Chronic diastolic stretch unmasks conduction defects in an in vitro model of arrhythmogenic cardiomyopathy

Abstract

We seek to elucidate the precise nature of mechanical loading that precipitates conduction deficits in a concealed-phase model of arrhythmogenic cardiomyopathy (ACM). ACM is a progressive disorder often resulting from mutations in desmosomal proteins. Exercise has been shown to worsen disease progression and unmask arrhythmia vulnerability, yet the underlying pathomechanisms may depend on the type and intensity of exercise. Because exercise causes myriad changes to multiple interdependent hemodynamic parameters, it is difficult to isolate its effects to specific changes in mechanical load. Here, we use engineered heart tissues (EHTs) with induced pluripotent stem cell (iPSC)-derived cardiomyocytes expressing R451G desmoplakin, an ACM-linked mutation, which results in a functionally null model of desmoplakin (DSP). We also use a novel bioreactor to independently perturb tissue strain at different time points during the cardiac cycle. We culture EHTs under three strain regimes: normal physiological shortening; increased diastolic stretch, simulating high preload; and isometric culture, simulating high afterload. DSP^R451G EHTs that have been cultured isometrically undergo adaptation, with no change in action potential parameters, conduction velocity, or contractile function, a phenotype confirmed by global proteomic analysis. However, when DSP^R451G EHTs are subjected to increased diastolic stretch, they exhibit concomitant reductions in conduction velocity and the expression of connexin-43 (Cx43). These effects are rescued by inhibition of both lysosome activity and ERK signaling. Our results indicate that the response of DSP^R451G EHTs to mechanical stimuli depends on the strain and the timing of the applied stimulus, with increased diastolic stretch unmasking conduction deficits in a concealed-phase model of ACM.

NEW & NOTEWORTHY We uncover a novel mechanotransductive mechanism that unmasks conduction defects in arrhythmogenic cardiomyopathy (ACM). Using a novel dynamic bioreactor and human-induced pluripotent stem cell (hiPSC)-derived engineered heart tissue (EHT), we apply strain at distinct intervals during the cardiac cycle. Imposing mechanical strain during diastole selectively slows conduction velocity. To our knowledge, this is the first time that the effect of mechanical loading during a specific part of the cardiac cycle has been shown to cause conduction deficits in ACM.

INTRODUCTION

Arrhythmogenic cardiomyopathy (ACM) is a leading cause of sudden cardiac death in young athletes, with high prevalence in specific geographical regions, such as certain areas of Italy (1, 2). Such a distribution suggests heritability, but family analysis reveals incomplete penetrance and many complex influencing factors. A majority of disease-associated mutations are found in genes encoding desmosomal proteins (3). Desmosomes are intercellular junctions that form the intercalated disk along with adherens junctions and gap junctions, providing adhesion between neighboring cardiomyocytes, rapid intercellular communication, and anchoring to the cytoskeleton (4). Desmosomes link intermediate filaments, consisting mainly of desmin, to the intercalated disks (5). Together, these connections create a unified network that spans the cardiomyocyte and provides mechanical strength and integrity (4, 6). Besides their structural role, desmosomes also act as hubs of cellular signaling, influencing pathways such as p53 and Wnt (4). Because demosomal proteins must withstand high loads that are continuously generated by contracting sarcomeres, mutations were thought to weaken these junctions, causing their failure under conditions of high mechanical loading. Subsequent decoupling of neighboring cardiomyocytes could potentially lead to cellular apoptosis and replacement of myocardium with fibrosis and fatty infiltration, creating an arrhythmogenic substrate (7–10). However, further investigations identified junctional abnormalities and conduction delays in the absence of macroscopic pathological remodeling suggesting a subcellular or molecular origin (10–12). This is consistent with the so-called concealed phase of ACM, in which arrhythmias and sudden death occur in the absence of overt structural or mechanical remodeling.

Variable penetrance of ACM implicates exogenous stressors as key factors that affect disease progression and/or unmask pathogenicity as is the case with sudden and unexpected death due to the abrupt emergence of ventricular arrhythmias (13). In particular, intense exercise has been associated with accelerated disease progression and worse outcomes, including the incidence of sudden cardiac death (14–16). In a heterozygous plakoglobin-deficient mouse model, exercise provoked ventricular arrhythmias and slowed conduction in the right ventricle in the absence of any remodeling or fibrofatty replacement of the myocardium (17). In the same model, reduction in ventricular load apparently rescued these abnormalities. Levels of the main ventricular gap junction protein connexin-43 (Cx43) were reduced in exercised heterozygous mutant mice, but were restored to control levels with reductions in volume (18). The effects of mechanical loading can also be seen in a separate model of ACM secondary to adenovirally mediated expression of mutant plakoglobin (2057del 2), in which 4 h of uniaxial stretch was sufficient to increase apoptosis and caspase-3 activity (19). Taken together, these results demonstrate that mechanical loading can directly affect disease progression.

