Transmural Heterogeneity and Remodeling of Ventricular Excitation-Contraction Coupling in Human Heart Failure

Abstract

Background

Excitation-contraction (EC) coupling is altered in the end-stage heart failure (HF). However, spatial heterogeneity of this remodeling has not been established at the tissue level in failing human heart. The objective is to study functional remodeling of EC coupling and calcium handling in failing and nonfailing human hearts.

Methods and Results

We simultaneously optically mapped action potentials (AP) and calcium transients (CaT) in coronary-perfused left ventricular wedge preparations from nonfailing (n = 6) and failing (n = 5) human hearts. Our major findings are: (1) CaT duration minus AP duration was longer at sub-endocardium in failing compared to nonfailing hearts during bradycardia (40 beats/min). (2) The transmural gradient of CaT duration was significantly smaller in failing hearts compared with nonfailing hearts at fast pacing rates (100 beats/min). (3) CaT in failing hearts had a flattened plateau at the midmyocardium; and exhibited a “two-component” slow rise at sub-endocardium in three failing hearts. (4) CaT relaxation was slower at sub-endocardium than that at sub-epicardium in both groups. Protein expression of sarcoplasmic reticulum Ca²⁺-ATPase 2a (SERCA2a) was lower at sub-endocardium than that at sub-epicardium in both nonfailing and failing hearts. SERCA2a protein expression at sub-endocardium was lower in hearts with ischemic cardiomyopathy compared with nonischemic cardiomyopathy.

Conclusions

For the first time, we present direct experimental evidence of transmural heterogeneity of EC coupling and calcium handling in human hearts. End-stage HF is associated with the heterogeneous remodeling of EC coupling and calcium handling.

Introduction

Congestive heart failure (HF) is one of the leading causes of death in Western countries.¹ Depressed contractility during congestive HF is associated with altered excitation-contraction (EC) coupling, in general, and calcium handling, in particular.^2–5

Most experimental studies of EC coupling in human hearts were conducted in isolated cells or muscle strips,^6–13 where anatomical differences could not be investigated. However, the anatomical location of the region from which cells are harvested could be very important.³ Animal studies suggest that transmural heterogeneities of EC coupling and intracellular calcium exist. For example, Cordeiro et al. observed that the latency to onset of contraction was shorter, and SR Ca²⁺ content is larger in epicardial cells as compared to endocardial cells in normal canine left ventricle.¹⁴ Investigating calcium handling, Laurita et al. showed that the recovery of intracellular calcium in canine left ventricle was slower in cells near the endocardium (ENDO) compared with cells near the epicardium (EPI).¹⁵ We have recently described spatial heterogeneity of action potential (AP) in human ventricle and its implication for the vulnerability to arrhythmias.¹⁶ However, spatial heterogeneity of EC coupling and intracellular calcium handling in human heart remains unclear.

It was suggested by a molecular study by Prestle et al. that the transmural heterogeneity of calcium handling was enhanced in the failing human hearts compared with nonfailing human hearts.¹⁷ In failing human hearts, the protein expression of the sarcoplasmic reticulum Ca²⁺-ATPase 2a (SERCA2a) was reduced significantly in the sub-ENDO compared to the sub-EPI,¹⁷ which might lead to the heterogeneous uptake of intracellular calcium and facilitate the induction of ventricular arrhythmias.^{15, 18, 19} In spite of the molecular evidence, it remains unknown if the heterogeneity of EC coupling and calcium handling is present and how it is functionally remodeled in heart failure. Furthermore, it is unknown if this remodeling could contribute to the increased ventricular arrhythmogenesis and mechanical dysfunction associated with human HF. We hypothesize that across the intact transmural wall there exists intrinsic heterogeneities of EC coupling and calcium handling and thus the susceptibility to remodeling during HF differs in different transmural layers of the left ventricle. To test this hypothesis, dual optical mappings of AP and calcium transient (CaT) were conducted in left ventricular (LV) wedge preparations from both failing and nonfailing human hearts.

