Cardiac Resynchronization Therapy Reduces Subcellular Heterogeneity of Ryanodine Receptors, T-Tubules and Ca²⁺ Sparks Produced by Dyssynchronous Heart Failure

Abstract

Background

Cardiac resynchronization therapy (CRT) is a major advance for treatment of patients with dyssynchronous heart failure (DHF). However, our understanding of DHF-associated remodeling of subcellular structure and function and their restoration after CRT remains incomplete.

Methods and Results

We investigated subcellular heterogeneity of remodeling of structures and proteins associated with excitation-contraction coupling in cardiomyocytes in DHF and after CRT. Three-dimensional confocal microscopy revealed subcellular heterogeneity of ryanodine receptor (RyR) density and the transverse tubular system (t-system) in a canine model of DHF. RyR density at the ends of lateral left ventricular cardiomyocytes was higher than in cell centers, while the t-system was depleted at cell ends. In anterior left ventricular cardiomyocytes, however, we found a similar degree of heterogeneous RyR remodeling despite preserved t-system. Synchronous heart failure (SHF) was associated with marginal heterogeneity of RyR density. We used rapid scanning confocal microscopy to investigate effects of heterogeneous structural remodeling on calcium signaling. In DHF, diastolic Ca²⁺ spark density was smaller at cell ends versus centers. After CRT, subcellular heterogeneity of structures and function was reduced.

Conclusions

RyR density exhibits remarkable subcellular heterogeneity in DHF. RyR remodeling occurred in lateral and anterior cardiomyocytes, but remodeling of t-system was confined to lateral myocytes. These findings indicate that different mechanisms underlie remodeling of RyRs and t-system. Furthermore, we suggest that ventricular dyssynchrony exacerbates subcellular remodeling in heart failure. CRT efficiently reduced subcellular heterogeneity. These results will help to explain remodeling of EC coupling in disease and restoration after CRT.

Cardiac resynchronization therapy (CRT) is an established clinical therapy for patients with moderate to severe heart failure (HF). CRT is based on biventricular pacing. It acutely improves left ventricular (LV) mechanical performance^1–3 and reduces myocardial oxygen consumption in patients with HF and intraventricular conduction delays.⁴ Several clinical trials demonstrated that CRT leads to a reduction of HF-related rehospitalizations and overall mortality in the majority of patients with dyssynchronous HF (DHF).^5–9 A clinical marker of DHF is prolonged QRS duration, which reflects interventricular delays of electrical activation. Consequently, mechanical contraction is regionally delayed, which aggravates weak cardiac performance in DHF patients.

While recent studies demonstrated remarkable remodeling of subcellular structures and function in DHF and partial restoration in response to CRT,^10–13 we are only starting to understand pathological remodeling and effects of CRT at the microscopic scale. This study focuses on remodeling of excitation-contraction (EC) coupling in DHF and the ability of CRT to restore structure and function at subcellular level. EC coupling is the mechanism by which an action potential at the membrane of a muscle cell initiates a cascade of events that result in release of Ca²⁺ from the sarcoplasmic reticulum (SR) and activation of contraction. Our current understanding of EC coupling in ventricular myocytes is organized around the concept of the couplon.¹⁴ A couplon comprises one or more voltage-gated L-type Ca²⁺ channels (LCCs) in the sarcolemma and a cluster of ryanodine receptors (RyRs) in the terminal cisternae of the SR.¹⁵ LCCs and RyRs are separated by a junction of ~12 nm.¹⁶ In ventricular myocytes, couplons are commonly associated with the transverse tubular system (t-system), which is a specialization of the sarcolemma penetrating into the interior of myocytes and in close proximity to the SR. The t-system facilitates rapid transmission of membrane voltage changes into the cell interior to enable synchronous activation of couplons. Our current knowledge of the mechanism by which couplons activate the Ca²⁺ transient is based on the classical work of Fabiato¹⁷ and insights into the subcellular arrangement of structures involved in EC coupling by microscopy.¹⁸ Couplons are activated when LCCs admit small quantities of Ca²⁺ into the junctional region and gate RyRs. Couplons are separated from one another by ~1 μm.¹⁹ When Ca²⁺ is released by the RyRs, it produces a local Ca²⁺ release event, or Ca²⁺ spark.^{20, 21} The spread of Ca²⁺ in the cytoplasm is governed by diffusion and buffering. Ca²⁺ transients are believed to result from the summation of the activation of many couplons.

