Molecular mechanisms and structural features of cardiomyopathy-causing troponin T mutants in the tropomyosin overlap region

Considering the known interaction between Tm and TnT, we determined if the mutations changed the affinity of TnT for Tm in vitro (Table 2), using bacterially expressed, purified proteins (SI Materials and Methods and Fig. S1A). Using MST, we measured the K_d of WT Tn complex binding to Tm to be 21 nM. Strikingly, the HCM mutants in regions 92–144, R92L, K124N, and R130C, significantly decreased the binding affinity for Tm (Fig. 1A and Table 2), while the DCM mutants R134G and R144W increased the affinity significantly (Fig. 1B). K97N, which lies close to the hotspot residue 92, increased the affinity slightly, but the change was not as great as that observed for R92L. Outside region 92–144, HCM mutants D86A and P80S did not significantly change the affinity of TnT for Tm (Fig. 1C and Table 2).

To examine the role of actin in binding of TnT and Tm, we performed MST using Tn WT and mutant complexes (SI Materials and Methods and Fig. S1B) and a constant 1:7 ratio of Tm to actin. The K_d for WT was tripled from 21 to 58 nM by the addition of actin (Table 3). All of the mutants between residues 92 and 144 altered the K_d and exhibited the same trend as observed with MST without actin (Fig. 2 A and B). Mutants P80S and D86A were exceptions, and did not follow the trend observed in the absence of actin. These mutants showed an increase in K_d in the presence of actin (Fig. 2C).

To analyze the effect of TNNT2 mutations on Ca²⁺ sensitivity of the system, we performed thin filament-regulated actomyosin ATPase assays with bovine cardiac myosin subfragment 1 (S1), measuring ATPase activity as a function of free Ca²⁺ concentration (Fig. 3). HCM substitutions R92L and K124N were sensitizing, as their pCa₅₀ (−log [Ca²⁺] at half-maximal activation) values were increased by 0.15 and 0.14, respectively, relative to WT. On the other hand, the DCM substitutions R134G and R144W reduced pCa₅₀ values by 0.06 and 0.11, respectively, relative to WT (Table 4). Many, but not all, of the HCM substitutions increased the maximal activation of S1 ATPase activity, while the DCM substitutions decreased it. The HCM substitution R130C only slightly increased pCa₅₀ but significantly increased the maximum ATPase activity (Table 4). The other three HCM substitutions, P80S, D86A, and K97N, did not significantly alter maximum ATPase activity. These trends are consistent with observations reported by other groups (30–33).

There is no crystal structure available for the TNT1 region in which most of the TnT mutants reside, including the ones that we have described here. Therefore, to understand the structural changes triggered by the mutants in which we are interested, we used an extensive template search protocol (SI Materials and Methods, Fig. S2, and Table S1) to model residues 1–200 of the TNT1 region. We extended the crystal structure coordinates (for residues 66–99 of TnT) obtained by Murakami et al. (15) to model TNT1. The best model on the basis of molecular probability density function (molpdf) scores (Table S2) had a folded-back architecture for TNT1 (Fig. S3A), with Ramachandran plots shown in Fig. S3 B and C. In this model, residues 92–144 of TnT form an α-helix and interact with the Tm overlap region (Fig. 4A, magenta). The model is available online (caps.ncbs.res.in/download/TnT/). We do not view the folded-back architecture as likely, as a nonfolded model had identical interactions with Tm, but with molpdf scores that were only slightly lower. Our focus in this study is on the TNT1–Tm interaction region, and our modeling data did not give strong predictions for noninteracting regions.

To model the effects of TNNT2 mutations, we introduced them into the model to detect possible changes in interactions and interface energy between TNT1 and Tm. The mutations were introduced using SYBYL software, and the interface energies were calculated using COILCHECK+ (34). The positions of pathogenic substitutions on the model are shown in Fig. 4. While HCM mutants D86A and K97N were similar to WT with energy values of −11.5 and −16 kJ/mol (Table 5), HCM mutants R92L and K124N were less stable, with energy values of +8.7 and +7.4 kJ/mol, respectively. HCM mutant R130C and DCM mutants R134G and R144W did not alter the stability of the model in silico, but changes in binding affinities were observed in vitro (discussed above). This is probably because these three residues lie at the end of the interacting region in our model. Because COILCHECK+ is designed to check the energetics of helix–helix interactions in coiled coils and P80S lies in the TnT loop region in our model, it was not included in the interface energy determinations.

To assess relationships between the changes in energy values predicted by our model and in vitro binding assays, we plotted correlations. The energy calculations from COILCHECK+ correlated well with the K_ds from MST experiments, with a Pearson’s coefficient (r) of 0.80 (Fig. 5A). We observed a moderate correlation between in silico predictions and Ca²⁺ sensitivity, with a Pearson’s r of 0.6 (Fig. 5B). We observed an excellent correlation between changes in binding affinity and Ca²⁺ sensitivity observed in vitro, with a Pearson’s r of 0.83 (Fig. 5C).

