HF with reduced EF and obesity/diabetes that likely exacerbate metabolic inflexibility.

Introduction:

Heart failure with preserved ejection (HFpEF) accounts for over half of all heart failure worldwide, conferring substantial morbidity and mortality, and has few effective treatments^1,2. While initially described in patients of advanced age with hypertension and ventricular hypertrophy, HFpEF has now become dominated by obesity and cardiometabolic disease². Emblematic of this transformation is that re-purposed diabetes (DM) drugs that block sodium-glucose transport are quite effective^3,4, whereas traditional hemodynamic and neurohormone targeting therapies are not. Such findings and recent preclinical model data are fueling efforts to better understand HFpEF metabolic defects and what might be done to ameliorate them^5–7.

The normal heart favors fatty acid (FA) as its primary fuel; however, in HF with a reduced EF (HFrEF)^5,8,9 and hearts subjected to sustained pressure-overload¹⁰, myocardial FA uptake and oxidation decline, myocardial levels of medium-long chain acylcarnitine (MLCA) required for FA oxidation are reduced⁶, and metabolism shifts to alternative fuels such as ketones^5,6,11–13. In both animal models of HFrEF and pressure-overload, or in human HFrEF, myocardial glucose uptake and/or levels are generally higher^{8–10,13–16}. This is thought to be beneficial as genetically blocking glucose utilization in pressure-overloaded mice is detrimental to the heart¹⁷. By contrast, obesity and DM are associated with reduced glucose uptake and utilization and declines in glycolytic enzymes as hearts rely more on FA oxidation^15,18.

Beyond data on net glucose uptake in HFpEF hearts which has been reported to be reduced¹⁹, virtually nothing is known about the myocardial glycolytic landscape in this syndrome. Whether an obesity/DM profile dominates, or features found in non-obese/DM HFrEF are more characteristic is unknown. One indication that HFpEF may differ from obesity/DM comes from myocardial metabolomics of human HFpEF that found MLCA levels and expression of genes for fatty acid uptake and oxidation are similarly depressed in HFpEF as in HFrEF⁶. Accordingly, the current study determined the glycolytic landscape of HFpEF myocardium using non-targeted and targeted metabolomics and protein expression and activity assays.

Methods:

Clinical source of myocardial tissue:

HFpEF myocardium was obtained from patients with HFpEF evaluated at the Johns Hopkins University HFpEF Clinic and prospectively diagnosed using consensus criteria². Details of the diagnostic criteria, patient clinical, echocardiographic, and hemodynamic characteristics have been reported⁶. Non-failing control tissue was provided from organ donors using tissue harvesting methods in both groups as previously described (see also Supplemental Methods).^6,20

Thirty-seven HFpEF biopsies and 21 control myocardial samples from the right ventricular mid-septum were studied. Metabolomics used both non-targeted and targeted assays. For the former, glycolytic intermediates: glucose, glucose-6-phosphate (G6P), fructose-1,6-biphosphate (F1,6bP), phosphoenolpyruvate (PEP), lactate, ribose-5P (R5P), sedoheptulose-7P (S7P), and Xylulose-5P (X5P) were identified, using n=13 controls, and n=24 HFpEF. Targeted metabolomics was then performed to capture metabolites missed in the non-targeted assay using n=16 non-failing and n=14 HFpEF. Eight controls and one HFpEF patient were shared between mass spectroscopy assays as there was insufficient tissue to run them in all patients. We found high concordance for the 5 metabolites measured in both assays (glucose, G6P, F1,6bP, PEP, and lactate, Figure S2), and show results from the non-targeted assay for these as they were derived from the larger HFpEF sample. Metabolites identified only in the targeted assay were fructose-6P (F6), glyceraldehyde-3P (G3P), 3-phosphoglycerate (3PG), pyruvate, acetyl-CoA, oxalacetate (OA), Glucosamine-6P, GlcNAc-6P, and UDP-GlcNAc.

Metabolite and protein extraction:

Details of myocardial tissue preparation, metabolite, and protein extraction, metabolomic assays, and protein analysis and quantitation are provided in Supplemental Methods.

Tissue glycogen and hexokinase activity assays:

Details are provided in Supplemental Methods.

Statistical Analysis:

For metabolomic and protein blot assays, data are shown normalized to the non-failing control mean. Plots show individual results and mean ± SEM. Raw values are displayed for assays with absolute units. Sample size, statistical results, and tests used for all analyses are provided in the figures or figure legends. Other statistical and analysis details are provided in Supplemental Methods.

