Three human protein constructs were used in the NMR study: 1) recombinant human aCys-cNTnC (C35S, C84S double mutant), 2) chimeric construct (gChimera) of the cNTnC-cTnI switch peptide complex, aCys-cNTnC_1-85—SSGGSSGGSSGG linker - cTnI_145-167 and 3) slow skeletal troponin I switch peptide (ssTnI). The protocol used to express and purify both cNTnC and gChimera in Escherichia coli was previously described (Cai et al., 2016). The ssTnI peptide was synthesized and purified by GL Biochem Ltd. (Shanghai).

RPI-194 was synthesized by Rane Pharmaceuticals, Inc. in Edmonton, Alberta, Canada. Chemical structure was confirmed by NMR. For each NMR titration experiment, recombinant ¹⁵N-labeled gChimera or cNTnC was dissolved in 500 µL NMR buffer (90% H₂O/10% D₂O) consisting of 100 mM KCl, 10 mM imidazole, and 0.5 mM 4, 4-dimethyl-4-silapentane-1-sulfonic acid as a chemical shift reference. Purified lyophilized forms of gChimera or cNTnC were dissolved in NMR buffer. Protein quantitation by acid hydrolysis followed by amino acid quantitation showed the lyophilized form was 54% pure protein by weight. The pH of each NMR sample was maintained at a slightly acidic pH ∼ 6.7 by adjusting with microliter quantities of either 1 M NaOH or 1 M HCl. An acidic pH is typically employed in NMR to slow down base-catalyzed solvent-amide exchange, improving the signal intensity for rapidly exchanging amide groups in the protein. Since the proteins of interest are not known to have native side chains that become ionized near pH 6.7 (typically histidine), the use of a slightly acidic pH should not impact the electrostatic surface of the protein at all.

RPI-194 compound was dissolved into d₆-DMSO to make a 68 mM stock solution, which was then diluted ten-fold in d₆-DMSO to perform titrations. For both cNTnC and gChimera, the starting concentration was 115 µM. RPI-194 was titrated to 0.1, 0.2, 0.3, 0.4, 0.6, 0.8, 1, 1.5, 2, 2.5, 3, 3.5, 4, 5, 6 and 8 equivalents of cNTnC. For gChimera-RPI-194 titration, RPI-194 was titrated to 0.2, 0.4, 0.6, 0.8, 1, 1.2 and 1.4 equivalents of gChimera. Each titration point was monitored by recording a two dimensional ¹H, ¹⁵N heteronuclear single quantum coherence (HSQC) spectrum. Dilution factors were applied at each titration point to calculate the final concentration of cNTnC and RPI-194 in the calculation of binding affinities.

Titration of ssTnI was performed against free cNTnC and against cNTnC complexed with RPI-194. A 10 mM stock concentration of ssTnI was made by dissolving it into d₆-DMSO. ssTnI was titrated to 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4, 5, 6, 8, 10, 13, 16, 20, 25 and 30 equivalents of cNTnC. For titrating ssTnI into cNTnC·RPI-194 complex, RPI-194 was first titrated into free cTnC (115 µM) until both protein and drug were 1:1 equivalent. Then ssTnI was titrated with 0.1, 0.2, 0.4, 0.6, 0.8, 1, 1.5, 2, 2.5, and 3 equivalents of cNTnC·RPI-194 complex (115 µM).

All titrations were performed on a Varian Inova 500 MHz NMR spectrometer equipped with triple resonance ¹H, ¹³C, ¹⁵N probe. 2D ¹H, ¹⁵N HSQC spectra were collected for each titration point at 30°C. All titration data were processed using NMRPipe (Delaglio et al., 1995). A MATLAB runtime-based two dimensional lineshape analysis program called TITAN was used to calculate the dissociation constant (K_D) from titration experiments (Waudby et al., 2016). First, the protein and ligand concentration of each titration point were specified in the program. Individual NMR spectra for each titration point were uploaded into TITAN, and a specific region of interest (ROI) was selected for peaks we selected for large chemical shift perturbations. TITAN only considers chemical shift changes within the selected ROI for fitting and calculating dissociation constants. Simplified two-state binding models were used to calculate binding dissociation constants and on/off rate constants.

