Etiology-Dependent Impairment of Diastolic Cardiomyocyte Calcium Homeostasis in Heart Failure With Preserved Ejection Fraction.

BACKGROUND: Whereas heart failure with reduced ejection fraction (HFrEF) is associated with ventricular dilation and markedly reduced systolic function, heart failure with preserved ejection fraction (HFpEF) patients exhibit concentric hypertrophy and diastolic dysfunction. Impaired cardiomyocyte Ca2+ homeostasis in HFrEF has been linked to disruption of membrane invaginations called t-tubules, but it is unknown if such changes occur in HFpEF.
OBJECTIVES: This study examined whether distinct cardiomyocyte phenotypes underlie the heart failure entities of HFrEF and HFpEF.
METHODS: T-tubule structure was investigated in left ventricular biopsies obtained from HFrEF and HFpEF patients, whereas cardiomyocyte Ca2+ homeostasis was studied in rat models of these conditions.
RESULTS: HFpEF patients exhibited increased t-tubule density in comparison with control subjects. Super-resolution imaging revealed that higher t-tubule density resulted from both tubule dilation and proliferation. In contrast, t-tubule density was reduced in patients with HFrEF. Augmented collagen deposition within t-tubules was observed in HFrEF but not HFpEF hearts. A causative link between mechanical stress and t-tubule disruption was supported by markedly elevated ventricular wall stress in HFrEF patients. In HFrEF rats, t-tubule loss was linked to impaired systolic Ca2+ homeostasis, although diastolic Ca2+ removal was also reduced. In contrast, Ca2+ transient magnitude and release kinetics were largely maintained in HFpEF rats. However, diastolic Ca2+ impairments, including reduced sarco/endoplasmic reticulum Ca2+-ATPase activity, were specifically observed in diabetic HFpEF but not in ischemic or hypertensive models.
CONCLUSIONS: Although t-tubule disruption and impaired cardiomyocyte Ca2+ release are hallmarks of HFrEF, such changes are not prominent in HFpEF. Impaired diastolic Ca2+ homeostasis occurs in both conditions, but in HFpEF, this mechanism for diastolic dysfunction is etiology-dependent.

Cardiac pump function depends on both adequate left ventricular contractile force to eject blood (systolic function) and efficient relaxation and compliance to ensure proper filling (diastolic function). Heart failure in its classical form—heart failure with reduced ejection fraction (HFrEF)—is associated with marked systolic dysfunction and ventricular dilation. However, in ≈50% of heart failure patients, ejection fraction (EF) is normal, a condition referred to as heart failure with preserved ejection fraction (HFpEF). In these patients, concentric hypertrophy and high filling pressure are typical features (1). Mortality rates for HFpEF patients are comparable to those for HFrEF, and are estimated at ≈25% over 3 years (2). However, established therapies for HFrEF have consistently proven ineffective in HFpEF, suggesting that distinct pathophysiology underlies these 2 conditions (3). Indeed, several lines of evidence support this notion. In addition to differences in left ventricular remodeling during HFrEF and HFpEF, HFrEF progression involves activation of the renin-angiotensin-aldosterone system and stretch-mediated signaling pathways, whereas inflammation and endothelial dysfunction are believed to be key mediators of HFpEF (3). Important differences in collagen deposition and titin phosphorylation, which predominantly affect diastolic function, are also described (4). Although impaired Ca2+ signaling has been causally linked to cardiomyocyte dysfunction in HFrEF (5), it is unclear if such changes occur in HFpEF. Such insight is essential, because Ca2+ homeostasis critically regulates all stages of the cardiac cycle (6).

Considerable evidence from animal models suggests that invaginations of the sarcolemmal membrane, called t-tubules, are disrupted during HFrEF (7). This structural disorganization includes a reduction in t-tubule density and loss of a uniform striated pattern, promoting delayed and de-synchronized Ca2+ transients. Slowed and less powerful cardiomyocyte contraction results. Accumulating data have linked these detrimental changes in cardiomyocyte structure/function to heightened cardiac workload during HFrEF, and specifically indicated that high wall stress is a key trigger for t-tubule degradation in this state (5,8). It is as yet unclear whether such mechanisms occur in human HFrEF, although limited data suggest that t-tubular structure may be similarly altered as in animal models (9,10).

In HFpEF, t-tubule structure remains to be examined. We presently show that concentric remodeling and maintained wall stress in human HFpEF are associated with high t-tubule density, as these structures proliferate during cellular hypertrophy. Accordingly, in rat models of diastolic dysfunction and HFpEF, we observed that cardiomyocyte Ca2+ release is generally maintained. These observations are in stark contrast to the degradative t-tubule remodeling and deficient Ca2+ release observed in HFrEF. Although impaired diastolic Ca2+ homeostasis also occurs in HFrEF, we show that similar slowing of Ca2+ removal occurs in diabetic HFpEF, but not ischemic and hypertensive disease modalities. These findings indicate that there are critical etiology-dependent differences in mechanisms of diastolic dysfunction in HFpEF.

Methods

An expanded Methods section is included in the Supplemental Appendix. Human left ventricular tissue was obtained using protocols in agreement with the Declaration of Helsinki and approved by The Regional Committee for Medical Research Ethics (Permit numbers s-07482a, 2010/2226, and 2017/715). Tissue from patients was obtained with their informed written consent, while tissue from explanted healthy hearts not suitable for transplantation due to surgical reasons was obtained with explicit approval from the regional ethics committee.

