Changes in the dynamics of the cardiac troponin C molecule explain the effects of Ca2+-sensitizing mutations

Cardiac troponin C (cTnC) is the regulatory protein that initiates cardiac contraction in response to Ca2+. TnC binding Ca2+ initiates a cascade of protein–protein interactions that begins with the opening of the N-terminal domain of cTnC, followed by cTnC binding the troponin I switch peptide (TnISW). We have evaluated, through isothermal titration calorimetry and molecular-dynamics simulation, the effect of several clinically relevant mutations (A8V, L29Q, A31S, L48Q, Q50R, and C84Y) on the Ca2+ affinity, structural dynamics, and calculated interaction strengths between cTnC and each of Ca2+ and TnISW. Surprisingly the Ca2+ affinity measured by isothermal titration calorimetry was only significantly affected by half of these mutations including L48Q, which had a 10-fold higher affinity than WT, and the Q50R and C84Y mutants, each of which had affinities 3-fold higher than wild type. This suggests that Ca2+ affinity of the N-terminal domain of cTnC in isolation is insufficient to explain the pathogenicity of these mutations. Molecular-dynamics simulation was used to evaluate the effects of these mutations on Ca2+ binding, structural dynamics, and TnI interaction independently. Many of the mutations had a pronounced effect on the balance between the open and closed conformations of the TnC molecule, which provides an indirect mechanism for their pathogenic properties. Our data demonstrate that the structural dynamics of the cTnC molecule are key in determining myofilament Ca2+ sensitivity. Our data further suggest that modulation of the structural dynamics is the underlying molecular mechanism for many disease mutations that are far from the regulatory Ca2+-binding site of cTnC.

Familial hypertrophic cardiomyopathy (FHC) 4 is the inherited form of hypertrophic cardiomyopathy (HCM), the most common cause of sudden cardiac death in young athletes (1), with a prevalence of 1 in 200 individuals (2). There is a growing list of over 1000 mutations that have been associated with HCM, primarily in genes that code for sarcomeric proteins such as the cardiac troponin (cTn) complex (3) (4,5). FHC is difficult to diagnose because it can be clinically asymptomatic prior to sudden cardiac death. The cTn complex is composed of three proteins: cTnC, the Ca 2ϩ sensing component; cTnI, the inhibitory subunit, and cTnT, that tethers the cTn complex to the cardiac thin filament (6). Mutations in cTnC have a pronounced functional effect because the sequence of cTnC is highly conserved throughout vertebrates (7).
In cardiac contraction, the cytosolic Ca 2ϩ concentration fluctuates between 100 nM in diastole and 400 -1000 nM during systole (8,9). When Ca 2ϩ binds to the regulatory N-terminal domain of cTnC (N-cTnC), a conformational change exposes a hydrophobic region on the surface, which binds to the "switch" region of cTnI. The Ca 2ϩ signal ultimately permits actomyosin cross-bridge formation and force production (6). Sequence substitutions in cTn components demonstrably affect the Ca 2ϩ sensitivity of force production in myofibrils, skinned cardiomyocytes, and trabeculae (10 -16). The N-cTnC-Ca 2ϩ interaction has been measured with fluorescent probes such as anilino-napthalenesulfote iodoacetamide (17)(18)(19), in which a shift in the dynamic equilibrium between populations of open and closed cTnC is reported in response to the addition of Ca 2ϩ . These experiments produced different results for the isolated N-TnC compared with experiments that include the cTn complex and cardiac thin filament proteins actin and tropomyosin (19,20). By understanding the thermodynamic basis of the function of N-cTnC, we can explain this variation and explore the specific effects of disease-associated mutations.
