Familial hypertrophic cardiomyopathy (HCM), a leading cause of sudden cardiac death, is primarily caused by mutations in sarcomeric proteins. The pathogenesis of HCM is complex, with functional changes that span scales, from molecules to tissues. This makes it challenging to deconvolve the biophysical molecular defect that drives the disease pathogenesis from downstream changes in cellular function. In this study, we examine an HCM mutation in troponin T, R92Q, for which several models explaining its effects in disease have been put forward. We demonstrate that the primary molecular insult driving disease pathogenesis is mutation-induced alterations in tropomyosin positioning, which causes increased molecular and cellular force generation during calcium-based activation. Computational modeling shows that the increased cellular force is consistent with the molecular mechanism. These changes in cellular contractility cause downstream alterations in gene expression, calcium handling, and electrophysiology. Taken together, our results demonstrate that molecularly driven changes in mechanical tension drive the early disease pathogenesis of familial HCM, leading to activation of adaptive mechanobiological signaling pathways.
Hypertrophic cardiomyopathy (HCM) is the leading cause of sudden cardiac death in people younger than 30 yr. HCM is characterized by hypertrophy of the left ventricular wall and interventricular septum, myocyte disarray, fibrosis, and diastolic dysfunction. HCM is also associated with marked alterations in cardiomyocyte function, including changes in electrophysiology, contractility, and calcium handling (Harvey and Leinwand, 2011). Large-scale sequencing of families has revealed that HCM is caused by mutations in sarcomeric proteins involved in cardiac contraction, including troponin T (Watkins et al., 1995).
The disease presentation in HCM is quite complex, with functional differences seen at scales ranging from molecules to tissues; however, the molecular triggers that drive the disease pathogenesis are alterations in the abundance, stability, and/or functionality of mutant protein (Greenberg and Tardiff, 2021). This initial trigger activates downstream adaptive and maladaptive processes, some of which can take years to decades to manifest, including ventricular remodeling, and eventually symptomatic disease. Given the inherent complexity of HCM, it has been challenging to link the molecular and cellular phenotypes and to dissect the initial biophysical trigger from secondary adaptive processes.
To better understand the initial molecular insult and its connection to cellular dysfunction in early HCM disease pathogenesis, we examined a pathogenic mutation in troponin T, R92Q (Fig. 1 A), identified in several unrelated families, that causes pronounced ventricular hypertrophy and a relatively high incidence of sudden cardiac death (Watkins et al., 1995). R92Q has been studied in model systems, including feline (Marian et al., 1997) and rat (Rust et al., 1999) cardiomyocytes, rabbit skeletal myofibrils (Morimoto et al., 1998), quail myotubes (Sweeney et al., 1998), and transgenic mice (Tardiff et al., 1999). These studies have reached conflicting conclusions about the effects of the mutation, in part due to phenotypic differences between species. For example, the widely studied transgenic mouse model of R92Q (Tardiff et al., 1999) recapitulates some, but not all, aspects of the human disease phenotype. Elegant experiments by the Tardiff laboratory showed that the disease presentation in mice depends on the myosin heavy chain isoform expressed, with different phenotypes seen when using the faster (MYH6) isoform found in mouse ventricles or the slower (MYH7) isoform found in human ventricles (Ford et al., 2012). This highlights the need to study the mutation in humanized systems.
Troponin T is part of the troponin complex, which, together with tropomyosin, regulates the calcium-dependent interactions between myosin and the thin filament that power muscle contraction. Biochemical (McKillop and Geeves, 1993) and structural (Lehman et al., 1994) measurements demonstrated that tropomyosin can lie in three states along the thin filament (Fig. 1 B). In the absence of calcium, tropomyosin lies in the blocked position and inhibits the binding of force-generating actomyosin cross-bridges. When calcium binds to troponin C, tropomyosin shifts to the closed position. The tropomyosin can then be pushed into the open position by either thermal fluctuations or myosin weak binding. Once weakly associated with the thin filament, myosin isomerizes into the strong binding state, generating force. The amount of force developed will be proportional to the number of strongly bound, force-generating myosin cross-bridges.
Three models have been put forward to describe the initial molecular insult that drives the R92Q disease pathogenesis (Fig. 1 B): (1) R92Q could affect the cycling kinetics of myosins bound to the thin filament (Ford et al., 2012); (2) R92Q could increase the calcium affinity of troponin, leading to altered calcium buffering by myofilaments that directly disrupts calcium homeostasis (Schober et al., 2012; Ferrantini et al., 2017; Robinson et al., 2018); or (3) R92Q could alter the distribution of positions assumed by tropomyosin along the thin filament, changing the number of bound myosin cross-bridges (McConnell et al., 2017). The mechanistic differences among these models have important implications for the design of therapeutic strategies.
Here, we set out to identify the initial molecular insult in troponin T caused by the R92Q mutation and link the molecular defect to observed derangements in cellular function. To do this, we developed a human R92Q model in gene-edited human induced pluripotent stem cell–derived cardiomyocytes (hiPSC-CMs). We show that the initial biophysical insult is altered positioning of tropomyosin along the thin filament, which directly affects cellular tension, leading to secondary adaptive changes in calcium homeostasis, gene expression, and electrophysiology. Our results implicate mechanobiological signaling as a primary driver of HCM disease pathogenesis.
Materials and methods
Protein modification and purification
Cardiac myosin and actin were purified from cryoground porcine ventricles (Pelfreez) as previously described (Clippinger et al., 2019). S1 myosin was prepared by chymotrypsin digestion as previously described (Clippinger et al., 2019). Recombinant human cardiac tropomyosin, troponin I, troponin T, and troponin C were expressed in Escherichia coli and purified from BL21-CodonPlus cells (Agilent) as described previously (Clippinger et al., 2019). Purified tropomyosin was reduced in 50 mM dithiothreitol (DTT) at 56°C for 5 min and ultracentrifuged to remove aggregates immediately before being used in each assay. The R92Q mutation was introduced into troponin T using QuickChange Site-Directed Mutagenesis (Agilent), and the presence of the mutation was verified by sequencing.
For the studies of calcium binding, we used IAANS (6-((4-((2-iodoacetyl)amino)phenyl)amino)-2-naphthalenesulfonic acid)–labeled troponin C (Davis et al., 2007). IAANS was custom synthesized by Toronto Research Chemicals. Troponin CT53C was labeled with fivefold molar excess IAANS dye overnight, and the reaction was quenched with DTT. Excess dye was dialyzed out with four dialysis buffer changes of 1 mM DTT, 0.01% NaN3, 50 µM CaCl2, 1 mM MgCl2, 3 M urea, 1 M KCl, 5 mg/liter tosyl phenylalanyl chloromethyl ketone (TPCK), 5 mg/liter tosyl-L-lysyl-chloromethane hydrochloride (TLCK), and 0.3 mM PMSF (Davis et al., 2007). The IAANS-labeled troponin CT53C was then purified over a MonoQ column and complexed with troponins T and I as done previously (Clippinger et al., 2019).