Patient inclusion and sampling of tissue

Left ventricular tissue was acquired from HFpEF patients undergoing elective coronary artery bypass graft surgery (11). HFpEF was diagnosed based on heart failure symptoms, according to the New York Heart Association (NYHA) functional classification, with the presence of an EF above 50% and diastolic dysfunction (eʹ <0.09 m/s [12]) (Table 1, Supplemental Table 1). Epicardial wedges were obtained from nonischemic areas of the apex.

Table 1

Patient Characteristics and In Vivo Cardiac Parameters

Pathology	HFpEF Control (Bypass Patients)	HFpEF	HFrEF Control (Donor Hearts)	HFrEF
Patient characteristics and standard echocardiography parameters
No. of patients	6	20	4	11
Age, yrs	65 ± 7.9	68 ± 12.1	59 ± 6.6	38 ± 15.2∗
Male/female ratio	5/1	7/13	3/1	6/5
NYHA functional class	N/A	1.9 ± 1.1	N/A	3.5 ± 0.5
NT-proBNP, pg/ml	<125	319 ± 302	<125	1,141 ± 1,183
BMI, kg/m2	27 ± 2.8	31 ± 5.9	23.5	24 ± 4.1
DM	0	8	0	0
Samples analyzed	LV apex nonischemic	LV apex nonischemic	LV free wall	LV free wall
Blood pressure (cuff), mm Hg	141 ± 26 /81 ± 12	134 ± 21/73 ± 10	≈ 120/80	111 ± 35/51 ± 19
EF, %	58.3 ± 5.7	59.9 ± 14.5	≈ 53.0	18.5 ± 5.1
LVIDs, cm	3.2 ± 0.6	2.7 ± 0.6	≈ 3.1	7.0 ± 1.0
LVIDd, cm	4.8 ± 0.6	4.2 ± 0.7	≈ 4.2–5.9	7.6 ± 0.9
IVSd, cm	1.0 ± 0.2	1.2 ± 0.3∗	≈ 0.6–1.0	0.8 ± 0.2
LVPWd, cm	0.9 ± 0.2	1.1 ± 0.3∗	≈ 0.6–1.0	0.7 ± 0.1
Diastolic function
E, m/s	0.72 ± 0.20	0.72 ± 0.23	≈ 0.76 ± 0.17	0.62 ± 0.15
A, m/s	0.64 ± 0.12	0.85 ± 0.19∗	≈ 0.6 ± 0.17	0.34 ± 0.23
E/A	1.13 ± 0.15	0.81 ± 0.20∗	≈ 1.37 ± 0.51	2.77 ± 0.99
eʹ, m/s	0.09 ± 0.01	0.06 ± 0.02∗	≈ 0.09 ± 0.03	0.06 ± 0.01
E/eʹ	8.57 ± 3.21	13.09 ± 4.26∗	≈ 6.6 ± 2.0	13.00 ± 5.66
Deceleration time, ms	218.3 ± 29.3	278.7 ± 64.9∗	≈ 160–220	125.5 ± 43.3
Wall stress
Curvature, cm−1	0.66 ± 0.17	0.79 ± 0.19	≈ 0.65	0.29 ± 0.05
Surface tension, mm Hg/cm	218.5 ± 42.6	177.9 ± 45.5	≈ 186	383.4 ± 141.0
Systolic wall stress, mm Hg/cm2	170.6 ± 43.6	160.0 ± 56.7	≈ 162.4	532.0 ± 149.0

Values are n or mean ± SD, unless otherwise indicated. All parameters of donor hearts are literature values from Galderisi et al. (33), Caballero et al. (34), and Otto et al. (35). NT-proBNP levels were not measured in control patients, but are assumed to be <125 pg/ml (1).

A = late ventricular filling velocity; BMI = body mass index; DM = diabetes mellitus; E = early ventricular filling velocity; eʹ = peak mitral annular velocity during early filling; EF = ejection fraction; LVIDs = left ventricle inner dimension at end systole; LVIDd = left ventricle inner dimension at end diastole; IVSd = interventricular septum at end diastole; LVPWd = left ventricular posterior wall thickness at end diastole; N/A = not applicable.

Indicates significant difference from control.

HFrEF tissue was acquired from the left midventricular free wall of explanted hearts from dilated cardiomyopathy patients undergoing cardiac transplantation.

As we observed important regional differences in t-tubule organization across the heart (Supplemental Figure 1), biopsies from HFpEF and HFrEF hearts were compared with regionally matched control subjects. HFpEF patients were compared with epicardial biopsies from nonfailing patients undergoing coronary bypass surgery, whereas HFrEF biopsies were compared with samples from the left ventricular free wall of healthy hearts. Interobserver and intraobserver variability were assessed for all reported echocardiographic parameters (Supplemental Table 2 and Supplemental Figures 2 to 5).

Rat models of HFpEF and HFrEF

Three different models of diastolic dysfunction and HFpEF were used to address the heterogeneous nature of this disease. Rats exhibiting predominant diastolic dysfunction following induction of a small myocardial infarction (MI) (mean size = 9.1 ± 6.4% of left ventricular wall) were used as an ischemic model, and hypertensive HFpEF was modeled by Dahl-Salt Sensitive (Dahl SS) rats fed a high salt diet (13). Combination hypertensive and diabetic HFpEF was modeled with obese ZSF1 rats (14). Rats with marked systolic dysfunction following induction of a large MI (mean infarction size = 38.6 ± 6.0% of left ventricular wall) were used to model HFrEF (8).