All steady-state fluorescence measurements were obtained on a SpectraMax i3x multi-mode microplate reader at 15°C. RPI-194 has intrinsic fluorescence with peak excitation and emission wavelengths at 335 and 470 nm, respectively. The same cNTnC and gChimera proteins (0–297 µM) used for NMR were titrated into a solution containing 5 µM RPI-194, 50 mM HEPES, 150 mM KCl, 5 mM MgCl₂, 1 mM DTT and 10 mM CaCl₂, and the intrinsic fluorescence spectrum of RPI-194 was monitored for changes. Binding affinities were calculated using GraphPad Prism version 9.0.2 (San Diego, California, United States).

Recombinant human cTnC (T53C, C35S, and C84S), cTnI and cTnT were used to reconstitute the cardiac troponin complex for stopped flow fluorescence studies. Expression, purification, production, and labeling of cTnC T53C with 2-(4′-(iodoacetamido) anilino) naphthalene-6-sulfonic acid (IAANS) were previously published (Davis et al., 2007). Expression and purification of recombinant cTnI, cTnT and reconstitution of the cardiac troponin complex (cTnC cTnI·cTnT) were as previously described (Davis et al., 2007).

Calcium release rates of IAANS-labeled, reconstituted cardiac troponin complex as a function of RPI-194 concentration were measured in a stopped-flow spectrometer (Applied Photophysics model SX.18 MV). IAANS excitation and emission were monitored at 330 nm and 420–470 nm, respectively. The calcium release rate was monitored by mixing calcium saturated (500 μM Ca²⁺) cardiac troponin complex with a stopped flow buffer containing calcium chelating solution (EGTA 10 mM, 10 mM MOPS and 150 mM KCl, pH 7.0) with a dead mixing time ∼ 1.24 ms. EGTA (10 mM) was used to sequester calcium from reconstituted cardiac troponin complex (0.3 µM) in the absence or presence of RPI-194. Increasing concentrations of RPI-194 were used in this reaction. P.J. King data analysis software, developed by Applied Photophysics (Leatherhead, Surrey, UK), was used to analyze stopped flow data. It uses a nonlinear Levenberg–Marquardt algorithm for data fitting.

Heart, soleus, and tibialis anterior muscles were isolated from each of 8 male Sprague-Dawley rats, ranging in age from 6–9 months. The rats were euthanized (anesthesia induced by isoflurane, followed by rapid cardiectomy) in accordance with a protocol approved by Institutional Animal Care and Use Committee of Ohio State University. The soleus and anterior tibialis muscles were immediately placed in cold relaxing solution, and fiber bundles were prepared and stored in relaxing solution containing 50% glycerol (v/v) at −20°C (Reiser et al., 2013). A single large cut was made through the free wall of both ventricles of the heart, which was then placed in ice-cold relaxing solution with 1% Triton X-100 for 30 min (Tikunova et al., 2019). The heart was removed from this solution, gently compressed, and blotted and transferred to cold glycerinating solution (Reiser et al., 2013).

Single trabeculae were isolated and studied as previously described (Tikunova et al., 2019). Briefly, a trabecula was mounted in the experimental chamber that was controlled at 15°C (Tikunova et al., 2019). In the chamber, one end of the trabecula was attached to a motor and another end was attached to a transducer. The trabecula was set to the resting striation spacing, the equivalent of sarcomere length. Striation spacing was determined using a camera mounted on the microscope and the SPOT image analysis software (https://www.spotimaging.com) (Tikunova et al., 2019). The distance spanned by ∼ 20 striations was measured to calculate resting sarcomere length. Fiber width and depth were measured, and fiber cross-sectional area (CSA) was calculated, assuming an ellipsoidal cross section. The average resting sarcomere length of the twenty-four trabeculae that were studied was set to 2.07 ± 0.02 μm. Each trabecula was then subjected to two series, A and B, of activations. Series A was always without RPI-194. Series B was with 0, 20, 50 or 100 µM RPI-194.