In vitro motility assays
In vitro motility assays were conducted using thin filaments containing R92Q troponin T as previously described (Clippinger et al., 2019). Data for WT troponin T are from Clippinger et al. (2019). Briefly, enzymatically inactive full-length porcine cardiac myosin was removed by cosedimentation with phalloidin-stabilized F-actin in the presence of ATP. Flow cells were loaded with 1 vol (50 µl) 200 nM myosin, 2 vol 1 mg/ml BSA, 1 vol 1 µM F-actin, 2 vol KMg25 (25 mM KCl, 4 mM MgCl2, 1 mM EGTA, 1 mM DTT, and 60 mM MOPS, pH 7.0) + 1 mM MgATP, 4 vol KMg25, and 1 vol 40 nM rhodamine-phalloidin–labeled thin filaments. After loading 2 vol activation buffer (KMg25 with 4 mM MgATP, 1 mg/ml glucose, 192 U/ml glucose oxidase, 48 µg/ml catalase, 2 µM troponin and tropomyosin, and 0.5% methyl cellulose), flow cells were imaged for 20 frames. Each buffer was balanced to give the desired free calcium, free magnesium, and ionic strength using MaxChelator (Bers et al., 2010). Individual motile filaments were manually tracked using the MTrackJ plugin in Fiji ImageJ (Schindelin et al., 2012), and each point shows the average and standard deviation of the speed from three separate experiments.
Stopped-flow transient kinetic measurement of KB and ADP release
The rate of ADP release from myosin bound to regulated thin filaments (20°C) was measured as previously described. (Clippinger et al., 2019) The average and standard deviation of the rate of at least four experiments was calculated and the P value was derived using a two-tailed Student’s t test.
Fluorescence titrations to measure KW, KT, and n
A SX-20 stopped-flow fluorometer was used for all fluorescence titrations. The values of KW (the equilibrium constant for myosin weak binding), KT (the equilibrium constant between the closed and open states), and n (the cooperativity) were determined for R92Q and WT using fluorescence titrations as previously described (Barrick et al., 2019; Clippinger et al., 2019). The WT data are from Clippinger et al. (2019). Hypothesis testing and uncertainty estimation were done using established techniques (Efron, 1979; Press, 1992; Martin, 2007). Briefly, the data were resampled using a bootstrapping algorithm, resampled data were fit to calculate 95% confidence intervals, and P values were calculated from the overlap regions between the distributions of bootstraps (Efron, 1979; Press, 1992; Martin, 2007; Barrick et al., 2019).
Measurement of calcium binding to troponin C
Computational modeling of sarcomeric contractility
To simulate the effects of the experimentally determined changes in equilibrium constants on force production, we used the computational model developed by Campbell et al. (2010), based on McKillop and Geeves (1993), as we have done previously (Clippinger et al., 2019). Briefly, in this model, nine sarcomeres are simulated, where the equilibrium constants between states and a coupling constant describing cooperativity define the probability of switching between biochemical states. The steady-state force is calculated from the equilibrium distribution of states at a given calcium concentration. Our biochemical experiments demonstrated that the primary change at the molecular scale with the mutation is an increase in KB, such that KB (R92Q) = 2.56 × KB (WT). To simulate the WT, we used the default model parameters. To simulate the mutant, we decreased the reverse rate constant that defines KB, so that KB (R92Q) = 2.56 × KB (WT). To simulate the force per sarcomere in response to a calcium transient for the WT, we used the default calcium transient. To simulate the response of R92Q, we changed KB as described above and we reduced the amplitude of the default calcium transient to 67% of its value to match our measurements in hiPSC-CMs.
Stem cell line derivation
R92Q stem cells were derived and the quality control was performed using procedures described in depth previously (Clippinger et al., 2019). Briefly, the parent human BJ fibroblast stem cell line (ATCC; BJFF.6) was reprogrammed to stem cells by the Genome Engineering and iPSC Center (GEiC) at Washington University in St. Louis. Two independent isogenic stem cell lines with the R92Q hTNNT2 point mutation were also generated at the GEiC using the CRISPR-Cas9 system (Jinek et al., 2012; Fig. S2). The oligo used to generate the guide RNA (gRNA) was 5′-CCTTCTCCATGCGCTTCCGGNGG-3′, and the mutation was introduced by homology-directed repair. This gRNA was selected to minimize off-target effects. The presence of the homozygous mutation was verified by sequencing. Karyotype (G-banding) analysis was performed by Cell Line Genetics (Fig. S2). Mycoplasma testing and immunofluorescence staining for pluripotency markers were performed by the GEiC (Fig. S3).
Stem cell and hiPSC-CMs culture
Stem cell culture and differentiation to hiPSC-CMs were done as previously described (Clippinger et al., 2019). Briefly, stem cells were maintained in feeder-free culture. To differentiate the stem cells to hiPSC-CMs, we used small-molecule manipulation of WNT signaling (Lian et al., 2012; Lian et al., 2013). hiPSC-CMs were enriched using metabolic selection (Sharma et al., 2015). All functional experiments were conducted at least 30 d after the initiation of differentiation. Experiments were conducted using two independently derived cell lines for the R92Q mutant. All experiments were repeated using at least two independent differentiations.
Microcontact patterning of hiPSC-CMs on glass and hydrogels
Fabrication of rectangular (7:1 aspect ratio) polydimethylsiloxane stamps for micropatterning of hiPSC-CMs on both glass and 10-kPa hydrogels was done as previously described (Ribeiro et al., 2015; Clippinger et al., 2019). Cells were patterned onto 10-kPa polyacrylamide hydrogels containing stamped Geltrex (Thermo Fisher Scientific) in rectangular patterns as reported previously (Clippinger et al., 2019; Fig. S3).
Quantification of troponin I isoform expression by Western blotting
Cell lysates from multiple differentiations were extracted from wild type and mutant hiPSC-CMs using protein lysis buffer containing 20 mM HEPES, pH 7.0, 150 mM NaCl, 0.5% 3-(3-cholamidopropyl)diethylammonio-1 propanesulfonate, and one protease inhibitor cocktail tablet (Roche complete mini EDTA-free tablets). 30 µg of each protein lysate was loaded per well of an Any kD Mini-PROTEAN TGX Precast Protein Gel (Bio-Rad), separated by electrophoresis, and transferred to polyvinylidene difluoride membranes (Sigma). Membranes were blocked with 5% milk and incubated overnight with primary antibodies. One membrane was incubated with mouse anti-GAPDH antibody (DSHB; DSHB-hGAPDH-2G7; 3 µg/ml) and mouse anti-slow skeletal troponin I antibody (Santa Cruz Biotechnology; sc-514899; 1:700). The other membrane was incubated with mouse anti-GAPDH antibody (DSHB; DSHB-hGAPDH-2G7; 3 µg/ml) and rabbit anti-cardiac troponin I (Santa Cruz Biotechnology; sc-15368; 1:500). After washing off the primary antibodies, the membranes were incubated with anti-rabbit IgG, HRP-linked antibody (Cell Signaling; 7074S; 1:2,000) and/or anti-mouse IgG, HRP-linked antibody (Cell Signaling; 7076S; 1:2,000). The proteins were detected using the SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific) and imaged on a MyECL Imager (Thermo Fisher Scientific). Blot intensity was quantified by densitometry using ImageJ (Schindelin et al., 2012; Fig. S4). P values were calculated using a Mann–Whitney U test.