Methods

Experimental Protocol

The study was approved by the Washington University Institutional Review Board. Both failing (n = 5) and nonfailing (n = 6) human hearts were optically mapped in this study. For Western blot assay, we used tissue from 18 hearts. Patient information is shown in Table 1.

Table 1.

Clinical characteristics of the studied hearts.

Patient Information
#	Group	Gender	Age	Diagnosis	Experiment
1	Nonfailing	Male	55	Death from stroke	Mapping*, WB†
2	Nonfailing	Female	59	Anoxic brain injury post cardiac arrest	Mapping, WB
3	Nonfailing	Male	53	Intracranial hematoma	Mapping, WB
4	Nonfailing	Male	56	Intracranial hematoma	Mapping, WB
5	Nonfailing	Female	47	Brain death due to anoxia	Mapping, WB
6	Failing	Female	65	Ischemic cardiomyopathy	Mapping, WB
7	Failing	Male	63	Ischemic cardiomyopathy	Mapping, WB
8	Failing	Male	49	Idiopathic cardiomyopathy	Mapping, WB
9	Failing	Female	54	Idiopathic cardiomyopathy	Mapping, WB
10	Failing	Female	54	Idiopathic cardiomyopathy	Mapping, WB
11	Nonfailing	Female	50	Brain death from anoxia	Mapping
12	Nonfailing	Female	66	Brain Death from hemorrhaging	WB
13	Failing	Male	61	Ischemic cardiomyopathy	WB
14	Failing	Male	64	Ischemic cardiomyopathy	WB
15	Failing	Female	49	Ischemic cardiomyopathy	WB
16	Failing	Male	50	Ischemic cardiomyopathy	WB
17	Failing	Male	47	Idiopathic dilated cardiomyopathy	WB
18	Failing	Female	44	Idiopathic cardiomyopathy	WB
19	Failing	Male	70	Idiopathic dilated cardiomyopathy	WB

^*Mapping indicates the conduction of optical mapping experiments

^†WB indicates the conduction of Western blotting assay

The isolated LV wedge preparation has been described in our previous paper.¹⁶ Briefly, a piece of LV wedge from the scar-free post-lateral LV free wall perfused by the left marginal artery (Figure 1A) was isolated and cannulated. Tissue was immobilized by blebbistatin (10~20 µM, Tocris Bioscience, Ellisville, MO) to suppress motion artifacts in optical recordings.²⁰ The tissue was co-stained with RH237 and Rhod-2 AM for simultaneous mapping of AP and CaT. Representative recordings of voltage and calcium collected at different layers of the LV wedge are shown in Figure 1C. The definition of quantified parameters are shown in Figure 1D. Signals from an array of pixels spanning the whole field of view are presented in the Supplemental Figure 1 to demonstrate the uniform quality of recordings.

Figure 1.

Left ventricular wedge preparation and optical recordings of action potentials (AP) and calcium transients (CaT). (A) An explanted nonfailing human heart. The region indicated by white rectangle was dissected and cannulated for wedge preparation. (B) The left ventricular wedge preparation from the same heart. (C) Pseudo-ECG (p-ECG) and representative optical recordings of AP and CaT from locations within sub-endocardium (sub-ENDO), midmyocardium (MID), and sub-epicardium (sub-EPI), which are indicated by the black stars shown in the panel B. (D) Terminology. Left: superimposed AP and CaT with illustrations of AP duration at 80% repolarization (APD80), CaT duration at 30% and 80% recovery (CaTD30 and CaTD80). Right: Close-up view of upstrokes (thin lines), and the derivatives (thick lines, labeled as dF/dt) with illustrations of AP-CaT delay and 10%–90% rise time of CaT.