Results:

Clinical characteristics of study groups

Demographics of the control and HFpEF cohorts are provided in Table S1, and comprehensive features measured only in the HFpEF cohort only provided in Table S2. HFpEF patients had a systolic blood pressure and mean pulmonary artery pressure of 141±17 and 26±11 mmHg, respectively, pulmonary capillary wedge pressure of 18±8 mmHg, LV ejection fraction of 63±6%, and median NT-proBNP of 332 pg/ml. These are similar to values in prior reports from these cohorts^6,20. The HFpEF population was generally obese (71% with body mass index (BMI)>30 kg/m², 43% BMI≥35); non-failing controls were less obese. HFpEF patients also had more type II DM (54% vs 19%, P=0.013) with 33% having a HbA1C ≥ 6.5%. Non-targeted and targeted metabolomics included shared but also different patients, but clinical features for the patients used in each assay were statistically similar (Tables S3, S4).

Proximal glycolysis intermediates and their generating enzymes

Figure 1A shows initial steps in glycolysis, with arrows identifying significant changes in HFpEF myocardium. Glucose increased (+74.8%, [36, 114], mean, [95% CI]) associated with greater expression of non-insulin dependent glucose transporter GLUT1 (+175%, [175,272]) but not insulin-dependent GLUT4 (Fig 1B, 1C). In the first irreversible step, hexokinases (HK1, HK2) convert glucose to G6P; and all were reduced significantly in HFpEF (HK1 −23% and HK2 −52% G6P −78% [−86, −70]; Figure 1D, 1E). Hexokinase activity significantly declined by −25%. The subsequent metabolite, F6P was at control levels suggesting possible engagement of a bypass pathway. However, the following and irreversible step producing F1,6bP by phospho-fructokinase (PFKM, PFKP) showed marked reductions in the metabolite (−91% [−97, −86]) and both kinases (−40%) (Fig 1E, 1F).

Figure 1. Changes in proximal glycolysis in myocardium from human HFpEF.

A) Proximal glycolytic pathway with arrows identifying statistically significant changes in HFpEF versus control. B) Glucose and protein expression of GLUT1 and GLUT4 for both groups, and C) representative immunoblots for GLUT1, GLUT4 protein expression. D) G6P level and protein expression of HK1 and HK2, and myocardial HK activity. E) Representative immunoblots for HK1, HK2, PFKP, and PFKM. F) F6P, F1,6bP levels and summary protein expression for PFKP and PFKM. For all summary figures, data are shown normalized to control, sample size for each group is provided below the x-axis, and P values are for Mann Whitney test between groups. Abbreviations: G6P: glucose-6-phosphate, F6P: fructose-6-phosphate, F1,6bP: fructose 1,6 biphosphate, G3P: glyceraldehyde 3-phosphate, 1,3bPG – 1,3 bi-phosphoglycerate, 3PG: 3-phosphoglycerate, 2PG: 2-phosphoglycerate, PEP: phosphoenolpyruvate, E4P: erythrose 4-phosphate, S7P: sedoheptulose 7-phosphate, R5P: ribulose 5-phosphate, X5P: xylulose 5-phosphate, HK – hexokinase, PFK: phosphofructokinase, GAPDH: GAPDH – glyceraldehyde 3-phosphate dehydrogenase, PGK: phosphoglycerate kinase, PK pyruvate kinase, GYS1: glycogen synthase 1, PYG: glycogen phosphorylase, PDH: pyruvate dehydrogenase, PC: pyruvate convertase, A-CoA - acetyl-CoA, G6PD: glucose 6-phosphate dehydrogenase.

Pentose phosphate, hexosamine, and polyol pathway metabolites

The pentose phosphate pathway (PPP) uses G6P as substrate and regulates oxidant/reductive balance and nucleotide synthesis⁵. It can also bypass suppressed proximal glycolysis to provide alternative sources for F6P and G3P (Figure 2A). Several PPP metabolites were significantly lower in HFpEF, including R5P (−39%, [−62,−15]), X5P (−44% [−66,−21]) and S7P (−92%, [−94,−86]). the ]), the latter capable of feeding back into glycolysis by converting to F6P (Figure 2A, 2B). G3P found in both PPP and glycolytic pathways was at control levels (Figure 2C).

Figure 2.