Traction force microscopy
Traction force microscopy was conducted on 10-kPa hydrogels with cells patterned on rectangular patterns as previously described (Clippinger et al., 2019) and analyzed using the computational tool developed in Ribeiro et al., 2017. Data were analyzed, and 95% confidence intervals of the mean were calculated as described previously (Barrick et al., 2019; Clippinger et al., 2019). Briefly, the data were resampled using a bootstrapping algorithm to calculate 95% confidence intervals, and P values were calculated from the overlap regions between the distributions of bootstraps (Efron, 1979; Press, 1992; Martin, 2007; Barrick et al., 2019). Videos of contracting cells (Video 1) and sample traces derived from those videos (Fig. S5) are provided. Two independently derived cell lines were examined (Fig. S6).
Measurement of calcium transients in live cells
Live-cell imaging was conducted using the ratiometric fluorescent calcium indicator dye Fura Red AM (Thermo Fisher Scientific). The use of a ratiometric dye is important since the mutation could affect the uptake of dye into the cells, and the ratiometric dye normalizes the calcium-induced changes in fluorescence to the total amount of dye taken up by the cell. hiPSC-CMs were patterned on hydrogels as described above. After 5–7 d on the patterns, the cells were loaded with 10 µM Fura Red AM dye and 0.01% Pluronic F-127 (Invitrogen/Thermo Fisher Scientific) in RPMI-B27 with insulin media for 20 min at room temperature. The cells were washed twice and incubated with Tyrode’s solution (1.8 mM CaCl2, 135 mM NaCl, 4 mM KCl, 1 mM MgCl2, 5 mM glucose, and 10 mM HEPES, pH 7) for 15–20 min at 37°C to allow de-esterification of the dye. Calcium transients were recorded with a Nikon A1Rsi confocal microscope in line scan mode using a 40× objective and the Ex2Em1 microscope setting. Fura Red AM loaded cells were excited at both 405 nm and 488 nm, and the emission fluorescence signal was collected at 595 nm. Line scans were acquired at a sampling rate of 512 pixels × 1.9 ms per line (total 10,000 lines per recording). Each cell was recorded along with a line scan of the background fluorescence outside the cell area.
Analysis of calcium transients
The calcium transient fluorescence counts were converted to ratios using Fiji software (Schindelin et al., 2012). The averaged background fluorescence was subtracted from each recording and a montage was created from the image stacks. The ratio of fluorescence at 405 nm/488 nm was then calculated in Excel. The resulting ratiometric calcium fluorescence traces were then analyzed using a custom MATLAB script to calculate the amplitude of the calcium transient. Traces with fewer than three peaks were not analyzed. Briefly, the data were smoothed over a 100-point sliding window using a Savitsky–Golay filter. The locations of peaks and minima in the fluorescence signal were determined using a peak-finding algorithm. Statistical significance was tested using a two-tailed Student’s t test.
Measurement of the expression of transcripts encoding key calcium-handling proteins
hiPSC-CMs were grown on Matrigel-coated (Corning) 10-kPa PrimeCoat elastic substrate culture dishes (ExCellness Biotech; 35 mm) for 10 d. Total RNA was isolated using RNeasy Mini Kit (Qiagen) with on-column DNase I treatment according to the manufacturer’s instructions. cDNA was generated using iScript Reverse Transcription Supermix (Bio-Rad) according to the manufacturer’s instructions. Quantitative PCR (qPCR) reactions were performed in triplicate using iTaq Universal SYBRGreen Supermix (Bio-Rad) and the ViiA 7 System (Applied Biosystems). Primers for all genes were obtained from IDT PrimeTime qPCR primers. Primer product numbers from IDT are listed in Table S1. Three separate biological samples were evaluated for both WT and R92Q homozygous hiPSC-CMs. The relative levels of mRNA were calculated using the comparative threshold cycle (ΔCt) method (Livak and Schmittgen, 2001). GAPDH and HPRT1 were used as endogenous controls, and Rox dye present in the master mix was used to normalize background fluorescence. ΔCt values are presented in Table S2. The statistical significance of differences in ΔCt values was evaluated using a Mann–Whitney U test.
Cellular electrophysiological studies
Whole-cell current- and voltage-clamp recordings were obtained at room temperature (∼22–24°C) from hiPSC-CMs plated on hydrogel-coated coverslips using a Dagan 3900A (Dagan Corporation) amplifier interfaced to a Digidata 1332A A/D converter (Axon) and the pClamp 10.3 software (Axon). For current-clamp recordings, recording pipettes contained 135 mM KCl, 5 mM K2ATP, 10 mM EGTA, 10 mM HEPES, and 5 mM glucose (pH 7.2; 310 mOsm). The bath solution contained 136 mM NaCl, 4 mM KCl, 2 mM MgCl2, 1 mM CaCl2, 10 mM HEPES, and 10 mM glucose (pH 7.4; 300 mOsm). For recordings of voltage-gated Ca2+ currents (ICa), pipettes contained 5 mM NaCl, 90 mM Cs CH3SO3, 20 mM CsCl, 4 mM MgATP, 0.4 mM Tris-GTP, 10 mM EGTA, 10 mM HEPES, and 3 mM CaCl2 (pH 7.2; 310 mOsm), and the bath solution contained 20 mM NaCl, 110 mM TEA-Cl, 10 mM CsCl, 1 mM MgCl2, 1 mM CaCl2, 10 mM HEPES, and 10 mM glucose (pH 7.4; 300 mOsm). In all experiments, pipette resistances were 2–3 MΩ.
Electrophysiological data were acquired at 10 or 100 kHz, and signals were low pass filtered at 5 kHz before digitization and storage. After the formation of a gigaohm seal (>1GΩ) and establishment of the whole-cell configuration, brief (10 ms) ±5-mV voltage steps from a holding potential (HP) of −70 mV were presented to allow measurements of whole-cell membrane capacitance (Cm), input resistance (Rin), and series resistance (Rs). Mean ± SEM Cm values were 32 ± 2 pF and 47 ± 2 pF (P < 0.001) in WT (n = 96) and R92Q (n = 67) hiPSC-CMs, respectively. Rin values were 1,665 ± 125 MΩ and 1,551 ± 162 MΩ (P > 0.05) in WT (n = 96) and R92Q (n = 67) hiPSC-CMs, respectively. Whole-cell Cm and Rs values were electronically compensated by 85%. Voltage errors resulting from the uncompensated Rs were always <2 mV and were not corrected. Leak currents were always <50 pA and also were not corrected.