The force versus pCa relationship was measured in six trabeculae for each concentration of RPI-194, first without (series A), then with (series B) RPI-194 (100 mM stock dissolved in DMSO) (Tikunova et al., 2019). RPI-194 was added to all of the solutions to which the trabeculae were exposed during the measurements of the force/pCa relationship: pCa 9.0 solution, HDTA pre-activating solution and each of the maximal (pCa 4.0) and submaximal activating solutions. The trabeculae were soaked in pCa 9.0 solution with RPI-194 at 15°C for 30 min before initiating the second series of force measurements. We reported previously that DMSO had no effect on the force/pCa relationship (Reiser et al., 2013). We initially determined, in three skeletal muscle fibers, that the control (no added compound) force/pCa relationship is essentially identical when measured twice in a given preparation. The trabecula was treated with series of pCa solutions as previously described (Reiser et al., 2013). The force versus pCa data were fit as previously described (Black et al., 2000; Tikunova et al., 2002; Tikunova et al., 2019).

The sarcomere length in slow and fast fibers was measured using the Fast Fourier Transform in ImageJ (https://imagej.nih.gov/ij/). The fiber type (slow or fast) of each studied skeletal muscle fiber was determined from an analysis of the myosin heavy chain isoform composition using SDS-PAGE, as described (Tikunova et al., 2018). The maximal velocity of shortening (V_o) was measured, using the slack test (Edman, 1979), in slow and fast fibers when activated in pCa 4.0 solution (every third activation in the force/pCa measurements series). Thus, two V_o determinations were made in each slow and fast fiber, first without, then with, RPI-194 (series A and B).

A total of 24 slow fibers, 24 fast fibers and 24 trabeculae were studied. The group size for each concentration of RPI-194 was six and each preparation was studied at one concentration. An analysis of variance (ANOVA) Tukey’s post-hoc test was used when comparing the effects of all four concentrations (0, 20, 50 and 100 µM) of RPI-194 in each preparation (slow fibers, fast fibers or trabeculae) or when comparing effects of a given concentration of RPI-194 in all three preparations (slow fibers, fast fibers and trabeculae). Student’s t-test was used to assess the significance of differences in maximal shortening velocity in fast and slow fibers at a given concentration of RPI-194. Pearson’s correlation was used to test for a relationship between the shift in pCa₅₀ and the change in V_o induced by RPI-194.

Adult ventricular cardiomyocytes were isolated and perfused as previously described (Sah et al., 2002). Contractility assays from isolated cardiomyocytes were conducted as previously described (Crackower et al., 2002). Briefly, a Grass S44 stimulator with a pulse duration of 3 milliseconds at 1 Hz was used to stimulate cardiomyocytes. Myocyte contraction was tracked at 240 Hz using a video edge detector. In addition, myocyte steady state contractions at 1 Hz and a subsequent equilibrium period for 4 min at 240 Hz were recorded. We determined fractional shortening, shortening rate (+dL/dT) and relaxation rate (dl/dT) in the isolated cardiomyocytes.

All animals used in isolated working heart perfusion experiments were treated in accordance with the guidelines of the Canadian Council of Animal Care and approved by the University of Alberta Health Sciences Animal Welfare Committee. All animal experiments were conducted on male C57BL/6 mice (7–10 weeks) obtained from Charles River Laboratories (Wilmington, MA, United States) and regularly fed with chow diet (Harlan Teklad, Madison, WI, United States). Animals were anesthetized with 60 mg kg⁻¹ isoflurane administered through the peritoneum. Isolated working heart perfusions were performed as previously described (Kuang et al., 2004). Rapidly excised hearts were immediately placed on an ice-cold Krebs-Henseleit solution. A recirculating perfusate solution was used for the isolation of working hearts. It consisted of a modified Krebs-Henseleit solution (100 ml) which was a mixture of 1.2 mM KH₂PO₄, 1.2 mM MgSO4, 2.5 mM CaCl₂, 4.7 mM KCl, 25 mM NaHCO₃ and 118 mM NaCl. The perfusate was supplemented with 1.2 mM palmitate prebound to 3% bovine serum albumin and 5 mM glucose as energy substrates. Glycolysis and glucose oxidation rates were calculated from the perfused heart by adding a small amount of radiolabeled [5-³H] glucose and [U-¹⁴C] glucose in the Krebs-Henseleit solution (Kovacic et al., 2003; Kuang et al., 2004). The perfusate was continuously supplied with a gas mixture of 95% O₂, 5% CO₂. Cardiac output, cardiac work, heart rate and peak systolic pressure were also assessed from the perfused hearts.