We hypothesized that different forms of mechanical loading applied to desmoplakin (DSP)-mutant versus wild-type (WT) cardiomyocytes would unmask differential effects on electrical function. Specifically, we surmised that a tonic increase in diastolic stretch would only disrupt desmosomes that are already weakened by mutant, but not WT, desmoplakin. We also hypothesized that under isometric conditions, mutant desmosomes would be spared because contractile forces are shared across redundant cytoskeletal elements, despite a lack of physiological shortening. To test our hypotheses, we used a dynamic bioreactor system that allowed us to apply precise mechanical loads chronically to engineered heart tissues (EHTs) in culture (20). Using an induced pluripotent stem cell cardiomyocyte (iPSC-CM) line expressing a previously identified ACM-linked DSP mutation (21), we compared the effects of increased diastolic stretch (to simulate preload) versus isometric culture (to simulate increased systolic resistance, or afterload). Our results suggest that only diastolic stretch leads to a loss of gap junction protein expression and a decrease in conduction velocity. Remarkably, culture of EHTs under isometric conditions, relative to physiological shortening, does not cause functional perturbations or significant differences in pathway regulation of desmosomal, sarcomeric, structural, or ion channel pathways at the proteomic level. The diastolic stretch-specific defects arise from lysosome-mediated degradation of Cx43, which is partially mediated through the stretch-sensitive ERK pathway. Interestingly, the stress caused by isometric culture and its lack of shortening do not cause similar conduction defects. Similar observations have been made in recent papers demonstrating the necessity of “dynamic” culture to expose disease in other models of ACM (22). However, this work, to our knowledge, is the first to specifically implicate diastolic mechanical loading on an electrophysiological phenotype in a human iPSC-CM (hiPSC-CM) model of arrhythmogenic cardiomyopathy.

MATERIALS AND METHODS

Generation of Human-Induced Pluripotent-Derived Cardiomyocytes

Human-induced pluripotent stem cells were purchased from Coriell Institute for Medical Research (GM23338). Genetic modification was performed as previously described to generate an isogenic line homozygous for the R451G mutation in desmoplakin (21). Undifferentiated iPSCs were maintained in mTeSR (StemCell Technologies) on Matrigel-coated plates (Corning). Cardiomyocyte differentiation was performed as previously described using biphasic WNT signaling (23) with beating cardiomyocytes observable by days 14–18 of differentiation. All cell cultures were routinely tested for mycoplasma by PCR and iPSC cells were used no more than 10–12 passages after thawing.

Fabrication of Engineered Heart Tissues

Decellularized porcine myocardium, prepared according to previous protocols, was used as an extracellular matrix (ECM) substrate for seeding of human iPSC-CMs (23). hiPSC-CMs were dissociated to single cells with TryplE (Thermo Fisher) before being enriched with human adult cardiac fibroblasts at a ratio of 9:1. Cells were seeded at a density of 1.5 million cells per scaffold. EHTs were then cultured isometrically in RPMI + B27 until mechanical loading and testing.

A Dynamic Bioreactor for Chronic Loading in Culture

Dynamic loading was achieved using a preexisting bioreactor system (20). A three-dimensional (3-D)-printed holder with a flexible stainless-steel spring arm allowed EHTs to undergo length changes during culture (Fig. 1A). A voice coil actuator (BEI), controlled using a custom-programmed microcontroller (MSP430, Texas Instruments), allowed EHTs to chronically undergo preset length transients. The microcontroller also facilitated the synchronization of electrical stimuli with the imposed mechanical loading. EHTs were paced at a frequency of 1.5 Hz and 10 V while undergoing defined shortening/stretching protocols, allowing for independent modulation of strain through the cardiac cycle. EHTs were matured under isometric conditions for 2 wk, then subjected to electromechanical loading for 24 h in the dynamic bioreactor.

Figure 1.

Engineered heart tissues (EHTs) are constructed from human-induced pluripotent stem cell cardiomyocytes (iPSC-CMs) seeded on decellularized porcine myocardium. A: EHTs are made using iPSC-CMs expressing wild-type desmoplakin (DSP) or R451G desmoplakin. After seeding onto decellularized porcine myocardium, EHTs are cultured under chronic dynamic culture which results in EHTs experiencing varying lengths and loads during the cardiac cycle. B: immunoblots for desmoplakin reveal DSP^R451G results in significantly reduced levels of desmoplakin in EHTs. Student’s t test; error bars means (SD); control EHT, n = 5; DSP^R451G EHT, n = 3.

Functional Characterization of Engineered Heart Tissues

After 24 h of culture under various training regimes, EHTs were unloaded from the dynamic bioreactor onto a custom muscle testing rig for functional characterization. Testing was performed in Tyrode’s solution, containing (in mM) 140 NaCl, 5.4 KCl, 1 MgCl₂, 25 HEPES, 10 glucose, and 1.8 CaCl₂ (pH 7.3). EHT length was controlled by micromanipulators (Siskiyou) and force twitches were collected using a force transducer (WPI) attached to one side of the EHT. Contraction characteristics including time to peak force, maximum force achieved, and time from peak force to half-maximal relation were calculated from twitch recordings. All twitch recordings were measured at culture length (10 mm). Pacing was applied 100 ms after the beginning of the recording, a delay of 10% of the basic cycle length of 1 s (1 Hz), and kinetic parameters were measured relative to the stimulus. Action potentials were collected using a photomultiplier tube after staining with 10 μM di-4-ANEPPS (Thermo Fisher) and 17.1 μM blebbistatin to prevent motion artifacts. Action potential characteristics including maximum depolarization rate, time to peak voltage, and action potential duration were calculated from Savitzky-Golay filtered transients. All data processing was performed using Matlab 2016 (The MathWorks, Natick, MA). Conduction velocity was calculated by measuring the activation time (time at maximum action potential upstroke) as the EHT was point stimulated using a bipolar electrode. The point electrode was attached to a computer-controlled stage and EHTs were stimulated at various points along the length of the tissue. Conduction velocity was defined as the slope of the line between local activation time and distance from the point stimulus.