Statistics

Statistical differences were tested using the Student’s t-test or 1-way analysis of variance, as applicable. The p values presented in this report have not been adjusted for multiplicity, and therefore inferences drawn from these statistics may not be reproducible. A p value ≤0.05 was considered statistically significant. Linear regression analysis was performed to calculate r2 values using SigmaPlot (SyStat Software Inc., San Jose, California). A standard outlier test (first or third quartile ± 1.5 times interquartile range) was performed for the Western data. All data are presented as mean ± SD.

Results

Echocardiographic and hemodynamic analysis

In HFpEF patients, echocardiographic analysis revealed marked diastolic dysfunction in comparison with control subjects, as defined by lowered E/A ratios and eʹ values, increased E/eʹ measurements, and prolonged deceleration times (Table 1). HFrEF patients exhibited both systolic dysfunction, indicated by markedly reduced EF, and diastolic dysfunction.

Geometric assessment revealed hypertrophic remodeling in HFpEF patients, as indicated by a thickened ventricular wall with greater curvature, and a trend toward reduced ventricular diameter (Table 1). Using these values and estimates of left intraventricular systolic pressure, we calculated that surface tension and wall stress were maintained in HFpEF hearts. In contrast, HFrEF hearts exhibited ventricular dilation and had elevated surface tension and wall stress (Table 1). NT-proBNP levels were elevated in both patient groups, but to a lesser degree in HFpEF. This finding is consistent with reports that elevated wall stress is the main trigger for neuropeptide release (1).

T-tubule density is increased in HFpEF, but reduced in HFrEF

Representative t-tubule images from control and HFpEF patients are shown in Figure 1A, with 3-dimensional reconstructions and 2-dimensional cross sections presented in the upper and middle panels, respectively. Enlargements of indicated regions are shown in the lower panels, together with distance maps illustrating the distance from each point in the cell to the nearest t-tubule or surface membrane. Overall t-tubule density was increased in HFpEF; an effect that was similar across NYHA functional classes (Figure 1B; see Supplemental Table 1 for partitioned patient data). Increased t-tubule density is suggestive of adaptive remodeling, and indeed included increased fractions of transversely-oriented elements (Figure 1B). Interestingly, when HFpEF patients were subdivided according to severity of diastolic dysfunction (Figure 1C, Supplemental Table 1), t-tubule density was positively correlated with E/eʹ ratio (p = 0.007) and deceleration time (p = 0.01) and negatively correlated with eʹ (p = 0.001) and E/A ratio (p = 0.03). Thus, HFpEF patients with the poorest diastolic function had the highest t-tubule levels. Separate analysis of t-tubule density in HFpEF patients with and without diabetes mellitus (Supplemental Table 1) revealed less marked t-tubule changes in diabetic patients, as density measurements were not significantly different from control subjects (Figure 1D). Whether divided by NYHA functional class, diabetic status, or diastolic function, HFpEF patient biopsies exhibited shorter average distance to the nearest membrane (Figure 1E), because t-tubule density and intracellular distance were negatively correlated (Figure 1F). These observations are consistent with adaptive t-tubule remodeling in HFpEF.

Figure 1

T-Tubule Density Is Increased in Patients With HFpEF

T-tubules were examined in biopsies taken from nonischemic regions of the myocardium in patients undergoing coronary bypass. (A, top) Confocal images of WGA-stained cryosections, with indicated regions enlarged below. (Bottom) Distance from each cytosolic point to the nearest t-tubule or surface membrane. Scale bar = 10 μm. Overall t-tubule density was increased in heart failure with preserved ejection fraction (HFpEF), both when patients were divided by New York Heart Association (NYHA) functional class (B) or severity of diastolic dysfunction (C). T-tubule density was not significantly altered in patients comorbid for diabetes mellitus (DM) (D). Distances to nearest membrane were shorter in HFpEF (E) and negatively correlated with t-tubule density (F). ∗p < 0.05 vs. control. ncells: control = 59 (4 hearts), NYHA functional class I = 70 (4 hearts), NYHA functional class II = 74 (5 hearts), NYHA functional class III = 27 (3 hearts), E/eʹ <14 = 109 (6 hearts), E/eʹ >14 = 61 (6 hearts), nondiabetes mellitus = 86 (6 hearts), diabetes mellitus = 85 (6 hearts). Data are mean ± SD.

Distinct changes in t-tubule structure were observed in HFrEF. In agreement with elevated in vivo wall stress (Table 1), ventricular tissue imaged from transplant patients revealed severe remodeling and lower t-tubule density, due to loss of transverse elements (Figures 2A and 2B). This finding was particularly prominent in the most severely ill patients, as indicated by NYHA functional class (Figure 2B, see Supplemental Table 3 for partitioned patient data). T-tubule loss during HFrEF resulted in longer intracellular distances to the nearest membrane (Figure 2A lower panels, Figures 2C and 2D). Thus, HFpEF and HFrEF patients exhibit opposite patterns of t-tubule remodeling, which further diverge with increasing disease severity.