The LV wedge preparations were paced at the ENDO at twice the diastolic pacing threshold. Dynamic restitution protocol was conducted in which pacing was started at a basic cycle length (BCL) of 1500 ms and it was gradually decreased until the ventricular functional refractory period was reached. Two Ag/AgCl electrodes were placed near the ENDO and EPI surfaces respectively to measure the pseudo-ECG. For more details of the Methods (tissue preparation, optical mapping system, and Western blot), please see the Online Data Supplement.

Data Analysis

All signals were low-pass filtered at 60Hz. The voltage-calcium delay was defined as the delay between the upstrokes of AP and CaT (Figure 1D). Each upstroke was defined at (dF/dt)_max, where F is the voltage or calcium fluorescent signal.²¹ AP duration (APD) was measured as the time from the upstroke to 80% repolarization (i.e., APD80, Figure 1D). Similarly, the CaT duration (CaTD) was measured as the time from the upstroke to 30% and 80% recovery (i.e., CaTD30 and CaTD80, Figure 1D). The 10–90% rise time of CaT was measured as the time from 10% CaT (close to the baseline) to 90% CaT (close to the peak, Figure 1D). Relaxation of CaT was quantified by the time constant (τ) of a single exponential fit of the CaT tail, i.e., the time from the minimum of d(CaT)/dt to the resting level of CaT. Sub-EPI was defined as the region within 2mm from the epicardial surface (See Figure 2B on the right); midmyocardium (MID) was the 2mm-wide midmyocardial layer; and sub-ENDO was the region within 2mm from the endocardial surface.

Figure 2.

Voltage-calcium (AP-CaT) delay. (A) Activation maps for AP and CaT from a failing human heart (#10). (B) Map for AP-CaT delay and the anatomic definition of sub-ENDO, MID and sub-EPI from the same heart. (C) Summarized results for AP-CaT delay at the sub-ENDO, MID, and sub-EPI at a basic cycle length (BCL) of 1500ms in failing (F, n=5) and nonfailing (NF, n=6) hearts.

Statistical Analysis

For statistical analysis, we used ANOVA. Specifically, we fit a linear mixed effects repeated measures model, where the patient was a random effect and other factors (failing/nonfailing, tissue layers, and basic cycle lengths [BCL]) were fixed effects. We compared by the small-sample-size corrected version of Akaike information criterion (AICc) the fit of models that allow heterogeneous variance among levels of the failing/nonfailing by tissue layer interaction and repeated measures correlation among tissue layers. Contrasts were used to test the significance of differences between the failing and nonfailing groups within different tissue layers (sub-ENDO/MID/sub-EPI). Bonferroni adjustment was used to account for multiple comparisons. Detailed specifications of statistical analysis for individual figure are provided in the Online Data Supplement. P value less than 0.05 was considered statistically significant. Values were given as means ± S.D.

Results

Voltage-Calcium Delay (AP-CaT Delay)

To quantify the EC coupling, the delay between the AP upstroke and CaT rise was measured. As expected, the upstroke of the AP was always followed by the rise of CaT (Figure 1D). To quantify the transmural heterogeneity, this delay was measured and averaged at all three tissue layers (sub-ENDO, MID, and sub-EPI). Figure 2A&B is a representative example displaying AP and CaT activation maps as well as the voltage-calcium delay. This delay is summarized in Figure 2C for the BCL of 1500 ms. We observed a transmural gradient of this delay within the failing group. That is, the delay was significantly larger at sub-ENDO than sub-EPI (P = 0.015, see Figure 2B as an example).

APD and CaTD

We quantified APD, CaTD, and the difference between the two (i.e., CaTD - APD). Figure 3 shows one example from a nonfailing heart (Figure 3A) and one from a failing (Figure 3B) heart. Maps of APD80 and CaTD80 are shown for both hearts. It should be noted that the color scales for APD and CaTD maps are different. It can be seen that APD80 mildly increased in this failing heart compared with that in the nonfailing heart, while CaTD80 was increased in a more substantial manner. Because of the disproportionate prolongation of CaTD relative to APD, (CaTD80 - APD80) in the failing heart was larger than that in the nonfailing heart.