The function of cTnC and other Ca 2ϩ -sensing EF-hand proteins has been described as a balance between the opposing forces that push the cTnC molecule open and those that keep it closed (21). When Ca 2ϩ binds N-cTnC, it creates a strain on the molecule, which is alleviated when N-cTnC changes conformation to better accommodate the presence of the ion; however, the energetic cost of the unfavorable exposure of a hydrophobic cleft provides a thermodynamic incentive to keep the N-cTnC molecule closed (21). The balance between these forces can be disrupted by sequence substitutions that alter the ability of N-cTnC to tolerate the conformational strain imposed by Ca 2ϩ binding or substitutions that modify the hydrophobic cleft. The TnI switch peptide (TnI SW ) binds to TnC and stabilizes the open TnC conformation by occluding the hydrophobic cleft from the aqueous environment. The structural effects of Ca 2ϩ binding have been examined through NMR and X-ray crystallographic data (22)(23)(24)(25), including for the HCM-associated TnC mutant L29Q (24,26). These structures and data from MD simulations have demonstrated minimal effects of sequence substitutions on the static structure but a greater effect on the dynamics of the protein (27)(28)(29)(30).
In this study, we report the ITC-derived Ca 2ϩ -binding affinity of N-cTnC mutant constructs and their calculated effects on the dynamics, Ca 2ϩ interaction, and TnI SW interaction strengths. The N-cTnC mutations selected for analysis in this work are the FHC-associated mutations A8V (31-33), L29Q (12,19,35,36), A31S (13), and C84Y (14); the engineered Ca 2ϩsensitizing mutation L48Q (11,15,16,37,38); and the dilated cardiomyopathy-associated mutation Q50R (39) (Fig. 1). The C84Y and Q50R mutations each conferred Ca 2ϩ affinities 3-fold higher than WT, whereas the L48Q Ca 2ϩ affinity was 10-fold higher than wild type. The combination of MD simulation techniques with ITC explains the molecular etiology of these mutations in terms of the energy landscape of the conformational change. The mutations that favor the open conformation of TnC indirectly increase the Ca 2ϩ affinity of the isolated N-cTnC molecule (40 -42). We propose that mutations that increase the Ca 2ϩ sensitivity of the myofilament destabilize the closed conformation of N-cTnC, stabilize the open conformation of N-cTnC, and/or promote association with the TnI SW peptide. The results presented in this work demonstrate that many N-cTnC mutations affect myofilament Ca 2ϩ sensitivity by affecting the molecular motions that govern the regulation of cardiac contraction.

ITC
The interaction between TnC and Ca 2ϩ was endothermic for each of the TnC constructs except L48Q, which was exother- Figure 1. A, structure of the core domain of the Tn complex (PDB code 1J1E) with each of the residues that were selected for this study highlighted. The troponin complex proteins are colored white (cTnC), gray (cTnT), and black (cTnI). B, the isolated N-cTnC domain bound to Ca 2ϩ used in the PMF simulations. Helices and mutation sites are labeled. C, the N-cTnC domain bound to the TnI SW and Ca 2ϩ is the system used in the TnI SW binding simulations. D, a schematic of the N-cTnC construct. The helices, Ca 2ϩ -binding loop, and the residues being examined in this study are labeled.
mic. In each case the stoichiometric ratio (N) of Ca 2ϩ binding to N-cTnC was 1:1, indicating that the regulatory site II was exclusively titrated during these experiments. Thermodynamic parameters are listed in Table 1, and ITC isotherms are shown in Fig. 2. At 25°C each of the K d values was within error from the WT, with the exceptions Q50R and C84Y, which each had K d values approximately one-third of WT, and L48Q, in which the K d was one-tenth of WT. The ⌬S values were lower than WT for A31S, L48Q, Q50R, and C84Y. The ⌬H values were less favorable than WT for A31S, Q50R, and C84Y but more favorable than WT for the L48Q construct.

Melting points
The melting points for all Apo TnC constructs were ϳ65°C, with the exceptions of A8V at 58.5°C and L48Q at 42.5°C (supplemental Table S2).