Figure 2

T-Tubule Density Is Reduced in Patients With HFrEF

(A) T-tubule structure was examined in explanted hearts of heart failure with reduced ejection fraction (HFrEF) patients, and compared with healthy hearts deemed unsuitable for transplant. Colored images indicate distance from each pixel to the nearest t-tubule or surface membrane. Scale bars = 10 μm. (B) Overall t-tubule density was decreased in HFrEF. Intracellular distance to nearest membrane (C) was negatively correlated with t-tubule density (D). ∗p < 0.05 vs. control. ncells: control = 102 (4 hearts), New York Heart Association (NYHA) functional class III = 65 (3 hearts), NYHA functional class IV = 122 (3 hearts). Data are mean ± SD.

Increased t-tubule density during HFpEF is due to both dilation and proliferation

Increased t-tubule density in HFpEF could be due to either larger t-tubule dimensions or growth. To examine this issue, dSTORM super-resolution microscopy was applied to study t-tubule geometry in greater detail. Obtained images from HFpEF, HFrEF, and their respective control subjects are shown in the top panels of Figures 3A and 3B, with corresponding binarized masks presented below. T-tubule widths were increased in HFpEF (Figure 3C). To determine whether this tubular dilation in HFpEF patients could account for the higher t-tubule density observed in confocal recordings (Figure 1B), these images were skeletonized. This processing reduced t-tubule density to control values, indicating that denser signals in HFpEF cardiomyocytes stemmed from t-tubule widening (Supplemental Figure 6A). It should be considered, however, that t-tubule density is calculated relative to cell size, and that HFpEF cardiomyocytes were markedly hypertrophied (Supplemental Figure 6B). Thus, to maintain the density of the t-tubule skeleton in HFpEF, cardiomyocytes must proliferate their t-tubules. Indeed, t-tubule area measurements (i.e., not normalized to cellular area) were increased in HFpEF, even after skeletonization (Supplemental Figure 6C), indicating that t-tubule growth occurs in parallel to cellular growth. Thus, during HFpEF t-tubules both proliferate and widen.

Figure 3

T-Tubules Are Dilated in HFpEF Patients Without Additional Collagen Infiltration

dSTORM super-resolution images from HFpEF (A) and HFrEF (B) patient hearts presented alongside respective control subjects, as raw and binarized images (top and bottom panels). (C) T-tubule widths were more variable in HFpEF than control subjects (p = 0.03), and mean measurements were increased. (D) T-tubule widths in HFrEF hearts were also more variable (p = 0.01), due to a subset with broadened geometries. ntubules = 211 (7 hearts), 162 (6 hearts), and 166 (3 hearts) in control, HFpEF, and HFrEF, respectively. (E) Collagen deposition within t-tubules examined by colabelling with WGA and collagen I and III antibodies. An increased fraction of t-tubules exhibited collagen staining in HFrEF. Less collagen signal/t-tubule was observed in HFpEF. ncells: control = 35 (4 hearts), HFpEF = 50 (6 hearts), HFrEF = 27 (3 hearts). ∗p < 0.05 vs. control. Data are mean ± SD. Abbreviations as in Figures 1 and 2.

In HFrEF hearts, mean t-tubule width was unaltered from control values (Figure 3D), although these measurements were more variable than in control subjects (SD = 118 ± 5 nm vs. 102 ± 4 nm; p = 0.01), due to a subset of t-tubules with dilated geometries. Skeletonized t-tubule density remained lower than control values (Supplemental Figure 6D), and un-normalized measurements showed that t-tubules did not proliferate to match the cellular hypertrophy (Supplemental Figures 6E and 6F).

Recent evidence has indicated that collagen can infiltrate t-tubules (10), and indeed, immunofluorescence labeling identified collagen in approximately 70% of tubules in control hearts (Figure 3E). In HFpEF hearts, a similar fraction of t-tubules contained collagen (Figure 3E), and overall intracellular collagen density remained unchanged (collagen density: control = 5.7 ± 0.4% of cross-sectional area occupied, HFpEF = 6.3 ± 0.4%; p = 0.22). Because t-tubules were dilated in HFpEF, collagen to t-tubule area ratios were lower than in control subjects (Figure 3E). Partitioning HFpEF patients according to diabetic status revealed that similar fractions of t-tubules contained collagen in diabetic and nondiabetic individuals (0.70 ± 0.01 vs. 0.65 ± 0.03, respectively; p = 0.22), and that collagen to t-tubule area ratios were also comparable (0.64 ± 0.04 vs. 0.60 ± 0.03; p = 0.27, diabetic: ncells = 24 from 3 hearts, nondiabetic: ncells = 26 from 3 hearts). By contrast, a greater fraction of t-tubules in HFrEF hearts exhibited collagen deposition than in control subjects (Figure 3E), further supporting maladaptive t-tubule remodeling in these patients.

