Severe Obesity in Human HFpEF Alters Contractile Protein Function and Organization

Abstract

Heart failure with preserved ejection fraction (HFpEF) causes substantial morbidity and mortality and has few effective therapies. Its phenotype has changed over time, with morbid obesity and metabolic defects supplanting hypertension and cardiac hypertrophy. We reveal that cardiomyocytes from patients with severe obesity and HFpEF have very depressed contractile reserve, including reduced calcium- and length-stimulated tension, power, and myosin activation compared to less-obese HFpEF and non-failing (NF) controls ±obesity, but similar to advanced HF with reduced EF. Myocyte defects correlate with body mass index and exercise hemodynamics in patients with HFpEF but not NF and appear reversible upon weight loss. Increased troponin-I phosphorylation at Thr181 occurs only in HF+obesity contributing to sarcomere dysfunction. Weight reduction and sarcomere enhancers may offer benefits in HFpEF with obesity.

INTRODUCTION

Heart failure (HF) affects over 64 million individuals worldwide and remains a major cause of morbidity and mortality. About half of this population have hearts with depressed resting contractile function and cardiac dilation, resulting in a reduced ejection fraction (HFrEF). The remaining people have normal appearing contraction at rest and non-dilated hearts with a preserved EF (EF≥50%; HFpEF) (1). This latter form of HF is increasing in prevalence and has similar morbidity and mortality to HFrEF. While there are many treatments for HFrEF, few exist for HFpEF, making it a major unmet medical need (1). When HFpEF was identified in the 1980s, most patients were elderly and exhibited chronic hypertension, left ventricular hypertrophy, and elevated diastolic pressures, often displaying myocardial hypercontraction with an increased EF (2, 3). Data at the myocyte level were scant but what was reported from endomyocardial biopsies was preserved calcium-activated tension and greater resting myocyte stiffness mostly attributed to titin alterations (3–6). This differed from myocytes from HFrEF, which displayed depressed calcium- and length-dependent tension, increased calcium sensitivity (7–9), and reduced maximal power (10), supporting the paradigm that diastole was the primary defect in HFpEF.

However, HFpEF has substantially changed over the past two decades, as hypertension and ventricular hypertrophy became more treatable and morbid obesity and metabolic syndrome more common as co-morbidities (11). Obesity is a leading predictor of overall HF evolution (12), particularly with HFpEF (13). Patients with obesity and HFpEF have lower exercise capacity with elevated right heart pressures and volumes during exertion (14, 15), and their myocardium exhibits excess lipid droplets, swollen/distorted mitochondria, and sarcomere structural disruption (16). Currently, only three therapies have improved chronic outcome in HFpEF: non-steroidal mineralocorticoid receptor antagonists, sodium glucose cotransporter 2 inhibitors, and glucagon-like peptide-1 (GLP-1) receptor agonists ±glucose-dependent insulinotropic polypeptide (GIP), the latter also targeting obesity (17–20). These are effective in patients with or without diabetes mellitus, and their success is shifting attention to the role of obesity in HFpEF, how it alters underlying pathophysiology, and what this means for effective therapeutics. A core question is whether and how the obesity HFpEF phenotype impacts sarcomere structure and function. Preliminary findings showed that activated tension is reduced in cardiomyocytes from patients with severe obesity and HFpEF (21), suggesting a shift from a solely diastolic phenotype. Similar behavior was not found in rodent models of HFpEF at myocyte or myofibrillar levels (22, 23), indicating that human tissue studies are needed to address the question. Accordingly, we comprehensively assessed sarcomere/myocyte and myofibrillar structure and function in human myocardium from patients with obesity and HFpEF and identified potential underlying mechanisms as avenues for future therapy.

RESULTS

Sarcomere dysfunction in HFpEF primarily segregates with body mass index (BMI)

HFpEF clinical presentations are heterogeneous. To identify those features most related to sarcomeric structural/functional properties, we assayed 27 different functional and structural properties of demembranated cardiomyocytes extracted from mid-septal endocardial biopsies from 80 deep-phenotyped patients with HFpEF (table S1, (24)). These features included calcium- and length-dependent tension, resting-myocyte stiffness, crossbridge activation/relaxation kinetics, sarcomere structure and activation, and contractile reserve (fig. S1). These data were entered into an unsupervised K-means cluster algorithm that identified two subgroups as the optimal data-fit: G1-HFpEF (n=30) and G2-HFpEF (n=50) with no overlap in a PCA plot (Fig. 1A, non-failing (NF) myocardium included in PCA in fig. S2A). PC-1 most prominently separated the subgroups, and its main components were maximum calcium activated tension, myosin crossbridge attachment rate, and thick filament (myosin) activation (fig. S1). Nearly all clinical features including invasive hemodynamics, echoDoppler measures of diastolic function, left ventricular hypertrophy, left atrial size, and many other parameters, were very similar between the two groups (Fig. 1B, table S1). The most prominent difference was a substantially greater BMI in patients with G2-HFpEF (Fig. 1C, 43±8 vs 31±6 kg/m², p=4×10⁻¹³). This group also had significantly lower NTproBNP (Fig. 1B, table S1) which negatively correlated with BMI, P=7.6e-6, r=−0.47), a slightly higher hemoglobin A1C, and a greater incidence of sleep apnea (also associated with obesity) (table S1). Patients with G2-HFpEF were more often treated with the loop-diuretic furosemide, used for HF fluid management (table S1). This analysis agnostically reveals severe obesity is a primary predictor of disparate sarcomere function among HFpEF patients.

Figure 1. Human sarcomere structural and functional data clustering analysis separates patients that differ primarily by body mass index.

A) Optimized K-means clustering analysis based solely on 27 sarcomere-features (Fig. S1, right) identified two HFpEF subgroups: G1-HFpEF (n=30) and G2-HFpEF (n=50). Ellipses are covariance zones between PC1 and PC2. B) Post-hoc comparisons of primary clinical properties of G1 vs G2-HFpEF. EF – Ejection Fraction; SBP – Systolic Blood Pressure; Sex-Adjusted Left Ventricular (LV) Mass /Height^1.7 (LV mass index(88)); E/e’ (E wave from mitral inflow on Doppler, e’ from tissue Doppler; E/A; Left Atrial (LA) Diameter; N-terminal Brain Natriuretic Peptide (NT-proBNP) log transformed; mean pulmonary artery pressure (mPAP); pulmonary capillary wedge pressure (PCWP); ratio of right atrial pressure (RAP) to PCWP (RAP/PCWP); pulmonary vascular resistance (PVR); cardiac index (CI). C) Body mass index (BMI) for both HFpEF groups. P-values in panels B and C based on non-parametric Mann-Whitney test.

Cardiomyocytes of G2-HFpEF with severe obesity have contractile depression with modest stiffening at rest

Figure 2 shows individual sarcomere functional features for the two HFpEF subgroups in contrast to non-failing controls (n=30, 60% with a BMI>35, table S2), and a group with advanced HFrEF (n=15, table S3). Activation of acto-myosin crossbridges for force generation and muscle contraction requires calcium and can be quantified by the sigmoidal isometric tension-calcium relationship (Hill equation). Myocyte maximal calcium activated tension (T_max) was nearly 50% lower in G2-HFpEF with severe obesity, similar to that for HFrEF, whereas less obese G1-HFpEF had T_max similar to NF (Fig. 2A, parameter summary fig. S2B). Tension-Ca²⁺ relations in NF controls ±severe obesity were superimposable (fig. S2C), so the G2-HFpEF behavior was not due to obesity alone but required the presence of HFpEF. These NF relations are identical when measured in myocytes from the septum or left ventricular free-wall (fig. S2C) (25) Tension at physiological systolic calcium concentrations was also lower in G2-HFpEF (Fig. 2A). At diastolic calcium (<1 μM), tension exceeded NF in G1-HFpEF but no other groups. Fig. 2B shows the same data normalized to maximal tension. Compared to NF, both HFpEF and HFrEF were left-shifted, indicating calcium sensitization – e.g. less calcium required to achieve 50% maximal activation. Both HFpEF groups also had a flatter slope in the mid-range of the tension-calcium curve, indicating reduced cooperativity for crossbridge formation and contractile activation; this was not observed in HFrEF (Fig. 2B, fig. S2B). T_max was also most reduced in G2-HFpEF myofibrils (Fig. 2C, fig. S2D), which correlated with myocyte T_max (fig. S2E), indicating the behavior was intrinsic to the fibrils and preserved following their isolation. Myocyte cross sectional area (fig. S2F) and sarcomeric protein isoforms (fig. S3) were similar among the groups and their stoichiometry preserved, making total fibrillar content or isoform switching unlikely confounders.

Figure 2. Myofilament dysfunction in severely obese HFpEF is similar to HF with reduced EF.

A) Tension-calcium relations fit to 3-element Hill equation; (2W-ANOVA, † P=0.02, * P<10⁻⁴, § P<0.003 NF vs G1-HFpEF; # P<10⁻¹³ NF vs G2-HFpEF, ‡ P<10⁻¹⁰, ¶ P=10⁻⁵ G2 vs G2 HFpEF). B) Same relations normalized to maximal tension plot on expanded scale. Arrow shows left shift at 50% activation. C) Myofibrillar calcium-stimulated tension-time plots and plateau from NF versus both HFpEF groups with rapid Ca²⁺ switch. D) Left: Tension-power relations, Right: summary data for maximal power (PWR_max) and area under the tension-power curve (AUC). E) Myosin attachment (2πb) and detachment (2πc) rates for NF and HFpEF groups from tension-time curve after step rise in strain. Myocytes perfused with 3.8 μM Ca²⁺ at starting SL= 2.1 μm. F) Early relaxation time (t_rel(Slow)) in myofibrils after rapid switch from activating to Ca²⁺=0. G) Resting tension-SL relations in Ca²⁺=0 for each group. H) Area-under tension-sarcomere length curves (AUC) shown in panel G. I) Relative contributions to rest tension of weakly bound (BDM sensitive) crossbridges, high salt-sensitive component (mostly titin), and residual factors. J) Component of rest tension-SL relations attributed to high-salt protein dissolution. K) AUC before and after BDM blocking crossbridge formation. L) Estimation of stochastic crossbridge contributions by high frequency stiffness (K_relative). M) Viscoelasticity assayed by creep (τ_σ) and relaxation (τ_ε) time constants.