TnC ؉ Ca 2؉ simulations
Each of the simulations diverged from the original coordinates (supplemental Fig. S1). The representative structures are very similar with a total backbone root mean square deviation of 1.9 Å. The local backbone dynamics are similar over 100 ns (supplemental Fig. S1). The mutations produce small backbone perturbations compared with WT in their respective local regions and have backbone root mean square deviation values that differ only by 1.2-2.4 Å. The mutations can; however, substantially disrupt the packing of interacting side chains for mutated residues that are not solvent-exposed: A8V, L48Q, Q50R, and C84Y. The L29Q and A31S substitutions affect fewer residues, but each introduces a hydrogen bond absent from the WT structure (Fig. 3). Compared with WT, the A8V model has a modified interaction between the N and D helices, specifically a hydrophobic interaction with that involves residues Tyr-5, Ala-8, Val-9, and Leu-12 on the N-helix and Leu-78, Val-79, and Val-82 on the D helix. In the L29Q model, Gln-29 is solvent-exposed, but the side chain amide is within hydrogen bonding distance of the Ile The Ca 2ϩ coordination geometry and distances are similar for all constructs with the exception of A8V, which has tighter Ca 2ϩ coordination geometry (supplemental Table S3). This is likely correlated with the potential of mean force (PMF) results discussed in detail below.
The average number of H-bonds in each mutant structure is the same over the course of each simulation, as are the average hydrophobic solvent-accessible surface (h-sasa) in the closed and open forms (supplemental Table S4); however, monitoring the A/B interhelical angle (Fig. 4) and h-sasa (supplemental Fig. S3) over repeated 1-s simulations demonstrates the frequency with which the TnC protein exposes the hydrophobic patch. The L29Q construct is least frequently open, followed in order by C84Y Ͻ A8V Ͻ WT Ͻ Q50R ϽA31S, and ϽL48Q. One of the replicated L29Q simulations had very high AB interhelical angles for a time; this was due to a transient loss of secondary structure, although it did not affect the exposed hydrophobic sasa (supplemental Fig. S2). The interhelical angle data can be expressed as the probability of observing the open conformation of the protein. An angle of 110°has been described as the maximum angle that can be considered open. The proportion of frames in which the N-cTnC molecule had an open AB interhelical angle was 0 for the A8V, L29Q, and C84Y constructs, 0.004% for the Q50R construct, 0.008% for WT, 0.016% for A31S, and 0.07% for the L48Q N-cTnC (Fig. 5). By defining cutoff angles for the transition between closed and open at 130°a nd 110°, we can determine the probability of observing the N-cTnC molecule in the closed state (AB interhelical angle Ͼ130°) and gain insight on the relative stability of the closed conformation (Fig. 5). The A8V construct has the lowest closed probability at 58%, followed by C84Y at 65%, WT and L48Q at 76%, Q50R at 86, A31S at 88%, and finally L29Q at 92%.

TnC ؉ Ca 2؉ ؉ TnI SW simulations
The effects on the structural interactions between the TnC molecule and TnI SW are subtle when averaged across the simulations (Fig. 6), and as expected, the majority of the specific contacts are maintained when each of the N-cTnC mutants are compared with WT N-cTnC. The TnI SW remains in contact with the TnC molecule for the duration of the simulations, but Table 1 Thermodynamic parameters derived from ITC For all mutants, n ϭ 3.

Basis of TnC calcium-sensitizing mutations
the specificity of that contact is somewhat different in response to certain mutations. For example: L29Q, A31S, and Q50R make more common contacts between the N-terminal region of TnI SW and the N-helix of TnC than WT, whereas L48Q and C84Y make fewer contacts than WT in this region. These results suggest that alterations in this binding interface may be due to modified interactions between the TnI SW and other regions of the protein near the mutation sites.
Interaction energies between TnC and the TnI SW were estimated by molecular mechanics/Poisson Boltzmann surface area (MM/PBSA) calculations. The energies reported are much larger than could be reasonably expected, although this method has been used as a means to score the relative strength interactions with some success (43). The values are similar for each of the constructs, with the exception of the A31S mutant, which has a weaker calculated binding interaction (Table 2).

Discussion
The measurements described in this study have provided novel information about the molecular basis of N-cTnC function in the regulation of cardiac muscle contraction. Muscle contraction begins when N-cTnC binds to Ca 2ϩ , opens, and interacts with TnI SW . The cTnC molecule also interacts with the cardiac-specific N-terminal extension of TnI (44), which responds to the phosphorylation of TnI residues 22 and 23 (19).