Alterations of Ca²⁺ transients are a common feature of cardiac myocytes from HF hearts and thought to contribute to the progression of HF.²² Recent studies investigated the connection between alterations of Ca²⁺ transients and remodeling of the t-system. The studies suggested that remodeling of the t-system in ventricular myocytes is a consequence of tachycardia-induced HF,^23–25 infarction²⁶ and DHF.¹¹ Our previous work on isolated myocytes revealed regional depletion of t-system in canine DHF models.¹¹ In particular, lateral LV myocytes were affected by DHF, but not myocytes from the anterior LV wall. We demonstrated that t-system depletion in DHF was accompanied by increased occurrence of non-junctional RyR clusters. A remarkable recent finding is that t-system depletion is reversible. Partial restoration of t-system was found after CRT,¹¹ SERCA2a gene therapy²⁷ and mechanical unloading of the heart.²⁸ In those studies t-system restoration was associated with restoration of Ca²⁺ transients, which emphasizes the crucial role of the t-system for efficient excitation-contraction coupling.

Beyond those studies remarkably little is known about the mechanisms and the degree of remodeling and restoration of the t-system. In particular, little is known about the subcellular heterogeneity of remodeling of the t-system and associated proteins such as RyRs as well as the effects of this structural remodeling on cell function. Also, we have only sparse information on restoration of structures and function at the subcellular scale. Here, we studied an animal model of synchronous heart failure (SHF) based on right atrial (RA) pacing and a model of DHF based on right ventricular (RV) pacing to test the hypothesis that ventricular dyssynchrony accentuates the extent and heterogeneity of t-system and RyR remodeling in HF. While both models are based on rapid pacing and lead to congestive HF, the DHF model additionally reproduces clinical findings of interventricular delays of electrical activation and delayed mechanical contraction of the lateral LV wall. Furthermore, we used an animal model of CRT to test the hypothesis that CRT effectively reduces the extent and heterogeneity of remodeling after DHF. We applied three-dimensional (3D) confocal microscopy and image analysis to provide insights into the spatial distribution of the t-system and RyRs in cardiac myocytes from animal models of SHF, DHF and CRT. Segmentation of myocytes in images of cardiac tissue allowed for 3D reconstructions of cells and quantitative analyses of subcellular distributions. We investigated for the first time the subcellular heterogeneity and degree of remodeling of t-system and RyRs in different models of HF and after CRT. Additionally, we measured Ca²⁺ spark density in isolated ventricular myocytes using dual-labeling with fluorescent dyes and rapid scanning confocal microscopy, which allowed acquisition of two-dimensional image sequences at high spatiotemporal resolution. We developed image-processing methods for analysis of these image sequences to investigate at the subcellular scale if structural heterogeneity is accompanied by functional heterogeneity.

Methods

Animal Models of HF and CRT

All procedures involving the handling of animals were approved by the Animal Care and Use Committees of the Johns Hopkins University and the University of Utah. Protocols complied with the published Guide for the Use and Care of Laboratory Animals published by the National Institutes of Health.

We generated canine models of SHF, DHF and CRT. These models have been described and characterized in detail previously.^{11, 12, 29–32} In brief, adult mongrel canines (25–30 kg) were used as control (CTRL) and models of SHF, DHF and CRT. DHF was induced by either 6 weeks of RV pacing or left bundle branch ablation followed by 6 weeks of RA pacing. The CRT model was identical to the DHF model for the first 3 weeks, followed by 3 weeks of biventricular pacing. SHF animals underwent RA pacing for 6 weeks. The pacing rate for SHF, DHF and CRT animals was 170–200 bpm. Our methods for monitoring, tissue collection and cell isolation are described in the Supplemental Data.

Tissue Preparation, Fluorescent labeling and 3D Confocal Microscopy

Our approach for preparation and labeling of tissue is detailed in the Supplemental Data. We used wheat germ agglutinin (WGA) conjugated to a fluorophore for labeling of the sarcolemma, t-tubules and the interstitial space. RyRs were labeled with a monoclonal antibody (MA3-916, ThermoFisher Scientific, Waltham, MA, USA). After labeling, tissue sections were placed on a glass slide, embedded in Fluoromount-G (Electron Microscopy Science, Hatfield, PA, USA) and covered with a #1.5 coverslip. Three-dimensional image stacks were obtained from labeled tissue sections using a Zeiss LSM 5 Live Duo confocal microscope (Carl Zeiss, Jena, Germany) with a 63× oil immersion objective (numerical aperture: 1.4). Alexa Fluor 488 and 555 were excited using a 488 nm and 543 nm laser, respectively. Emitted light was band-pass filtered at 505 to 530 nm and long-pass filtered at 560 nm, respectively. Typical stack dimensions were 1024×1024×240 voxels at a resolution of 0.1×0.1×0.1 μm.