Contractile function was also assayed by tension-power relations, maximal power (PWR_max) occurring when contractility optimally matches afterload (26). Skinned myocytes were allowed to shorten at a fixed tension level, shortening velocity measured (fig. S4A), and PWR determined from the tension-velocity product and plot versus tension (Fig. 2D, left). PWR_max, the area under the tension-PWR relation (related to total energy expenditure) (Fig. 2D, right), and tension at PWR_max (fig. S4B) were all reduced in G2-HFpEF with severe obesity, matching levels found in advanced HFrEF, whereas G1-HFpEF and NF were similar. These parameters were also similar regardless of obesity in NF controls (fig. S4C).

Active myocyte tension depends in part on the time myosin is strongly bound to actin. Abrupt shortening and re-lengthening of an activated myocyte breaks crossbridges that then reform. The resulting tension redevelopment rate (K_tr) measures overall crossbridge cycling (sum of attachment and detachment rates) and was slower in G2-HFpEF but preserved in G1-HFpEF versus NF (fig. S4D). We also measured myocyte responses to a step increase in length to estimate myosin attachment (2πb) and detachment (2πc) rates (fig. S4E) and gain insight into force development and duration and relaxation. Attachment was faster in G1-HFpEF but slower in G2-HFpEF relative to NF, whereas detachment was similarly slower in both HFpEF groups (Fig. 2E). The combination of faster attachment but slower detachment in G1-HFpEF could explain why K_tr is preserved in G1-HFpEF, whereas both rates and thus K_tr were reduced in G2-HFpEF. This could impair both contraction and relaxation in myocytes from HFpEF with severe obesity, whereas crossbridges in less-obese HFpEF form quickly but stay attached longer that could augment contraction at the cost of diastolic function.

Relaxation parameters were also quantified in individual myofibrils to study responses to rapid calcium reduction from activating to zero. This provides more precise measures of relaxation kinetics since calcium change is not diffusion limited as it is in myocytes. Following calcium removal, tension declines in two phases: an initial slow linear phase (combination of thin filament inactivation and myosin detachment) and subsequent fast exponential phase (residual crossbridge dissociation/elastic recoil) (fig. S2D). Duration of the initial phase (t_rel(slow)) was longer (its rate, k_rel(slow), slower) in both HFpEF groups versus NF (Fig. 2F, fig. S4F). The initial myofibrillar k_rel(slow) rate directly correlated with detachment rate 2πc in myocytes (fig. S4G). These data imply that impaired relaxation is intrinsic to the contractile elements and not due to slower calcium reuptake or diffusion constraints, implicating impaired crossbridge detachment and thin filament deactivation as direct contributors to diastolic dysfunction.

Prior HFpEF myocyte studies in non-obese hypertensive patients with ventricular hypertrophy reported resting tension-sarcomere length relations with greater stiffness versus controls (3, 6, 27). We found similar increases in G1-HFpEF, but these were blunted in G2-HFpEF myocytes (unchanged in HFrEF, all vs NF; Fig. 2G–H). There was no difference in these relations between non-obese versus obese NF control myocytes (fig. S4H). Resting stiffness is in part a function of stochastically (weakly) bound crossbridges at low Ca²⁺ (blocked by 2,3-Butanedione monoxime, BDM) and titin stiffness (removed along with some other proteins by high salt incubation) (28, 29). Both were greatest in G1-HFpEF myocytes (Fig. 2I–2K). The disparity in resting stiffness between groups was largely eliminated by incubation with the active catalytic subunit of protein kinase A (fig. S4I), which could reflect depressed PKA-phosphorylation of titin (30, 31). Resting stiffness related to myosin-heads weakly bound to actin at zero calcium was also assayed by low-amplitude, high frequency length-perturbation and found greater in G1-HFpEF (Fig. 2L). Cardiomyocyte viscoelasticity raises tension with faster stretch and contraction rates. Microtubules and titin primarily determine this property (32), with titin being most dominant in our preparation as microtubules depolymerize and exit the cell with myocyte skinning (33). Viscoelasticity was greater in both HFpEF groups (Fig. 2M) unlike in HFrEF. Thus, myocytes from severely obese HFpEF patients have markedly depressed myocyte sarcomere contractility similar to that found in end-stage HFrEF and slowed cross bridge kinetics compared to less-obese HFpEF and NF controls. Resting stiffness in this group is closer to NF, although rate-dependent stiffness is similar in both. These behaviors are not attributable to having severe obesity itself.

Myocytes from severely obese G2-HFpEF patients have disrupted sarcomere structure and reduced thick filament activation.

The pseudo-crystalline arrangement of myofilaments enables sarcomeric structural analysis by small angle X-ray diffraction. The ratio of the first two equatorial diffraction intensities (I_1,1/I_1,0) reflects thick and thin filament hexagonal spacing and can help elucidate crossbridge association with thin filaments (Fig. 3A). A lower ratio indicates that myosin heads are primarily arranged along the myosin backbone, away from actin (OFF state), whereas a higher ratio indicates a shift towards actin (ON state) (34). I_1,1/I_1,0 was 43% lower in G2-HFpEF myocardium (as in HFrEF) indicating that a larger proportion of myosin heads are in the structural OFF state vs NF controls, whereas G1-HFpEF and NF were similar (Fig. 3B). Inter-filament (lattice) spacing (d_1,0), was greater in G2-HFpEF and reduced in G1-HFpEF versus NF (Fig. 3C). Neither parameter was significantly different between non-obese and severely obese NF control tissue, therefore these data were combined. The mean radius from the center-of-mass of myosin heads (R_m) to the myosin backbone can be estimated from the distance from the X-ray diffraction meridian to the first maximum on the first myosin layer line (MLL1) (34) (Fig. 3D). This was lower in all HF groups (fig. S5A), but when normalized to lattice spacing, R_m/d_1,0 was significantly reduced only in G2-HFpEF (p≤8e-4 vs other groups, Fig. 3E). This is also compatible with a higher proportion of myosin heads being in the OFF state in this group, whereas in the less obese G1-HFpEF subgroup, reduced R_m is proportional to reduced d_1,0. To test whether greater myosin-backbone separation (lattice spacing) might contribute to reduced calcium-stimulated tension, spacing was compressed by exposing myocytes to hyperosmolar 3% high molecular weight dextran (35, 36). This increased T_max in G2-HFpEF but not G1-HFpEF (Fig. 3F), the latter already having a more compressed lattice. Dextran did not alter T_max in NF cells, but induced calcium sensitization as expected (fig. S5B). Myofibrillar and lattice disarray can also adversely impact contractile force and kinetics and be assayed by X-ray diffraction (37). The thick filament lattice was more ordered in G1-HFpEF, perhaps due to increased electrostatic constraint from a more compressed lattice but was similar to NF in the other groups (Fig. 3G). Myofibrillar disarray was greater in G2-HFpEF and HFrEF but not G1-HFpEF relative to NF (Fig. 3H) and could also contribute to reduced myocyte force production and its propagation among myofibrils in severely obese HFpEF.

Figure 3. Thick filament structure/function disengagement in severely obese HFpEF.

A) Schematic showing lattice structure responsible for the two primary equatorial diffraction intensities, I_1,1 and I_1,0, and the first myosin layer line (MLL1). Representative images from each group shown to the right. B) I_1,1/I_1,0 in each patient group. C) Thick filament lattice spacing, (d_1,0). D) Schematic for R_m (center-of-mass of myosin heads from thick filament backbone) and 2/$\sqrt{3}$ d_1,0. E) $\sqrt{3}$/2 R_m/d_1,0 for the four groups. F) Tension-calcium relationship before and after incubation with 3% high molecular weight dextran containing solution to compress the lattice spacing. This impacted G2- but not G1-HFpEF relations. G) Lattice disarray measured by standard deviation of Gaussian fit of I_1,0 in the equatorial dimension (width σ). H) Myofibrillar disarray measured by standard deviation of Gaussian of I_1,0 reflection in angular dimension (angle σ). I) Percent super-relaxed (SRX) myosin heads determined by fluorescent mant-ATP assay.

In the absence of calcium, myosin adopts two primary biochemical states of ATP hydrolysis: a disordered relaxed state (DRX) with fast ATP hydrolysis, and a super-relaxed state (SRX) with slower ATP hydrolysis (38). The probability of crossbridge formation upon activation is much higher for DRX vs SRX myosin, making the latter effectively unavailable for force generation. The differences in ATP hydrolysis rates are used to estimate the relative proportion of SRX vs. DRX myosin (39). We found %SRX in G2-HFpEF greater than in G1-HFpEF myocytes (72% vs 50%, Fig. 3I) but similar to HFrEF. %SRX was also greater in severely obese NF controls, indicating this feature is impacted mostly by obesity and not presence of heart failure. Overall, these data show dysregulation of thick filament structure and myosin inactivation in severely obese G2-HFpEF consistent with depressed activated tension and less resting stiffness versus less obese G1-HFpEF.