In current-clamp recordings, spontaneous action potentials were recorded on establishing the whole-cell configuration. To record evoked action potentials, small (approximately −10 to −100 pA) current injections were made to hyperpolarize the membrane potential to −80 mV and to stop spontaneous firing. Individual action potentials were then evoked by brief (4 ms) depolarizing current (600 pA) injections. In voltage-clamp experiments, whole-cell ICa, evoked in response to depolarizing (300 ms) voltage steps to test potentials between −45 and +15 mV (in 5-mV increments at 1-s intervals) from a HP of −50 mV, were recorded.
Analysis of electrophysiological data
Electrophysiological data were compiled and analyzed using Clampfit 10.3 (Axon) and GraphPad (Prism). Cm values were determined by integration of the capacitive transients recorded during ±5-mV voltage steps from −70 mV. Current amplitudes in each cell were normalized to the Cm, and current densities are reported (pA/pF). All data are presented as means ± SEM. The statistical significance of observed differences between WT and R92Q hiPSC-CMs was evaluated using two-tailed Student’s t test or two-way ANOVA; P values are presented in Table S3.
Online supplemental material
Fig. S1 shows the affinity of calcium binding to the troponin complex at 20°C and 37°C. Fig. S2 shows the generation of gene-edited hiPSC-CMs. Fig. S3 shows immunofluorescence images of stem cells and hiPSC-CMs. Fig. S4 shows the relative expression of fetal (slow skeletal) and adult (cardiac) troponin I isoforms in hiPSC-CMs. Fig. S5 shows representative traces of hiPSC-CMs spontaneously beating on 10-kPa polyacrylamide gels measured by traction force microscopy. Fig. S6 shows comparison of the two independently derived R92Q clones examined using traction force microscopy, revealing no significant differences between the clones. Table S1 displays quantitative reverse transcription PCR (RT-qPCR) gene names and primers. Table S2 lists RT-qPCR measurements of the expression of key calcium-handling genes. Table S3 presents values from the electrophysiological experiments in this study. Video 1 shows the displacement of beads by WT and R92Q hiPSC-CMs contracting on a 10-kPa polyacrylamide gel. This video was used to generate the representative traces in Fig. S5.
Determination of the molecular mechanism of R92Q
At the molecular scale, the initial insult that drives the disease pathogenesis is mutation-induced alterations in protein function. Therefore, we set out to determine the molecular mechanism of the R92Q mutation in troponin T. WT and R92Q human troponin T were expressed and reconstituted into functional troponin complexes for biochemical and biophysical measurements. All assays were conducted using recombinant human tropomyosin and troponin complex. β-Cardiac ventricular myosin (MYH7) and cardiac actin were purified from porcine hearts. The porcine β-cardiac myosin isoform has 98% sequence identity with adult human β-cardiac ventricular myosin and similar biophysical properties, including indistinguishable ATPase kinetics and mechanics measured in the optical trap (Deacon et al., 2012; Greenberg et al., 2014; Sung et al., 2015).
We examined the effect of the R92Q mutation on thin filament regulation using an in vitro motility assay. In this assay, fluorescently labeled reconstituted regulated thin filaments are translocated over a bed of myosin in the presence of ATP and varying concentrations of calcium. The speed of translocation was measured as a function of calcium concentration, and normalized data were fitted with the Hill equation, as previously described (Greenberg et al., 2010). R92Q-regulated thin filaments show a shift in the pCa50 toward activation at submaximal, but physiologically relevant, calcium concentrations (P < 0.001; Fig. 2 A). There is no change in cooperativity, as determined by the Hill coefficient (P = 0.76).
The R92Q mutation does not change myosin detachment kinetics or calcium binding affinity
The shift toward submaximal calcium activation observed for R92Q in the in vitro motility assay stems from changes in the function of the troponin T protein. Given the role of troponin T in regulating calcium-dependent muscle contraction, three models have been proposed to explain the molecular mechanism of the R92Q mutation (Fig. 1 B): (1) R92Q could affect the cycling kinetics of myosins that are bound to the thin filament (Ford et al., 2012). In this model, one would expect to observe a change in the amount of time that myosin remains bound to the thin filament during cross-bridge cycling in the mutant. (2) R92Q could increase the calcium affinity of the troponin complex, leading to altered calcium buffering by myofilaments that directly disrupts calcium homeostasis (Schober et al., 2012; Ferrantini et al., 2017; Robinson et al., 2018). In this model, one would expect to observe an increased binding affinity for calcium in the troponin complex containing R92Q. (3) R92Q could alter the distribution of positions assumed by tropomyosin along the thin filament, leading to changes in the fraction of bound myosin cross-bridges (McConnell et al., 2017). In this model, one would expect to see changes in the equilibrium constants that define the positioning of tropomyosin along the thin filament. We set out to test these three models.
First, we tested whether the mutation affects the kinetics of myosin detachment from the thin filament by using stopped-flow kinetics to measure the rate of ADP release from actomyosin (i.e., the transition that limits actomyosin dissociation and unloaded sliding velocity), as we have done previously (Clippinger et al., 2019). We found that the rate of ADP release from myosin bound to regulated thin filaments is not affected by the R92Q mutation (P = 0.88; Fig. 2 B). Therefore, changes in myosin detachment kinetics cannot explain the shift toward submaximal calcium activation seen in the in vitro motility assay.
We measured whether the calcium binding affinity to the troponin complex is affected by the mutation. We used an IAANS-labeled form of troponin C to characterize calcium binding to the troponin complex (Robinson et al., 2007; Liu et al., 2012). The fluorescence intensity of this probe changes upon calcium binding to troponin C (Davis et al., 2007; Robinson et al., 2007; Liu et al., 2012; Williams et al., 2016). We used it to spectroscopically measure the affinity of calcium binding to regulated thin filaments (Fig. 2 C; Liu et al., 2012). We saw that the calcium concentration required for half-maximal activation is not significantly different for the WT and R92Q mutant proteins (P = 0.93). Similar results were seen at 15°C (Fig. 2 C), 20°C, and 37°C (Fig. S1). These results demonstrate that changes in the affinity of calcium binding to troponin C cannot explain the shift toward submaximal calcium activation seen in the in vitro motility assay (Fig. 2 A). Our finding of no change in calcium affinity for troponin C is consistent with some (Liu et al., 2012), but not all (Robinson et al., 2007), previous studies. The difference between our results and those obtained in previous studies showing changes in calcium affinity could come from differences in the calcium-sensing troponin constructs used, the actin proteins used (i.e., skeletal [Robinson et al., 2007] versus cardiac in the present study), or possibly other experimental conditions, such as the buffer composition.