Analysis of T-System and RyR Clusters in Image Stacks

A detailed description of the analyses is provided in Supplemental Data. In short, we applied methods for noise reduction, deconvolution and background removal to the image stacks. The WGA and RyR images were segmented using histogram-based thresholding. Segmented WGA images were used for semi-automatic segmentation of myocytes.³³ Segmented cells then served as masks to analyze RyR clusters and sarcolemma including t-tubules. Cells were divided into two regions: 0–10 μm and 10–40 μm from a longitudinal cell end. Analysis of each RyR cluster yielded sum of intensity and intensity-weighted centroid position. Clusters with intensities higher than 90% of all clusters in the same cell (90^th percentile) were defined as high-intensity clusters. To determine the distance of voxels to the nearest sarcolemma we calculated Euclidean distance maps from the segmented WGA images (Supplemental Figure 1). As a measure of t-system density we used the mean distance of intracellular voxels to the nearest sarcolemma (Δ_SL). Accordingly, high and low distances indicate low and high t-system densities, respectively. Distances of RyR cluster centroids to the nearest sarcolemma (Δ_RyR) were calculated. We analyzed RyR clusters in groups proximal (0–1 μm) and distal (> 1μm) to the sarcolemma. RyR cluster density (ρ_RyR) was computed by dividing the number of clusters by the volume of interest. Additionally, we normalized the total intensity of RyR clusters in each segmented cell and in subcellular regions to the corresponding volumes. Dividing intensity densities of subcellular regions by the intensity density of the whole cell yielded subcellular RyR intensity distribution (I_RyR).

Rapid Scanning Confocal Microscopy and Analysis of Ca²⁺ Sparks

Isolated LV lateral myocytes were incubated at room temperature in modified Tyrode’s solution containing 12.5 μM of the Ca²⁺ sensitive dye Fluo-4 AM (Invitrogen) for 20 min followed by incubation in 6.25 μM of the membrane staining dye Di-8-Anepps (Invitrogen) for 8 min. Cells were placed in a perfusion bath system and allowed to settle on the glass slide. The chamber was perfused with the modified Tyrode’s solution containing either 2 mM or 4 mM Ca²⁺ and held at 37˚C. Imaging was performed using a Zeiss LSM 5 Live Duo confocal microscope equipped with a 63× oil immersion objective. Only brick-shaped cells with clear striations were imaged. A diode laser emitting a wavelength of 489 nm was used to excite both dyes simultaneously. Emitted light was filtered using a dichromatic mirror and two bandpass filters of 505–610 nm and 560–675 nm for capturing Fluo-4 and Di-8-Anepps signals, respectively. Cells were conditioned with a train of at least 10 stimuli at 0.5 Hz. Image acquisition was triggered 460 ms after the final stimulus to measure Ca²⁺ in the diastolic phase. Sequences of 100 two-dimensional images were acquired at a size of 256×1024 pixels and a frame rate of 9.3 ms/frame. Pixel sizes were 0.1×0.1 μm. Details of the detection algorithm are provided in the Supplemental Data. Briefly, cells were segmented using histogram-based thresholding in Di-8-Anepps images (Supplemental Figure 2). After preprocessing of the Fluo-4 images (Supplemental Figures 2 and 3) sparks were detected in the cell interior using mode plus five times standard deviation of the intensity values. We measured the total number of sparks, number of sparks within 10 μm of the cell end, and the number of sparks between 10 and 40 μm from the cell end. Spark number was normalized by temporal duration and area used for detection, which yielded spark density (sparks/μm²/s).

Statistical Analyses

All results are presented as mean ± standard error. All statistical analyses were performed in Matlab R2012b (The Mathworks Inc., Natick, Ma, USA). We accounted for multiple measurements from an animal by averaging of data for each animal. We used one-way analysis of variance (ANOVA) followed by Tukey’s multiple-comparison for microstructural analyses. We applied the paired t-test for functional analyses. Significance of differences was defined as p<0.05.