A) Schematic of proximal glycolysis related ancillary pathways – pentose phosphate, hexosamine biosynthetic pathway, and polyol pathway. B) Pentose phosphate pathway intermediates, ribose-5P (R5P), xylulose (X5P), and Sedoheptulose-7P (S7P) in non-failing and HFpEF myocardium. C) glyceraldehyde-3-phosphate (G3P) levels. D) Hexosamine biosynthetic intermediates: glucosamine-6-phosphate, N-acetylglucosamine-6-phosphate (GlcNAc-66P), uridine diphosphate-N-acetylglucosamine (UDP-GlcNAc). E) Immunoblot assay for total protein O-GlcNAcylation in non-failing and HFpEF myocardium. F) Summary densitometry quantifying total protein O-GlcNAcylation. G) Levels of sorbitol and fructose in both groups. For all summary figures, data are normalized to controls, the sample size for each group is provided below the x-axis, and P values are for Mann Whitney test between groups.

The hexosamine biosynthetic pathway (BP) utilizes F6P to fuel N- or O-linked glycosylation and protein O-GlcNAcylation that impacts their functionality⁹. Metabolites in the HBP are shown in Figure 2D, and were also reduced in HFpEF myocardium, the largest decline being in the end-product UDP-GlcNAc (−80% [−87,−72]. However, this did not translate into reduced total protein O-GlcNAcylation that was similar between the groups (Figure 2E, 2F).

The polyol pathway converts glucose into sorbitol and fructose and is involved with redox chemistry (Fig 2A). Both metabolites increase with DM and thought to contribute to its complications^21,22, and also rise in human HFrEF myocardium¹³. Surprisingly, both sorbitol ( −78%, [−87, −69]) and fructose (−61%, [−83, −39]) were significantly lower in HFpEF (Figure 2G).

Downstream glycolytic and mitochondrial metabolites

More distal glycolytic intermediates are shown in Figure 3A. 3-phosphoglycerate (3PG) is a major node in glycolysis as it serves to fuel biosynthetic pathways. It was substantially reduced in HFpEF (−73%, [−93,−53]) as was protein expression of glyceraldehyde 3-phosphate dehydrogenase (GAPDH, −40%) and phosphoglycerate kinase-1 (PGK-1, −32%) both involved with 3PG generation (Figures 3B, S3A, S3B). Despite lower 3PG, the subsequent metabolite phosphoenolpyruvate (PEP) was at control levels. While its synthesizing enzyme enolase-1 (ENO1) was similarly expressed in both groups (Fig 3B, Fig. S3C), pyruvate kinase (PKM) that metabolizes PEP to pyruvate was significantly reduced (−40%, [−55,−27], Fig. 3C). Interestingly, the final product of glycolysis, pyruvate, was increased 11-fold [6.5, 15.8] in HFpEF (Figure 3D), while lactate was similar in both groups (Figure 3E).

Figure 3.

A) Schematic of glycolysis highlighting distal metabolites and associated proteins. B) Protein expression of GAPDH and PGK-1 and levels of 3-phosphoglycerate (3PG). C) Levels of phosphoenolpyruvate (PEP) and enolase 1 (ENO1). D) Protein expression of pyruvate kinase (PK) and representative immunoblots and summary densitometry. E) Pyruvate and lactate levels. For all summary figures, data are normalized to controls, the sample size for each group is provided below the x-axis, and P values are for Mann Whitney test between groups.

To further explore the potential impact of higher pyruvate, we examined enzymes and metabolites involved with its uptake and processing in mitochondria (Figure 4A). Mitochondrial carrier protein MPC1 was reduced in HFpEF (Figure 4B, 4C) which could limit uptake²³. Pyruvate dehydrogenase (PDH) converts pyruvate to acetyl-CoA in mitochondria and its activity varies inversely with phosphorylation by PDK4. Both phospho-PDH/total PDH ratio and expression of PDK4 were significantly lower in HFpEF (Fig 4D), yet acetyl-CoA levels were similar to control (Figure 3E, left). Pyruvate can also be converted to oxaloacetate (OA) that when combined with acetyl-CoA generates citrate used in the tricarboxylic acid cycle (TCA), or be converted to aspartate. Both OA and aspartate were higher in HFpEF (Figure 4E). However, downstream TCA metabolites malate, fumarate and succinate were generally lower (Fig S4) as we had previously reported⁶.

Figure 4.