Figure 3.

Representative maps of APD80 and CaTD80 at the basic cycle length (BCL) of 1500ms. (A) APD and CaTD maps from a nonfailing human heart. (B) APD and CaTD maps from a failing human heart.

APD80, CaTD80, and (CaTD80-APD80) for both failing and nonfailing groups are shown in Figure 4A for the BCL of 1500 ms and are summarized in Figure 4B for multiple BCLs. The transmural APD and CaTD gradient represented by the sub-ENDO and sub-EPI duration differences are shown in Figure 4C. Transmural APD gradients are present in both failing and nonfailing human hearts. Similar to our previous study,¹⁶ at slow heart rates, this gradient was less pronounced in the failing group compared with the nonfailing group (Figure 4A&C-left). CaTD80 (Figure 4A-middle) also exhibited gradients from the endocardium to the epicardium. Interestingly, this gradient (Figure 4C) was significantly smaller in the failing group compared with the nonfailing group at fast heart rates (e.g., 100 beats/min [bpm] or BCL=600 ms) but not at slow heart rates (e.g., 40 bpm or BCL=1500 ms).

Figure 4.

APD, CaTD and the duration difference (APD-CaTD). (A) APD80, CaTD80 and the duration difference at the basic cycle length (BCL) of 1500ms in failing (F, n=5) and nonfailing (NF, n=6) hearts at sub-ENDO, MID, and sub-EPI. (B) Dynamics of APD80, CaTD80 and the duration difference at various cycle lengths at sub-ENDO, MID and sub-EPI. The top row in panel B is for nonfailing hearts, and the bottom row in panel B is for failing hearts. (C) Transmural APD and CaTD gradients. The gradient is calculated as the difference between the values of sub-ENDO and sub-EPI.

The duration difference (CaTD80 - APD80) was significantly increased at sub-ENDO in the failing hearts during bradycardia (P = 0.022, Figure 4A-right). As the BCL decreased, the duration difference was significantly decreased within the failing group, while it remained unchanged in the nonfailing group (Figure 4B-right). Both APD80 and CaTD80 were decreased as the BCL was decreased (Figure 4B-left and middle).

M cell islands, which contain prolonged APDs and are surrounded by large APD gradients, were observed previously in nonfailing human hearts.¹⁶ In the majority of the hearts in this study, APD decreased gradually from ENDO to EPI without the presence of M cells. However, M cells were observed in one nonfailing heart, where we specifically searched for them (Figure 5). As shown in the map of APD80 (Figure 5B), the M-cell area was in the form of an isolated island rather than a continuous layer. This region exhibited a delayed repolarization (Figure 5E) and was surrounded by steep local APD gradients (Figure 5C). As shown in Figure 5D and 5E, the M-cell island had longer CaTD compared with neighboring regions. Other M-cell island parameters (AP-CaT delay, CaTD-APD, CaT rise time, CaTD30/CaTD80, and τ) were not different from neighboring mid-myocardium regions (Supplemental Figure 2).

Figure 5.

M cell island. (A) Activation propagation from ENDO to EPI. (B) APD map at the basic cycle length (BCL) of 1500 ms. An island of prolonged APD is evident in this map. (C) Map of APD gradient at the BCL of 1500 ms. This shows that the region of prolonged APD is surrounded by steeper APD gradients. (D) Map of CaTD at the BCL of 1500 ms. (E) From top to bottom: psudo-EKG (p-ECG), action potential (AP), and calcium transient (CaT) from locations (marked by asterisks in panel B) at sub-ENDO, sub-EPI and M-cell island.