Western Blot Analysis

Immediately after functional characterization, EHTs were snap-frozen on crushed dry ice. EHTs were then thawed and homogenized in RIPA buffer. Protein content was determined by bicinchoninic acid (BCA) assay (Sigma) and then analyzed via SDS-PAGE and Western blot analysis. Fluorescent signal was quantified using the LICOR Odyssey system, normalized to total protein (Li-Cor), and calculated as fold change compared with control EHTs. For assessment of protein phosphorylation, EHTs were immediately frozen after conditioning and did not undergo functional testing.

Confocal Microscopy

After functional characterization, EHTs were loaded back into isometric frames and fixed for 20 min in 4% paraformaldehyde. EHTs were permeabilized for 10 min in 0.1% Triton-X100 and blocked in 5% normal goat serum. EHTs were incubated under gentle agitation with primary antibodies overnight at 4°C. After washing in PBS + 0.1% Tween-20, EHTs were incubated with secondary antibodies for 1 h at room temperature before being mounted in Fluoroshield + DAPI (Thermo Fisher). Images were collected at ×63 with a Zeiss 880 confocal fluorescent microscope.

Antibodies

Antibodies used included N-cadherin (1:500, 610920; IF), connexin43 (1:500, ab11370; IF), (1:500, c6219; WB), desmoplakin (1:1,000, 25318-1-AP; WB), ERK (1:500, 9107S; WB), and pERK (1:500, 9101S; WB).

Statistical Analyses

Reported data represent means (SD). Comparisons of two groups were made using an unpaired Student’s t test, whereas comparisons of more than two groups were made with two-way analysis of variance (ANOVA). P values of <0.05 were considered significant. Outliers were excluded using Grubb’s method (α = 0.05). All statistical calculations were made with Prism 9 (GraphPad; Boston, MA).

Label-Free Proteomic Quantitation

Proteomic data were collected at the Yale University School of Medicine Keck MS and Proteomics Resources. Briefly, EHT lysates in RIPA buffer were created as described earlier, precipitated with methanol and chloroform, and subjected to tryptic cleavage. Peptides were run on an LTQ Orbitrap XL mass spectrometer. Data analysis was performed in R and pathway analyses executed with String version 12.0.

RESULTS

EHTs Expressing DSP^R451G Have Greatly Reduced Levels of Desmoplakin

Engineered heart tissues were constructed using a previously characterized cell line homozygous for a missense variant in the gene encoding for desmoplakin (NST00000379802.8; c.1596CGT>GGT; Chr. 6:7568521, GRCh38.p12) (21) (Fig. 1A). Our previous work identified the primary effects of this mutation as increased susceptibility of mutant desmoplakin to calpain-mediated degradation (21). In confirmation of these previous results, the homozygous EHTs expressing mutant desmoplakin (DSP^R451G) that we used in the present study exhibited marked depletion in desmoplakin expression compared with isogenic control EHTs expressing WT desmoplakin (7.7 ± 0.7 vs. 100 ± 18%; P < 0.01, Fig. 1B).

Isometric Culture, Compared with Physiological Shortening, Does Not Induce Contractile or Electrical Defects in EHTs Expressing Mutant Desmoplakin

In our model system, we simulated physiologic conditions by imposing dynamic shortening on EHTs during contraction and simulated a “high afterload” condition by culturing EHTs isometrically (Fig. 2A). All EHTs were cultured isometrically for 2 wk after cardiomyocyte seeding for baseline maturation. Then, to demonstrate the effects of isometric versus physiologic loading, EHTs were subjected to 24 h of mechanical conditioning in a dynamic bioreactor before action potential and force transient measurements (Fig. 2, A–C). All tissues were stimulated at time 0, plus a 10% delay relative to the basic cycle length (i.e., 100 ms for 1-Hz pacing); force measurements were taken at culture length (e.g., 0% stretch), which is the same for every tissue regardless of the mechanical conditioning that was applied. The measured twitches confirm that isometric loading increased peak force (Fig. 2A). Independent of loading conditions, DSP^R451G EHTs had greater upstroke velocity (63.8 vs. 51.8 s⁻¹, P < 0.05) and decreased time to peak depolarization (0.050 vs. 0.068 s, P < 0.05) compared with WT EHTs (Fig. 2, B and D). The application of isometric culture did not significantly alter action potential morphology for either WT or DSP^R451G EHTs (Fig. 2B). Both WT and DSP^R451G EHTs generated similar peak twitch force, contractile kinetics (time to peak force), and relaxation kinetics (time from peak force to 50%) regardless of loading condition (Fig. 2, C and E). When linear conduction velocity was measured in the longitudinal direction of the EHT (Fig. 2F), there was no significant decrease in conduction velocity because of the presence of mutant desmoplakin or loading condition (Fig. 2F). A similar baseline conduction velocity between different WT and DSP^R451G EHTs recapitulates our previous findings (21). For all comparisons of mechanical and electrical properties of EHTs, two-way ANOVA was used to identify the individual effects of genotype and mechanical load, whereas genotype-specific effects of mechanical load were evaluated using the interaction P value. These data support our hypothesis that isometric culture does not affect cardiomyocyte mechanics or action potentials, even in the setting of mutated desmosomes. As a result, mechanical strain that coincides with contraction, as in isometric culture, may be borne by nondemosomal structures within the cytoskeleton.