A) Downstream pathways for pyruvate metabolism in mitochondria, and glucose conversion to glycogen. B) Representative immunoblots for expression for mitochondrial pyruvate carrier-1 (MPC1), total pyruvate dehydrogenase (PDH), phosphorylated PDH (p-PDH), and pyruvate dehydrogenase lipoamide kinase isozyme 4 (PDK4). C, D) Summary densitometry for the same proteins. E) Acetyl-CoA, oxaloacetate (OA), and aspartate levels. F) Representative immunoblots for glycogen synthase (GYS) and glycogen phosphorylase-M (PYGM) and (PYGL) and G) summary densitometry for each protein. H) Glycogen content in the NF (n=9) and HFpEF (n=10) myocardium. Glycogen is normalized to total protein in the sample.

Increased glycogen in HFpEF myocardium

Increased myocardial glucose yet depressed proximal glycolytic and ancillary pathway intermediates suggested the potential for enhanced glucose storage as glycogen. We found glycogen synthase (GYS; generates glycogen from G6P) and glycogen phosphorylase (PYGM, PYGL; converts it back) both reduced in HFpEF (Figure 4F, 4G). This was associated with a >5-fold increase in myocardial glycogen content (Fig 4H).

Obesity and DM do not explain glycolytic disparities between HFpEF and Controls

Both obesity and diabetes are well known to be associated with reduced glycolysis^5,15, and as the HFpEF cohort was more obese and diabetic versus controls, analysis of their potential role to the glycolytic findings was important. We did this several ways. First, multiple regression analysis was performed that patient group, but also sex, age, BMI, and DM status as covariates; metabolite level or protein expression was the dependent variable. We found group differences in metabolites and proteins remained highly significant after adjusting for the other covariates (Table 1). The covariates had no significant impact generally, but there were some exceptions: females had higher hexokinase, phosphofructokinase and pyruvate kinase levels, diabetic patients tended to increased glycogen regulating proteins, and more obesity was negatively associated with PPP intermediates and positively with F6P and G3P.

Table.

Multiple regression analysis to test metabolite and protein expression disparities between groups that also accounted for the individual’s body mass index (BMI), history of diabetes (DM), sex, and age. The table provides the individual factor coefficient P-value from the multiple regression. A red color indicates the factor was negatively associated with a given metabolite or protein; a blue color indicates a positive association.

Metabolite	Group	BMI	DM	Sex (M)	Age
Glucose	0.25	0.26	0.38	0.42	0.57
G6P	0.0003	0.92	0.53	0.73	0.45
F6P	0.14	0.03	0.5	0.7	0.28
F1,6bP	7.20E-05	0.22	0.58	0.32	0.09
G3P	0.82	0.03	0.29	0.42	0.72
3PG	0.03	0.21	0.92	0.13	0.18
PEP	0.22	0.58	0.29	0.14	0.66
Pyruvate	0.0004	0.8	0.85	0.68	0.12
Lactate	0.050	0.1	0.72	0.19	0.25
R5P	0.0065	0.04	0.97	0.13	0.99
X5P	0.0049	0.03	0.77	0.16	0.86
S7P	0.0003	0.76	0.83	0.65	0.99
Fructose	0.024	0.52	0.55	0.21	0.61
Sorbitol	0.021	0.73	0.77	0.6	0.2
GlcNac6P	0.24	0.09	0.6	0.57	0.15
Glucosamine-6P	0.21	0.35	0.9	0.23	0.69
UDP-GlcNac6P	0.01	0.75	0.23	0.72	0.15
Protein	Group	BMI	DM	Sex (M)	Age
Glut1	1.24E-05	0.61	0.67	0.03	0.54
GLut4	0.16	0.07	0.06	0.72	0.22
HK1	0.0058	0.12	0.19	0.04	0.16
HK2	8.27E-05	0.92	0.07	0.02	0.78
PFKM	0.0014	0.97	0.31	0.008	0.29
PFKP	6.10E-05	0.25	0.10	0.005	0.27
GYS1	0.005	0.80	0.02	0.08	0.13
PYGL	0.0001	0.67	0.07	0.62	0.35
PYGM	4.00E-06	0.91	0.05	0.37	0.43
GAPDH	0.0003	0.17	0.09	0.36	0.75
PGK1	0.011	0.31	0.79	0.69	0.80
PKM	1.24E-05	0.78	0.03	0.028	0.23
MPC1	0.0059	0.24	0.23	0.94	0.48
PDK4	1.93E-05	0.74	0.92	0.83	0.67
P-PDH/Total	0.005	0.31	0.74	0.71	0.87

In a second approach, HFpEF patients with a BMI > 35 were removed from the analysis was a control with BMI<15 to generate a HFpEF subgroup that had matched BMI and %DM versus controls. Whether matched for these co-morbidities or not, the metabolic disparities in HFpEF vs control were very similar (Figure S5). Lastly, we removed any patient from either group that had a DM history or had received insulin and then repeated the comparisons. These results were also very similar to the primary analysis (Figure S6). Together, these analyses show that DM or BMI were not likely primary determinants for the marked differences in HFpEF glycolytic and ancillary pathway metabolites and their associated proteins.