The ATPase activity of human beta-cardiac myosin subfragment 1 (amino acids 1–843) containing a C-terminal green fluorescent protein tag (M2β-S1 GFP) was examined using the NADH coupled assay (Swenson et al., 2017; Tang et al., 2019). M2β-S1 GFP was produced using the C2C12 cell expression system and purified as previously described (Swenson et al., 2017; Tang et al., 2019). The ATPase activity was examined in the presence of 40 µM actin and varying RPI-194 concentrations with 1% DMSO present.

During systole, cardiac muscle contraction is triggered by the calcium dependent binding of the cTnI switch region to the regulatory cNTnC. A cardiac troponin activator drug promotes and stabilizes formation of this activated complex. We previously produced multiple variations of the cNTnC-cTnI chimera with different linkers (Pineda-Sanabria et al., 2014; Cai et al., 2016). In the current study we used a further refined version that we call “gChimera”, which utilizes a linker containing multiple Ser and Gly residues for maximum flexibility and solubility while maintaining charge neutrality (amino acid sequence shown in Figure 1A).

Based on our previous work, 3-chlorodiphenylamine was a promising starting compound to develop a cardiac troponin activator (Cai et al., 2016). Addition of hydrophobic substituents to the aryl rings of 3-chlorodiphenylamine tends to improve binding affinity but greatly reduces solubility, whereas more polar substituents are not well tolerated. We aimed to add at least one hydrophilic group to enhance solubility and specificity of binding. We designed a total of 54 3-chlorodiphenylamine-based compounds that were synthesized by Rane Pharmaceutical Inc., Edmonton, AB, Canada. The compounds were assessed for binding to gChimera by NMR, and we identified a compound, RPI-194, which has an additional p-benzoic acid in the ortho position of the aniline group, with a measured dissociation constant, K_D, of 24 µM (see Figure 1A). Linear migration of the NMR signals suggests 1:1 binding kinetics in the fast exchange regime. (“Fast” exchange means fast relative to the frequency differences between NMR signals in the two different states.) We also calculated the binding affinity by titrating gChimera into RPI-194, monitoring the steady state intrinsic fluorescence of the RPI-194 compound to yield a measured dissociation constant, K_D, of 14 µM (see Supplementary Figure S1).

RPI-194 binds to the isolated cNTnC domain with a lower affinity than gChimera (K_D = 300 µM), as measured by steady state fluorescence (see Supplementary Figure S1). Compared with gChimera, the isolated cNTnC domain lacks the cTnI switch region, which shifts it to an open conformation and binds to small molecules like RPI-194 via the side chains of residues Ile148 and Met153.

NMR titration experiments show a complex equilibrium when RPI-194 is titrated into the cNTnC domain, precluding binding affinity determination by NMR, as was done for gChimera. Prior to addition of RPI-194, the NMR spectrum of isolated cNTnC domain demonstrates signal broadening due to fast timescale conformational exchange between closed and open states (Sia et al., 1997; Spyracopoulos et al., 1997), undergoing a closed-to-open transition with a k_ex of about 30,000 s⁻¹ (Eichmuller and Skrynnikov, 2005), with the more open conformation representing a minor population of about 5% (Mckay et al., 2000; Paakkonen et al., 2000). Peaks shift and then rapidly disappear upon addition of RPI-194, indicating intermediate timescale binding, consistent with selective binding of RPI-194 to the less populated open state (see Figure 1B). As more RPI-194 is added, new NMR signals corresponding to RPI-194-cNTnC complex abruptly re-appear, but in some cases (for example, for residues G30 and G42) they appear in a different position than one would expect based on the start of the titration. This suggests a new conformational process occurring different from the initial 1:1 binding of cNTnC to RPI-194. Other peaks in the spectrum that do not shift become visibly reduced in intensity, suggesting a large increase in molecular weight consistent with dimerization.