Myocytes from severely obese G2-HFpEF patients have reduced sarcomeric contractile reserve

Two major mechanisms for augmenting myocyte contraction are increasing sarcomere length (SL) and/or calcium concentration. Length-dependent activation (LDA) was determined as the slope (stiffness) of calcium-activated SL-tension relations, analogous to end-systolic elastance in intact hearts. LDA was reduced in G1-HFpEF myocytes and even more so in G2-HFpEF myocytes, compared to NF (Fig. 4A–B, NF ±obesity having similar relations, fig. S5C). Crossbridge attachment rate increased with a rise in SL in NF and both HFpEF subgroups, but less so in G2- versus G1-HFpEF (Fig. 4C). Myosin detachment rate slowed with increasing SL in NF myocytes, behavior thought to augment contraction with length but was unaltered in both HFpEF groups (Fig. 4D). Attachment rates were also differentially impacted by increasing calcium: G1-HFpEF had a marked increase in 2πb versus NF, whereas this rate was unaltered in G2-HFpEF cells (Fig. 4E). Detachment rate (2πc) increased with higher calcium in NF but was unaltered in both HFpEF groups (Fig. 4F). Thus, myocytes from severely obese G2-HFpEF patients have limited length and calcium-dependent crossbridge attachment rates, while these rates are conserved or enhanced respectively, in G1-HFpEF. Both HFpEF groups lack normal length and calcium-dependent prolongation of crossbridge detachment rate that can contribute to reduced reserve function.

Figure 4. Sarcomere resting and length/calcium activated indices associate with clinical cardiac function and reserve.

A) Active length-tension relations (3.8 μM Ca²⁺). B) Stiffness (regression slope) of individual length-tension relations. C) Effect of SL increase on crossbridge attachment (2πb) and D) detachment (2πc) rate. *P<0.05, **P=0.0005 versus zero within group. E) Change in 2πb and F) 2πc upon increasing [Ca²⁺] from 1.7 to 3.8 μM at SL 2.1 μm. *P<0.05, **P=0.0005 versus Δ=0 within group for panels B-F. G) Heatmap of correlations between myofilament functional/structural properties and clinical exercise parameters. Abbreviations as in Fig. S1 and Table S1. H) Relations between BMI and T_max for NF controls (NF) and HFpEF (G1 and G2 with respective color coding). I) Similar plots and P-values for I_1,1/I_1,0, J) %SRX, and K) change in %DRX from an increase in sarcomere length from 2.1–2.4 μm. Regressions±95%CI, P-values for relations and if present slope differences are shown. L) Linear regression ±95%CI of ΔBMI after 15.5±4 months GLP1-RA treatment versus myocyte T_max measured post-treatment. Color coding of patients is from PCA analysis post-treatment (Fig. 1). M) T_max vs BMI for GLP1-RA treated patients measured at the same time. Shape symbol denotes responder (>2 kg/m² decline in BMI or non-responder to GLP1-RA treatment, color is group from PCA analysis. Regression is mean ±95% CI based on all remaining HFpEF subjects (Fig. 4H).

Sarcomere structure/function indices correlate to exercise reserve and BMI

While resting hemodynamic differences were minimal between the two HFpEF subgroups, we hypothesized that sarcomeric defects may have a more prominent correspondence to stress responses. To test this, patients undergoing right heart catheterization and ventricular biopsy also underwent supine bicycle exercise testing. System- and organ-level function with exercise were then related to sarcomere behaviors in the same subject. Fig. 4G shows a heatmap for positive and negative correlations between various sarcomere and clinical exercise-hemodynamic parameters. The color indicates direction and strength of the correlations all with P<0.05. Elevation of right heart pressures at matched exercise level, a marker of abnormal reserve, negatively correlated with measures of active sarcomere function and positively with thick filament inactivation (%SRX). Sarcomere PWR_max positively correlated with exertional capacity (Watts). The rise in pulmonary pressures for a given change in cardiac output, a predictor of adverse HF outcome (40), negatively correlated with myosin attachment rate. Between HFpEF group comparisons for exercise hemodynamics at rest and matched stage 2 are provided in table S4. G2-HFpEF had a greater rise in pulmonary systolic pressure but also stroke volume, cardiac output, and stroke work index compared to G1-HFpEF. While the latter might seem contrary to worse sarcomere function in G2-HFpEF, total ventricular afterload (assayed by effective arterial elastance (41)) also declined significantly in G2 but not G1 (p=0.02). By enhancing stroke volume, reduced afterload could itself increase cardiac output and stroke work. Cox regression analysis relating sarcomere parameters to a combined outcome of re-hospitalization for heart failure or death (within a 3-year follow-up period, median 2 years, fig. S6A) found decreased cooperativity (nH) and lower tension at a physiological systolic calcium level (2.5 μM) to be positive predictors. While BMI strongly predicted G1 vs G2 sarcomeric differences, it was not itself a predictor of rehospitalization/death outcomes. BMI has been reported to worsen such outcomes in other HFpEF cohorts (42). Together these data suggest that the identified abnormalities in sarcomeric structure and function may translate to adverse exercise hemodynamics and worse clinical outcomes.

As BMI was the dominant factor separating G1 and G2 HFpEF groups, we further examined its relationship to sarcomere parameters in each patient using BMI as a continuous independent variable. Figure 4H displays this for T_max, showing a strong negative correlation in the HFpEF cohorts, but no significant correlation in NF controls (p=0.003 for slope difference). This disparity was mirrored in T_2.5 – the tension developed at 2.5 μM calcium (fig. S6B). A similar pattern was found for I_1,1/I_1,0 ratio reflecting crossbridge engagement, with a negative correlation for HFpEF and a weakly positive one for NF controls (p=0.001 for difference, Fig. 4I). By contrast, the %SRX rose with BMI along the same regression relation in both NF and HFpEF groups (Fig. 4J). Increasing sarcomere length (for example, from 2.1 to 2.4 μm) recruits DRX from SRX to contribute to contraction (9, 43), and we also found this recruitment declined with greater obesity in both NF and HFpEF groups also with similar relationships (Fig. 4K). Thus, obesity itself appears to have an impact on myosin ATP turnover rates increasing baseline %SRX and decreasing length-sensitive conversion to DRX. BMI negatively correlated with PWR_max in HFpEF and negatively correlated with the crossbridge attachment rate in G2-HFpEF and NF. Other parameters showed disparities between HFpEF and controls that were independent of BMI (fig. S6B).

Among the HFpEF patients, 16 with a starting BMI averaging 38.5±6 kg/m² received weight-loss therapy (primarily GLP1-RA) for a median 1.5 years prior to their evaluation at our hospital when endocardial biopsies were first obtained. Fig. 4L shows the post-treatment T_max versus absolute ΔBMI for each patient, revealing a strong negative correlation. There was no correlation between T_max and the duration of GLP1-RA treatment (p=0.3; r²=0.07). The colors reflect group assignment post-treatment, showing that those who lost the most weight were generally sorted into G1-HFpEF. This suggests lowering weight may reverse sarcomere contractile defects. We next labeled responders as those with >2 kg/m² BMI decline and plotted T_max vs BMI after weight loss therapy, and the mean T_max-BMI regression ±95% confidence intervals derived from the remaining patients. The initial BMI was slightly less in the responders (fig. S6C), but for a given BMI reached after GLP-1RA therapy, the corresponding T_max fell above the lower 95% CI of the population regression in all but one responder. In non-responders, only 2 fell above this level (χ² p=0.01, Fig. 4M). This suggests it is not the absolute BMI but perhaps how it is achieved that predicts calcium-activated sarcomere function; those reaching a given BMI after weight loss are more likely to have better function.

Role of altered sarcomere protein phosphorylation in severely obese G2-HFpEF patients

Sarcomere protein function is prominently regulated by post-translational modifications (PTMs) and among them, phosphorylation remains the best studied and annotated. In particular, reduced phosphorylation by protein kinase A of cardiac troponin I (cTnI) and myosin binding protein C (MyBPC) is found in human HFrEF (44), the former principally responsible for increased Ca²⁺-sensitivity (44, 45) and the latter for reduced adrenergic mediated contractility, slowed crossbridge kinetics, and increased SRX myosin (46, 47). We found TnI phosphorylation at S23/S24 and MyBPC at S275, S284, and S304, all known PKA-targeted residues, markedly reduced in both HFpEF groups (fig. S7A–B). Incubation with active PKA reduced myocyte Ca²⁺ sensitivity (right shifted the tension-calcium curve) in both HFpEF groups but did not alter it in NF (fig. S7C), consistent with basal phosphorylation disparities. PKA did not alter T_max in any group.

Since PKA modifications were similar in the HFpEF subgroups and unassociated with T_max change, we broadened the analysis using a phospho-protein gel (ProQ Diamond) that revealed myofilament protein phosphorylation was generally greater in G2>G1-HFpEF (Fig. 5A–B). Data from a prior HFpEF myocardial proteomics analysis (48) revealed reduced expression of phosphatases including the catalytic subunits of PP1a and PP2a (and their regulatory subunits), though this was similar in G1 and G2-HFpEF subgroups (Fig. 5C). We next incubated myocytes with each recombinant phosphatase. PP1a reverses PKA-modified residues (49, 50) and increased Ca²⁺ sensitivity in NF cardiomyocytes (fig. S7D), but had no impact on the tension-Ca²⁺ relation in myocytes from either HFpEF subgroup (Fig. 5D). Since PKA-cTnI phosphorylation is primarily responsible for this shift, these results are also consistent with pre-existing lower cTnI S23/S24 phosphorylation in both HFpEF groups (fig S7A, S7B).

Figure 5. Sarcomere hyperphosphorylation contributes to reduced tension in severely obese HFpEF.