HFpEF subgroup analysis and metabolic/clinical associations

As the HFpEF patients themselves had considerable variance in BMI and hemoglobin A1C, we further tested if these parameters were associated with the most prominently altered glycolytic intermediates. Both non-targeted and targeted MS analyses were combined to yield n=37, restricting analysis to the metabolites in both assays (glucose, G6P, F1,6bP, PEP, and lactate). Each data set was normalized to its respective controls, data combined, and then split into upper or lower median for BMI or HbA1C, or history of DM (Figure 5A). G6P, F1,6bP, and PEP were significantly lower in the group with the highest BMI, whereas there were no differences when patients were split based on HbA1C or DM history. Indeed, average HbA1C was identical in both the lower and higher BMI groups (mean 6.5 for both), the two variables having no correlation (P=0.4, r²=0.02). Figure 5B displays BMI versus F1,6bP or PEP for each patient plot as continuous variables; both had significant negative correlations. Myocardial lactate positively correlated with systolic blood pressure and negatively with estimated glomerular filtration (Table S5). Relations between other clinical/hemodynamic indices and metabolites did not reach statistical significance.

Figure 5.

A) Comparison of glucose, G6P, F1,6bP, and PEP in HFpEF patients separated into those in the upper or lower half for BMI (≤ 34 kg/m², 28.9±3.7 mean±SD or >34, 50.4±9 (upper row), HbA1C ≤ 6.1% (5.5±4.4) or HbA1C > 6.1 (7.7±1.3) middle row, or history or not for diabetes (DM) bottom row. P values are for Mann-Whitney test. B) Regression analysis of Log-transformed metabolites F1,6bP, PEP versus body mass index (BMI), and lactate versus systolic blood pressure (SBP). C) Principal component analysis of NMF groups within HFpEF patients. D) myocardial glucose (M-Glucose), F1,6bP, PEP, BMI, blood (plasma) glucose (B-Glucose), and log NT-Pro-BNP) measured in the two NMF defined HFpEF Groups 1 and 2. P-values displayed are from Mann Whitney test. Group 1 (n=14), Group 2 (n=23).

Metabolic-clinical associations were tested in reverse using non-negative matrix factorization (NMF) to split HFpEF into two subgroups based solely on metabolites and then comparing clinical features of the cohorts. Figure 5C displays principal component analysis of the two derived NMF groups, the full list of clinical features in each group is provided in Table S6. Group 2 had higher glucose, lower PEP, and a trend to lower F1,6bP, and also greater BMI, lower blood glucose and NT-proBNP, and was more likely female (Fig 5D). The two NMF groups had no significant differences in cardiac morphometry, systolic and diastolic function, or hemodynamics (Table S6).

Discussion:

This study presents the first global landscape for glycolytic intermediates and their associated synthetic and catabolic enzymes in human HFpEF myocardium. We found substantial reductions in three major glycolytic intermediates: G6P, F1,6bP, and 3PG, each accompanied by reduced expression of their respective synthesizing enzymes. The first two involve glycolytic rate-limiting and irreversible steps and occurred despite increased glucose. Metabolites in the PPP and HBP that utilize proximal glycolytic intermediates as substrates were also reduced, whereas glycogen increased suggesting potential storage for the glucose. Pyruvate, the final glycolytic product, was higher in HFpEF myocardium than controls, but MCP1 was reduced and mitochondrial metabolites suggested limitations in the TCA cycle. Importantly, these differences could not be attributed to disparities in BMI, DM history, sex, or age between the control and HFpEF groups. Marked obesity was associated with even more depression of F1,6bP and PEP.

The current study uncovered many glycolytic abnormalities that borrow features of HFrEF and/or obesity/DM. The most comparable data set for HFrEF is found in Flam et al¹³ where metabolomics and proteomic data were obtained from human myocardium. Their results are paired to the present ones in Table S7, and there are many similarities. Both have reduced PFKM and PGK1 and their corresponding metabolites F1,6bP and 3PG, although the decline is more in HFpEF¹³. Also, both HFrEF and HFpEF have reduced PDK4 expression and corresponding PDH phosphorylation that would favor enhanced pyruvate conversion to acetyl-CoA. This is opposite what is found with DM and obesity²².