In the presence of the Tn complex, the cTnC Ca 2ϩ affinity increases by 10-fold over the isolated cTnC molecule. Improved affinity was attributed to the stabilized open conformation (5,45). Our hypothesis was that the Ca 2ϩ sensitizing mutations would directly increase Ca 2ϩ affinity on the isolated N-cTnC, whereas desensitizing mutations reduce the N-cTnC Ca 2ϩ affinity. Mutations were hypothesized to affect TnC function through directly modifying Ca 2ϩ coordination or modifying the energetic cost of the N-cTnC conformational change. Mutations in the Tn complex have been shown to modify Ca 2ϩ affinity indirectly through altering the interaction between N-cTnC and the TnI SW (5,45). We have explored this hypothesis through ITC measurements of Ca 2ϩ affinity and MD simulations that 1) assess the strength of the Ca 2ϩ -binding interaction; 2) describe the properties of the N-cTnC conformational change; and 3) measure the strength of the interaction between N-cTnC and the TnI SW peptide.
It is challenging to directly measure the N-cTnC-Ca 2ϩ interaction. Fluorescence-based experiments employ reporters such as anilino-napthalenesulfote iodoacetamide (17)(18)(19) or an F27W mutation (46). Upon titration with Ca 2ϩ , the fluorophore reports the N-cTnC conformational change as a measure of N-cTnC Ca 2ϩ affinity, whereas measurements made using stopped flow fluorospectroscopy report rates of Ca 2ϩ dissociation (15). The K d for WT TnC has been previously reported fluorometrically as 11.3 M for WT N-cTnC, which was lowered to 8 M with the L29Q mutation (12). Similarly, fluorescence titration of the N-cTnC-Ca 2ϩ interaction yielded midpoint values of 12.3 M for WT N-cTnC, 12.9 M for A8V, and 37.2 M for C84Y (5). Our ITC experiments, which consider the unmodified N-cTnC molecule, demonstrated no statistically significant differences in K d between WT and the FHC-associated mutants, with the exception of C84Y (Table 1). Our measurements of the dilated cardiomyopathy-associated mutant Q50R and engineered Ca 2ϩ -sensitizing mutation L48Q have shown 3-and 10-fold increases in affinity for Ca 2ϩ , respectively. Our results agree with the previously reported K d for full-length cTnC at 24 M (16), which decreased to 1.9 M in the L48Q construct (16). When compared with another ITC study of the WT N-cTnC-Ca 2ϩ interaction, the ⌬G values are ϳ3 kJ⅐mol Ϫ1 lower (47). Each of A31S, L48Q, Q50R, and C84Y produced increased ⌬H and decreased ⌬S values relative to WT, although the molecular basis of these changes is not necessarily the same. The A31S mutation stabilizes the loop between helices A and B with a hydrogen bond, which accounts for changes in both ⌬H and ⌬S. The L48Q, Q50R, and C84Y mutations are along the interface between the NAD and BC helical bundles, which likely reduces the entropic cost of the closed/open transition by introducing a polar residue into a hydrophobic region. Each of these substitutions also creates at least one new hydrogen bond that affects ⌬H (Fig. 3).
The increases in Ca 2ϩ affinity of L48Q and Q50R N-cTnC were attributed to the reduced cost of exposing the hydrophobic patch. Our ITC experiments reported lower ⌬S values for these mutants (Table 1), associated with the exposure of hydrophobic residues. The changes in ⌬S are consistent with our measurement of the A/B interhelical angle, as the molecule transitions into the open form; hydrophobic residues in the interface between the NAD helical bundle and the BC helical bundle are exposed. This has been demonstrated in another MD-based study of the L48Q and V44Q N-cTnC mutations (38). In the A31S mutation, the ⌬S value is lower than that of WT, perhaps because of the hydrogen bond formed by the serine, which reduces the mobility of the loop, a finding consistent with a previous exploration of A31S (13). The C84Y mutant disrupts the side-chain packing between Cys-84 and several residues between helices C and D and produces a more favorable open conformation (Fig. 3). Despite the overall structural similarity of the mutant constructs, there are changes in sidechain packing caused by each substitution compared with the WT model (Fig. 3). The melting points were lower for the A8V

Basis of TnC calcium-sensitizing mutations
and L48Q mutants, consistent with the observation that they spend the least time in the closed conformation. That the hydrophobic interactions disrupted by these mutations have an important role in the stability of the N-TnC molecule (Fig. 5).