A) Representative ProQ diamond phosphoprotein (top) and total protein (bottom) immunoblot from NF and HFpEF groups. MyBPC3: myosin binding protein C; TPM: tropomyosin; cTnT: Troponin T; cTnI: troponin I; cTnC: troponin C; RLC: regulatory light chain (all cardiac forms). B) Summary densitometry of proteins in Panel A; p/t – phospho/total. C) Heatmap for log₂-fold change of phosphatase and subunit proteins comparing HFpEF groups to NF. *P<0.05 vs. NF. Data from Jani et al. (48). D) Tension-calcium relations before and after treatment with active subunit of protein phosphatase-1a (PP1a). P-values: 2-way RMANOVA. E) Tension-calcium relations and F) summary data for T_max, before and after incubation with active subunit of protein phosphatase-2a (PP2a). Same replicates and analysis as panel D. F) Summary data for ΔT_max in each group. G) I_1,1/I_1,0 and d_1,0 for HFpEF groups before and after PP2a. H) Relation of mass spectrometry derived levels of p/t-cTnI Thr181 versus BMI for four patient groups. Regression and 95% CI shown; P and r values for NF versus HFpEF and HFrEF regressions shown, slope difference determined by covariance analysis. I) Tension-calcium relations in NF human myocyte after exchange with full recombinant human troponin protein complex containing cTnI that is either wild-type (WT), or has the point mutation T181D, or T181E. The other two components cTnT and cTnC are wild-type. All are dephosphorylated and remain after entry into the demembranated myocyte. *P<0.05 T181D vs WT and ‡P<0.05 T181E vs. WT, Tukey’s multiple comparison test.

By contrast, incubation with PP2a raised T_max by 37% in G2-HFpEF myocytes but did not alter it in G1-HFpEF or NF (Fig. 5E–F, fig. S7E). PP2a also increased T_max in advanced HFrEF (fig. S7F) which had lower myocardial protein expression versus NF as well (fig. S7G). PP2a expression was similar in NF non-obese and obese myocardium (fig. S7G). PP2a incubation also increased thick-filament activation (greater I_1,1/I_1,0) in G2-HFpEF myocardium without altering the lattice spacing (Fig. 5G). These findings support the idea that hyper-phosphorylated sarcomere proteins are likely involved with a lower calcium-tension response and thick filament in-activation, both ameliorated by PP2a in severely obese HFpEF.

To identify potential hyperphosphorylated residue candidates on HFpEF sarcomeric proteins, phospho-proteomics were performed (fig. S8A). While few changes were detected in part due to assay limitations and amount of sample, this confirmed lower MyBPC phosphorylation at S275 and S284, as detected by immunoblot and revealed changes in several other sarcomere proteins. The one residue with higher phosphorylation vs NF was cTnI at T181. In myocardium from HFpEF and HFrEF patients, pT181 positively correlated with BMI (P=3.4e-6, r=0.84), whereas in tissue from NF patients with or without obesity, pT181 remained low at higher BMI (slope difference to HF groups, P=2e-9, Fig. 5H). Phosphorylation at this site has been previously reported to increase modestly in human HFrEF and a canine model of HF (51), although its specific function was unknown. The residue resides in the highly conserved C-terminus but is replaced by a non-phosphorylatable isoleucine only in the Muridae (mouse, rat, etc.) mammalian subfamily (table S5). It is adjoined by 3 negatively charged residues and so may impact interactions with lysines on tropomyosin and actin and TnI. To test if this modification could contribute to the G2-HFpEF phenotype, demembranated NF human myocytes were incubated with fully reconstituted troponin complex including either a phosphomimetic recombinant cTnI T181D or T181E, or WT cTnI (fig. S8B). Since the proteins were generated by bacteria and then incorporated into demembranated cells that lacked kinases and other cytosolic proteins, there were no post-translational modifications present. Both phospho-mimetic mutants reduced T_max by 38% similar to G2-HFpEF (Fig. 5I), and reduced cooperativity and calcium sensitivity (left shift of tension-calcium relation, fig. S8C), mimicking behavior observed in G2-HFpEF myocytes.

DISCUSSION

This study identifies strong negative associations between severe obesity and sarcomere contractile and structural properties in human HFpEF that are nearly all absent in similarly obese NF controls. Specifically, myocytes from severely obese HFpEF patients displayed greatly diminished maximal myocyte power, calcium- and length-stimulated tension, crossbridge attachment rates, and thick-filament structural inactivation compared to obese G1-HFpEF and NF. Many of the changes are similar to those found with end-stage HFrEF. By contrast, diastolic resting/passive stiffness and myosin attachment rate are greatest in less obese HFpEF. Several prominent abnormalities are found in both HFpEF groups, including depressed crossbridge cooperativity and slower relaxation kinetics. Two properties are shared in very obese HFpEF and NF: a greater proportion of resting SRX myosin and impaired conversion of SRX to DRX with sarcomere lengthening. Table S6 summarizes our findings. Data also suggest that prominent sarcomere contractile defects may be reversed with weight loss therapy that has been shown to reduce heart failure hospitalization in HFpEF patients (18). Lastly, we show that PP2a enhances calcium-tension responsiveness in severely obese G2-HFpEF and that cardiac TnI-pT181, which increases with BMI only in HF (not NF) hearts and may play an important role to sarcomere dysfunction. Together, these results implicate an adverse interaction between severe obesity and the HFpEF syndrome, each required to generate most of the sarcomere functional and structural abnormalities observed.

Prior clinical studies have reported worse hemodynamics and ventricular function with exertion in patients with severe obesity versus those without (14). By contrast, we found no such differences in these parameters between G1 and G2 sub-groups, which were defined purely by sarcomere-structure/function yet revealed major disparities in BMI. However, in the prior study, non-obese controls had a mean BMI of 26 kg/m², whereas for G1-HFpEF patients the mean was already 31, and a BMI of 26 fell in the 20th percentile. Thus, the patient groups in our study had more obesity overall and differences in their global hemodynamics could in turn be less disparate. Our data show that such hemodynamics, and certainly EF, are poor predictors of underlying sarcomere abnormalities. There were only a few other factors differing between G1 and G2 HFpEF besides BMI: NT-proBNP, HbA1c, and prevalence of sleep apnea, all changing in directions concordant with severe obesity. An analysis of human HFpEF found obesity but not insulin resistance correlates with abnormal hemodynamics, heart remodeling, and dysfunction, indicating that obesity is the stronger driver of HFpEF disease severity (15).

It may seem paradoxical that such marked sarcomere deficits identified in severely obese HFpEF can co-exist with a normal range EF, particularly because many similar defects are found in late-stage HFrEF. However, EF prominently reflects chamber dilation, and therefore increased end-diastolic volume, whereas stroke volume in the numerator of EF is usually preserved until advanced stage HF. Other examples of heart disease with altered sarcomere function yet preserved EF are diabetic and diffuse ischemic disease (52–54), which may also contribute to HFpEF. Moreover, sarcomere defects were related to reserve more than rest function, notably in severely obese HFpEF, and this carried over to clinical exercise responses.

The tension-calcium dependence was similarly left-shifted (reduced EC₅₀) in both HFpEF groups and reversed by PKA incubation, consistent with resting hypo-phosphorylation of TnI at S23/S24. In G1-HFpEF, higher tension at low calcium could contribute to diastolic dysfunction. However, in G2-HFpEF, the accompanying marked decline in T_max and thus tension at intervening calcium levels offset the reduced EC₅₀. This is similar to changes in HFrEF, where calcium sensitization is observed yet net tension-calcium dependence is depressed (9). The presence of contractile depression in severely obese HFpEF was further supported by reduced PWR_max that occurred at a lower tension than in G1-HFpEF and NF controls, but similar to that in HFrEF (55). This suggests that elevated afterload in severely obese HFpEF will be less tolerated compared to HFpEF with lower BMI. Lastly, the lack of augmentation in myosin attachment and detachment rates with length or calcium stimulation in severely obese HFpEF reveals failed kinetic responses that would normally improve contraction and relaxation with preload, catecholamines, and exertional stresses. The slow tension decay in single myofibrils after rapid calcium reduction indicates that intrinsic defects in crossbridge dynamics and thin filament inactivation are also involved.

The higher SRX/DRX myosin ratio and lack of length-dependent SRX to DRX recruitment in patients with severe obesity and G2-HFpEF and NF controls with equal obesity was not anticipated and indicates that this level of obesity is itself altering myosin properties. The results also show it is not straight-forward to use SRX/DRX balance to predict net sarcomere function, as corresponding active tension-SL and tension-calcium dependencies were similar despite severe obesity in NF controls yet very reduced in G2-HFpEF. Another example is the small rise in resting stiffness in G2-HFpEF but not HFrEF or very obese NF controls, despite all having similarly elevated %SRX. One possibility is that while %SRX is a factor, other abnormalities such as structural and kinetic changes are required to produce the declines found in severely obese G2-HFpEF. The relationship between lattice spacing, sarcomere length, crossbridge formation, and %SRX is complex (35, 36, 56). Computational modeling suggests that the reduction in force in the length tension curve of muscle can be explained by increases in inter-filament lattice spacing with decreasing SL. Increased lattice spacing can also explain reduced activated force as found in HFpEF with severe obesity (57). The partial rescue of contractile force by osmotic compression of the lattice in G2-HFpEF supports this notion. A more compressed lattice as in G1-HFpEF could make stochastic crossbridges more probable, increase their attachment rate, and in turn diastolic tension, and all three were observed. How severe obesity impacts lattice structure (or %SRX) is unknown, but ultrastructural disorganization with interspersed lipid droplets (16), molecular signaling changes, and other factors may play a role.

Prominent features of hypertrophic cardiomyopathy (HCM) are diastolic dysfunction, preserved or enhanced EF, and impaired energetics(58), features often shared by HFpEF (59, 60). For HCM, insufficient SRX myosin is proposed as an underlying mechanism (58, 61) and myosin inactivators such as mavacamten (62) and aficamten (63) that stabilize SRX are now approved treatments. Whether they might also benefit HFpEF is being tested (64). However, we found only ~5% HFpEF hearts had below-normal %SRX and in the most obese HFpEF group, %SRX was above normal, and sarcomere contractility by multiple measures was very depressed, similar to HFrEF. Negative myokines that impede thick filament activation are not considered for HFrEF, which also has diastolic dysfunction and impaired energetics, and our data suggests they are unlikely to help most patients with obese HFpEF. It is true that the less obese HFpEF group had greater resting stiffness in part from stochastic crossbridges and that myosin inactivators that block this could perhaps improve diastolic function. However, this strategy would potentially further depress the already impaired systolic reserve. Even the G1-HFpEF group had depressed active length-dependent tension, so careful patient selection is warranted.