Plasmids encoding human cardiac isoforms of TnC, TnT, TnI, and α-Tm (with an Ala-Ser N-terminal dipeptide) were provided by James D. Potter, University of Miami Miller School of Medicine, Miami. TnT mutants P80S, D86A, R92L, K97N, K124N, R130C, R134G, and R144W were introduced by site-directed mutagenesis. All of the mutants were confirmed by sequencing.

Escherichia coli Rosetta cells were transformed, grown in terrific broth at 37 °C to an A₆₀₀ of ∼1, selected with ampicillin and chloramphenicol (each at 50 μg/mL), and induced with 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) at 18 °C overnight. The WT Tm was purified as described previously (30). WT and mutant TnT, TnI, and TnC were purified using protocols published previously, with some modifications (39). Cultures were grown overnight, pelleted, and resuspended in lysis buffer [6 M urea, 50 mM Tris (pH 7.5), 2 mM EDTA, and 1 mM DTT], along with PMSF and protease inhibitors. After lysis by sonication, the samples were clarified by spinning at 25,000 × g for 45 min at 4 °C. The lysates were then dialyzed into 50 mM Tris (pH 7.4), 6 M urea, 2 mM EDTA, and 1 mM DTT for TnT; 50 mM Tris (pH 7.8), 6 M urea, 1 mM EDTA, and 1 mM DTT for TnC; and 50 mM Tris (pH 7), 6 M urea, 2 mM EDTA, and 1 mM DTT for TnI. Lysates were further clarified by syringe filtration, and then loaded onto ion-exchange columns (HP-Q column for TnC and HP-SP column for TnT and TnI) and eluted with gradient of 0 to 600 mM KCl. For the second ion-exchange column, peak fractions for TnT collected from the previous column were dialyzed into 50 mM Tris (pH 7.8), 6 M urea, 1 mM EDTA, and 1 mM DTT, and passed onto an HP-Q column and eluted using a KCl gradient. For TnC, the peak fractions from the HP-Q column were dialyzed overnight into 50 mM Tris (pH 7.5), 1 mM CaCl₂, 1 mM MgCl₂, 50 mM NaCl, and 1 mM DTT. Solid ammonium sulfate was added to the dialyzed samples at a concentration of 0.5 M at room temperature (RT), and they were loaded onto a phenylsepharose column. The pure fractions were eluted with 50 mM Tris (pH 7.5), 1 mM EDTA, and 1 mM DTT. The peak fractions for TnI were then dialyzed overnight into affinity buffer [50 mM Tris (pH 7.5), 2 mM CaCl₂, 1 M NaCl, and 1 mM DTT]. TnC affinity columns were prepared by coupling pure TnC fractions with Affi-Gel (Bio-Rad) in 100 mM 3-(N-morpholino)propanesulfonic acid (MOPS; pH 7) and 10 mM DTT. The dialyzed TnI samples were then loaded into the TnC affinity column and eluted with 6 M urea. All of the Tns were dialyzed into storage buffer [20 mM imidazole (pH 7.5), 1 M KCl, 1 mM MgCl₂, and 1 mM DTT], flash-frozen, and stored at −80 °C. Representative purified proteins are shown in Fig. S1A.

WT and mutant Tn complexes were formed as previously reported (39). TnT, TnI, and TnC were mixed at a ratio of 1.3:1.3:1 and incubated on ice for 1 h. The complexes were then dialyzed into 10 mM MOPS (pH 7), 1 mM DTT, and 0.7–0.025 M KCl. Complexes were flash-frozen and stored at −80 °C. Actin was purified from chicken pectoral muscle acetone powder (43).

Myosin was purified from bovine left ventricles. Cardiac myosin S1 was further purified from bovine myosin by chymotrypsin digestion (at a concentration of 5 mg/mL) for 7 min at 25 °C. The reaction was stopped using 100 mM PMSF. The sample was then spun at 150,000 × g for 30 min. The supernatant contained S1 and was dialyzed into assay buffer. Representative SDS/PAGE of purified proteins is shown in Fig. S1B.

Purified α-Tm was dialyzed into 50 mM Tris (pH 7), 0.5 mM EDTA, and 5 M guanidine HCl overnight. Excess DTT (10 mM) was added to reduce disulfide bonds. Tm was labeled in 25 M excess with N-(5 fluoresceinyl) maleimide (5-FM; Sigma). The labeled protein was then dialyzed into MST buffer [50 mM Tris (pH 7.4), 150 mM NaCl, 10 mM MgCl₂, and 0.05% Tween 20] overnight at RT. The concentration of fluorescently labeled Tm was kept constant at 12.5 nM, and the concentration of the titrant Tn complex (WT and mutant) was varied from 0.06 nM to 2 μM. Tm and Tn complex were mixed together with MST buffer and incubated for 20 min at RT. Samples were then loaded into standard treated capillaries and analyzed with a Monolith NT.115 (NanoTemper Technologies GmbH). The laser duration was 30 s (MST power = 70, LED power = 80). The data were analyzed using GraphPad Prism software and plotted using the Hill equation:with X as the concentration of ligand, Y as specific binding, B_max as maximum binding, K_d as the dissociation constant, and h as the Hill coefficient.