Remodeling of t-tubule structure and Ca2+ homeostasis in HFpEF depends on etiology

Experimental rat models were used to further examine t-tubule structure and its functional consequences. In total, 3 different models of diastolic dysfunction and HFpEF were used to address the diverse etiologies that underlie this condition in patients. As an ischemic model, we used rats with diastolic dysfunction (reduced E/A) but maintained systolic function (preserved EF) following induction of a small MI (Supplemental Table 4). In comparison with sham-operated control subjects, T-tubule density was increased in cardiomyocytes from post-MI hearts (Figure 4A, left panels), reminiscent of changes observed in ischemic HFpEF patients undergoing bypass surgery (Figure 1B). Dahl SS rats fed a high-salt diet and obese ZSF1 rats were respectively used to model hypertensive and combination hypertensive and diabetic HFpEF. In vivo characterization of these rats revealed marked diastolic dysfunction (reduced E/A, increased E/eʹ), concentric hypertrophy, increased left atrial diameter, and maintained EF (Supplemental Table 5), in comparison with control animals (Dahl SS with low salt diet and lean ZSF1 rats, respectively). In congruence with recent work (15,16), we observed unchanged t-tubule density in Dahl SS rats (Figure 4A). T-tubule levels were also maintained in the obese, diabetic ZSF1 rats, which paralleled measurements in diabetic HFpEF patients (Figures 1D and 4A).

Figure 4

T-Tubule Density and Ca2+ Release Are Predominantly Maintained in Preclinical Models of HFpEF

Isolated cardiomyocytes were examined from post–myocardial infarction (MI) Wistar rats with diastolic dysfunction, hypertensive Dahl/Salt Sensitive (Dahl SS) rats, and obese/diabetic ZSF1 rats. (A) Di-8-ANEPPS stains and mean measurements of t-tubule density (ncells: sham = 91 (3 hearts), post-MI = 136 (4 hearts), low-salt Dahl SS=30 (3 hearts), high-salt Dahl SS = 30 (3 hearts), lean ZSF1 = 77 (3 hearts), and obese ZSF1 = 66 (3 hearts). (B) Ca2+-transient recordings (fluo-4 AM), and mean measurements of synchrony of Ca2+ release and removal. ncells: sham = 38 (4 hearts), post-MI = 48 (4 hearts), low-salt Dahl SS = 23 (3 hearts), high-salt Dahl SS = 21 (3 hearts), lean ZSF1 = 49 (3 hearts), obese ZSF1 = 36 (3 hearts). ∗p < 0.05 vs. control or sham. Data are mean ± SD.

We next investigated Ca2+ handling in cardiomyocytes from the 3 rat models. In agreement with high t-tubule densities, myocytes generally exhibited maintained synchrony of Ca2+ release and removal (Figure 4B). One exception was the obese ZSF1 hearts, where a small but significant increase in Ca2+ release dyssynchrony was found. Post-MI rats with diastolic dysfunction showed indications of compensated Ca2+ homeostasis, with larger Ca2+ transients apparent at high stimulation frequencies (Figure 5A). Transients in these cells were also observed to decay more quickly, as Ca2+ removal rates by both sarco/endoplasmic reticulum Ca2+-ATPase (SERCA) and Na+-Ca2+ exchanger (NCX) were augmented. In contrast, we observed a modest reduction in Ca2+ transient magnitude in both hypertensive Dahl SS and obese ZSF1 cardiomyocytes, manifested at high stimulation frequencies (Figures 5B and 5C). In obese ZSF1 rats, we further observed an accompanying impairment of diastolic Ca2+ homeostasis (Figure 5C). This included slowed decline of Ca2+ transients, attributed to reduced activity of SERCA and NCX. SR content was also reduced in these cells (caffeine F/F0= 2.67 ± 0.17 vs. 3.22 ± 0.19 in control subjects; p = 0.04), but was unaltered in myocytes from hypertensive Dahl SS (F/F0 = 2.98 ± 0.21 vs. 2.92 ± 0.20 in control subjects; p = 0.85). Calibrated measurements of diastolic Ca2+ levels in obese ZSF1 cardiomyocytes also tended to be elevated at high stimulation frequencies versus control subjects (136.3 ± 3.4 nmol/l vs. 128.1 ± 3.4 nmol/l in control subjects at 4 Hz; p = 0.10).

Figure 5

Impairment of Diastolic Ca2+ Handling in HFpEF Is Etiology-Dependent

Representative 1- and 4-Hz Ca2+ transient recordings and measurements of transient magnitude and decay are presented for myocytes isolated from rats with post-MI diastolic dysfunction (A), hypertensive HFpEF (Dahl SS) (B), and diabetic HFpEF (ZSF1) (C). Caffeine-elicited transients were used to calculate rates of Ca2+ reuptake and removal from the cell, estimating SERCA and NCX activity (right panels). Only cells from diabetic ZSF1 HFpEF hearts exhibited slowed Ca2+ transient decline, as activity of both SERCA and NCX were reduced. ncells: sham = 38 (4 hearts), post-MI = 48 (4 hearts), low-salt Dahl SS = 23 (3 hearts), high-salt Dahl SS = 21 (3 hearts), lean ZSF1 = 49 (3 hearts), and obese ZSF1 = 36 (3 hearts). (D) Western blot data showing levels of SERCA, PLB, phosphorylated PLB (Ser16, Thr17), and NCX. nhearts: sham = 5, post-MI = 5, low-salt Dahl SS = 5, high-salt Dahl SS = 6, lean ZSF1 = 6, obese ZSF1 = 6. ∗p < 0.05 vs. control or sham. Data are mean ± SD. Abbreviations as in Figure 4.