Morphological Changes of CaT

There were two morphological changes in CaT in the failing hearts compared with that in the nonfailing hearts. Figure 6A shows two representative examples of CaT recorded at the sub-ENDO of a failing (top) and a nonfailing heart (bottom). The first change was a “two-component” rising phase, including an initial fast rising phase (labeled by I in Figure 6A) and a subsequent second slow rising phase (labeled by II in Figure 6A). This was observed in 3 out of 5 failing hearts but not in any of the nonfailing hearts; and it was only present at sub-ENDO. The second slow component resulted in a significant increase of the rise time of CaT (P < 0.001) from 26 ± 3 ms (nonfailing and two failing) to 49 ± 12 ms (three failing) at sub-ENDO. However, if a comparison is made between nonfailing and failing groups, this increase would not be statistically significant (See Figure 6B). The second morphological difference was a more flattened plateau of CaT within the failing group, which is reflected by an increased ratio of CaTD30 to CaTD80. Figure 6C shows that this ratio was significantly increased in failing hearts compared with nonfailing hearts at MID.

Figure 6.

Remodeling of CaT due to heart failure. (A) Representative traces of action potential (AP) and calcium transient (CaT) from a failing heart and a nonfailing heart. (B) 10–90% rise time of CaT at a BCL of 1500ms in failing (F, n=5) and nonfailing (NF, n=6) hearts at sub-ENDO, MID and sub-EPI. (C) Ratio of CaTD30 and CaTD80 at a BCL of 1500ms at sub-ENDO, MID, and sub-EPI.

Relaxation of CaT

The time constant of CaT relaxation reflects the rate of Ca²⁺ reuptake from the cytoplasm by SERCA2a and Na⁺/Ca²⁺ exchanger. We observed a gradient of the relaxation time constant of CaT from ENDO to EPI in both failing and nonfailing hearts. Figure 7A&B shows representative examples of the time constant measurement (τ) in a failing human heart. It is evident that τ at sub-ENDO was larger than that at sub-EPI. The difference between failing and nonfailing groups was not statistically significant although there was a trend of an increase in τ within the failing group (Figure 7C).

Figure 7.

Relaxation of calcium transient (CaT). (A) Map of CaT relaxation time constant (τ) from a failing human heart. (B) Representative traces of CaT (solid lines) from sub-ENDO (red) and sub-EPI (blue) and their corresponding single exponential fittings (dashed lines). (B) Summary of time constant (τ) of CaT relaxation in failing (F, n=5) and nonfailing (NF, n=6) hearts at sub-ENDO, MID, sub-EPI at a BCL of 1500ms.

Protein Expression of SERCA2a and Phospholamban

To determine the molecular mechanism of the observed gradient of τ presented above, we quantified the protein expressions of SERCA2a and phospholamban. In Figure 8, representative bands and the statistical summary are shown for SERCA2a (Figure 8A) and phospholamban (Figure 8B). We divided our samples into three groups (See Table 1) including nonfailing, failing with ischemic cardiomyopathy, and failing with nonischemic/idiopathic cardiomyopathy. Each group was subdivided into sub-ENDO and sub-EPI. For SERCA2a, there was a significant difference between sub-ENDO and sub-EPI (p < 0.001, Figure 8A; interaction between tissue layers and patient groups was not significant [P = 0.295]). SERCA2a expression at sub-ENDO in the ischemic group was significantly lower as compared to that in the nonischemic group (P = 0.023, Figure 8A). For phospholamban, we did not observe any differences between sub-ENDO and sub-EPI, nor among any of the three groups (Figure 8B).

Figure 8.

Protein expressions of SERCA2a and phospholamban. Representative examples of Western blots (top) and normalized protein expression (bottom) are shown for SERCA2a (A) and phospholamban (B). NF (n=6) represents the group of nonfailing hearts; Ischemic-F (n=6) represents the group of failing hearts with ischemic cardiomyopathy; Nonischemic-F (n=6) represents the group of failing hearts with nonischemic/idiopathic cardiomyopathy.