The initial biophysical insult of R92Q is increased thin filament activation due to repositioning of tropomyosin along the thin filament
To test whether the shift in calcium sensitivity can be explained by a change in the distribution of positions assumed by tropomyosin along the thin filament (Fig. 3 A), we measured the equilibrium constants that define the fraction of thin filament regulatory units in each state (McKillop and Geeves, 1993; Barrick et al., 2019). The equilibrium constant between the blocked and closed states, KB, was determined by rapidly mixing fluorescently labeled regulated thin filaments together with myosin and then measuring the rate of myosin binding (seen as quenching of the fluorescence signal) in the presence and absence of calcium (see Materials and methods for details). At low calcium, when tropomyosin is primarily in the blocked state, the rate of myosin binding to the thin filament is slower than at high calcium, when the blocked state is less populated. The ratio of the rates of binding at low and high calcium were used to calculate KB (Eq. 1 and Fig. 3 B). As can be seen from the fluorescence transients, the rate of myosin binding to regulated thin filaments is similar for the WT and R92Q mutant proteins at high calcium (pCa 4); however, at low calcium (pCa 9), the rate of binding for the mutant is much faster than for the WT, consistent with lower population of the blocked state. When we calculate KB, we see that it is significantly larger in the mutant compared with the WT (P = 0.003), meaning that population of the more inhibitory blocked state is reduced while population of the closed state is increased. The increased KB value means that, at low calcium levels, the thin filament will be more activated in the mutant, consistent with the in vitro motility measurements (Fig. 2 A).
Next, we considered whether the mutation affects the equilibrium constant for the transitions between the closed and open states, KT, or the equilibrium constant between the open and myosin weakly bound states, Kw. To do this, we performed titrations of fluorescently labeled regulated thin filaments with increasing concentrations of myosin and measured the quenching of the fluorescence as the myosin binds to the regulated thin filaments (Fig. 3 C). These data, analyzed using a modification of the method of McKillop and Geeves (McKillop and Geeves, 1993; Barrick et al., 2019), show that there are no significant differences in KT between WT and R92Q (Fig. 3). There is a small increase in Kw; however, its magnitude is insufficient to explain the shift in the in vitro motility assay. This demonstrates that the primary molecular defect in R92Q is partial activation of the thin filament at low calcium levels due to reduced population of the inhibitory blocked state. Taken together, our results demonstrate that the initial molecular insult resulting from the R92Q mutation is decreased population of the thin filament blocked state, leading to increased thin filament activation and myosin-driven force generation.
Computational modeling corroborates the assertion that altered tropomyosin positioning with R92Q is sufficient to explain the shift toward submaximal calcium activation
To model the effects of the initial molecular insult identified in our biochemical studies, we utilized a widely used computational model of thin filament activation that was originally developed by Campbell et al. (Campbell et al., 2010). This model is based on the McKillop and Geeves three-state formalism that was also used to analyze our data (McKillop and Geeves, 1993). In this model, the equilibrium constants for transitions between thin filament states are inputted, and the model predicts several parameters, including the force per sarcomere as a function of calcium. When we used the default parameters of the model but proportionally increase the value of KB to match the fractional change seen in our biochemical experiments with the mutant (Fig. 4 A), we find that this change alone is sufficient to produce the shift toward submaximal calcium activation observed in the in vitro motility experiments (Fig. 2 A).
Generation of gene-edited stem cell–derived cardiomyocytes
To examine the effects of the R92Q mutation in human cells, we used CRISPR-Cas9 to generate two independent hiPSC lines that are homozygous for the R92Q mutation (Fig. S2). While there are several splice isoforms of troponin T expressed during development, the primary isoform expressed in hiPSC-CMs is the adult isoform (Cai et al., 2019). All of these isoforms contain R92, so all of the troponin T expressed in the cell has the mutation (Fig. S2). Homozygous lines were used to facilitate direct correlation of the molecular insult with alterations in cellular function. Heterozygous lines would better mimic the disease seen in humans but would contain complex mixtures of WT and mutant proteins, confounding the correlation of the molecular and cellular results. Both WT and R92Q hiPSCs were derived from the same parent line and are therefore isogenic except for the mutation. We previously showed, by whole-exome sequencing of the parent line, that it has no known variants associated with cardiomyopathy (Clippinger et al., 2019). Gene-edited hiPSCs are pluripotent, as assessed by immunofluorescence (Fig. S3 A), and have normal karyotypes (Fig. S3 B). hiPSCs were differentiated to hiPSC-CMs through temporal modulation of WNT signaling (Lian et al., 2012), and our efficiency of differentiation using this procedure is >90% (Clippinger et al., 2019). Both WT and R92Q cells express similar ratios of fetal (slow skeletal) to adult (cardiac) troponin I isoforms (P = 0.49; Fig. S4).
R92Q hiPSC-CMs generate increased force, power, and contraction speed compared with WT cells
To test whether R92Q hiPSC-CMs show the altered contractility seen in some model systems, we measured the contractility of single hiPSC-CMs using traction force microscopy. hiPSC-CMs were seeded onto rectangular extracellular matrix patterns on polyacrylamide hydrogels of physiological stiffness (10 kPa; Clippinger et al., 2019). This patterning on physiological stiffness hydrogels promotes hiPSC-CM maturation and sarcomeric alignment (Ribeiro et al., 2015). Patterned hiPSC-CMs show sarcomeres that align preferentially along the long axis of the cell (Fig. S3 B), and these cells contract primarily along that single axis (Video 1). The force, speed of contraction, and power were calculated from the displacement of beads embedded in the hydrogel (Fig. S5; Ribeiro et al., 2017). Data were plotted as cumulative distributions of single cells to account for cell-to-cell variability (Clippinger et al., 2019). R92Q hiPSC-CMs generate more force and power and have a higher contractile speed compared with the WT (Fig. 5). There were no significant differences observed between the two independently derived R92Q clones that we examined (Fig. S6).
Intracellular calcium transients are reduced in R92Q cells
While the initial molecular insult driving the disease pathogenesis is alterations in tropomyosin positioning, which increases molecular contractility, changes in contractility can lead to downstream activation of adaptive and maladaptive pathways. Previous studies using transgenic mice with troponin T mutations also showed altered cardiomyocyte calcium handling (Rice et al., 2010; Schober et al., 2012; Coppini et al., 2017; Ferrantini et al., 2017; Robinson et al., 2018). To examine calcium dynamics in hiPSC-CMs, cells were patterned onto rectangular ECM patterns on 10-kPa hydrogels and loaded with the ratiometric fluorescent calcium indicator dye Fura Red. Line scans of the fluorescence of spontaneously beating cells were collected at 1.9-ms intervals. As can be seen, hiPSC-CMs display well-defined calcium transients (Fig. 6 A); however, the amplitudes of the transients are lower (P < 0.002) in R92Q (0.56 ± 0.13; n = 18) compared with WT (0.84 ± 0.11; n = 19) cells. Although the low signal-to-noise ratio of the dye on the hydrogel substrate precluded us from determining whether there were differences in diastolic calcium, as has been reported previously (Schober et al., 2012; Robinson et al., 2018), the data presented demonstrate that despite generating increased force, R92Q hiPSC-CMs have reduced calcium transient amplitudes compared with WT cells.