Figure 2.

Isometric culture does not provoke genotype-specific changes in engineered heart tissue (EHT) electrical or contractile properties. A: representative length transients and twitches experienced by EHTs in culture under control conditions (left) or isometric conditions (right). B: representative normalized action potentials from EHTs with/without DSP^R451G and cultured under either loading condition. Action potentials were normalized to maximum signal strength. C: representative twitches from EHTs with/without DSP^R451G and cultured under either loading condition. D: isometric culture, relative to physiological shortening, did not induce genotype-specific changes in the maximum rate of depolarization (dV/dt_max), time to maximum action potential, or action potential duration (APD₉₀). DSP^R451G EHTs had faster depolarization kinetics irrespective of loading condition (*P < 0.05). E: no contractile differences were observed as a result of culture condition or expressed desmoplakin (DSP). F: conduction velocity was calculated by measuring a fixed spot on the EHT and varying the position of a point stimulus. Conduction velocity was calculated as the slope between the position of the point stimulus and the time it took the electrical wave front to reach the measurement point. Neither loading conditions nor genotype affected conduction velocity. All measurements: wild-type (WT) EHTs under physiological shortening, n = 5; WT EHTs under isometric culture, n = 6; DSP^R451G EHTs under physiological shortening, n = 8; DSP^R451G EHTs under isometric culture, n = 7. Data were analyzed using two-way ANOVA and each point represents 1 EHT. EHTs were stimulated at time 100 ms. Error bars represent means (SD).

Increased Diastolic Stretch Results in Prolonged Action Potentials

We next investigated the effects of diastolic stretch on DSP^R451G EHTs. As previously mentioned, EHTs were grown isometrically for 2 wk, then subjected to 24 h of mechanical conditioning. Here, we compared physiological shortening (Fig. 2A) to diastolic stretch, in which the EHT was allowed to shorten but was also stretched by an additional 10% during relaxation. In other words, mechanical stretch was imposed after contraction and before the next pacing stimulus, to simulate high preload. Under these conditions, EHTs exhibited increased action potential duration irrespective of genotype (APD₉₀, 350 vs. 320 ms, P < 0.05). The modest delay in action potential repolarization as reflected by prolonged APD₉₀ occurred in the absence of changes to the action potential upstroke (Fig. 3, B and D). Although DSP^R451G EHTs reached significantly higher peak forces compared with control EHTs (695.5 vs. 444.0 μN, P < 0.001), this occurrence is likely due to batch-to-batch variation in iPSC-CM differentiation efficiency as it was not accompanied by changes in twitch kinetics and was not observed in other experiments (Fig. 2E). Moreover, there was no interaction between genotype and mechanical load on measured contractile kinetics, suggesting that the application of diastolic stretch does not differentially affect force production in a WT or desmosome-null EHT (Fig. 3, C and E).

Figure 3.

Diastolic stretch provokes reduced conduction velocity in engineered heart tissues (EHTs) expressing R451G desmoplakin (DSP). A: representative length transients and twitches experienced by EHTs in culture under control shortening conditions (left) or diastolic stretch (right). B: representative normalized action potentials from EHTs with/without DSP^R451G and cultured under either loading condition. Action potentials were normalized to maximum signal strength. C: representative twitches from EHTs with/without DSP^R451G and cultured under either loading condition. D: diastolic stretch did not induce genotype-specific changes to action potential shape. EHTs from either genotype exposed to diastolic stretch had elongated action potentials. *P < 0.05; wild-type (WT) EHTs under control loading, n = 13; WT EHTs under diastolic stretch, n = 12; DSP^R451G EHTs under control loading, n = 11; DSP^R451G EHTs under diastolic stretch, n = 12. E: no genotype-specific changes to contraction were observed as a result of increased diastolic stretch. DSP^R451G EHTs from either loading condition generated stronger twitches than control EHTs. As this was not observed in previous or subsequent experiments it may represent batch-batch variability. *P < 0.05; WT EHTs under control loading, n = 13; WT EHTs under diastolic stretch, n = 12; DSP^R451G EHTs under control loading, n = 10; DSP^R451G EHTs under diastolic stretch, n = 12. F: significant interaction between mechanical loading and genotype in conduction velocity was observed. *P < 0.05, WT EHTs under control loading, n = 13; WT EHTs with diastolic stretch, n = 11; DSP^R451G EHTs under control loading, n = 11; DSP^R451G EHTs with diastolic stretch, n = 12. G: results from a repeated experiment demonstrate conduction velocity is significantly reduced in DSP^R451G EHTs exposed to stretch compared with WT EHTs (WT EHTs under diastolic stretch, n = 4; DSP^R451G EHTs under diastolic stretch, n = 7). Data were analyzed using two-way ANOVA and each point represents 1 EHT. EHTs were stimulated at time 100 ms. All error bars indicate means (SD).

DSP^R451G EHTs Have Reduced Conduction Velocity after Experiencing Diastolic Stretch

Although conditioning EHTs with diastolic stretch failed to provoke genotype-dependent effects on twitch force or action potential morphology, we did observe a significant interaction between genotype and loading on conduction velocity (P < 0.05) (Fig. 3F). Diastolic stretch provoked a trend toward increased conduction velocity in WT EHTs but reduced conduction velocity in DSP^R451G EHTs. We repeated this experiment using a two-group design, focusing just on WT and mutant EHTs undergoing diastolic stretch conditioning, to confirm the reproducibility of these results and the genotype-specific effect of diastolic stretch. Action potential propagation was significantly slower in DSP^R451G EHTs compared with control EHTs (10.5 ± 1.1 vs. 17.9 ± 1.7 cm/s; P < 0.01) (Fig. 3G). These results indicate that in the setting of vulnerable desmosomes secondary to loss of DSP expression, conduction deficits are unmasked when mechanical strain is applied to tissues during diastole.