Results

Subcellular Heterogeneity of Remodeling of RyRs and T-System in DHF

We first analyzed RyR clusters and t-tubules in myocytes in LV lateral tissue from control animals (Figure 1). All images were preprocessed, which improved signal-to-noise ratio and resolution (Figure 1A–D versus 1E–H). Control cells exhibited a dense t-system (Figure 1E) and homogenous intracellular RyR density (Figure 1F). This was also visible in 3D reconstructions of sarcolemma and RyR clusters in a segmented cell (Figure 2A–C). RyR clusters of high intensity were found evenly distributed in all cell regions. We measured the mean intracellular distance to the sarcolemma (Figure 2D), the mean distance of RyR clusters to the sarcolemma (Figure 2E), the mean density of RyR clusters (Figure 2F) and the normalized intensity of RyR signal (Figure 2G) in regions at a distance of 0–10, 10–20, 20–30, and 30–40 μm to a cell end. These quantitative measures indicate homogeneity of t-system and RyRs within LV lateral cells in control animals.

Figure 1.

Preprocessing of confocal microscopic images. XY sections at a depth of 27 μm from a 3D image stack (102.4×102.4×35μm) of control tissue. (A–D) Unprocessed data. (E–H) Preprocessed data. Green: RyRs, Blue: WGA. White squares: zoom-in regions shown in (C,D) and (G, H). Scale bars: 20 μm.

Figure 2.

Visualization and analysis of control cell. (A–C) 3D visualizations of 4 μm thick sections through example control cell. (A) Longitudinal section, transverse section (B) 5 μm from cell end and (C) 40 μm from cell end. Red bars indicate section planes. Blue: sarcolemma. Green and red spheres: RyR clusters. Clusters of high intensity (90^th percentile) are shown in red. Scale bars: 10 μm. Scale bar in (B) applies to (C). Analysis of (D) mean intracellular voxel distance to the nearest sarcolemma, (E) mean RyR cluster distance to nearest sarcolemma, (F) RyR cluster density and (G) RyR intensity per volume normalized to mean value of the whole cell. These parameters were analyzed at distances 0–40 μm from the left cell end in bins of 10 μm.

In contrast to control, DHF tissue from lateral LV tissue showed a pronounced loss of t-tubules near longitudinal cell ends together with striking heterogeneity in RyR density (Figure 3). RyR intensities were considerably higher near cell ends as compared to cell centers. In this overview image, most cells were affected and exhibited this heterogeneity at least at one cell end. A high-resolution 3D image stack is shown in Figure 4A–C. Cell ends almost completely lacked t-tubules and exhibited increased RyR intensities. This was confirmed when inspecting 3D reconstructions of sections through a representative DHF cell (Figure 4D–F). T-tubules were virtually absent within 10 μm from the cell end, while RyR clusters of high intensity aggregated in this region. Quantitative data corresponded to these findings. The mean intracellular distance to the sarcolemma was increased within 10 μm from the cell end versus other regions (Figure 4G). The mean distance of RyR clusters to the sarcolemma decreased slightly from the cell end to center (Figure 4H). The increase of density and intensity of RyR clusters at the cell end was remarkable (Figure 4I and J, respectively). These findings suggest substantial subcellular heterogeneity of t-tubule and RyR cluster remodeling in LV lateral DHF myocytes.

Figure 3.

Confocal microscopic image of DHF tissue from lateral LV wall. (A) T-system density decreases at cell ends. (B) RyR clusters exhibit increased intensities near cell ends. Scale bar: 40 μm.

Figure 4.

Visualization and analysis of DHF cell. (A–C) Confocal microscopic images of extracellular space (blue) and RyR clusters (green) in DHF tissue. (B) and (C) are zoom-ins from the region indicated by the white rectangle. Note the increase in RyR intensity and t-system depletion near cell ends. (D–F) 3D visualizations of 4 μm thick sections through example DHF cell. (D) Longitudinal section, (E) transverse section 5 μm from cell end and (F) transverse section 40 μm from cell end. Red bars indicate section planes. Blue: sarcolemma. Green and red spheres: RyR clusters. Clusters of high intensity (90^th percentile) are shown in red. All scale bars: 10 μm. Scale bar in (B) applies to (C). Scale bar in (E) applies to (F). Analysis of (G) mean intracellular voxel distance to the nearest sarcolemma, (H) mean RyR cluster distance to nearest sarcolemma, (I) RyR cluster density and (J) RyR intensity per volume normalized to mean value of the whole cell.

Partial Restoration in Response to CRT

We next investigated whether CRT was able to reduce the DHF-associated subcellular heterogeneity of remodeling. Our model of CRT is based on application of the DHF model for three weeks followed by three weeks of resynchronization by rapid biventricular pacing. The RyR and t-system density in myocytes from LV lateral tissue of CRT animals appeared more homogeneous than in DHF (Figure 5A–C), indicating partial structural restoration. Three-dimensional reconstructions of longitudinal and transverse sections through an example cell are presented in Figure 5D–F, showing low t-system densities in close proximity to the cell end. However, large regions devoid of t-tubules as found in DHF were not observed. Similarly, RyR clusters of high intensity still appeared in higher concentration near the cell end, but to a lesser degree than in DHF. Quantitative analyses confirmed partial restoration of subcellular homogeneity in this cell (Figure 5G–J).