Etiology-dependent differences in diastolic Ca2+ handling were linked to distinct expression patterns for Ca2+ handling proteins. In post-MI rats with diastolic dysfunction, SERCA expression tended to be reduced, but PLB expression was significantly decreased and phosphorylation at Thr17 was increased (Figure 5D). Thus, greater SERCA activity observed in these cells (Figure 5A) was linked to decreased inhibition by PLB. Maintained SERCA activity in hypertensive Dahl SS rats (Figure 5B) was linked to reduced SERCA expression and maintained PLB levels, but increased PLB phosphorylation at both Ser16 and Thr17 (Figure 5D). In contrast, reduced SERCA activity in obese ZSF1 cardiomyocytes (Figure 5C) was associated with lower SERCA expression without compensatory changes in PLB expression or phosphorylation (Figure 5D). Furthermore, reduced NCX activity in these cells was linked to lower protein expression (Figure 5D). These data indicate that altered expression and regulation of SERCA and NCX impair diastolic Ca2+ handling in diabetic HFpEF, but not in ischemic or hypertensive disease modalities.

Finally, we compared cardiomyocyte structure and function in rats which had developed clear signs of HFrEF following induction of a large MI (Supplemental Table 4). In agreement with our findings in human HFrEF, and typical for rodent HFrEF (17,18) we observed reduced t-tubule density in these animals, with fewer transverse and more longitudinal elements (Figure 6A). As expected, lower t-tubule density in HFrEF was accompanied by markedly dyssynchronous Ca2+ release (Figure 6B). However, diastolic Ca2+ homeostasis was also impaired in these cells, as transients declined dyssynchronously (Figure 6B) with markedly prolonged kinetics (Figure 6C). At higher pacing frequencies, there was also greater build-up of diastolic [Ca2+] (HFrEF = 183.2 ± 7.5 nmol/l, sham = 151.1 ± 7.0 nmol/l at 4 Hz; p < 0.01). Reduced Ca2+ removal was associated with lower SERCA expression and activity (Figures 6C and 6D), without compensatory expression or phosphorylation of PLB (see Supplemental Figures 7 to 11 for complete Western blots). Thus, diastolic dysfunction in HFrEF and diabetic HFpEF are similarly linked to loss of SERCA expression and activity.

Figure 6

T-Tubules and Ca2+ Homeostasis Are Disrupted in HFrEF

(A) Di-8-ANEPPS staining revealed lower t-tubule density in post-MI rats with HFrEF (ncells: sham = 57 [4 hearts] and HFrEF = 92 [6 hearts]). (B) Correspondingly, Ca2+ transient recordings (fluo-4AM) revealed desynchronized Ca2+ handling in HFrEF (ncells: sham = 19 [3 hearts] and HFrEF = 23 [3 hearts]). (C) Slowed Ca2+ decline in HFrEF cells was linked to decreased SERCA activity (ncells: sham = 19 [3 hearts] and HFrEF = 23 [3 hearts]). (D) Western blotting revealed reduced SERCA expression, with unchanged expression and phosphorylation of PLB. nhearts: sham = 6, HFrEF = 5. ∗p < 0.05 vs. sham. Data are mean ± SD.

Discussion

We presently investigated alterations in cardiomyocyte t-tubule structure and Ca2+ homoeostasis during HFpEF, in comparison with the more extensively studied condition of HFrEF. In contrast to marked t-tubule loss typically observed in HFrEF, HFpEF patients and rats exhibited high t-tubule density, linked to proliferation and/or broadening of these structures, without collagen infiltration. Although Ca2+ release was markedly desynchronized in HFrEF, the robust presence of t-tubules in HFpEF cardiomyocytes ensured that systolic Ca2+ homeostasis was generally maintained. However, we observed an etiology-dependent impairment of diastolic Ca2+ homeostasis in HFpEF. Indeed, diabetic but not hypertensive or post-ischemic disease modalities were linked to slowed Ca2+ removal via SERCA and NCX. In this regard, diastolic Ca2+ impairments in diabetic HFpEF were reminiscent of decreased SERCA functionality endemic in HFrEF. Collectively, our data suggest that HFpEF and HFrEF are respectively associated with predominantly compensatory and decompensatory remodeling of cardiomyocyte substructure, but that abnormal diastolic Ca2+ homeostasis can contribute to impaired myocardial relaxation in both conditions depending on disease etiology.

Insights into t-tubule remodeling during disease

The present analyses allow new interpretation of how t-tubules remodel during disease. Most previous publications, including those from our own group, have reported changes in t-tubule density, where t-tubule area measurements are normalized to cell size. We presently show that this type of presentation can be misleading, because changing t-tubule geometry affects density measurements. Furthermore, it is tempting to interpret lower t-tubule density observed in HFrEF as t-tubule “loss.” Our present results rather suggest that the t-tubule frame (skeletonized area) is maintained in HFrEF, and that the apparent loss of t-tubules is due to a lack of adaptive remodeling to meet the developing hypertrophy. Cardiomyocytes in HFpEF patients, however, grow new t-tubules as the cells enlarge. What are the signals that allow t-tubule proliferation during HFpEF, but not in HFrEF? We did not observe changes in mRNA levels of proposed t-tubule regulators such as junctophilin-2, BIN1, or MTM1 in either condition (Supplemental Figure 12). However, our present and recent (8) findings suggest that elevated wall stress in HFrEF may prevent adaptive subcellular remodeling. Because we have observed a strikingly similar structure in developing cells, we have proposed that a re-expression of fetal genes during HFrEF may be involved (18).