However, HFpEF also differed from both DM and HFrEF in several aspects. The PPP and polyol pathways typically have elevated metabolite levels in the latter syndromes, that is considered linked to their pathophysiology^9,21. They were consistently reduced in HFpEF. The reduced PPP metabolites could be a result of higher flux through the pathway serving to bypass proximal glycolytic inhibition to restore F6P and 3PG. This can only be proven by labeled flux studies that are hard to do in humans. The hexosamine pathway and in particular O-GlcNAcylation is also reportedly elevated in HFrEF²⁴ and DM/obesity²¹, yet its major metabolite UDP-GlcNAc was lower in HFpEF with and without accounting for covariates. However, overall protein O-GlcNAcylation was unchanged from controls, and this could result from several factors. First, UDP-GLcNAc is used in both O- and N-glycosylation and both can be differentially regulated. Second, O-GlcNAcylation is regulated by the activity of its primary regulating O-GlcNAc-transferase and O-GlcNAcase enzymes that can rebalance net O-GlcNAcylated protein levels despite altered UDP-GLcNAc²⁵.

Despite reduced proximal glycolytic intermediates, pyruvate was higher in HFpEF myocardium, similar to what we reported previously⁶. There are several potential explanations. One is greater uptake of pyruvate from arterial blood as has been recently reported in HFpEF patients based on coronary sinus-arterial blood differences¹⁹. This study also found greater cardiac uptake of OA that was also higher in HFpEF myocardium. Higher OA can inhibit the electron transport chain and redirect metabolites to fuel hypertrophy²⁶. Another cause for greater pyruvate is less utilization by mitochondria. Depressed MPC1 expression could contribute²³ as well as impaired TCA cycle activity suggested by lower downstream TCA intermediates. Proof of the fate of pyruvate in HFpEF will require studies with hyperpolarized ¹³C-pyruvate magnetic resonance spectropscopy²⁷ or injecting ¹³C-substrates and then taking biopsies to track their fate.

The current findings come at a time when sodium-glucose transporter-2 inhibitors are becoming widely used in HFpEF patients given the strong evidence for improved clinical outcomes^3,4. While originally developed to treat DM, their benefits in HFpEF are virtually the same in patients with or without DM. Their mechanisms remain unclear and maybe off target as recent studies have found cardiac beneficial effects preserved in mice that genetically lacked SGLT2^28,29. Whether SGLT2 therapy ameliorates glycolytic or other metabolic pathway abnormalities⁶ remains to be determined.

Our study has several limitations, mostly notably that snapshot metabolomics were measured so flux rates cannot be ascertained. The biopsies were from the same cardiac location but HFpEF tissues were more rapidly frozen than controls. While the cardioplegic solution contained glucose, it was unlikely to artificially elevate glycolytic intermediates to make HFpEF data appear relatively lower. This follows from prior studies that found reduction or no change in glycolytic metabolites placed in cold cardioplegia for similar durations³⁰, and the rat study described in Supplemental Methods that used the identical cardiac isolation and processing protocol to controls, found metabolites mostly unaltered. Any decreases in metabolites in controls would only reduce the disparities we found. The metabolites were normalized to total protein and so combine cellular and extracellular material. But this composition would apply to all analytes derived from a given piece of tissue and so could not explain consistent different patterns among them. Lastly, the HFpEF population that presents to Johns Hopkins Hospital reflects a majority female, obese, African American cohort, and whether the findings will apply to populations with different sex and race/ethnicity characteristics with less obesity remains to be determined. Importantly, our study population is very impacted by HFpEF yet has historically been under-represented in pathophysiology studies and clinical trials.

In conclusion, human HFpEF myocardium has enhanced glucose, markedly reduced proximal glycolytic, PPP, and HSP metabolites, yet elevated pyruvate. Protein expression of the enzymes involved with critical rate-limiting glycolytic steps (HK, PFK, and PK) are depressed. Many features are quite similar to HFrEF rather than DM, and neither obesity nor DM were primary factors underlying the disparities. Greater obesity does associate with even more depressed key intermediates such as F1,6bP. These data in conjunction with prior results further highlight metabolic inflexibility in HFpEF and support its therapeutic targeting.