Increased myocyte resting stiffness has been previously reported in HFpEF studies of non-obese patients and ascribed to reduced phosphorylation and isoform changes in titin (3, 65). Such stiffening was found in the less obese G1-HFpEF but was muted in severely obese HFpEF. In the absence of calcium, G1-HFpEF myocytes had more myosin in the weakly bound state, which is non-force-generating but does contribute to increased diastolic stiffness. This may be related to the reduction in lattice spacing, which in turn would provide a potential mechanism for faster myosin attachment and slower myosin detachment rates in G1-HFpEF. The role of stochastic crossbridges was greater in the former group, which could relate to faster attachment but slower crossbridge detachment rates and to reduced lattice spacing. Many factors control diastolic and systolic function in myocytes in situ, including calcium cycling, adrenergic stimuli, and extracellular fibrosis, and all are either removed or rendered ineffective (including microtubules) in the demembranated myocyte preparation used in our study. Integrating myocyte-dependent abnormalities shown here with these factors is important but remains difficult to study in human HFpEF biopsies.

The finding that both reducing lattice spacing with dextran and dephosphorylating proteins with PP2a augmented T_max in G2-HFpEF myocytes, yet PP2a did not narrow lattice spacing, reveals that more than one mechanism is involved. The discovery of pT181 on TnI as a relevant modification that increases with BMI but only in myocardium from HF patients is intriguing. The local sequence of cTnI around T181 is KKEDTEKE, thus the threonine is already surrounded by negatively charged residues. The other surrounding residues are highly conserved, as is the entire C-terminus of cTnI in the H4 helix (66) that does not interact with other troponins but does with actin and tropomyosin thereby sharing responsibility for making actin binding sites available to myosin (67). Prior evidence supports a potential link between elevated pT181 and impaired sarcomere function. One study found increased pT181 in HFrEF patients, and even more prominently in a canine model of dyssynchronous HF induced by rapid right ventricular free wall pacing (51). In that model, both the increased pT181 on cTnI and associated reduced T_max were reversed by biventricular pacing (51, 68). Moreover, mutations in residues proximal to pT181 (such as K178E) are associated with restrictive cardiomyopathy, and fibers exchanged with these mutants show reduced tension and cooperativity as well as increased calcium sensitivity (69). Although the kinase responsible for pT181 remains unknown, these findings suggest that targeting this modification may improve contractile performance in severely obese HFpEF. Our results also suggest other modifications may be involved as there appeared to be greater phosphorylation in other sarcomere proteins in G2-HFpEF besides cTnI (Fig 5B,5C). One example is phosphorylation of MyBPC, as prior studies have found a mutant with pseudo-phosphorylation at S275D and S304D combined with phosphosilenced S284A results in a marked decline in T_max (47). However, we found all three sites hypo-phosphorylated in both HFpEF groups, so this is not the mechanism. Uncovering precise hyperphosphorylated residues on this and other proteins not identified in our non-targeted phospho-proteomics will likely require targeted approaches.

Our study has some limitations. We used flash frozen myocardial biopsies, as the unpredictable availability of tissues and need to transport them long distances for structural studies precluded use of fresh tissue. However, freezing is reported to have minimal effect on skinned myocyte structure and function (70). The HFpEF biopsies were frozen in liquid nitrogen within minutes after removal, whereas the control and HFrEF hearts underwent in vivo cold cardioplegic arrest in situ and were placed in iced buffer prior to tissue sampling and liquid nitrogen freezing. We have previously reported that myocyte active and resting mechanics are the same regardless of which procedure is used (22). The HFpEF biopsies were obtained from the RV side of the intraventricular septum. While this muscle acts as part of the LV, direct comparisons of LV and RV remain lacking. Animal models have had limited value in this regard as the most common ones do not recapitulate sarcomere behavior found in humans (22). As previously noted, a prior study of RV and LV myocyte tension-calcium dependence in NF-myocytes (25) found them to be superimposable (fig. S2C). Not all clinical or myocyte indices could be assayed in every patient. For most assays, the core 80 patients were used, but for some, additional patients were used. As these patients were extrinsic to the algorithm used to define G1 and G2 (c.f. Fig. 1), their group assignment was accomplished using a logistic regression predictor trained on the core 80 patients that had 99% accuracy in that group. Paired clinical and myocardial biopsy data before and after GLP1-RA treatment were not acquired, and such analysis awaits a formal prospective clinical study. Our study population was nearly all obese as this has become a dominant HFpEF presentation. The NF controls also included a similarly severely obese cohort, and we made these comparisons throughout the analysis. The G2-HFpEF group reflected severe obesity (Class III), which is already common in the United States and increasing globally. Lastly, while BMI has limitations as an obesity metric, and waist/height and waist/hip circumference ratios have been proposed as alternatives (71), in HFpEF patients with a BMI>30, all of these metrics are concordant (15).

In conclusion, the obesity pandemic has intersected with HFpEF and the result at the sarcomere level is a constellation of mostly activation/contractile defects associated with structural abnormalities that predict limited exertional reserve. This is not due to severe obesity itself, but the confluence of this morbidity with a syndrome that ultimately presents as heart failure with an EF≥50%. Identification of specific sarcomere protein phosphorylation targets is a potential avenue to counter underlying defects and should help design treatments for obese and severely obese HFpEF.

MATERIALS and METHODS

Study Population and Tissue Procurement.

Patients (n=108) referred to the Johns Hopkins HFpEF Clinic between June 2014 and February 2023 meeting clinical consensus criteria (59) for the diagnosis of HFpEF, who further had hemodynamic assessment by right heart catheterization with endomyocardial biopsy, were included. The primary diagnostic criteria were clinical signs and symptoms of HF, an ejection fraction (EF) ≥50%, and hemodynamic evidence of elevated left-sided filling pressures (pulmonary capillary wedge pressure ≥15 mmHg or ≥25 mmHg with exertion). Of these, 80 formed the primary subgroup cluster analysis cohort and source of most subsequent structural and functional analyses. Biopsies from the remaining 28 patients were used for some x-ray diffraction, myofibrillar, and immunoblot studies for which permeabilized cardiomyocyte mechanics could not also be obtained. For these patients, sub-group identification was assigned using a logistic model trained on 7 clinical parameters (body mass index, diabetes status, systolic blood pressure, LV mass index, HbA1c, and RV diastolic pressure), agnostically identified by iterative model building. This assignment model achieved 99% (79/80) accuracy on the core group of 80 HFpEF patients with cardiomyocyte functional data.

Endocardial biopsies were performed under a research protocol approved by the Johns Hopkins Institutional Review Board, and informed consent provided by all patients. HFpEF myocardial tissue was obtained by mid-septal endomyocardial biopsies (~2–3 mg) from the RV side of the septum as previously described (24, 72). They were rapidly placed in either OCT (Tissue-Tek) (73) and then liquid nitrogen or directly placed in liquid nitrogen. Non-failing control hearts (n=30) were obtained from brain-dead organ donors whose hearts had no identifiable disease but were not used for transplantation, most often due to advanced age, as described (9). Of these, 18 had a body mass index exceeding 35 (mean 39, severe obesity), and the remaining ones were all less than 31 (mean 25). Explanted HFrEF hearts from patients undergoing cardiac transplantation (n=15) were also studied. For these whole heart NF and HFrEF explants, hearts were retroperfused with cold cardioplegic solution in vivo, extracted, and immediately placed in cold calcium-free buffer, and myocardium samples (~ 1 g) from the same mid-septum region used for HFpEF biopsies were isolated and placed into liquid nitrogen.

Unsupervised Machine Learning Cluster Analysis.

Twenty-seven functional and structural analytical metrics of sarcomere function and structure were obtained (fig. S1) and entered into the analysis. As not every parameter was obtained in every patient due to practical and tissue availability limitations, K-nearest neighbor imputation was used to account for missing data. A total of 6 cluster models were compared: 1) K Means Clustering, 2) Gaussian Mixture Modeling, 3) Dirichlet Process Model, 4) Hierarchical Clustering, 5) Spectral Clustering, and 6) Non-Negative Matrix Factorization using Python sklearn. The model with the lowest Bayesian information criterion (BIC) was considered optimal and was a 2-group K means clustering model. For some sarcomere structural (X-ray diffraction), myofibrillar, and molecular analyses, biopsies were obtained from an additional 28 HFpEF or NF patients for whom we did not obtain all the parallel permeabilized cardiomyocyte mechanics. For these tissues, patient group identification (G1 or G2) was determined using a logistic model as described above. Kmeans cluster groups results were visualized using principal component analysis.

Invasive Resting and Exercise Hemodynamics.

Patients underwent right heart catheterization with a thermodilution fluid-filled catheter (Edward Lifesciences Corp, Irvine, California) advanced via the right internal jugular vein under ultrasound guidance. Right atrial, right ventricular, pulmonary artery, and pulmonary capillary wedge pressures were measured at end-expiration. Pressures were analyzed offline. Cardiac output was determined by the thermodilution technique from a mean of 3 consecutive measurements with <10% variance. Patients were then positioned into a supine bicycle ergometer, and after rest data measured, were exercised starting at 15 Watts (stage 1) and increasing by 10 Watt increments every 2 minutes until symptom limitation or a maximum of 12 minutes. All patients reached Stage 2 (25W), and thus data from this stage and maximal achieved were used in the analysis.

Myocyte Demembranization.