Figure 5.

Visualization and analysis of CRT cell. (A–C) Confocal microscopic images of the extracellular space (blue) and RyR clusters (green) in CRT tissue. (B) and (C) are zoom-ins from the region indicated by the white rectangle. RyR and t-system remodeling are less pronounced than in DHF. (D–F) 3D visualizations of 4 μm thick sections through an example CRT cell. (D) Longitudinal section, (E) transverse section 5 μm from cell end and (F) transverse section 40 μm from cell end. Red bars indicate section planes. Blue: sarcolemma. Green and red spheres: RyR clusters. Clusters of high intensity (90^th percentile) are shown in red. All scale bars: 10 μm. Scale bar in (B) applies to (C). Scale bar in (E) applies to (F). Analysis of (G) mean intracellular voxel distance to the nearest sarcolemma, (H) mean RyR cluster distance to nearest sarcolemma, (I) RyR cluster density and (J) RyR intensity per volume normalized to mean value of the whole cell.

Statistical Analyses of RyR and T-System Remodeling

We applied the same analyses used for the examples to a group of control, DHF and CRT cells. Extracted features binned by increasing distance from the cell end are presented in Supplemental Figure 4. Because t-system loss and increased RyR intensities in DHF occurred primarily near cell ends, we compared t-system and RyR properties in regions 0–10 μm with regions 10–40 μm from the cell end.

The mean intracellular distance to the nearest sarcolemma indicated high t-system density in control, while DHF cells exhibited a pronounced loss of t-tubules at cell ends (Figure 6A). CRT was not able to completely restore this measure to levels of control, but the intracellular distance to the nearest sarcolemma was similar at cell ends and centers. In our analysis of the mean distance of RyR clusters to the sarcolemma (Figure 6B), we found increased values in DHF and CRT versus control near cell ends (0.67±0.4 and 0.66±0.09 vs 0.47±0.03 μm, respectively). The distance was similar in cell centers of DHF, CRT and control cells (0.52±0.03, 0.52±0.06 and 0.51±0.04 μm, respectively). DHF cells exhibited significant subcellular heterogeneity of mean RyR-sarcolemma distance.

Figure 6.

Statistical analysis of myocytes from control (CTRL), DHF and CRT animals. Error bars indicate standard errors of the mean. Parameters were analyzed 0–10 μm and 10–40 μm from the cell end. (A) Mean intracellular voxel distance to nearest sarcolemma. (B) Mean RyR cluster distance to nearest sarcolemma, (C) RyR cluster density and (D) RyR intensity per volume normalized to mean intensity of the whole cell. Parameters were analyzed 0–1 μm and ≥1 μm from the sarcolemma. (E) RyR cluster density and (F) RyR intensity per volume normalized to mean value of the whole cell. Brackets mark statistical significance between experimental groups, asterisks between corresponding bins of groups.

RyR cluster densities ranged from 0.75 to 0.91 μm⁻³ and our analysis did not indicate significant subcellular heterogeneity in DHF and CRT (Figure 6C). In contrast, analyses of normalized RyR intensities (Figure 6D) revealed that DHF was associated with a substantial increase at cell ends (134±10% of cellular mean) and a decrease in centers (92±2%) versus control (110±6% and 98±1%, respectively). CRT cells presented RyR distributions similar to those of DHF.

Our data show that DHF is associated with pronounced heterogeneity of remodeling of t-tubules and RyR clusters at cell centers and ends. We next sought to determine whether similar changes occur when grouping RyR clusters by their distance to the sarcolemma (Supplemental Figure 5). We therefore analyzed cluster density and intensity within 0–1 μm and at a distance ≥1 μm to the sarcolemma. Cluster densities did not differ between groups or regions (Figure 6E). However, when analyzing RyR cluster intensities (Figure 6F), DHF cells exhibited decreased intensities 0–1 μm (91±2% of cell mean) and strikingly increased intensities ≥1 μm from the sarcolemma (144±12%). A similar, but less pronounced distribution was found in CRT cells (95±1% and 124±6%, respectively). RyR cluster intensity in control cells was almost equally distributed at distances 0–1 μm and ≥1 μm from the sarcolemma (101±2% and 99±8%, respectively). In DHF and less pronounced after CRT, we also found increased RyR intensities ≥1 μm from the sarcolemma in cell centers. RyR intensities in control, however, were homogeneous throughout the cell (Supplemental Figure 6A, B). In all experimental groups, the majority of RyR clusters was found within 1 μm of the sarcolemma (Supplemental Figure 6C).