Discussions

In the present study, we conducted for the first time the simultaneous mapping of both voltage and calcium in LV wedge preparations from failing and nonfailing human hearts. We found that HF-induced remodeling consists of (1) increased differences of AP and CaT durations at sub-ENDO during bradycardia (40 bpm), (2) decreased transmural CaTD gradients at fast pacing rates (100 bpm), (3) a slow component of rise and a dome-shaped plateau in CaT, and (4) a lowered level of SERCA2a expression at sub-ENDO in failing human hearts with ischemic cardiomyopathy. We also found that there existed transmural gradients of CaTD80, CaT relaxation time constant (τ), and protein expression of SERCA2a in both failing and nonfailing human hearts.

Implications from CaT Morphology Changes

There was a two-component rise of CaT at sub-ENDO in three of the failing hearts, with an initial fast component followed by a slow second component. This was previously observed in isolated myocytes from failing human and canine hearts.^{2, 21} Piacentino et al. suggested that this might result from increased Ca²⁺ entry during the AP plateau due to less calcium-mediated inactivation of L-type calcium currents and increased activity of the Na⁺/Ca²⁺ exchanger in the reverse mode (Ca²⁺ influx).² The same mechanism could also explain the apparent dome shape of CaT observed in the failing hearts.

The morphological changes of CaT could also result from dyssynchronous Ca²⁺ release within a cell. The delayed release of Ca²⁺ in defective regions might be responsible for the slow component of rise and subsequent dome shape of CaT observed in the failing human hearts. Confocal line scan recordings in whole failing rat hearts revealed that the release of Ca²⁺ at some part within a cell does not occur at the time of initial depolarization but a short time after the depolarization.²² It is possible that the normal Ca²⁺ release corresponds to the first fast rising phase of CaT; and delayed Ca²⁺ release corresponds to the second slow component of CaT.

Interestingly, this slow secondary rise of CaT was observed only in sub-ENDO region in 60% of failing hearts. Further studies will be required to investigate the mechanisms of this remodeling at both molecular and tissue structural levels.

APD, CaTD, and (CaTD – APD)

The decay of CaT was markedly prolonged in isolated cells from failing human hearts.⁹ In contrast, CaTD in failing hearts was not statistically different from nonfailing hearts in our study. This is likely due to differences between isolated cell and tissue preparations as well as differences in the pacing cycle length. O’Rourke et al. showed that CaTD was 3-fold longer in myocytes from failing hearts at BCL of 6 seconds, but was not significantly different at 1-second interval when compared with myocytes from nonfailing hearts.²¹ Another possible explanation for this discrepancy is the afterload dependence of CaT, which indicates that mechanical work and metabolic demand is crucial for inducing the pathological regulation and morphological changes of CaT.¹² Since mechanical work was inhibited in our study by blebbistatin to eliminate motion artifacts, changes of CaT therefore might not be as evident.

(CaTD - APD) was significantly increased at the sub-ENDO at a slow heart rate (40 bpm) in failing hearts as compared to nonfailing hearts (Figure 4A-right). That is, CaT significantly outlasts AP and is elevated during phase 3 of the AP. This difference in duration was previously proposed to promote late phase 3 early afterdepolarization (EAD) by the strong recruitment of electrogenic Na⁺/Ca²⁺ exchanger currents.^{23, 24} Though EAD was not observed in this study, we speculate that this might contribute to the enhanced arrhythmogenesis in HF by promoting EADs under conditions such as metabolic inhibition.

Transmural CaTD Gradient

The transmural gradient of CaTD at fast heart rates (100 bpm) was significantly smaller in the failing group as compared to the nonfailing group. This might have important physiological implications relevant to the mechanical dysfunction of failing human hearts.