Demembranated (skinned) cardiomyocytes were prepared by taking frozen biopsies or 5–10 mg pieces of excised heart myocardium and immediately incubating them in 0°C isolation buffer (5.55 mM Na₂ATP, 7.11 M MgCl₂, 2 mM EGTA, 108.01 mM KCl, 8.91 KOH, 10 mM imidazole, 10 mM DTT) with 0.3% Triton X-100 in the presence of protease (Sigma-Aldrich, MO) and phosphatase inhibitors (PhosSTOP, Roche, Germany) (74). Tissue was homogenized at low speed (7500 rpm, OMNI Digital Programmable Homogenizer, Kennesaw, GA) for 6–8 seconds. Skinning involved incubation of homogenate in 0.3% Triton X-100 for 20 minutes at 4 °C. Cardiomyocytes were then affixed to a force transducer-length controller (Aurora Scientific, Canada) using ultraviolet-activated adhesive (Norland, NJ), and placed into relaxing buffer at ~22°C (5.95 mM Na₂ATP, 6.41 mM MgCl₂, 10 mM EGTA, 100 mM BES, 10 mM CrP, 50.25 mM Kprop, protease inhibitor (Sigma-Aldrich, St. Louis, MO), 1 mM DTT). Sarcomere length (SL) was measured by Fourier transformation of digital images (IPX-VGA210, Imperx, FL) and adjusted by micro-manipulators (Siskiyou, CA).

Isometric Tension-Calcium Relationship.

Tension-calcium relations were measured at a fixed 2.1 μm sarcomere length (SL). Calcium was varied by bathing isolated cardiomyocytes in 6 different preset solutions of 0.0 to 46.8 μM Ca²⁺ [46.8 μM solution composition: 5.95 mM Na₂ATP, 6.20 mM MgCl₂, 10 mM Calcium-EGTA, 100 mM BES (N,N-Bis(2-hydroxyethyl)-2-aminoethanesulfonic acid), 10 mM creatine phosphate - CrP, 29.98 mM Kprop, protease inhibitor (Sigma-Aldrich, St. Louis, MO), 1 mM DTT]. Force developed within each solution was measured at steady state. Tension was force dividing by myocyte cross sectional area (CSA=π/5·d², where d is the measured myocyte diameter). Tension versus -log[Ca²⁺] were fit to the three-element Hill equation: T = T_max [Ca²⁺]^nh/(EC₅₀^nh+ [Ca²⁺]^nh), yielding maximum isometric tension (T_max, mN/mm²), calcium sensitivity (EC₅₀, μM), and cooperativity (Hill coefficient, n_h). Subsets of myocytes were exposed to additional conditions: 3% high molecular weight (T-500) dextran solution for 10 minutes (Fisher Scientific), 0.125 U/mL catalytically active protein kinase A (PKA, Sigma Aldrich), and 0.30 U/mL of either active protein phosphatase 1a (PP1a, Sigma Aldrich, 0.30 U/mL) or 2a (PP2a, Sigma Aldrich, 0.30 U/mL). All experiments with kinases and phosphatases were conducted following a 30-minute incubation.

Isolated Myofibril Studies.

Tissue was permeabilized in 1% Triton X-100 overnight at 4 °C, after which samples were rinsed in isolation buffer and homogenized 3–6 times for 5 sec with an Omni PDH 700D at 15,000 RPM. The resulting myofibrils were kept on ice and used within 3 days. Myofibrils were mounted between a piezo length controller and a cantilever force sensor in a 17 °C temperature-controlled bath and stretched to a sarcomere length of 2.1 μm. Cantilever stiffness was determined by cross-calibration with AFM probes of known compliance (MikroMasch). Myofibrils were activated and relaxed by rapidly switching between two continuously flowing streams of calcium-free or activating (100 μM Ca²⁺) solutions, emanating from a single, double-barreled pipette attached to a fast-step motor (Warner Instruments). Light was directed by the microscope condenser onto the force probe cantilever, and the image of the cantilever tip was projected via microscope optics (40× objective, 1.6× intermediate magnification) between two light-sensitive areas of a split photodiode (Hamamatsu). The current output difference between the two photodiodes was proportional to the deflection of the cantilever, which in turn was proportional to myofibrillar force. Data were collected and analyzed using custom LabView software (National Instruments). Active force (F_MAX) was calculated as the difference between baseline force and the absolute force prior to relaxation. The rate constants for tension development (k_ACT) and for the final “fast” exponential phase of relaxation (k_REL) were determined by fitting the data to a single exponential function using the Levenberg–Marquardt nonlinear least squares algorithm. The duration of the slow relaxation phase (t_Slo) was measured from the onset of solution change to the beginning of the exponential force decay, detected with an automated linear-fit residual algorithm. The rate constant of the slow relaxation phase (k_Slo) was calculated as linear slope normalized to T_max.

Velocity/Power and Tension Relationships.

Tension-velocity and tension-power (PWR) relationships were obtained in myocytes starting at SL=2.1 μm in 3.8 μM [Ca²⁺] and then subjecting cells to 5–10% stepwise reduced tension starting at T_max, all at the same [Ca²⁺]. Velocity of cell shortening was measured over 75 ms immediately after the change in tension using linear regression to account for internal viscous load and resulting exponential dependence between length and time (75). Force transducer output was used for all measurements to correct for errors in servo-control. Data were fit to the hyperbolic Hill equation (T+a)(V+b) = b*(T_3.8+a) to determine V_max and PWR=Tb((T_3.8+a)/(T+a)-1) and peak power, PWR_max, using MATLAB (Mathworks, 2020). The area under the power-tension curve was numerically integrated by the trapezoidal method.

Cardiomyocyte Contractile Kinetics.

The rate constant of tension redevelopment after acute crossbridge disruption (k_tr), hyperbolic tension-velocity dependence, and tension-power relations(75–77) were measured. K_tr was acquired at 3.8 μM [Ca²⁺], SL = 2.1 μm. After steady-state force was achieved, SL was reduced by 20% for 20 ms followed by re-stretch to the initial length using servocontrol. Force redevelopment data were fit to a mono-exponential with the Levenberg–Marquardt nonlinear least squares algorithm. The rate constant of this exponential is k_tr.

Crossbridge kinetics, parameterized as $2\pi b$ and $2\pi c$, were acquired using a step response stretch activation protocol as described previously (78). After steady state force was achieved, a small length step (2% cell length, <0.15% sarcomere length) was applied, and force measured for 7 seconds at 2000 Hz. The resulting force tracing was normalized such that $\overline{F}=(\mathrm{F}-{\mathrm{F}}_{\text{max}})/{\mathrm{F}}_{\text{step}}$, where $\overline{F}$ is the normalized force, ${\mathrm{F}}_{\text{max}}$ is the steady state force, and ${\mathrm{F}}_{\text{step}}$ is the force immediately after the step response. The application of low strain perturbations to isolated cardiomyocytes is known to result in a three-phase tension waveform, consisting of the A process (complex modulus), B process (proportional to myosin attachment rate), and the C process (proportional to myosin detachment rate). The resulting force trace takes the form $\overline{F}\left(t\right)=A+B{e}^{-2\pi bt}-C{e}^{-2\pi ct}$, from which myosin attachment ($2\pi b$) and detachment ($2\pi c$) rates were acquired. This protocol was performed at two different sarcomere lengths, 2.1 and 2.4 μm, and two calcium concentrations, 1.7 and 3.8 μM.

Length-Tension Relationships and Titin Stiffness.

Both active and resting length-tension relationships were obtained. The active length tension relationship was acquired at 3.8 μM Ca²⁺. The myocyte was set to a fixed sarcomere length and exposed to calcium. Steady state tension was measured. Sarcomere length was then increased in 0.1 μm increments from 2.0 to 2.5 μm. Active tension was the difference of the tension developed, and the tension required to increase sarcomere length. The resting length-tension relationship was obtained by a similar protocol but in the absence of calcium and in 0.1 μm increments from 2.0 to 2.6 μm sarcomere length.

Passive tension was determined by measuring the length-tension relationship after exposure to 30 mM 2,3-butanedione monoxime (BDM), a myosin inhibitor, for 10 minutes. The difference of resting tension and passive tension was taken to the be the tension associated with stochastic crossbridges. Myocytes were also exposed to a high salt environment, 0.6 M KCl for 30 minutes followed by 1.0 M KCl for 30 minutes and length-tension relationships remeasured (28). This disrupts titin but also myosin and actin, and the relative contributions vary depending on orientation, with titin dominating longitudinal stiffness (29). While traditionally interpreted to reflect titin stiffness, some ambiguity remains though titin is still considered a dominant contributor to high-salt sensitive stiffness changes (29).

High Frequency Stiffness and Viscoelasticity.

High frequency stiffness was obtained by subjecting cells to 0.1% sinusoidal length change at 1 kHz for 5 seconds (79). This was performed before and after treatment with 30 mM 2,3-butanedione monoxime (BDM). The resulting force signals were demodulated by fast Fourier transform and ratio of their magnitude before and after BDM calculated. High Frequency stiffness was one minus this ratio.

Resting cardiomyocyte viscoelasticity was quantified using step force (creep) and step length (stress relaxation) protocols under relaxing conditions. In the absence of calcium, myocytes can be considered standard solids with a single elastic element, determined from the force-SL curve, in parallel with an elastic and viscous element in series. For the creep test, a force clamp was applied at a load equivalent to a 14% SL change (2.1 to 2.4 μm), while for the relaxation test, a 14% SL change (2.1 to 2.4 μm) was applied. The rate constant of either the length signal (creep) or force signal (relaxation) in a mono-exponential fit was the creep (τ_σ) or relaxation (τ_ε) time constant, respectively.

Small Angle X-Ray Diffraction.