Unbiased Proteomic Screen of DSP^R451G EHTs

To determine whether changes in applied mechanical load, such as isometric culture or diastolic stretch, affect DSP-mutant tissues at the protein level, we performed label-free quantitation of lysates from DSP^R451G tissues. Globally, there were relatively few differentially expressed proteins across the conditions (Fig. 4B), and gene ontology analysis for enriched biological pathways did not reveal sarcomeric, cytoskeletal, desmosomal, or ion channel differences among the loading conditions (for proteomic data see Supplemental Figs. S1–S3; all Supplemental figures are available at https://doi.org/10.6084/m9.figshare.24250345). Of note, these findings are consistent with the functional phenotypes described in Figs. 2 and 3, wherein changes in mechanical loading do not affect contractile properties or most electrical properties. Moreover, even a targeted comparison of selected desmosomal proteins (DSP, PKP2, JUP, CTNNB1, CTNND1, and DSG2), compared using t tests with correction for multiple comparisons (false discovery rate cutoff of q = 1%), did not reveal major changes in desmosomal protein expression as a function of applied mechanical load (Fig. 4A). However, given that our mechanical perturbation of mutant EHTs was only 24 h in this study, it is not surprising that the proteomic changes were modest. Moreover, a lack of overt changes in disease-related pathways may suggest that our model system captures the concealed phase of arrhythmogenic cardiomyopathy, before large structural and functional perturbations occur.

Figure 4.

Summary of proteomic data obtained from label-free quantitation of desmoplakin (DSP)-mutant engineered heart tissues (EHTs) under different loading conditions. For proteomic runs, n = 3 for each loading condition [physiological shortening, isometric culture (high afterload), and diastolic stretch (high preload)]. A: comparison of normalized log abundances of a priori selected desmosomal proteins (DSP, desmoplakin; PKP2, plakophilin; JUP, plakoglobin; CTNNB1, beta-catenin; CTNND1, p120-catenin; DSG2, desmoglein). No difference in abundance was found among pairwise comparisons with physiological culture as the control group [multiple t tests with Benjamini–Hochberg correction for multiple comparisons; cutoff, q = 0.01. Error bars represent means (SD)]. B: Venn diagram of differentially expressed proteins between three mechanical loading conditions. C: gene-ontology analysis using STRING v.12.0 of top five enriched pathways between loading conditions, with physiological shortening as the control in each comparison. For the top comparison between isometric culture and physiological shortening, the top four and top sixteenth entry were tabulated, because the GO terms for Nos. 5–15 were essentially identical.

Nevertheless, a consistent electrophysiological finding in our study was a selective decrease in conduction velocity in DSP-mutant tissues cultured under diastolic-stretch conditions (Fig. 2F vs. Fig. 3, F and G). Our proteomic screen did not reveal peptide coverage for gap junction proteins, such as Cx43, so we sought to characterize differences in this protein using Western blot analysis to elucidate a possible mechanism by which differential mechanical loading results in a conduction velocity deficit.

Reductions in Conduction Velocity Are Accompanied by Changes in the Levels of Connexin43

Previous work implicates slowed conduction in the concealed phase of ACM on loss of gap junction protein connexin-43 (Cx43) (24). Therefore, we investigated whether the slowed electrical propagation observed in our system could be attributed to reduced Cx43 levels. Western blotting did not show significant loss of total Cx43 when EHTs were subjected to isometric culture relative to physiological shortening (Fig. 5A), corresponding to maintained conduction velocity observed in that loading regime (Fig. 2F; uncropped blots in Supplemental Figs. S1–S6). However, there was a significant interaction between genotype and mechanical loading condition on Cx43 levels (P < 0.01), with diastolic stretch leading to a significant drop in Cx43 among DSP^R451G EHTs (Fig. 5B). In addition, stretched WT EHTs trended toward increased Cx43 compared with unstretched WT EHTs (1.33 vs. 1, SE of difference = 0.18). A separate set of samples specifically comparing WT and DSP^R451G EHTs demonstrated a significant reduction of Cx43 levels in stretched DSP^R451G EHTs compared with stretched control EHTs (0.60 ± 0.08 vs. 1 ± 0.12, P < 0.05) (Fig. 5C).

Figure 5.

Connexin-43 (Cx43) protein levels are reduced in DSP^R451G engineered heart tissues (EHTs) conditioned under increased diastolic stretch. A: isometric culture, simulating increased afterload relative to physiological shortening, did not affect Cx43 protein levels in wild-type (WT) EHTs or DSP^R451G EHTs. B: diastolic stretch and DSP^R451G had significant interaction on the levels of Cx43. WT EHTs under control loading, n = 7; WT EHTs under diastolic stretch, n = 5, DSP^R451G EHTs under control loading, n = 7; DSP^R451G EHTs under diastolic stretch n = 6. C: separate set of samples confirmed the effects of DSP^R451G in the different responses to diastolic stretch, as levels of Cx43 were significantly lower in stretched DSP^R451G EHTs compared with stretched WT EHTs (WT EHTs under stretch n = 4, DSP^R451G EHTs under diastolic stretch n = 7). Data in A and B were analyzed using two-way ANOVA; data in C were analyzed by Student’s t test; each point represents 1 EHT. Error bars indicate means (SD).