Small-angle X-ray scattering has been used to investigate tertiary protein contacts in the apo and Ca 2ϩ -bound states of the A8V and A31S mutants. Despite minimal structural changes in these mutants, this technique can be used to uncover poten- tially lower stability resulting from salient local changes in FHC mutants (48).
The strengths of the charge-charge interactions that govern the direct Ca 2ϩ -N-cTnC interaction are proportional to the distance between the Ca 2ϩ ion and the coordinating oxygen atoms (supplemental Table S2). The PMF-derived ⌬G of Ca 2ϩ binding of the L48Q mutation was most similar to the WT protein, and the remaining mutations each yielded stronger Ca 2ϩ interactions (Fig. 7). Notably the A8V interaction was the strongest, and the Ca 2ϩ coordination was the tightest in 100-ns simulations; however, this was not found in the PMF calculation based on the 1-s simulation (Fig. 7). The absolute values of the interaction energies are overestimated because of the parameterization of Ca 2ϩ in the simulation; however, the results are useful as a relative measure of the change in free energy of binding (49).
In skeletal TnC, the affinity for Ca 2ϩ is inversely related to protein stability. It is reasonable to assume that similar mechanisms govern the cardiac isoforms of this protein (50). The affinity of N-cTnC for Ca 2ϩ is set by a balance between the conformational strain induced by Ca 2ϩ binding that is acting to open the N-cTnC molecule and the energetic cost of exposing a hydrophobic cleft (21). Introduction of a polar amino acid into the hydrophobic cleft reduces the cost of opening N-cTnC by destabilizing the closed conformation or stabilizing the open conformation (41). Ca 2ϩ binding causes a change in the dynamic equilibrium of the populations of open and closed TnC molecules. When Ca 2ϩ is bound, 20 -27% of the N-cTnC population is open (33,51). Previous MD simulations have estimated the ⌬G of the WT cTnC open/closed transition at 33.5 kJ⅐mol Ϫ1 (27). The N-cTnC conformational change necessarily precedes the TnI SW interaction, and therefore changes in the favorability of the conformational change are related to the probability of N-cTnC binding to the TnI SW (27).
Monitoring the A/B interhelical angle through 1-s simulations has revealed that mutations affect the conformational dynamics of N-cTnC (Fig. 5). A previous study had defined the N-cTnC molecule as "open" with an AB interhelical angle Ͻ110° (27). We define a closed structure as any with an AB interhelical angle Ͼ130°, similar to the NMR structure of N-cTnC (22). This allows for the quantification of the relative stability of the closed state (Fig. 5). The open conformation is found most frequently in simulations of the L48Q construct, followed by the A31S mutation. Work by Marques et al. (48) shows that A31S may cause greater exposure in the primed state of the full-length TnC (i.e. when Ca 2ϩ is bound to sites III/IV). A similar degree of openness was observed for the remaining constructs, a finding corroborated the h-sasa (supplemental Fig. S3). The L29Q mutant was the least open; interhelical angles below 115°were not observed and had the highest closed probability. The closed state was least frequently observed in the A8V construct, which is consistent with paramagnetic NMR data showing that A8V cTnC opened more readily than WT N-cTnC (33) and contribute to the increased affinity between A8V TnC and TnI SW (31). These data suggest that the molecular etiology is different for each mutation, despite producing a similar disease phenotype.
The ⌬G of N-cTnC and TnI SW interaction is similar to WT for each of the N-cTnC mutants with the exception of A31S, which is ϳ75% of the WT ⌬G ( Table 2). As with the PMF calculations, the ⌬G values derived from MM/PBSA calculations provide insight into the relative strengths, but the absolute values are not expected to correspond to more computationally expensive calculations or experimentally derived measurements (43). The A31S N-cTnC has a modified interface with TnI SW , which favors closer interactions with the N-terminal region of TnC and longer distance interactions with remainder of N-cTnC, particularly in the vicinity of the mutation (Fig. 6). Relative to WT, each mutant has a slightly different interaction with TnI SW but did not produce a change in interaction energy. We have previously observed a similar effect in the zebrafish TnC-TnI SW interaction in which TnI substitution had a greater effect on the interaction than TnC substitution (52). There is fluorescence-based evidence that both the L48Q (53) and A8V (31) N-cTnC mutations increase the affinity of N-cTnC for the TnI SW ; our results suggest that this affinity change may be due, in part, to the ability of these mutants to open more readily than WT, which generates more opportunity for TnC-TnI interaction.