The transmural gradient of CaTD was 72 ± 20 ms and 81 ± 16 ms at a slow rate (40 bpm) for failing and nonfailing groups during ENDO pacing, respectively (Figure 4C-right). The corresponding conduction time from the ENDO to EPI was 49 ± 13 ms and 30 ± 5 ms (40 bpm). Therefore, the transmural gradient of the time at 80% of CaT relaxation from ENDO to EPI (ENDO to EPI CaTD80 difference - conduction time) was 23 ± 15 ms and 51 ± 19 ms. The positivity of these values indicates that the sequence of relaxation of CaT was from EPI to ENDO for both failing and nonfailing groups at 40 bpm. This sequence is the same as the transmural sequence of myofiber relaxation measured in vivo in normal canine hearts during sinus rhythm.²⁵

At fast heart rates this sequence was maintained in nonfailing human hearts (as expected) but was reversed in failing human hearts. At 100 bpm in the failing group, the transmural gradient of CaTD was 25 ± 11 ms and conduction time increased to 57 ± 13 ms at (Figure 4C-right). Therefore, the transmural gradient of the time at 80% of CaT relaxation was −29 ± 15 ms, the negativity of which indicates that the sequence of CaT relaxation for the failing group was from ENDO to EPI. This reversed sequence of relaxation at fast heart rates could be associated with poor mechanical function and might be one of the mechanisms underlying the higher risk for primary composite endpoint in HF patients with higher heart rates.²⁶

Heterogeneous Calcium Handling

The protein expression of SERCA2a was significantly lower in ischemic failing hearts than the nonischemic failing hearts, at sub-ENDO but not at sub-EPI. This indicates that the protein level of SERCA2a is dependent on both the etiology of HF and the anatomic location of the myocardium. Previously, SERCA2a was found to be significantly down-regulated in failing human hearts in some studies,^27–30 while it was not in other studies.^31–37 According to our results, the inconsistent observations might be related to the anatomic inconsistency within and across studies, and due to the etiology dependence of down-regulation.

In both failing and nonfailing hearts, the protein level of SERCA2a was less abundant at sub-ENDO than that at sub-EPI (Figure 8A). This difference was consistent with previous observations in canine and human hearts.^{15, 17} The lower expression of SERCA2a was suggested to lead to the larger relaxation time constant of CaT in canine hearts.¹⁵ Our results suggest that this causal relationship might also exist in human hearts.

No significant increase of the relaxation time constant was observed in failing human hearts in this study. Since our failing group consisted of two hearts with ischemic cardiomyopathy and three hearts with nonischemic/idiopathic cardiomyopathy for the functional part of this study, the lack of statistical significance could be explained by different etiologies of HF. Indeed, the time constant in hearts with ischemic cardiomyopathy was longer than in hearts with nonischemic/idiopathic cardiomyopathy (i.e., sub-ENDO: 170 ms vs. 136 ms; MID: 152 ms vs. 124 ms; sub-EPI: 137 ms vs. 114 ms). Due to limited samples, future studies are needed to test this hypothesis. Nevertheless, this hypothesis is supported by a study in isolated myocytes from human hearts with end-stage HF, which showed that relaxation of CaT and contraction was significantly slower in hearts with ischemic cardiomyopathy than that with dilated cardiomyopathy.³⁸ This difference might be explained by the significant down-regulation of SERCA2a protein expression in ischemic failing group but no change within the nonischemic failing group as shown in this study.

M Cell Island

We have recently reported¹⁶ that the M cells were present in the form of spatially discrete and isolated islands rather than a continuous layer in 3 out of 5 nonfailing human hearts. Moreover, M cells were not observed in failing human hearts due to nonhomogeneous APD prolongation and decreased transmural APD gradient.¹⁶ In order to compare failing and nonfailing hearts under the same pattern of APD distribution, we did not concentrate on searching for M cell islands in nonfailing hearts. This explains why we only present M cells in one nonfailing heart (Figure 5). In this experiment, we specifically searched for M cells which were in the form of isolated islands rather than a continuous layer, as was previously described.¹⁶ The regions above and below the M cell island had continuous APD gradient from ENDO to EPI (similar to Figure 3). To compare failing and nonfailing hearts under the same pattern of APD distribution, data only in the region without M cells (e.g., upper part of Figure 5B) were used in the statistical analysis.