Lysosome Inhibition Prevents Stretch-Induced Decreases in Conduction Velocity

Cx43 turnover in the heart is mediated by lysosome degradation (25, 26). Knockout of desmoplakin in neonatal rat ventricular cardiomyocytes (NRVMs) results in a loss of Cx43 at the intercalated disks, hypothesized to be due to increased degradation of Cx43, reversible with lysosome inhibition (27). Although we did not observe loss of Cx43 in unstressed DSP^R451G EHTs, we hypothesized that a similar mechanism may be responsible for Cx43 loss associated with increased diastolic stretch. To test this, we treated DSP^R451G EHTs with 20 μM chloroquine, a lysosome inhibitor, for 24 h during diastolic stretch mechanical conditioning. Broad lysosome inhibition appeared to have toxic effects, as chloroquine decreased conduction velocity in nonstretched DSP^R451G EHTs (11.8 ± 1.1 vs. 29.7 ± 9.0 cm/s). However, there was a significant interaction between chloroquine treatment and diastolic stretch on conduction velocity (Fig. 6A). In contrast to unstretched DSP^R451G EHTs, chloroquine treatment increased conduction velocity in prestretched DSP^R451G EHTs (21.6 ± 5.3 vs. 13.2 ± 1.6 cm/s; ANOVA interaction, P < 0.05). These changes occurred in the absence of changes in action potential morphology between groups (Fig. 6A). Confocal microscopy revealed a reduction of Cx43 at the intercalated disk (marked by N-cadherin) in untreated stretched DSP^R451G EHTs, matching the observed effects on conduction velocity (Fig. 6B). Chloroquine appeared to increase cytoplasmic Cx43 in nonstretched DSP^R451G EHTs, and potentially to an even greater extent in Cx43 in stretched DSP^R451G EHTs. Despite the fact that chloroquine promotes cytoplasmic localization of Cx43, and contributes to cellular toxicity, it is possible that enough Cx43 is still localized to intercalated disks to promote a partial rescue in conduction velocity. Overall, these results suggest that mutant desmoplakin sensitizes gap junctions to diastolic stretch-induced lysosome degradation.

Figure 6.

Lysosome inhibition mitigates stretch-induced reductions in conduction velocity. A: 20 μM chloroquine treatment increases conduction velocity in DSP^R451G engineered heart tissues (EHTs) condition under high preload over a background effect of reduced conduction velocity due to off-target effects of lysosome inhibition. *P < 0.05, error bars represent means (SD); unstretched DSP^R451G EHTs, n = 3; unstretched DSP^R451G EHTs with chloroquine, n = 3; stretched DSP^R451G EHTs n = 3, stretched DSP^R451G EHTs with chloroquine, n = 4. Chloroquine treatment had no significant effects on action potentials. Data in A were analyzed using two-way ANOVA. B: representative confocal microscopy images. Connexin-43 (Cx43) is detected in the red channel and N-cadherin is detected in the green channel. Arrows indicate reduced Cx43 signal at the intercalated discs in stretched DSP^R451G EHTs. Chloroquine treatment appears to increase Cx43 at the intercalated disc, as well as in the cytoplasm in stretched DSP^R451G EHTs. Images were taken at ×63 magnification; scale bars represent 20 μm.

Inhibition of ERK Activation Partially Blunts the Effects of Stretch on DSP^R451G EHTs

Cx43 contains multiple phosphorylation sites that regulate trafficking and turnover, some of which may be phosphorylated by the stretch-sensitive ERK pathway to mediate lysosome-mediated degradation (28). Previous work also suggests that loss of desmoplakin is associated with elevated ERK activation (27, 29). Although we did not observe increased ERK activation at baseline, ERK was slightly activated by diastolic stretch across both genotypes compared with nonstretched EHTs of each genotype (1.17 vs. 1.01, P < 0.05) (Fig. 7A). This effect was blocked by 50 µM PD98059 during stretch conditioning (1.33 vs. 0.83, P < 0.05) (Fig. 7B). Although stretched DSP^R451G EHTs once again showed significantly lower levels of Cx43 compared with nonstretched DSP^R451G (0.72 vs. 1, P < 0.05), this difference was abrogated by PD98059 (0.82 vs. 1, N.S.) (Fig. 7C).

Figure 7.

ERK inhibition partially restores connexin-43 (Cx43) levels. A: levels of p-ERK normalized to t-ERK. Stretch induces increased levels of p-ERK equally in DSP^R451G engineered heart tissues (EHTs) and wild-type (WT) EHTs. *P < 0.05, n = 6 all groups. B: 50 μM PD98059 significantly reduced stretch-induced ERK activation (*P < 0.05; n = 9 all groups). C: significant loss of Cx43 associated with high preloads was mitigated through PD89059 treatment (*P < 0.05; n = 9 all groups). Cx43 abundance in PD98059 treated, stretched, DSP^R451G EHTs was not significantly different from Cx43 level in either nontreated group. Comparisons in A and B were made with two-way ANOVA, whereas comparisons in C were made with a one-way ANOVA followed by Sidak’s comparison. Error bars represent means (SD).