This work provides insight into how the dynamics of N-cTnC can govern the interactions with each of Ca 2ϩ and cTnI, which in turn influence sarcomeric Ca 2ϩ sensitivity. The engineered L48Q mutation has the most salient effect on the N-cTnC ITC-derived Ca 2ϩ binding and on the dynamics of the N-cTnC molecule. A finding that has been corroborated, experimentally (16,53), in silico (38) and in vivo (11,37). The L48Q mutation has a large disruptive effect on the hydrophobic interactions that maintain N-cTnC in the closed conformation, which creates increased Ca 2ϩ affinity through a modification of the thermodynamic landscape of the conformational change and allows the TnI SW to bind to N-cTnC more readily (53). Through these molecular changes, the L48Q mutation can produce a positive inotropic effect or, with higher sarcomere incorporation and ␤-blocking drugs, can produce hypertrophy in

Basis of TnC calcium-sensitizing mutations
murine models (11,37). The L29Q mutation was similar to WT in each of our measurements; However, L29Q has been shown to affect length-dependent activation and modify the response to phosphorylation of serine 22/23 of cTnI (19,26,54,55). The A8V and C84Y mutations are near the interface with the N-helix of TnI (Fig. 1) and may have a more pronounced effect on the orientation of N-cTnC in the Tn complex (56).
Our results support the model for the molecular mechanism of Ca 2ϩ binding that is dictated by the favorability of the conformational change and the stability of the TnI-TnC interaction (42). Mutations modify the structural dynamics of TnC rather than the regulatory Ca 2ϩ -binding site. The changes are observed in the relative favorability of the protein conforma-tions that transduce the contraction signal (Fig. 8). This can increase Ca 2ϩ sensitivity of contraction by destabilizing the closed conformation of N-cTnC, stabilizing the open conformation, or stabilizing the TnI SW interaction. These changes lead to an increase in the Ca 2ϩ buffering capacity of the myofilament that may increase the duration of the Ca 2ϩ transient and is consistent with observations of the greater capacity of the Tn complex to bind Ca 2ϩ than the isolated N-cTnC molecule (16,20,45). The complimentary use of MD, ITC, and fluorometric techniques provides detailed information about the molecular etiology of cTnC mutations. A complete understanding of the molecular and thermodynamic basis for myofilament Ca 2ϩ sensitivity will inform the risk stratification of Distance (nm) Figure 6. Average distance between cTnC and cTnI residues. The mutated TnC residue in each plot is indicated by a gray bar. The structures to the right are representative structures of independent simulations and indicate the differences in the orientation and variability of the cTnI SW peptide across replicates for each mutant. The TnI SW is colored as a spectrum from blue at the N terminus to red at the C terminus. The calculated ⌬G of interaction is maintained across mutations despite differences in the interaction distance profiles, which suggests a nonspecific interaction. The A31S mutant has a ⌬G of interaction with the TnI SW ϳ25% lower than WT (Table 2), perhaps because of shorter interaction distances with the N-terminal region of N-cTnC but longer interaction distances in the vicinity of the A31S mutation and C-terminal portion of the TnI SW .
existing and novel FHC-associated mutations and the selection of appropriate therapeutics and enable rational drug development that specifically target or compensate for the unique deleterious effects of these disease mutations.