In summary, control cells were more homogeneous at the subcellular scale than DHF and CRT cells. DHF led to a significant loss of t-tubules and increased RyR intensities at cell ends versus centers, causing pronounced subcellular heterogeneity. Increased RyR intensities were found particularly in regions ≥1 μm from the sarcolemma. CRT was able to partially reverse these structural changes.

RyR Remodeling Despite Preserved T-System in Anterior Cells in DHF

In previous work we found that subcellular remodeling of the t-system is regional, i.e. cells isolated from the lateral LV wall of DHF animals exhibited a higher degree of t-system depletion than cells from the anterior wall.¹¹ To shed light on this regional heterogeneity of subcellular remodeling, we additionally studied cells from the anterior LV wall obtained from DHF animals. An example cell is shown in Supplemental Figure 7. The mean intracellular distance to the nearest sarcolemma did not differ from control proximal and distal to the cell end (Supplemental Figure 8A). In contrast to lateral cells, DHF in anterior cells was not associated with heterogeneous t-system remodeling. However, the distribution of RyR intensity in anterior DHF cells was heterogeneous as observed in lateral cells (Supplemental Figure 8D, F). This indicates that RyR remodeling is independent of t-system remodeling in anterior cells.

Preserved T-System and Marginal RyR Remodeling in SHF

To investigate the role of dyssynchrony in remodeling, we analyzed lateral cells from animals in SHF induced by RA pacing. Cells from this group did not show t-system remodeling or changes in the relationship between t-system and RyRs versus control (Supplemental Figures 9 and 10). However, we found changes in RyR intensity as observed in DHF, although to a lesser degree. RyR intensity at cell ends was higher than at cell centers in SHF.

Subcellular Heterogeneity of Ca²⁺ Sparks

To investigate whether subcellular heterogeneity of structural remodeling was associated with heterogeneity of Ca²⁺ release events, we acquired Fluo-4 images from isolated myocytes using rapid scanning confocal microscopy. Processing of the Fluo-4 images (Figure 7A) enabled the detection of Ca²⁺ release events during the diastolic phase of ventricular myocytes. We applied filtering, attenuation correction and pixel-wise self-ratioing on the image data to improve signal quality. Results of this processing are presented for a single pixel (Figure 7B, C) and region with a spontaneous Ca²⁺ release event (Figure 7D, E). In Figure 7F–H, we present detected sparks localized in the Di-8-Anepps images. We detected the cell orientation and the most distal point of the cell. This enabled the creation of distance maps from the cell end. Subsequently, specific regions of interest at varying distances were compared. Figures 7F, 7G, and 7H show the investigated regions and centroids of detected sparks mapped onto the Di-8-Anepps images of an exemplary control, DHF and CRT cell, respectively. Example Fluo-4 and Di-8-Anepps images of these cells are presented in Supplemental Figures 2 and 3. Supplemental Videos 1, 2 and 3 illustrate the raw and processed Fluo-4 image sequences in control, DHF and CRT cells, respectively. While spark density was homogeneously distributed in the control and CRT cell, the DHF cell exhibited a reduced spark density at the end versus center.

Figure 7.

Analysis of subcellular heterogeneity of diastolic Ca²⁺ sparks using rapid scanning confocal microscopy. (A) Representative Fluo-4 images from control cell with a spark marked by the red box. (B) Trace of Fluo-4 signal over time for pixel located at center of red box marked in A. (C) Trace of processed Fluo-4 signal in B. (D) Image of Fluo-4 signal in region outlined by red box in A. (E) Processed image of region in E. (F–H) Di-8-Anepps images overlaid with red circles centered on all detected sparks. (F) Control cell with a spark density of 0.020 sparks/s/μm² near the end and 0.0253 sparks/s/μm² in the center of the cell. (G) DHF cell with a spark density of 0.0227 sparks/s/μm² near end and 0.0418 sparks/s/μm² in the center of the cell. (H) CRT cell with spark density 0.0166 sparks/s/μm² near the end and 0.0167 sparks/s/μm² in the center of the cell.