Tm was labeled with 5-FM. The change in the distribution of Tm fluorescence upon heating was measured as a function of the concentration of Tn complex. Since migration of an individual molecule differs from migration of a molecule bound to ligand, the change in distribution of Tm fluorescence was used to determine the ratio of free Tm to Tm bound to TnT (or bound to TnT + actin). F_cold and F_hot are the fluorescence measurements before and after heating, respectively. The ratio of F_hot/F_cold gave the normalized fluorescence, F_norm. Plotting F_norm against the logarithmic concentrations of serially diluted ligand (Tn complex) gave sigmoidal binding curves. The binding parameters were determined from these plots.

Freshly prepared myosin S1, actin, Tm, and Tn complex were dialyzed into ATPase buffer [20 mM imidazole (pH 7.5), 10 mM KCl, 3 mM MgCl₂, and 1 mM DTT]. Regulated thin filaments were made by incubating F-actin, Tm, and Tn complex at a molar ratio of 7:2:3. The 10× pCa buffers were prepared with 20 mM EGTA, 40 mM nitrilotriacetic acid, and varying concentrations of CaCl₂, which was calculated as reported previously (44). Each well contained 7 μM actin, 2 μM Tm, 3 μM Tn complex, and 0.6 μM bovine S1 in ATPase buffer. The reaction was started by adding 2 mM ATP and stopped with 20% SDS and EDTA. The S1 ATPase activity was measured using the Fiske–Subbarow method (45). S1 in the absence of actin was used as a baseline. The data were fit to the Hill equation in GraphPad Prism:in which y_max is the maximum activity, y_min is the minimum activity, and n_H is the Hill coefficient.

The sequence of human cardiac TnT was retrieved from the UniProt database (accession no. P45379) (46), and the first 200 residues were used to model the N-terminal tail domain. A preliminary search for homologs to serve as templates for modeling was performed using BLAST and PSI-BLAST (47) with an E-value cutoff of 10⁻⁴ against the PDB database and four iterations. No homologs were found for the TnT N-terminal tail region except for the Tm–TnT junction protein complex (PDB ID code 2Z5H), which includes an additional 10 residues of skeletal TnT.

Fold recognition, threading, and profile-profile–based comparison methods were used to find distant homologs of known structure to use as templates for modeling. pGenTHREADER (48) and FORTE (49) were used for fold recognition (Table S1). MODBASE (50), a database consisting of protein models derived through comparative/homology modeling, was also searched to identify template structures. In all searches, the full-length protein sequence, as well as the N-terminal region alone, was separately used as a query (Fig. S2).

The MODELER (51) package and multiple templates were used for modeling the N terminus of TnT. The templates found through multiple search approaches were consolidated such that the whole query sequence was covered. Alignments were checked, and missing residues in the PDB file of the templates were accounted for in the alignments. The top 100 models that satisfied spatial restraints were generated for each query, and the model with the best molpdf score was selected. This selected model was then energy-minimized to convergence using SYBYL software. The parameters used for minimization included the TRIPOS force field, 1,000 iterations using Powell’s gradient, and were terminated at a convergence of 0.05 kcal⋅mol⁻¹⋅Å⁻¹. The distance-dependent cutoff was kept as 1, and the nonbonded interaction cutoff was kept as 8 for all rounds of minimization. The crystal structure for the Tm-skeletal TnT fragment region (PDB ID code 2Z5H) was used, in combination with the generated TNT1 model, to model the structural complex of the Tm overlap region and the N-terminal region of TnT. The models were energy-minimized using SYBYL.

The models generated were tested for global and local sequence-structure compatibility using RAMPAGE (52), VERIFY3D (53), PROSAII (54), HARMONY (55), and ANOLEA (56), all of which provided global scores (Table S2). RAMPAGE, VERIFY3D, and ANOLEA provide local goodness scores as well; however, these scores lack universal cutoffs, so our best model was selected based on comparative scoring among the multiple models. Finally, CHAHO (34) was applied to the best models to test the interaction between the clusters of charged residues present in the TNT1–Tm interaction zone.

Mutations were introduced in the basic model (Fig. 4) using PyMOL after energy minimization. The change in energy for each mutant was calculated using our in-house program COILCHECK+ (34). COILCHECK+ determines the sum of energies of hydrophobic, van der Waals, and electrostatic interactions, as well as inter- and intramolecular hydrogen bonds for the inputted coiled coil. The sum of these energies was used to calculate interprotomer energy and then divided by number of interface residues to obtain energy per residue.