Experimental procedures
The codon-optimized gene sequence for human TNNC1 was cloned into the pET21a(ϩ) vector (Novagen). A stop codon was introduced at position 90 by following the Phusion site-directed mutagenesis protocol (Thermo). Protein expression, purification, melting point determination, and ITC were carried out using protocols modified from previous work (30). The N-cTnC construct corresponds to the human TNNC1 gene (Uniprot ID P63316), which was cloned into the pET-21aϩ expression vector (Novagen). The codon corresponding to residue 90 was mutated to a stop codon, and individual mutations were introduced with the Phusion site-directed mutagenesis kit (Thermo). Mutated constructs were sequenced and transformed into the BL21(DE3) host strain. Overnight cultures were grown in lysogeny broth supplemented with 50 g/ml ampicillin at 37°C overnight with shaking at 250 RPM, 1% subcultures were grown for 3 h followed by induction with 1 mM isopropyl ␤-D-1-thiogalactopyranoside and a further 3 h of growth. The cells were harvested by centrifugation and resuspended in 50 mM Tris-HCl, pH 8.0, 1 mM PMSF, 5 mM EDTA, and a cOmplete TM protease inhibitor tablet (Roche). The cells were sonicated on ice at 80% amplitude with 30-s pulses separated by 30 s. The lysate was centrifuged at 30,000 ϫ g for 30 mins, and the supernatant was decanted. The protein was purified with a fast-flow DEAE column (GE Healthcare). The column was equilibrated with 50 mM Tris-HCl, pH 8.0, 1 mM DTT, and 5 mM EDTA, and the protein was eluted with a 180-ml NaCl gradient up to 0.55 M. Fractions containing the TnC protein were retained and concentrated to 5 ml with an Amicon centrifugal concentrator with a molecular mass cutoff of 3 kDa (Millipore). The protein was further purified with a HiPrep 26/60 Sephacryl S-100 column (GE Healthcare) equilibrated with 50 mM Tris-HCl, pH 8.0, 100 mM NaCl, and 1 mM DTT. The fractions that contain the pure N-cTnC protein were pooled, concentrated, and stored at Ϫ80°C.
ITC buffer contained 50 mM HEPES, pH 7.2, and 150 mM KCl, protein samples were dialyzed three times against 2 liters of ITC buffer and diluted to 200 M. In successive dialysis steps, 15 mM ␤-mercaptoethanol (BME) and 2 mM DTT, 15 mM of BME, and 2 mM of BME were added to the buffer. An extinction coefficient of 1490 M Ϫ1 cm Ϫ1 and a molecular mass of 10.1 kDa were used to measure protein concentration. The Ca 2ϩ solution was diluted from a 1 M Ca 2ϩ solution (Sigma) into the buffer from the final dialysis step to a final concentration of 4 mM. Ca 2ϩ was titrated into the protein solution by a single 0.4-l injection, followed by a series of eighteen 2-l injections at 2-min intervals while stirring at 1000 rpm. The experiments were carried out at 25°C. The heat of dilution of Ca 2ϩ was accounted for by subtracting the average of the final three data points from the titration curve. Analysis was performed with Origin 8.0 (OriginLab, Northampton, MA).
To measure the melting point of the proteins, the samples were first dialyzed four times against 2 liters of MT buffer (10 mM HEPES, pH 7.5, 150 mM KCl, 3 mM MgCl 2 , 2 mM EGTA) and combined with 2.5 l of 100-fold diluted SYPRO orange (Thermo) to a final concentration of 3 mg/ml. The temperature was increased from 4 to 95°C at 5-s intervals using a CFX96 Touch real-time PCR system (Bio-Rad). The melting point was determined at the midpoint of the unfolding transition, which is indicated by the peak of the first derivative curve.
Equilibrium MD simulations of WT and mutant TnC were performed as previously described (30,52) with the exception that all calculations in this study were performed at 300 K. Structural models of the mutant constructs were generated with the Swiss model workspace (57), The N-cTnC ϩ Ca 2ϩ models used the NMR-derived structure of human N-TnC as a template (PDB code 1AP4) (22), and the models of the human N-TnC in complex with TnI SW were based on the WT NMR derived structure of N-cTnC in complex with Ca 2ϩ and cTnI residues 147-163 (PDB code 1MXL) (23).