DISCUSSION

In this work, we present a novel mechanoelectrical feedback mechanism underlying defective conduction in an iPSC-derived model of arrhythmogenic cardiomyopathy. Specifically, we demonstrate that application of mechanical stretch during myocyte relaxation, or diastolic stretch, causes decreased Cx43 abundance and slowed conduction velocity in DSP^R451G EHTs. These deficits were partially reversible with broad lysosome inhibition (Fig. 6). Notably, tissues that were subjected to other mechanical conditions, including normal physiological shortening or isometric culture, did not exhibit altered conduction or Cx43 expression (Fig. 2). These contrasting observations highlight the importance of considering the nature and timing of the mechanical load stressor when considering arrhythmic vulnerability in DSP-linked ACM. Specifically, isometric culture enhances actomyosin-based force generation (Fig. 2A), which results in greater force transmission to adherens junctions (30). Conversely, diastolic stretch, which by definition is applied when sarcomeres are relaxed, likely relies on force transmission across other structural elements such as desmosomes. In DSP^R451G tissues, whose desmosomes are already vulnerable due to desmoplakin haploinsufficiency, diastolic stretch likely triggers further destabilization of cellular adhesion, degradation of Cx43 at the intercalated disk, and depressed conduction velocity.

Our homozygous DSP^R451G ACM model effectively acts as a desmoplakin depletion model, consistent with previous work implicating this mutation in increased susceptibility to calpain-mediated degradation (21). Homozygous DSP deletion is embryonically lethal, and cardiac-specific deletion impairs cardiac morphogenesis, with high levels of embryonic lethality (31, 32). Heterozygous cardiac-specific deletion of DSP in mice causes fibro-fatty replacement of cardiomyocytes, biventricular failure, and early death (33). In contrast with these findings, and despite dramatic loss of desmoplakin protein, our model maintains contractile function without gross morphological changes (21), only developing conduction defects in the presence of mechanical stretch applied during diastole. As a result, our model resembles in some way the “concealed” phase of ACM, wherein there are no overt changes in tissue structure or mechanical function, but rather, primary effects on myocardial conduction vis-à-vis Cx43 downregulation that may form an early arrhythmogenic substrate.

Our proteomic findings, which, to our knowledge, are the first to examine the effects of mechanical loading on disease progression in a DSP-mutant model, are relatively consistent with the notion of a “concealed” phase of ACM. First, relative to our protein coverage (3,549 proteins in the physiological-shortening group; 3,366 in the diastolic stretch group; and 3484 in the isometric-culture group), the number of differentially expressed proteins is relatively low (Fig. 4B). Next, gene-ontology analysis of biological processes in DSP-mutant tissues reveals relatively few differences between physiological loading and diastolic stretch, with the most-highly enriched pathways including cell cycle processes and nucleotide metabolism (Fig. 4C). Notably, in our comparison of isometric culture and physiological shortening, many pathways regarding metabolism are upregulated in the physiological-stretch group (gene ontology terms “ATP biosynthetic process,” false discovery rate (FDR) 0.00044; ATP metabolic process, FDR 4.15 × 10⁻⁵; oxidative phosphorylation, FDR 0.00012), which recapitulates our previous work implicating the importance of contractile work on mitochondrial function (20). However, no pathways regarding sarcomeric, desmosomal, ion channel, or gap junction expression, were different with mechanical loading. It is possible that either 1) 24 h of mechanical conditioning is not long enough to elicit large-scale proteomic changes or 2) DSP-mutant tissues must be further stressed using interventions like tachypacing in combination with diastolic stretch to see a larger phenotype. We were surprised to not find coverage of Cx43 and Nav1.5 in our proteomic data, but we believe that the hydrophobicity of the membrane-spanning regions of these proteins, and relatively small amount of lysate (50–75 μg) from hiPSC-EHTs may have been a technical limitation. Regardless, when taking a more granular approach using Western blotting for total Cx43 levels, the effect of differential mechanical loading is reproducible, with conduction deficits and reduction of Cx43 only seen in the diastolic-stretch group. Of note, despite extensive trials of commercial phospho-specific antibodies for Cx43 (especially phospho-Ser368 and phospho-Ser279/282), we were not able to see signals for phosphorylated versions of this protein.

Our proposed mechanism of DSP loss is broadly consistent with those reported in other studies. In one such study, cardiac-specific DSP deletion caused loss of Cx43 and concomitant dissociation of the desmosome complex (31), although electrophysiological deficits occurred independent of structural changes at the desmosome. In another study, loss of DSP caused lysosome-mediated degradation of Cx43, which was completely reversed by ERK inhibition (27). Our results showed only partial rescue of Cx43 and conduction velocity (CV) with ERK inhibition, suggesting involvement of ERK-independent mechanisms in the targeting of Cx43 for degradation. Studies using iPSC cardiomyocytes derived from patients with ACM (19, 34–38) usually require adipogenic differentiation protocols, such as nonphysiological activation of peroxisome proliferator-activated receptor (PPAR)-α and -γ, to recapitulate aspects of ACM, but our model is notable in that conduction slowing is created solely by modulating diastolic stretch, rather than altering metabolism. Our results are largely consistent with the recent publication by Bliley et al. (22), which also reports the necessity of dynamic loading in vitro to observe ACM disease progression in mutant DSP cardiomyocytes. However, we did not observe the changes in twitch force and kinetics in our WT preparations that they noted, perhaps because of differences in EHT structure and in the nature of dynamic culture loading. The system devised by Bliley et al. compared isometrically cultured EHTs with EHTs tensioned by a polydimethylsiloxane (PDMS) strip under bending load. The result of PDMS loading was a simultaneous increase in preload (stretch) and decrease in afterload of the EHT relative to the isometrically conditioned tissues. In contrast, the mechanically actuated bioreactor used in these studies allowed us to simulate isolated changes in either preload or afterload while maintaining other parameters constant. In addition, the work done by this group examines more chronic changes in disease progression, because dynamic mechanical loading is applied for 2 wk, whereas our studies focus on the acute phase (24 h) to identify the very first perturbations in electrical function. Taking both studies together, the data seem to suggest that dynamic culture is necessary to produce a phenotype, and that diastolic stretch in particular exacerbates the effects of DSP mutations, with concealed electrical phenotypes arising after 24 h in our system, and mechanical deficits arising after 2 wk in the system by Bliley et al.