As shown in Figure 5D, CaTD within the M cell island was prominently longer compared to the neighboring region. However, CaTD within the M cell island (698±9 ms) was comparable to that at the sub-ENDO (711±15 ms, also see Figure 5D), while APD within the M cell island (649±7 ms) was longer than that at sub-ENDO (618±13 ms, also see Figure 5B). This difference is similar to the observation made in the canine study by Cordeiro et al.¹⁴ Interestingly, other parameters related to EC coupling and calcium handling (such as AP-CaT delay, CaT rise time and CaT relaxation time constant) were not distinctly different from the surrounding region (Supplemental Figure 2). The role of M cell islands in the contraction and their nature in nonfailing hearts remains unclear.

Limitations

1. Nonfailing donor hearts are not necessarily representative of healthy hearts.³⁹ However, none of the donors have a history of HF, and were thus the best controls available for this functional study.

2. Due to technical limitations, only a limited transmural surface of LV with good perfusion was mapped. Due to the anatomical heterogeneity of the heart itself, caution should be taken to extrapolate the results to the whole heart.

3. Due to the limited access to functional human hearts, the number of hearts for each group is small, and might compromise the statistical significance of potential differences. Because all of the failing hearts with different cardiomyopathy were grouped together for the functional data analysis, only changes common to different etiologies of cardiomyopathy could be revealed. Changes unique to individual cardiomyopathy could be masked.

4. Several other important aspects of EC coupling were not examined in this study, such as the SR calcium content and the EC coupling gain. Regional differences of these parameters need to be resolved in future studies.

5. Application of blebbistatin in our study liberated ATP from mechanical contraction and thus allowed ample supply of ATP to electrophysiological processes. Pathological changes could thus be less evident in the absence of metabolic disturbance that could be unmasked by mechanical work.

6. As shown in neonatal rat myocyte cultures, the use of high-affinity dyes including Rhod-2 may overestimate CaTD, which was about twice as large as APD.^{40, 41} This was not likely the case in our measurements because CaTD in normal human hearts was comparable to APD. Also, the main focus of the present study was on regional differences of CaTDs rather than on their absolute values. Thus, potential systematic errors, if they existed, were likely to be subtracted or minimized.

Summary

In conclusion, we present for the first time the simultaneous transmural mapping of AP and CaT in LV wedge preparations in both failing and nonfailing human hearts. Our results demonstrate the transmural heterogeneity of EC coupling and calcium handling, and the transmurally heterogeneous remodeling of these properties due to HF.

Clinical Relevance.

Excitation-contraction coupling (ECC) is a complex process mediated by a network of proteins that control ionic currents, cell signaling, calcium handling, and sarcomeric mechanics. The two hallmarks of ECC preceding mechanical contraction are the transmembrane action potential and intracellular calcium transient. Numerous studies in animal models have significantly advanced our understanding of the fundamental mechanisms of ECC at the molecular, cellular, and whole heart levels. However, the extrapolation of these findings to clinical practice is limited by the lack of functional human data. Our study provides for the first time in medical history such functional data: simultaneous recordings of action potentials and calcium transients from the human left ventricular tissue of nonfailing and failing hearts. We found that the nonfailing human left ventricle has a transmural gradient of calcium transient kinetics, with the calcium transient being longer at the endocardial versus epicardial layers of the ventricular wall. Decrease of this gradient in the failing heart reversed the normal sequence of relaxation. We also found that heart failure leads to significant changes in calcium dynamics manifesting as a biphasic calcium entry into the cytosol. Downregulation of SERCA2a depends on etiology and anatomic location, which might explain the inconsistency in the literature on the subject. Transmural heterogeneity indicates that anatomic location must be considered in studies of molecular remodeling of ECC. Our study confirms some of the earlier findings in animal models, but contradicts the others. Thus, extrapolation of findings from animal models to humans should be done with caution.