The structural models were simulated using GROMACS 4.6.5 (58), and the AMBER99sb-ILDN force field (59); the models were placed in a periodic, cubic simulation system and solvated with the TIP3P water model (60); and the charges were neutralized with the addition of K ϩ or Cl Ϫ ions. The composition of each system is listed in supplemental Table S1. The systems were energy minimized using the steepest descent algorithm to a tolerance of 10 kJ⅐mol Ϫ1 ⅐nm Ϫ1 followed by conjugate gradient minimization for 10,000 steps. 100 kJ⅐mol Ϫ1 ⅐nm Ϫ1 restraints were placed on every protein and Ca 2ϩ atom, and the system was simulated for 1 ns to allow the water to equilibrate around the protein.
The WT and mutant N-cTnCϩCa 2ϩ systems were then simulated for either 100 ns or 1 s total time, and the WT and mutant N-cTnCϩcTnI SW ϩCa 2ϩ constructs were simulated for 100 ns total (Fig. 1, B and C). All of the simulations were performed with Berendsen pressure coupling (61) with a T of 0.1. V-rescale temperature coupling (62) with a P of 4.0, PME (particle mesh Ewald) electrostatics (63) with a grid spacing of 0.12 nm and interpolation order of 6, and the Verlet cutoff scheme was used with a 1.0-nm cutoff (64). Bond lengths were constrained with the LINCS algorithm (65).
Clustering was carried out over the backbone and C␤ atoms of each construct using the Daura algorithm (66). The degree of the open/closed N-cTnC conformational change and protein

Basis of TnC calcium-sensitizing mutations
stability were assessed through measurements of the solventaccessible surface area, the interhelical angles, and the number of hydrogen bonds, which were calculated using g_sas (67), Interhlx (68), and g_hbond (58), respectively. The g_hbond program used a cutoff radius of 3.5 Å and a 30°angle to define a given hydrogen bond.
Umbrella sampling and PMF calculations were performed as described previously (30). The Ca 2ϩ was extracted from the N-cTnC molecule by restraining the ␣-helical C␣ atoms with a force constant of 1000 kJ⅐mol Ϫ1 ⅐nm Ϫ1 and restraining the Ca 2ϩ ion in the Y and Z dimensions with a force constant of 1000 kJ⅐mol Ϫ1 ⅐nm Ϫ1 . A constraint pulling force in the X direction as applied at 0.01 Å per second until the Ca 2ϩ ion was 5 nm from the N-cTnC molecule. The conformations for umbrella sampling were extracted from the resulting trajectory at distance intervals of 0.5 Å between 0 and 1 nm, every 1 Å between 1 and 2 nm, and every 2 Å between 2 and 5 nm. Umbrella simulations were run with the same parameters as the pull simulations, with the "pull rate" parameter set to 0, and were unrestrained aside from a single restraining potential between the center of mass of the N-cTnC molecule and the center of mass of the Ca 2ϩ ion. These simulations were run for 30 ns each, and a potential of mean force was calculated with the weighted histogram analysis method through the use of g_wham (69), and errors were estimated with 5000 bootstraps of the weighted histogram analysis method calculation. To further explore the outlier A8V mutation, the process was repeated using the center struc-ture of the most populous cluster from the five replicated 1-s simulations.
Interaction energies between the TnC models and TnI SW were calculated with g_mmpbsa (34) 100 ps apart over the last 10 ns of each equilibrium simulation of the TnI SW -N-cTnC complex. The MM/PBSA calculations used the non-linear Poisson-Boltzmann equation, and calculations were performed at 300 K, with a solvent dielectric constant of 80, and a probe radius of 1.4 Å. Contact maps of the interacting surfaces between N-cTnC and the cTnI SW were calculated over the final 50 ns of each simulation and were based on measurements made with g_mindist (58).

Figure 8. Schematic of the energetic landscape of N-cTnC activation.
N-cTnC is shown as a cartoon. Ca 2ϩ is a blue circle, and the TnI switch peptide is represented as a red ellipse. Lower energy states are more favorable. The orange arrows represent the resistance to the conformational change caused by the hydrophobic cleft. The blue arrows indicate conformational strain introduced by Ca 2ϩ binding. The Ca 2ϩ -bound, open conformation relieves the conformational strain while occluding the hydrophobic cleft and is therefore the most favorable conformation. Mutations that affect the relative stabilities of these states will modify the probability of transitions between them and increase or decrease the Ca 2ϩ sensitivity of the myofilament.