Importantly, we do not believe that the increased t-tubule density observed in nondiabetic HFpEF patients undergoing bypass surgery is a direct result of latent ischemia, because we observed similar t-tubule changes in rat hearts distal from the infarction. Furthermore, chronic ischemia in the absence of HFpEF has rather been linked to reduced t-tubule density (19). Somewhat different subcellular remodeling was observed in diabetic HFpEF patients and rats, as t-tubule densities were merely maintained at control levels. Less adaptive t-tubule remodeling during diabetes is perhaps not unexpected, because diabetic hearts exhibit abnormal expression and activity of the t-tubule regulators caveolin-3 and phosphoinositol-3 kinase (20). In addition, the sarcolemmal lipid composition of cholesterol, fatty acids, and phosphoinositides is altered during diabetes (20), which may attenuate adaptive t-tubule remodeling. Certainly, we expect that diabetes can be associated with fully detrimental t-tubule remodeling when coupled with ventricular dilation and progression toward HFrEF, as previously reported in db/db mice (21).

In addition to lowered t-tubule density in HFrEF, we anticipate that collagen deposition within these structures is also maladaptive. The presence of fibrillar collagen isoforms I and III is expected to stiffen t-tubular membranes, which may prevent normal t-tubule deformation, mechanosensing, and maintenance. Although it has been proposed that collagen deposition may promote t-tubule dilation (10), we observed that only a subset of t-tubules were dilated in HFrEF patients. In HFpEF, on the other hand, we observed significant t-tubule enlargement despite the fact that overall intracellular collagen was unchanged. Thus, collagen accumulation is not a prerequisite for dilatory t-tubule remodeling. Interestingly, although it is well-established that the diabetic heart is fibrotic (3,4), our data show that collagen deposition does not extend to the t-tubules in these patients. This finding suggests that fibrosis may be differently regulated within the t-tubule lumen than elsewhere in the extracellular matrix.

Functional implications

We observed striking differences in Ca2+ handling between HFpEF etiologies. Ca2+ transient measurements in the 3 rat models revealed compensated Ca2+ homeostasis in post-MI rats, relatively normal Ca2+ handling in hypertensive Dahl SS rats, and impaired Ca2+ removal in diabetic HFpEF (obese ZSF1) rats (Figure 5). Our data show that this remarkable range of phenotypes is intimately linked to differences in SERCA activity. Interestingly, although SERCA expression was reduced in all 3 HFpEF modalities, compensatory changes in PLB expression and/or phosphorylation only occurred in post-MI and hypertensive HFpEF hearts. No such compensation occurred in diabetic ZSF1 rats, as lowered SERCA expression translated into lower pump activity, slowing of Ca2+ reuptake, and a tendency for greater Ca2+ build-up at higher frequencies. These observations are consistent with an increased “active,” Ca2+-dependent stiffness that contributes to impaired diastole in diabetic HFpEF. This deficit is likely tied to the hyperglycemic milieu of the diabetic heart. Indeed, although SERCA reduction is a common feature in diabetes, hyperglycemia also triggers oxidative stress and increased O-GlcNAcylation, which further reduce pump activity (22). Interestingly, Dutta et al. (23) observed decreased PKA activity in the diabetic heart (23), which may critically abrogate the compensatory phosphorylation of PLB that we observed in non-diabetic HFpEF.

Measurements of NCX-mediated Ca2+ extrusion also revealed key differences between HFpEF etiologies. Although ischemic and hypertensive disease modalities showed augmented NCX activity, exchanger expression and activity were reduced in diabetic HFpEF. Because NCX plays an important role in setting resting [Ca2+], this observation is consistent with an increase in tonic, active stiffness previously reported in HFpEF patients (24). Decreased NCX function in diabetic HFpEF may also be linked to intracellular Na+ accumulation, due to increased expression/activity of the Na+-glucose cotransporter-1 (25) (Supplemental Figure 13). By comparison, maintained or even augmented Ca2+ handling observed in other HFpEF phenotypes suggests that diastolic dysfunction in these patients stems rather from passive stiffening of the ventricle following collagen deposition and/or titin alterations (3,4,26).

Our results contribute to a growing consensus that functional deficits in HFrEF critically involve detrimental t-tubule remodeling. Although it is established that these structural changes impair Ca2+ release, t-tubule organization also controls the synchrony of Ca2+ removal by NCX (6). Consistent with this view, we presently observed that t-tubule disruption in HFrEF was associated with more dyssynchronous Ca2+ removal and slowed Ca2+ transient decline (Figures 6B and 6C), which likely contributes to diastolic dysfunction in HFrEF. By comparison, we observed t-tubule proliferation in HFpEF, and our recent data indicate that newly grown t-tubules contain L-type Ca2+ channels, which trigger Ca2+ release (18). Such compensations may not fully compensate Ca2+ homeostasis, however, because we observed modest reductions in Ca2+ transient magnitude in hypertensive Dahl SS and obese ZSF1 cardiomyocytes, with the latter including a mild desynchronization of Ca2+ release. Importantly, recent investigations have noted that although EF is maintained in HFpEF, longitudinal strain is reduced (27). We postulate that modest impairment of cardiomyocyte Ca2+ release may be an underlying mechanism.