To see whether the reduced population of the blocked state, observed in our molecular studies (Fig. 4 B), is sufficient to explain the hypercontractility seen in cells despite the reduction in calcium transient amplitude, we used the same computational model described earlier to predict the force per sarcomere in response to a calcium transient. In the modeling, the amplitude of the calcium transient for R92Q was reduced to 67% of the value seen in the WT, as observed in our cellular measurements (Fig. 6 A). As above, we proportionally increased KB for the mutant to match the relative difference seen in our biochemical experiments. Consistent with our cellular experiments, the model predicts that the mutant will generate more force in response to a calcium transient than the WT, despite having a smaller amplitude calcium transient (Fig. 4 B). It should be noted that hiPSC-CMs express complex mixtures of fetal and adult protein isoforms and that, as a result, there could be some subtle differences between the absolute forces predicted by the model, which is based on the equilibrium constants measured for the adult proteins, and the experimentally measured forces in hiPSC-CMs.
R92Q cells show alterations in expression of calcium-handling genes
The observed changes in calcium handling could come from a variety of sources, including changes in transcription, protein expression, and/or posttranslational modifications of proteins that regulate calcium homeostasis. To explore a possible role for transcriptional remodeling downstream of the initial molecular insult, we performed RT-qPCR analyses of the expression of transcripts encoded by key genes involved in the regulation of calcium homeostasis in cardiomyocytes (Fig. 6). Specifically, we examined the expression levels of transcripts encoding phospholamban (PLN), sarcoendoplasmic reticulum calcium-ATPase (ATPA2), voltage-gated calcium channel subunits (CACNA1C, CACNA1G, and CACNA1H), inositol triphosphate receptor (ITPR2), calsequestrin (CASQ2), calcium-calmodulin dependent kinase 2 (CAMK2D), sodium-calcium exchanger (SLC8A1), and the ryanodine receptor (RYR2). We found marked up-regulation of CASQ, CAMK2D, and SLC8A1 and down-regulation of CACNA1H in R92Q compared with WT hiPSC-CMs (Figs. 6 B, Table S1, and Table S2), demonstrating that the expression levels of key genes associated with calcium handling are altered in R92Q hiPSC-CMs.
R92Q cells show altered action potentials and reduced inward calcium current densities
The observed reductions in the calcium transients observed in spontaneously beating R92Q cells could reflect changes in transmembrane calcium influx. To determine directly if membrane excitability is altered in R92Q cells, we obtained whole-cell current-clamp recordings of spontaneous action potentials in WT and R92Q mutant hiPSC-CMs patterned onto rectangular ECM patterns on 10-kPa hydrogels (Fig. 7, A and B). Analyses of the data obtained in these experiments revealed that the maximum diastolic potential (the most negative membrane potential achieved between action potentials in spontaneously firing cells) is more depolarized in R92Q hiPSC-CMs than in WT hiPSC-CMs (Fig. 7 B). In addition, the frequency of spontaneous action potential firing is higher, upstroke velocities (i.e., the rate of membrane depolarization) are lower, and action potential durations at 50% repolarization (APD50) are shorter in R92Q hiPSC-CMs compared with WT cells (Fig. 7 B and Table S3).
To better understand the mechanisms contributing to the reductions in the APD50 seen in spontaneously beating R92Q cells, we examined the waveforms of evoked action potentials of hiPSC-CMs hyperpolarized to a membrane potential of −80 mV. Although similar hyperpolarizing currents were required to render R92Q and WT hiPSC-CMs electrically silent and similar currents were required to evoke action potentials in WT and mutant cells, the durations of evoked action potentials are significantly shorter in R92Q cells than in WT cells (Fig. 7, C and D).
Additional voltage-clamp experiments were conducted to determine directly if voltage-gated inward calcium current densities were altered in R92Q compared with WT cells. With outward potassium currents blocked, we recorded whole-cell voltage-gated calcium currents evoked on membrane depolarization in WT and R92Q hiPSC-CMs. As illustrated in Fig. 7 E, these experiments revealed that inward calcium current densities are markedly reduced in R92Q compared with WT hiPSC-CMs (Fig. 7 F).
Here, we elucidated the molecular and cellular consequences of the R92Q mutation in troponin T. We show that the initial molecular insult that drives disease pathogenesis is increased thin filament activation at physiologically relevant micromolar calcium levels due to destabilization of the tropomyosin blocked state. We demonstrate computationally and experimentally that this increased activation is consistent with the increased mechanical force produced by hiPSC-CMs. We show that altered mechanical forces lead to downstream changes in the expression of calcium-handling genes, altered calcium transients, and alterations in cellular electrophysiology. Taken together, our results highlight the role of mechanobiology in driving the early disease pathogenesis.
Defining the primary molecular driver of the disease pathogenesis
Previous studies demonstrated that R92Q mutant protein is expressed and properly integrated into sarcomeres, suggesting that the driver of the disease is changes in protein function rather than haploinsufficiency (Marian et al., 1997; Sweeney et al., 1998; Tardiff et al., 1999; Yanaga et al., 1999). Our results demonstrating that R92Q causes a shift toward submaximal calcium activation (Fig. 2 A) are consistent with some (Morimoto et al., 1998; Yanaga et al., 1999; Szczesna et al., 2000; Chandra et al., 2001; Robinson et al., 2002; Ford et al., 2012), but not all (Sweeney et al., 1998; Rust et al., 1999), previous measurements in muscle fibers and biochemical assays using noncardiac protein isoforms. The shift toward submaximal calcium activation could come from changes in actomyosin dissociation kinetics, the affinity of calcium binding to troponin C, and/or the positioning of tropomyosin along the thin filament (Fig. 1 B).
Our results demonstrate that R92Q does not affect the calcium binding affinity of troponin C or the kinetics of actomyosin dissociation (Fig. 2); however, the mutation increases the equilibrium constant between the blocked and closed states (Fig. 3 B). This change favors the closed state over the blocked state, lowering the energy barrier for thin filament activation at physiologically relevant calcium concentrations (pCa 5–7). Our computational modeling (Fig. 4) demonstrates that the increase in this equilibrium constant is sufficient to explain the shift toward submaximal calcium activation seen in our in vitro motility measurements (Fig. 2 A). Our data demonstrate that the initial insult that drives the early disease pathogenesis is reduced population of the inhibitory blocked position of tropomyosin at low calcium, which would lead to increased thin filament activation.