Our results are also consistent with previous literature on the effect of mechanical loading on Cx43 expression. In particular, cyclic mechanical stretch has been shown to increase Cx43 expression at the intercalated disk (39) in a focal adhesion kinase (FAK) and VEGF-dependent manner (40). Our WT data show a similar relationship, with a trend toward increased Cx43 expression in EHTs cultured under cyclic loading conditions (Fig. 2F, right), regardless of changes to preload or afterload. A separate study that reported decreased Cx43 secondary to mechanical stimulation used an isometric preparation (41), suggesting that cyclic length changes, rather than mechanical load alone, are associated with increased Cx43 levels. Ultimately, our data show that in the setting of functional DSP knockdown secondary to the R451G mutation, Cx43 levels are decreased, rather than increased, in the presence of mechanical stretch, with diastolic stretch particularly exacerbating this observation.

Although our model highlights an important mechanotransductive aspect to ACM conduction deficits, it does not recapitulate the entire mechanism, and we wish to acknowledge some technical limitations. For example, both biopsy samples and experiments in cultured myocytes implicate decreased expression of Nav1.5 channels on arrhythmogenesis in ACM (42, 43). Our action potential recordings bore no indication of changes that would be consistent with this loss of Nav1.5 density. Whether sparing of Na-channel density might be a feature of DSP-null mutations is not known. Regarding our inhibition of Cx43 degradation using chloroquine, the broad effects of this inhibitor may call into question the specificity of our results. However, our findings are consistent with previous use of chloroquine in a DSP-null context (27); chloroquine was shown to potentiate Cx43 expression to a similar extent as leupeptin, E-64, and ALLN in NRVMs (44); and in a model that examined Cx43 in the context of breast cancer cells, chloroquine and leupeptin resulted in similar increases in protein expression (45). More broadly, many factors besides Cx43 expression can account for decreased conduction velocity, such as its phosphorylation state. We were not able to identify S279/S282-phosphorylated Cx43, which corresponds to Cx43 that has been targeted for lysosomal degradation, but would expect that diastolic stretch would increase phosphorylation at this site (28). Beyond Cx43 expression and phosphorylation, in silico models suggest that even a 50% loss of gap junctions should only lower CV by 11% (46, 47). The relative immaturity of our tissues, which have small cardiomyocytes, nascent gap junctions, and correspondingly slower CV compared with adult myocardium, may account for this discrepancy (47). Despite these acknowledged limitations, the data shown here support an intriguing new paradigm that DSP mutation pathogenicity is modulated by the timing, within the cardiac cycle, of applied mechanical loads.

We believe that translation of our findings to an in vivo model would help further elucidate the pathomechanisms of ACM and yield clinical insight. Independently manipulating hemodynamic parameters in a mouse model (48), analogous to our modulation of diastolic stretch, would delineate the nature of mechanical loads that are associated with worse disease progression. This may help clarify unanswered clinical questions about why ACM-related sudden death is unmasked by vigorous exercise. Though exercise physiology is much more complex than we can model in our system, the underlying changes in hemodynamic parameters, like preload and afterload, are analogous to our modulation of tissue strain over the cardiomyocyte contraction-relaxation cycle. If our work can be recapitulated in vivo, it may uncover mechanistic insights that explain how different types of hemodynamic loading differentially affect disease progression. We hope that such mechanistic insights will not only enhance our understanding of ACM and desmosomal diseases but also lead to better-informed clinical decision-making and potential treatments.

DATA AVAILABILITY

Data will be made available upon reasonable request.

SUPPLEMENTAL DATA

Supplemental Figs. S1–S6: https://doi.org/10.6084/m9.figshare.24250345.

GRANTS

This work was supported in part by National Science Foundation Grant 1653160 (to S.G.C.); National Institutes of Health (NIH) Grants R01 HL163092 (to S.G.C and F.G.A), R01 HL149344 (to F.G.A.), and R21 HL165147 (to F.G.A); and NIH Medical Scientist Training Program Grant 5T32GM136651-02 (to I.G.).

DISCLAIMERS

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

DISCLOSURES

S.G.C. has equity ownership in Propria, LLC, which has licensed technology used in the research reported in this publication. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

R.N., I.G., and S.G.C. conceived and designed research; R.N., I.G., and P.S. performed experiments; R.N., I.G., P.S., and S.G.C. analyzed data; R.N., I.G., and S.G.C. interpreted results of experiments; R.N. and I.G. prepared figures; R.N., I.G., and S.G.C. drafted manuscript; R.N., I.G., F.G.A., and S.G.C. edited and revised manuscript; I.G. and S.G.C. approved final version of manuscript.