Small angle X-ray diffraction patterns (43) were acquired at the BioCAT beamline 18ID at the Advanced Photon Source of Argonne National Laboratory in Illinois and at Center for High-Energy X-ray Sciences (CHEXS) in Ithaca, NY (X-ray beam energy 12 keV, 0.1022 nm wavelength, incident flux 10¹³ photons per second) (80). The X-ray beam was focused to 150×30 μm and 250×250 μm at the sample position. Flash-frozen tissue pieces (10–15 mg) were cut in ice-cold pCa 8 relaxing solution (2.25 mM Na2ATP, 3.56 mM MgCl₂, 7 mM EGTA, 15 mM sodium phosphocreatine, 91.2 mM potassium propionate, 20 mM imidazole, 0.165 mM CaCl₂, and protease inhibitor cocktail pH 7.0) and skinned for 1 hour in relaxing solution containing 1% Triton X-100 and 15 mM BDM at room temperature. Myocyte strips were dissected to a length of ~4×0.2 mm and affixed to aluminum T-clips at both ends. Preparations were suspended between two hooks in a customized chamber with two Kapton windows in the X-ray path. The preparation was lengthened to SL 2.1 μm by monitoring light diffraction patterns from a helium-neon laser (633 nm). Radiation damage was minimized by moving the X-ray exposure region for each pattern. X-ray patterns were collected at 2.1 μm SL at pCa 8. A MarCCD 165 detector (Rayonix Inc., Evanston IL) with 1 s exposure time was used for acquisition. Between two and four patterns were collected for each condition per patient. A subset of muscle bundles were treated with 0.30 U/mL PP2a for 1 hour, with patterns taken before and after treatment.

Data were analyzed using the open-source MuscleX v1.5 software package, developed at BioCAT(81). Equatorial reflections were measured using the “Equator” protocol in the software as described previously (43). Briefly, the intensity trace along the equation is summed across the image and integrated to a one-dimensional intensity reflection. The 1,0 and 1,1 reflections were modeled as Gaussian function superimposed on a smooth background. Lattice spacing was calculated by measuring the distance between the 1,0 reflection and the origin. For pattern visualization, “Quadrant Folding” in MuscleX v1.5 was used to visualize the equatorial and meridional reflections. The intensities of reflections were measured by the Project Traces protocol, which involves integrating the pattern to a one-dimensional projection and identifying the peak position as the centroid of the intensity distribution in a diffraction peak. This was fit to a Gaussian function, which gives the integrated intensity. The distance from the first maximum in the intensity distribution on the first myosin layer line to the center (r) is inversely related to the radius to the center of mass of the crossbridges (Rm), where intensity is modelled as a J3 Bessel function with the argument 4.2 = 2πrR_m.

Measures of myofibrillar and lattice disarray were obtained from x-ray diffraction as defined previously (37). In brief, the (1,0) intensity was radially integrated and fit to a Gaussian to determine width σ, a measure of lattice disarray. The standard deviation in the angular direction was used to calculate angle σ, which is a measure of myofibrillar disarray.

Myosin ATP Turnover Kinetics.

The percentage of SRX myosin (%SRX) was determined with a single nucleotide turnover assay (47, 82, 83) applied to skinned single cardiomyocytes. Myocytes were set to 2.1 μm SL in relaxing buffer, washed in rigor buffer (relaxing buffer without ATP or CrP) for 1 minute, incubated for 1 minute in rigor buffer with the fluorescent ATP analog 25 μM 2’-/3’-O-(N’-Methylanthraniloyl) adenosine-5’-O-triphosphate (a.k.a. mant-ATP, Enzo Life Sciences, Axxora LLC, Framingham, NY), and moved to room temperature relaxing buffer. As mant-ATP (excitation 352–402 nm, emission 417–444 nm) was hydrolyzed, fluorescence was acquired with a photomultiplier tube (Horiba / PTI 814 Photomultiplier Detection System) continuously at 100 Hz for 1000 seconds using a Nikon Eclipse Ti2 inverted microscope.

The acquired fluorescence decay from the pulse chase protocol is bi-exponential: the initial rapid phase is driven by DRX myosin ATPase activity of DRX myosin while the slow phase is driven by the myosin ATPase rate of SRX myosin). Calculating the relative contribution of each phase can in turn determine the proportion of DRX versus SRX myosin in a cell. The raw fluorescence decay signal was filtered with a second-order Savitzky-Golay filter, normalized, and fit to I = 1 – P₁(1-e^−t/T1) – P₂(1-e^−t/T2). P₁ and T₁ measure the fraction and rate of myosin with fast ATP turnover, respectively, while P₂ and T₂ measure the fraction and rate of myosin with slow ATP turnover. The proportion of SRX myosin is then determined by 2⋅P₂, while the percentage of DRX myosin is 1-(2⋅P₂). Background noise was limited using the IonOptix Cell Frame Adapter (CFA, Westwood, MA). Background was subtracted by measuring average photomultiplier tube voltage output in the surrounding relaxing buffer at the end of the assay for each cardiomyocyte. A subset of myocytes was subjected to stretch (2.1 to 2.4 μm). Paired fluorescence decay measurements were taken before and after each of these perturbations.

Western Blotting.

Lysates from endomyocardial biopsies and non-failing control tissue samples were generated by homogenization and methanol-chloroform precipitation of proteins. Lysates were clarified by high-speed centrifugation (16000 g, 1 minute) and assayed from protein concentration (BCA assay, Thermofisher). Extracts were loaded into Bio-Rad Criterion TGX Gels (4–20% Gradient), transferred to nitrocellulose membranes. Gels were imaged (Odyssey, Li-Cor) and band intensity quantified (ImageStudio) and normalized to total protein (BioRad, Total Protein Stain). Details for the antibodies used are provided in table S7.

Troponin Expression and Purification.

cDNA sequence of human cardiac troponin I (cTnI) in pET3D vector was used to generate phosphomimetics cTnI-T181E and cTnI-181D using the QuickChange II XL site-directed mutagenesis kit (Agilent Technologies). Clones were then sequenced using University of Arizona Genetics Core and verified using SnapGeneViewer. cTnI plasmids were then transformed into BL21 competent cells (Sigma) and streaked on Luria Broth +ampicillin agar plates with overnight incubation at 37°C. Single colonies from each plate were inoculated in 7 mL of Luria Broth at 37°C shaking at 250 rpm for 7–8 hours. Starter culture was then inoculated into 2L of ZYP medium (1% tryptone, 0.5% yeast, 0.5% (w/v) glycerol, 0.05% glucose, and 0.2% lactose) with 5% 20X P-buffer (1 M Na₂HPO₄, 1 M KH₂PO₄, and 0.5 M (NH₄)2SO₄), 1 mM MgSO₄, and ampicillin and grown overnight at 37°C shaking at 250 rpm. Large cultures were centrifuged at 4000 rpm for 20 minutes and bacterial pellets were collected and resuspended in 50 mL of Sp-sepharose buffer (6 M urea, 50 mM Tris, 2 mM EDTA, and 1 mM DTT, pH 7.0). The suspended pellets were subsequently frozen, thawed, sonicated, and centrifuged at 17,000 rpm for 45 min. The supernatant was then collected and loaded into a column with Sp-sepharose resin at 1.3 mL/min. The Sp-sepharose column was then washed with 300 mL of Sp-sepharose buffer and protein was eluted via an increasing concentration of KCl (0–0.6M KCl). Presence of cTnI was confirmed in elution tubes using Coomassie staining of sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gel. Tubes were pooled together and dialyzed against 2×2L of TnC affinity buffer (50 mM Tris, 2 mM CaCl₂, 0.5 M KCL, 1 mM DTT, pH 7.5). Protein was then recovered from dialysis and loaded into TnC affinity column, washed with TnC affinity buffer, and eluted in the presence of increasing urea and EDTA concentration (0–6 M and 0–3 mM, respectively).

His-tag TnC cDNA in pET-6 vector was obtained from Vector builder and subsequently transformed and grown up as described with cTnI constructs. Bacterial pellets were then collected and resuspended in Q-sepharose buffer (6 M urea, 20 mM Tris, 1 mM EDTA, 1 mM DTT, pH 7.8). The suspended pellet was then frozen, thawed, sonicated, and centrifuged at 17000 rpm for 45 minutes. The supernatant was then loaded into a column of Q-Sepharose resin. His-tag TnC was then eluted using an increasing concentration of KCl (0–0.6 M KCl). Protein presence was determined via Coomassie stained SDS-page gels. The fractions with cTnC protein were dialyzed against 4L of Phenyl-Sepharose A buffer (50 mM Tris, 1 mM CaCl₂, 1 mM MgCl₂, 50 mM NaCl, 1 mM DTT, pH 7.5) for four subsequent dialysis changes for at least 8 hours at 4C. During the dialysis process the room temperature phenyl Sepharose column was regenerated with 5x the volume of de-gassed 30% isopropanol (in ddH2O) followed by 500 mL of ddH2O. The phenyl-sepharose column was then pre-equilibrated with room temperature, degassed Phenyl Sepharose A buffer (50 mM Tris, 1 mM CaCl₂, 1 mM MgCl₂, 50 mM NaCl, 1 mM DTT, pH 7.5) with the addition of 0.5 M ammonium sulfate. The protein was recovered from the dialysate and allowed to warm to room temperature. Once the protein was at room temperature, solid ammonium sulfate was added to a concentration of 0.5 M. The protein was then loaded into the phenyl-sepharose column. The column was then washed with 500 mL of Phenyl Sepharose A buffer with 0.5 M ammonium sulfate. The cTnC was then eluted with 500 mL of Phenyl Sepharose C buffer (50 mM Tris, 1 mM EDTA, 1 mM DTT, pH 7.5).

cDNA of cTnT in pET-3D plasmid was transformed into Rosetta competent E. coli cells. The cells were then streaked and grown up in the same manner as described for the cTnI constructs. Bacterial pellets were then collected and resuspended in Sp-sepharose buffer. The suspended pellet was then frozen, thawed, sonicated, centrifuged, and purified through the Sp-sepharose column as described for cTnI protein. cTnT was then dialyzed into 2×2L of Q-sepharose buffer. Protein was then loaded into a column of Q-sepharose resin and then washed with 300 mL of Q-sepharose buffer. cTnT was then eluted using an increasing concentration of KCl (0–0.6 M KCl).