Similar changes occur when the drug trifluoperazine is titrated into calmodulin, a protein homologous to troponin C (Feldkamp et al., 2010; Waudby et al., 2016). Each homologous domain of calmodulin binds two molecules of trifluoperazine, and this promotes association of the N-terminal domain with the C-terminal domain through hydrophobic interactions. We propose that at high concentrations of RPI-194, one or two molecules of RPI-194 bind and stabilize the open conformation of the cNTnC domain, which then has a tendency to dimerize. Physiologically, the cNTnC domain does not dimerize because cTnC is tethered to fixed positions along the thin filament. Moreover, the cNTnC domain is predominantly in the closed state unless the cTnI switch region is bound. Thus, while the behaviour of free cNTnC domain in the presence of RPI-194 (and many other compounds) is interesting in terms of its tendency to dimerize, it is not physiologically relevant, except to note that RPI-194 does bind to calcium-saturated cNTnC domain in the absence of cTnI switch region, though binding is substantially more effective once the cTnI switch region is bound. This suggests that RPI-194 is more effective at stabilizing the activated troponin complex once it is formed, rather than promoting the formation of the activated complex.

To examine the impact of RPI-194 on binding of TnI switch peptide to cNTnC, we titrated TnI switch peptide into isolated cNTnC domain. Cardiac muscle troponin C (cTnC) is the same isoform as slow skeletal muscle troponin C (ssTnC), though slow skeletal muscle possesses different isoforms of troponin I (ssTnI) and troponin T (ssTnT). We used the ssTnI switch peptide instead of the cardiac isoform because of its superior solubility. The cTnI switch peptide readily precipitates out of solution when titrated into solutions containing cTnC and RPI-194, making determination of binding constants unreliable.

When ssTnI switch peptide (as opposed to small molecule RPI-194) is titrated into cNTnC, signals that were broad at the start of the titration progressively become narrower as the switch peptide shifts the cNTnC conformational equilibrium to a fully open state (see Figure 2A), indicating fast kinetics of binding. This is in marked contrast to when RPI-194 is titrated into cNTnC, in which peaks that were broad at the beginning of the titration become broadened beyond detection as RPI-194 is added. Thus, the ssTnI switch peptide appears able to bind cNTnC via a rapid induced fit mechanism, whereby it stimulates the transition of cNTnC from a closed to an open state, whereas RPI-194 binds via conformational selection, requiring a stochastic transition to the open state prior to binding. Using a two-dimensional lineshape analysis tool, TITAN, we calculated the ssTnI switch peptide binding affinity for cNTnC (K_D 652 µM) that is weaker than that previously determined for the corresponding cTnI switch peptide (K_D 154 µM) (Li et al., 2002).

We then titrated ssTnI switch peptide into cNTnC domain in the presence of RPI-194 (Figure 2B). The presence of one equivalent of RPI-194 significantly enhances the binding of ssTnI switch peptide to cNTnC. The binding affinity of ssTnI switch peptide for the cNTnC: RPI-194 complex is K_D 22 µM as calculated by TITAN, which is significantly tighter than the value of 652 µM determined for cNTnC-ssTnI binding in the absence RPI-194, over an order of magnitude change. Improved binding of the TnI switch peptide is consistent with previous NMR studies of the cardiac troponin activator dfbp-o (Robertson et al., 2010; Lindert et al., 2015).

We used stopped-flow fluorescence of IAANS-labeled troponin C to measure the impact of RPI-194 on calcium release rates in troponin. RPI-194 binding to reconstituted heterotrimeric cardiac troponin complex slowed the rate of calcium release from 38 s⁻¹ to 13 s⁻¹ (Figure 3), with an apparent dissociation constant, K_D, of 6 μM, in agreement with NMR and steady state fluorescence measurements. This is consistent with our NMR studies demonstrating that RPI-194 stabilizes the calcium-saturated activated troponin complex.

At baseline and in the absence of drug, the intrinsic calcium sensitivity of skeletal muscle fibers is about the same for limb slow (pCa₅₀ 6.28 ± 0.03) and fast (6.27 ± 0.03) muscle fibers (see Figure 4 and Table 1). The calcium sensitivity of cardiac trabeculae is significantly lower (5.68 ± 0.02). Slow skeletal muscle has a higher calcium affinity than cardiac muscle, even though both muscle types utilize the same cTnC/ssTnC isoform.