Study limitations

Mean NYHA functional class designation was lower in HFpEF than HFrEF for our patient cohorts (NYHA functional class 1.9 vs. 3.5; p < 0.01). However, this difference is in accordance with previous studies; 77% of HFpEF patients in the PARAGON-HF (Prospective Comparison of ARNI with ARB Global Outcomes in HF with Preserved Ejection Fraction) trial were classified as NYHA functional class II, and 73% of HFrEF patients were class III or above in PIONEER-HF (Comparison Of sacubitril/valsartan versus Enalapril on effect on nt-pRo-bnp in patients stabilized from an acute Heart Failure episode) (28,29). Importantly, when we partitioned patients according to NYHA functional class, we observed that within class III individuals, t-tubule densities were elevated in HFpEF but trended lower in HFrEF. This finding and data from our multiple animal models support that t-tubule remodeling is distinct in these 2 conditions. In future work, we aim to obtain tissue from less-severely ill HFrEF patients undergoing bypass surgery. We anticipate that such patients will exhibit less marked t-tubule remodeling than the NYHA functional class III to IV patients currently investigated, because t-tubule changes are widely believed to be an important driver of HFrEF progression (5).

Importantly, the HFpEF patients examined were not overtly ischemic during echocardiographic imaging, because recordings were performed while patients were at rest, and since patients undergoing emergency surgery were excluded. Although latent ischemia can promote diastolic dysfunction, we do not believe that this was a major contributor to functional deficits observed in our patient group. Indeed, previous examination of a similar coronary bypass population showed no effects of revascularization on diastolic function, with no change in E/eʹ during bypass surgery follow-up (30).

We have attempted to match our HFpEF and HFrEF datasets with appropriate control groups, to account for regional differences in t-tubule density (Supplemental Figure 1). Despite this approach, we cannot exclude that t-tubule remodeling in the left ventricular free wall, where samples were taken for HFrEF patients and control subjects, is different than in the apex, where HFpEF hearts and their control subjects were sampled. Nevertheless, we believe this to be unlikely, because previous work has revealed similar patterns of t-tubule loss across HFrEF hearts (31).

As we collected human tissue from different institutions, our analyses were limited to frozen samples. Even when fresh human tissue is available for cell isolation, we and others have experienced that cardiomyocyte experiments are hampered by low viability and Ca2+ overactivity (32). Thus, for insight into live cell function we turned to rat models of HFpEF and HFrEF, and striven to include an array of HFpEF etiologies to reflect the spectrum of phenotypes found in patients. Although these models may not fully reflect human pathophysiology, key characteristics of human HFpEF are recapitulated, including concentric hypertrophy, predominance of diastolic dysfunction, and in the case of the ZSF1 model, confirmation of a diabetic phenotype both in vivo and in isolated cardiomyocytes (Supplemental Figure 13).

Conclusions

The present study has shown that HFpEF and HFrEF are associated with strikingly different remodeling of t-tubule structure and cardiomyocyte Ca2+ homeostasis (Central Illustration). HFrEF is associated with reduced t-tubule density, including significant infiltration of collagen, and markedly impaired Ca2+ release. In contrast, t-tubule density and Ca2+ release are maintained or even increased in HFpEF depending on disease etiology. However, impaired diastolic Ca2+ homeostasis occurs in both conditions, but within HFpEF is limited to diabetic individuals, where reduced SERCA activity contributes to impaired myocardial relaxation. These findings support the notion that HFpEF includes a nonhomogenous group of patients with dissimilar pathophysiologies, requiring tailored treatment strategies.

Central Illustration

Heart Failure With Preserved and Reduced Ejection Fraction Exhibit Distinct Changes in Cardiomyocyte T-Tubule Structure and Etiology-Dependent Impairment of Diastolic Ca2+ Homeostasis

Ventricular dilation and elevated wall stress in heart failure with reduced ejection fraction (HFrEF) promote lower cardiomyocyte t-tubule density in human patients. In rats this is accompanied by desynchronization of Ca2+ release. In contrast, heart failure with preserved ejection fraction (HFpEF) hearts exhibit concentric remodeling and maintained wall stress, linked to high t-tubule density and robust Ca2+ release. Impaired diastolic calcium handling occurs in both conditions, but within HF with preserved EF is limited to diabetic individuals, as shown in rat models. Thus, there are critical etiology-dependent differences in mechanisms for diastolic dysfunction in HF with preserved EF.

COMPETENCY IN MEDICAL KNOWLEDGE: Although HFrEF is characterized by disruption of cardiomyocyte t-tubules and impaired release of Ca2+, these abnormalities are not prominent in patients with HFpEF. In HFpEF, disrupted diastolic Ca2+ homeostasis is etiology-dependent.

TRANSLATIONAL OUTLOOK: Given the limited efficacy in patients with HFpEF of drugs developed for HFrEF, future research should focus on developing specific therapies tailored to the diverse pathophysiology and cardiomyocyte phenotypes associated with HFpEF.

Author Disclosures

This study was supported by the European Union’s Horizon 2020 Research and Innovation Programme (Consolidator grant to Dr. Louch) under grant agreement No. 647714. Additional support was provided by The South-Eastern Norway Regional Health Authority, Anders Jahre’s Fund for the Promotion of Science, the Research Council of Norway, the Norwegian Institute of Public Health, Oslo University Hospital, the University of Oslo, the K.G. Jebsen Center for Cardiac Research, Norway, the European Union Projects No. FP7-HEALTH-2010.2.4.2-4 (‘‘MEDIA-Metabolic Road to Diastolic Heart Failure’’), and the Marsden Fund administered by the Royal Society of New Zealand (UOO1501). The authors have reported that they have no relationships relevant to the contents of this paper to disclose.