Using this approach we performed a statistical analysis of sparks in control, DHF and CRT cells bathed in modified Tyrode’s solution containing either 2 or 4 mM Ca²⁺ (Figure 8 and Supplemental Figure 11). We did not find significant changes in overall spark density for cells in either 2 or 4 mM Ca²⁺ (Figure 8A, B). However, cells from DHF animals exhibited a significantly lower spark density in regions 0–10 μm versus 10–40 μm from cell ends. This heterogeneity was not visible in cells from either control or CRT animals (Figure 8D, E), indicating restoration of homogeneity of sparks after CRT.

Figure 8.

Statistical analyses of diastolic spark density. Overall spark density in control (CTRL), DHF and CRT cells in a bathing solution containing (A) 2 mM and (B) 4 mM Ca²⁺ was similar. Analysis of spark density with regions less than 10 μm and between 10 and 40 μm from the cell end in a bathing solution containing (C) 2 mM and (D) 4 mM Ca²⁺ revealed subcellular heterogeneity in DHF cells. Asterisks mark statistical significance between corresponding bins of groups.

Discussion

Our study revealed previously unknown subcellular heterogeneity of structural and functional remodeling in ventricular myocytes in HF. Three-dimensional analysis of confocal microscopic images of tissue sections from DHF myocytes indicated that t-system density and RyR distribution is heterogeneous at the subcellular scale. Both, DHF and SHF cells showed heterogeneous RyR distributions, but only lateral DHF myocytes exhibited t-system depletion at cell ends. The t-system was not affected in anterior DHF and lateral SHF myocytes. In subsequent studies using rapid scanning confocal microscopy we found that the subcellular heterogeneity of t-system and RyR distribution is associated with heterogeneous Ca²⁺ spark densities. After CRT subcellular homogeneity of structures and function was, in part, restored.

Our finding of subcellular heterogeneity of the density of RyRs in HF is novel. Our studies revealed increased density and fluorescence intensity of RyR clusters at cell ends versus cell centers in DHF and SHF. In both HF models, the intensity of RyR clusters not associated with sarcolemma was higher than proximal to sarcolemma. This effect of HF was not restricted to cell ends (Supplemental Figure 6A), but also occurred in cell centers (Supplemental Figure 6B). In contrast, in control cells the intensity of RyR clusters was not affected by the distance to sarcolemma. This suggests that differences in optical properties or immunostaining do not underlie heterogeneous RyR intensities found in DHF cells. Instead, our study indicates that in HF cells the number of RyR channels is higher in non-junctional clusters than in couplons. Accordingly, RyR distribution along the longitudinal cell axis was especially heterogeneous in lateral DHF cells. Cell ends, where t-system density was low, presented a higher percentage of non-junctional RyR clusters. Since these clusters exhibited increased intensity, the overall RyR intensity at cell ends was increased. In agreement with this finding, SHF and anterior DHF cells exhibited an increase of RyR intensity ≥1 μm from the sarcolemma although the t-system was normal (Supplemental Figures 8 and 10). This suggests that RyR remodeling is associated with rapid pacing underlying our HF models. However, subcellular heterogeneity of RyR signal intensity was higher in DHF than in SHF indicating that ventricular dyssynchrony exacerbates the remodeling.

Previously, significant reduction of RyR mRNA and protein expression has been reported in LV myocytes from DHF and CRT animals.¹⁰ The reduction was found in both anterior and lateral cells. Our analysis of RyR associated fluorescence revealed a ≈40% increase of RyR expression at ends versus centers of lateral DHF cells (134±10% vs 92±2%) (Figure 6D). Assuming an overall decrease of RyR protein expression in DHF lateral cells by ≈20%,¹⁰ this would translate into an increase of RyR expression to 107% at cell ends and a decrease to 74% at cell centers versus control. Similarly, RyR reduction in DHF cells can be explained by reduction of junctional RyRs (Figure 6F). Our data suggest that RyR protein expression is reduced to 73% within 0–1 μm and increased to 115% in regions ≥1 μm from the sarcolemma. This reduced RyR expression within 0–1 μm from the sarcolemma may reflect decreased junction size, which was suggested to underlie delayed stochastic couplon activation and reduced amplitude of Ca²⁺ transients.³⁴

It is well established that t-system depletion is a feature of HF. Our previous study on isolated ventricular myocytes suggested regional heterogeneity of t-system depletion. In particular this study indicated that t-system depletion is pronounced in the lateral LV region in DHF, but we did not investigate subcellular heterogeneity of t-system depletion. However, it has been observed in isolated cells that t-system depletion in tachypacing induced HF occurs preferentially at cell ends.^{23, 24} In agreement with those studies, remodeling of t-system in our DHF model appears to be regional and restricted to ends of myocytes. These findings indicate that ventricular dyssynchrony is a major contributor of this type of regional subcellular remodeling.