The R92Q mutation has been studied in many model systems, including quail myotubes (Sweeney et al., 1998), transfected rat cardiomyocytes (Rust et al., 1999), skinned rabbit muscle fibers (Morimoto et al., 1998), transgenic mice (Tardiff et al., 1999), and transfected cat cardiomyocytes (Marian et al., 1997). While these studies have greatly advanced our understanding of the mutation, they also demonstrated that the effects of the mutation depend on the model system used. Studies using transgenic mice demonstrated that disease presentation depends on whether proteins with biophysical properties similar to human isoforms are used (Ford et al., 2012). The use of all cardiac proteins with biophysical properties similar to human proteins is important for studies of thin filament mutations, since the activation of the thin filament depends on both myosin and calcium binding (Fig. 3 A). The most commonly used myosins in studies of R92Q mutation have been murine ventricular myosin (MYH6–92% sequence identity with human ventricular myosin) and rabbit fast skeletal muscle myosin (MYH2–82% sequence identity with human ventricular myosin), both of which have very different biophysical properties than human ventricular myosin (VanBuren et al., 1995; Malmqvist et al., 2004; Debold et al., 2007; Deacon et al., 2012). Our molecular studies used human cardiac troponin and tropomyosin and porcine cardiac myosin and actin. Porcine cardiac actin is identical in sequence to human cardiac actin. Porcine cardiac myosin (MYH7) is 98% identical in sequence to the human isoform and displays biochemical kinetics, mechanical step sizes, and load-dependent kinetics that are indistinguishable from human cardiac myosin (Deacon et al., 2012; Greenberg et al., 2014; Sung et al., 2015).
Interestingly, the R92 residue is in the region of troponin T that interacts with tropomyosin (Yamada et al., 2020). Two other HCM mutations have been identified at R92, R92W, and R92L, leading to the suggestion that R92 is a hotspot for HCM mutations. While there are no atomic-resolution structures of this region of the thin filament, structural studies have shown that troponin T plays a role in stabilizing the tropomyosin blocked state (Tobacman et al., 2002; Johnson et al., 2019; Madan et al., 2020). Molecular dynamics simulations demonstrated that mutations at R92 can change the distance between troponin T and tropomyosin (Manning et al., 2011), and biochemical experiments showed that R92L decreases the affinity of troponin for tropomyosin (Gangadharan et al., 2017). These two mechanisms are not mutually exclusive, and we speculate that R92Q reduces the coupling between troponin conformation and tropomyosin positioning, leading to destabilization of the blocked state. It is worth noting that we have focused on the effects of a troponin T mutation on calcium-based activation due to the central role of troponin in calcium-dependent thin filament regulation. Recent work has suggested that the N-terminal region of myosin binding protein C can play a role in activating the thin filament in the C-zone (Mun et al., 2014; Risi et al., 2018; Napierski et al., 2020). It will be interesting to investigate whether the R92Q mutation also affects the interactions of myosin binding protein C with the thin filament.
Recent studies of HCM-causing mutations in thick filament proteins, including β-cardiac myosin, myosin binding protein C, myosin regulatory light chain, and myosin essential light chain, have demonstrated that many of these mutations disrupt the autoinhibited super relaxed state of myosin, leading to the recruitment of more cross-bridges and thus to hypercontractility (Spudich, 2015; McNamara et al., 2016; Alamo et al., 2017; Nag et al., 2017; Adhikari et al., 2019; Sitbon et al., 2020). It has been proposed that increased cross-bridge recruitment correlates with the hyperdynamic cardiac function seen in HCM (Alamo et al., 2017). Our studies with R92Q, a thin filament mutation, demonstrate a similar net effect of increased cross-bridge recruitment at physiologically relevant calcium levels, suggesting that altered recruitment of cross-bridges in HCM as a common theme for both some thin and thick filament mutations.
It has been proposed that R92Q causes an increase in troponin C’s calcium affinity, which would affect the buffering of calcium by myofilaments, disrupting calcium homeostasis (Schober et al., 2012; Robinson et al., 2018). While our cellular data reveal disrupted calcium homeostasis, our molecular experiments show no change in the affinity of calcium for R92Q troponin, demonstrating that disrupted calcium homeostasis is a downstream consequence of the initial molecular insult. Our results are consistent with the notion that the development of HCM correlates with changes in tension rather than calcium handling (Davis et al., 2016).
Connecting the molecular and cellular phenotypes in R92Q
R92Q hiPSC-CMs show both an increase in force production (Fig. 5 A) and a reduction in the amplitude of the calcium transient (Fig. 6 A). These seemingly conflicting findings can be reconciled by our computational modeling (Fig. 4 B), which predicts that the shift toward submaximal calcium activation observed at the molecular scale can lead to cellular hypercontractility, despite the reduction in the amplitude of the calcium transient. This hypercontractility occurs at physiologically relevant (micromolar) concentrations of calcium (Bers, 2002). Therefore, the observed cellular hypercontractility is consistent with the proposed molecular mechanism.
At the cellular level, we see disrupted calcium homeostasis with R92Q, which is downstream of the initial contractile insult. Calcium homeostasis in the myocardium is a complicated process that depends on many factors, including gene expression and posttranslational modifications of signaling and contractile proteins (Bers, 2002). While a complete dissection of this mechanism is beyond the scope of the current study, our results provide insights into potential transcriptional mechanisms. We observed changes in the expression of several genes involved in calcium handling (Fig. 6 B), including calsequestrin (CASQ2), calcium-calmodulin kinase (CAMK2D), the sodium-calcium exchanger (SLC8A1), and a voltage-gated calcium channel subunit (CACNA1H). Interestingly, overexpression of CASQ2 or CAMK2D in transgenic mice causes heart failure and arrhythmogenesis (Sato et al., 1998; Zhang et al., 2003). While we recognize that changes in transcript expression do not always correlate with protein function, our data demonstrate that altered mechanics at the molecular level can drive changes in gene expression, showing a mechanobiological link between these processes in HCM. It is worth noting that some of the differentially expressed calcium-handling genes observed here in human cells are different from observations in the R92Q mouse model, which may be due to species differences (Coppini et al., 2017; Chowdhury et al., 2020). Interestingly, in mice, ablation of PLN can rescue the HCM phenotype, and it will be interesting to see whether similar results are observed in human cells (Chowdhury et al., 2020).
Our single-cell electrophysiological experiments reveal that action potential durations are shorter in R92Q, compared with WT cells, due in part to reduced inward L-type calcium current densities (Fig. 7). These changes would be expected to be arrhythmogenic and could contribute to the increased incidence of arrhythmias and sudden death in individuals harboring the mutation. We observe normal expression levels of the transcripts encoding L-type channel subunits, and therefore, the reduced current density could be due to alterations in signaling pathways and/or posttranslational modifications of channel subunits. The reduced inward calcium current densities (Fig. 7) would be expected to reduce calcium-induced calcium release from intracellular calcium stores, potentially contributing to the reductions in calcium transient amplitudes (Fig. 6 A).
Based on our observation that molecular hypercontractility drives downstream changes in cellular calcium handling and electrophysiology, we propose that mutation-induced changes in cellular tension alter mechanosensitive signaling pathways in cardiomyocytes (Saucerman et al., 2019). Consistent with this idea, we recently demonstrated that a dilated cardiomyopathy mutation in troponin T, ΔK210, affects molecular tension and mechanosensing, which helps to drive the disease progression (Clippinger et al., 2019). Similarly, increases in cardiac tension from external sources, including hypertensive disease and aortic stenosis, promote pathological hypertrophy. Deciphering the mechanobiological pathways in cardiomyocytes is an active field of research (Prosser et al., 2011; Robison et al., 2016; Saucerman et al., 2019). Hypercontractility of cardiomyocytes can impose aberrant stretch on fibroblasts, activating the transition to myofibroblasts (Saucerman et al., 2019) and contributing to the diffuse myocardial fibrosis frequently seen with HCM.