Protein presence and purity were all confirmed using Coomassie stained SDS page gels for all proteins.

Primers used for mutagenesis:

TnI-T181A-Forward GAAGGAGGACGCCGAGAAGGAAAAC

TnI-T181A-Reverse GTTTTCCTTCTCGGCGTCCTCCTTC

TnI-T181D-Forward GAAGGAGGACGACGAGAAGGAAAAC

TnI-T181D-Reverse GTTTTCCTTCTCGTCGTCCTCCTTC

TnI-T181E-Forward GAAGGAGGACGAAGAGAAGGAAAAC

TnI-T181E-Reverse GTTTTCCTTCTCTTCGTCCTCCTTC

Troponin Complex Reconstitution and Exchange.

All WT and variant (T181D, T181E) troponins were reconstituted in a 1:1:1: molar ratio of cTnI:cTnT:cTnC(His). Troponins then underwent stepwise dialysis for at least 8 hours with the following solutions: 30 mM MOPS, 1.25 mM MgCl₂, 1.25 mM CaCl₂, 1.5 mM DTT; pH 7 and (1) 6 M urea + 0.5 M KCl, (2) 4 M urea + 0.5 M KCl, (3) 2 M urea + 0.5 M KCl, and (4) 0 M urea + 0.5 M KCl. Following the final dialysis step, protein concentration was re-measured with a Bradford protein assay. For troponin exchange, 2 mM ATP was added to the final reconstitution solution during dialysis. Post dialysis, troponin was diluted with 30 mM MOPS to 1 μg/μL (final concentration: 0.3 M KCl, 30 mM MOPS). Permeabilized cardiomyocytes were pelleted, and pellets were washed once with 30 mM MOPS, 1.25 mM MgCl₂, 1.25 mM CaCl₂, 1.5 mM DTT; pH 7, 2 mM ATP. They were subsequently exchanged with 1 μg/μL troponin complex overnight (~12 hours) in troponin exchange buffer (0.3 M KCl, 30 mM MOPS). After the incubation, the cells were pelleted again and reconstituted in isolation buffer, and tension-calcium measurements obtained. At the end of the experiment, myocytes were pelleted and lysis buffer added. A western blot was run and probed for a His tag. Reconstitution efficiency was calculated using the ratio of the intensity of the His band to the total protein intensity at the location corresponding to cTnC. The average exchange efficiency was 71%.

Phospho-Proteomics.

Proteins from human hearts were denatured, alkylated, and digested with Trypsin. Tryptic peptides were labeled with isobaric mass tags (ThermoFisher) and following the manufacturer’s protocol. A small fraction was preserved for total protein identification, and phospho-peptides were enriched in the remainder of the sample using TiO2 then Fe-NTA (ThermoFisher). Peptides were loaded onto an Xbridge C18 HPLC column (Waters) and profiled with a linear gradient over 60 mins. The chromatographic performance was monitored by sampling the eluate with a diode array detector (1200 series HPLC, Agilent). Twelve concatenated fractions from total protein samples and six fractions from the phospho-enriched samples were run on an Ultimate 3000-nLC coupled to an Orbitrap Fusion Tribrid instrument (Thermo Fisher Scientific) equipped with a nanoelectrospray source, as described (84). Peptides were separated on an EASY-Spray C18 column (Thermo Fisher Scientific). Separation was achieved by 120 min linear gradient from 5% to 35% acetonitrile in 0.1% formic acid, at a flow rate of 300 nL/min. MS2 was triggered by selecting the most intense precursor ions above an intensity threshold of 1 × 10⁴ for collision induced dissociation (CID)-MS2 fragmentation. Mass filtering was performed by the quadrupole, followed by CID fragmentation in the linear ion trap with ~35% normalized collision energy in turbo scan mode. SPS ions were selected within the 300–2,000 m/z range (85). Raw data files were processed in Proteome Discoverer (Thermo Fisher Scientific) using the Sequest HT (Thermo Fisher Scientific) search algorithm against SwissProt homo sapiens database. Carbamidomethylation of cysteine, TMT 10-plex modification of lysine, and peptide N-terminus were set as static modifications. Deamidation of glutamine and asparagine; oxidation of methionine and proline; and phosphorylation of serine, threonine, and tyrosine were set as variable modifications. Up to two missed cleavages were permitted. The Percolator (University of Washington) algorithm was used to calculate the false discovery rate (FDR) of peptide spectrum matches (86, 87).

Targeted detection and quantification of phosphopeptides.

Cardiac tissue samples from twenty-two individuals were subjected to targeted detection and quantification of phosphorylation at Thr181 of cardiac troponin I (TNNI3), focusing on the EDT(phospho)EK motif, using the Complete360^® (Complete Omics, MD) workflow adapted for solid-tissue proteomics (bioRxiv, 2025; https://www.biorxiv.org/content/10.1101/2025.05.16.654403v1). This workflow builds upon the Complete360 framework originally developed for ultra-deep, clinical-grade quantitative profiling of complex biological matrices and was specifically modified to address the biochemical and analytical challenges associated with solid tissue samples.

Briefly, cardiac tissues were processed using a tissue-optimized lysis and digestion protocol designed to preserve post-translational modifications while minimizing proteolytic bias and sample-dependent variability. In contrast to plasma or other biofluids, solid tissues introduce increased matrix complexity, tissue-specific variable protein composition, and higher levels of endogenous background interference. To account for these factors, the Complete360 workflow incorporated tissue-specific matrix and noise modeling parameters, including empirically derived target-specific background noise thresholds, co-elution behavior across tissue matrices, matrix-dependent interference profiles, and tissue-informed target lists curated for human proteomes. These parameters were derived from large-scale solid-tissue proteomics datasets and integrated into the targeted detection strategy to enhance specificity and quantitative robustness.

For the Thr181 phosphopeptide target, additional assay-level customization was implemented. The EDT(phospho)EK-containing peptides are intrinsically short and relatively hydrophilic, properties that can compromise chromatographic retention, ionization efficiency, and quantitative reproducibility in complex tissue digests. To address these limitations, samples were chemically labeled with TMT10plex reagents (Thermo Fisher Scientific), a strategy used within the Complete360 framework to increase effective peptide hydrophobicity and improve chromatographic performance, ionization efficiency, and quantitative reproducibility for selected targets. This labeling strategy also facilitated consistent relative quantification across the cohort under a unified analytical framework. Complete360 targeted detection was then performed to simultaneously monitor both phosphorylated and non-phosphorylated TNNI3-derived peptides, including AYATEPHAK, KEDTEK, KEDT(phospho)EK, EDTEK, EDT(phospho)EK, KEDTEKENR, and KEDT(phospho)EKENR, across +2, +3, and +4 charge states. The inclusion of multiple peptide forms and charge states enabled cross-validation of detection fidelity and improved confidence in site-specific phosphorylation measurements.

Raw data were processed using the Complete360 CompletePeaking^® pipeline (Complete Omics, MD), which integrates targeted signal extraction with matrix-aware noise modeling and peak-profile evaluation. Chromatographic peak boundaries were manually validated and supporting fragment-ion series were inspected to confirm peptide identity and modification state. Phosphorylation occupancy at the Thr181 site was determined by calculating ratios of phosphorylated peptides to their corresponding non-phosphorylated counterparts, providing a quantitative measure of site-specific modification levels across cardiac tissue samples.

Statistical Analysis.

Results are expressed as mean±SD. Results are expressed as between-group comparisons of clinical data were compared with a Mann-Whitney test or Fisher exact test for continuous or categorical variables, respectively. Differences in survival between subgroups was performed with a logrank test. All tension-calcium, resting tension, and tension-velocity/power curves were compared using 2-way repeated-measures ANOVA with the Sidak multiple comparison test. For all cardiomyocyte analyses, data are presented are biological replicates (one point per patient), with technical replicates in each biopsy averaged before entry into analysis. The number of technical replicates varied among the various assessments depending on biopsy size and signal acquisition quality. We assayed 3–15 individual cells from each biopsy sample, the exact number varying depending on the isolation. The total number of these technical replicates for every group in every assay is provided in table S8. Fit parameters from these relationships, as well as equatorial intensity ratios, and ATP turnover proportions/rates were compared using a Brown-Forsythe or Welch’s ANOVA with Dunnett’s multiple comparison test. For myofiber studies where a large number of fibrils are required to power the analysis given individual variance, we performed the analysis using all technical replicates. The data were entered into a general linear mixed effects model to account for technical replicates from the same biopsy sample. For paired kinase/phosphatase studies, tension-calcium relationships were compared with 2-way repeated-measures ANOVA. Delta fit parameters with kinase incubation were compared using a 1-sample paired t-test vs 0. Analysis was performed using Stata 15.1 or Prism Version 9.0. All MATLAB used for analysis is publicly available from prior studies (9). Table S8 provides details for each figure and relevant panel regarding sample size and statistical tests used. The ROUT test was used to identify and eliminate outliers using a stringent Q-value of 0.1%.

Supplementary Material

List of Supplementary Materials:

Figs. S1 to S8

Tables S1–S9

Data, code, and materials availability:

While the human HFpEF biopsies cannot be made available given their scarcity, limited size, and ongoing studies, the authors welcome interaction with interested laboratories in forging research collaborations that could lead to sharing of such material. All of the individual data points presented in the dot-plot graphs (e.g Fig 1 panels) and those used to generate graphs of averaged relationships (e.g. Fig 2) are provided in worksheets in table S9. All new code for this manuscript is available on Zenodo (89) and original mRNA sequencing data used to generate Supplemental Figure 3D also at Zenodo (90).