RPI-194 has a greater effect on slow skeletal muscle than cardiac muscle. The effects of RPI-194 were tested at 20, 50 and 100 µM concentrations. Six cardiac trabeculae, six slow fibers and six fast fibers were studied at each concentration, and each trabecula/fiber was used to study one concentration of the compound. The shift in the pCa₅₀ (i.e., ∆pCa₅₀) is significant between each tested concentration of RPI-194 within each muscle group. At 50 μM RPI-194, the pCa₅₀ of slow fibers shifted +0.35 units, while the fast fibers and cardiac trabeculae shifted +0.14 and +0.16, respectively (Figure 4 and Table 1). Therefore, there is a larger calcium sensitizing effect of RPI-194 on limb slow fibers, compared to cardiac trabeculae and limb fast fibers. Although RPI-194 was designed and tested to bind to the cardiac/slow skeletal troponin complex, there is some cross-reactivity with fast skeletal muscle, but the effect is not nearly as large as that observed with tirasemtiv (+0.89) (Russell et al., 2012), which was designed specifically for fast skeletal muscle.

RPI-194 did not increase the maximum isometric force (P_o) generated under saturating calcium concentrations in any of the muscle types (Supplementary Table S1). In fact, upon addition of 20, 50, or 100 μM RPI-194, there was an initial 15% decrease in P_o for the slow muscle fibers and a 10% decrease for fast muscle fibers and cardiac trabeculae, with no apparent concentration dependence at the concentrations tested. There was a trend towards recovery in P_o with time seen at higher concentrations of RPI-194 in cardiac trabeculae. The reasons behind these phenomena are not known. The skinned muscle fiber experiments are described in more detail in the Supplementary Material.

There was a marked effect of RPI-194 on maximal shortening velocity in unloaded slow and fast skeletal muscle fibers, with velocity generally decreasing with higher concentrations of RPI-194. The overall slowing effect of RPI-194 was similar in slow and fast fibers, with the velocity V_o being reduced to about half in the presence of 100 μM RPI-194 (Table 1).

Shortening velocity is determined by the load-dependent rate of actin-myosin cycling. It is possible that RPI-194 slows the velocity of unloaded contraction, V₀, via a direct interaction with actin-myosin. We therefore proceeded to examine the effect of RPI-194 on human beta-cardiac myosin S1 ATPase activity in the presence of 40 µM actin. We found no effect whatsoever (see Supplementary Figure S2), suggesting that the reduction in V₀ by RPI-194 seen in slow and fast skeletal muscle is not due to direct binding to actin-myosin S1, but rather, an effect on another myosin domain or light chain, troponin, or some other unknown off-target effect.

The addition of RPI-194 to individual unloaded cardiomyocytes caused a decrease in observed contractility. 10 μM RPI-194 significantly increased resting sarcomere length, decreased fractional shortening, and decreased the velocity of contraction and relaxation (Figure 5). At a concentration of 100 μM, cardiomyocyte contractions ceased completely, with sarcomere lengths suggestive of a relaxed state (rather than a contracted state). The inhibition of contractility observed in cardiomyocytes contrasts with the increased calcium sensitivity of isometric contraction seen in skinned cardiac trabeculae. This raises the possibility that RPI-194 interferes with excitation-contraction coupling in living cells, for example, the inhibition of ion channels. When we attempted to measure calcium transients in cardiomyocytes using calcium-sensitive fluorophores, the strong intrinsic fluorescence of RPI-194 created too much background signal. It is therefore necessary to assess cardiac contractility in another system that includes intact cells.

We studied the impact of RPI-194 in a mouse isolated perfused working heart model, which was maintained at constant pressure. There was no consistent trend in heart rate or heart rate times peak systolic pressure product as RPI-194 was added to the system up to a maximum of 100 µM (Figure 6). Cardiac output and cardiac work increased with increasing RPI-194 levels, but the trend was not statistically significant, and similar changes could also be observed in controls over the course of 60 min (see Supplementary Figure S3 ). As RPI-194 was added, the observed increase in cardiac work was accompanied by higher rates of glucose utilization and oxygen consumption. Thus, it appears that metabolic pathways are generally intact in the presence of RPI-194. The severe inhibitory effect of RPI-194 observed in individual unloaded cardiomyocytes was not observed in isolated perfused working hearts, making it less likely that the inhibitory effect observed in cardiomyocytes is due to modulation of ion channels.