Implications for modeling and treating HCM
Modeling HCM is challenging since it is a complex disease where different adaptive and maladaptive pathways are activated as the disease progresses. Our goal was to connect the initial molecular insult with the early disease pathogenesis in human cells. We therefore used genome-edited hiPSC-CMs, which are excellent tools for dissecting the early disease pathogenesis (Lam and Wu, 2018). These experiments used isogenic cells, making it easier to decipher the direct consequences of the mutation on a controlled genetic background. R92Q hiPSC-CMs recapitulate some important aspects of HCM-induced changes in contractility (Tardiff et al., 1999; Robinson et al., 2002; Ford et al., 2012), altered electrophysiology (Coppini et al., 2017; Robinson et al., 2018), and calcium dysfunction (Rice et al., 2010; Ferrantini et al., 2017; Robinson et al., 2018) seen in other model systems. It should be noted that hPSC-CMs have several important limitations (see limitations section below). Limitations aside, hiPSC-CMs are powerful tools for studying the connection between the initial molecular insult and the early disease pathogenesis in human cells.
Our identification of altered cellular mechanics and downstream mechanobiological signaling pathways as key drivers of the disease has important implications for therapeutics. There is currently an outstanding need to develop new strategies to treat HCM. Our results suggest that targeting mechanobiological signaling pathways in cardiomyocytes could be useful for treating some forms of HCM.
We believe that cardiomyopathies are excellent candidates for a precision medicine approach (McNally and Mestroni, 2017; Fatkin et al., 2019; Lavine and Greenberg, 2020; Greenberg and Tardiff, 2021). Recently, there was a report of a HCM mutation in α-actinin linked to action potential prolongation attributed to increased calcium current density (Prondzynski et al., 2019), and the patient harboring this mutation was successfully treated with the L-type calcium channel blocker diltiazem. In R92Q, we observed a reduction in the calcium current density (Fig. 7), and therefore, a different therapeutic strategy would be necessary. Similarly, hiPSC-CMs harboring an HCM-linked mutation in troponin T, I79N, display changes in the action potential durations due to altered sodium–calcium exchanger function. Again, the cellular mechanism is distinct from that observed for R92Q, which is associated with changes in the inward calcium current density, potentially suggesting different therapeutic strategies (Wang et al., 2017; Wang et al., 2018). The marked differences between the functional effects of different HCM mutations highlight the need to understand the underlying derangements in molecular and cellular function when designing therapies.
In our molecular studies, we were interested in uncovering the initial molecular insult of the R92Q mutation in troponin T that drives the early disease pathogenesis. As such, we focused on the effects of the mutation on calcium-dependent thin filament activation. Our experiments and computational modeling used well-established methods (McKillop and Geeves, 1993; Campbell et al., 2010) that are tailored to capture the effects of changes in thin filament function. It is important to note that other models exist that incorporate higher-order sarcomeric structures beyond the thin filament, including thick filament–based regulation of contractility (e.g., the super-relaxed state of myosin in the thick filament that modulates the number of available cross-bridges, mutation-induced alterations in the force per cross-bridge, and myosin binding protein C’s role in thin filament activation). However, the potential impacts of the mutation on thick filament function, if present, are downstream of the initial molecular insult. That said, to a first-order approximation, our proposed molecular mechanism of mutation-induced changes in tropomyosin positioning is sufficient to explain the observed change in the calcium sensitivity of thin filament activation that we and others have observed.
Moreover, while our R92Q hiPSC-CMs recapitulate aspects of the early disease pathogenesis, they cannot fully capture the clinical phenotype for several reasons. First, hiPSC-CMs are developmentally immature, and they lack many of the physiological cues present in the myocardium (Lam and Wu, 2018; Musunuru et al., 2018). This immaturity includes expression of fetal genes (Fig. S4), spontaneous beating (Fig. 6), less negative resting membrane potentials (Fig. 7), and metabolic differences. As such, they do not capture aspects of clinical HCM, including fibrosis, myocyte disarray, and ventricular arrhythmias. While we employ several techniques to promote hiPSC-CM maturation, including plating the cells on physiological stiffness hydrogels, providing rectangular patterns to promote sarcomeric alignment, and aging cells at least 30 d before using them, these cells are still immature compared with adult cardiomyocytes (Ribeiro et al., 2015). Moreover, while patients are typically heterozygous for R92Q, our studies used homozygous cell lines to facilitate connecting the molecular and cellular phenotypes. It will be interesting to see how heterozygous cells differ from the homozygous cells used here. Despite these limitations, our gene-edited hPSC-CMs captured several aspects of the disease phenotype and enabled us to decipher the initial molecular insult from the activation of downstream adaptive and maladaptive signaling pathways.
We demonstrate that the initial insult of the R92Q mutation in troponin T is molecular hypercontractility at physiologically relevant calcium concentrations, which leads to alterations in mechanobiological signaling pathways that regulate calcium homeostasis, gene expression, and cellular electrophysiology. Taken together, our results demonstrate that mechanobiological adaptations are important in the early disease pathogenesis, and they suggest that targeting these pathways could open new avenues for treating HCM.
Olaf S. Andersen served as editor.
The authors thank Jonathan Davis for the troponin CT53C plasmid.
This work was supported by the National Institutes of Health (grant R01 HL141086 to M.J. Greenberg and grants R01 HL034161 and R01 HL142520 to J.M. Nerbonne), the March of Dimes Foundation (grant FY18-BOC-430198 to M.J. Greenberg), the Children’s Discovery Institute of Washington University and St. Louis Children’s Hospital (grant PM-LI-2019-829 to M.J. Greenberg), the Washington University Institute of Materials Science, and the Washington University Center for Cellular Imaging (grant CDI-CORE-2015-505 to M.J. Greenberg). S.R. Clippinger was supported through a National Institutes of Health institutional training grant (T32 EB018266).
The authors declare no competing financial interests.
Author contributions: S.R. Clippinger purified proteins and performed and analyzed the stopped-flow and fluorescence experiments. P.E. Cloonan performed and analyzed the traction force microscopy experiments with the stem cell–derived cardiomyocytes. W. Wang performed and analyzed electrophysiological experiments. L. Greenberg purified proteins, implemented the cell-based assays, performed and analyzed experiments with stem cell–derived cardiomyocytes, performed RT-qPCR measurements, and performed calcium imaging experiments. W.T. Stump designed tools for microcontact printing. P. Angsutararux performed in vitro motility assays. J.M. Nerbonne oversaw the electrophysiological experiments and analyzed data. M.J. Greenberg oversaw the project, performed simulations, generated mutant proteins, implemented biochemical assays, analyzed data, and drafted the manuscript. All authors contributed to the writing and/or editing of the manuscript.
A preprint of this paper was posted at bioRxiv on May 5, 2020.