Previously, we found reduced amplitudes and slowed decays of Ca²⁺ transients in DHF myocytes and restoration after CRT to levels similar as in control.^{10, 11} In this study, we provide insights into structural subcellular heterogeneity in DHF, which may underlie these alterations. According to our recent model of Ca²⁺ release³⁵ the detubulation and increase in non-junctional RyRs at cell ends would cause heterogeneous Ca²⁺ transients in DHF cells, in particular delayed onset times at ends versus centers of cells. This hypothesis is supported by a preliminary analysis of a DHF cell (Supplemental Data). The analysis also suggests reduced amplitude of the Ca²⁺ transients at cell ends as a further effect of heterogeneous microstructural remodeling in DHF.

Since diastolic Ca²⁺ sparks have been linked to arrhythmogenesis,³⁶ we investigated whether diastolic sparks are affected by the structural heterogeneity in DHF cells. Spark densities in control and CRT cells were homogeneous, but in DHF spark density was lower at cell ends than in cell centers. This agrees with a previous study in which regions void of t-tubules lacked sparks in normal and failing canine myocytes.³⁷ Possible causes of subcellular heterogeneity of spark density include heterogeneous SR Ca²⁺ load and different spark probabilities of junctional and non-junctional RyR clusters. Interestingly, increased RyR density at cell ends was not associated with increased spark density. Instead spark density was increased in regions with intact t-system and, thus, high ratio of junctional to non-junctional RyRs. A potential explanation for increased spark probability in couplons is that spontaneous Ca²⁺ releases from junctional RyRs into the dyadic cleft have a higher probability of triggering neighboring RyRs than Ca²⁺ release from non-junctional RyRs into the cytosol. Overall spark density was modulated by bath Ca²⁺ concentration, which can be explained by a positive relationship between bath Ca²⁺ concentration, SR Ca²⁺ load and spontaneous spark probability. However, our study did not indicate significant differences of overall spark density in control, DHF and CRT cells. This suggests that increased spark density in centers of DHF cells is counterbalanced by reduced spark density at cell ends.

While our studies revealed striking subcellular remodeling and restoration, the underlying mechanisms remain unclear. A potential mechanism is related to heterogeneity of mechanical properties of cardiac myocytes and their strain profiles during a heartbeat. We previously proposed that changes in strain profiles directly affect t-system structure and maintenance.^{11, 38} Here, we extend this hypothesis by suggesting that subcellular heterogeneity of strain profiles in myocytes lead to subcellular heterogeneous remodeling of t-system and RyRs. Based on principles of mechanics, for instance, Hooke’s law,³⁹ we expect larger stretch in those regions of strained myocytes, which have a smaller cross-sectional area. Indeed most cells taper towards their ends and thus have a smaller cross-sectional area at the end (e.g. Figure 2). In the cell population of our study, cross-sectional areas within 10 μm from the cell end were reduced to 62.8±1.2% of central cross-sectional areas (within 10–40 μm from the cell end). An alternative explanation of subcellular heterogeneity of strain profiles is subcellular heterogeneity of myocyte stiffness. Hooke’s law predicts larger stretch in those regions of strained myocytes, which have a smaller stiffness.

It is well established that DHF is associated with contraction in anterior and septal sites, but reciprocal stretch in lateral sites during early systole.^{10, 32} Opposite strains occur during late systole. Stretch during early systole in lateral sites is significantly larger than at end diastole. Based on our hypothesis we expect that lateral left ventricular myocytes are heterogeneously stretched during early systole, with pronounced stretch at the tapered cell ends that is larger than in the cell center. The pronounced stretch at the cell ends could directly destabilize t-system or interfere with its maintenance.

A similar argument based on mechanical principles can be made to suggest potential mechanisms for up-regulation of RyRs at cell ends in SHF and DHF in the context of a general RyR down-regulation in ventricular myocardium. Both, SHF and DHF are associated with increased left ventricular end diastolic blood pressure.¹¹ Based on our new hypothesis we expect that ventricular myocytes in SHF and DHF are heterogeneously stretched during diastole, with larger stretch at the tapered cell ends than in the cell center. We speculate that the increased stretch at cell ends provides a signal for local up-regulation of RyR expression. Alternatively, the locally increased stretch might interfere with protein degradation.

Limitations

Limitations of our animal models have been explained previously.^{10, 11, 13} Further limitations related to confocal microscopy, image analysis and tissue preparation are summarized in the Supplemental Data.