The intrinsically disordered C terminus of troponin T binds to troponin C to modulate myocardial force generation

Aberrant regulation of myocardial force production represents an early biomechanical defect associated with sarcomeric cardiomyopathies, but the molecular mechanisms remain poorly defined. Here, we evaluated the pathogenicity of a previously unreported sarcomeric gene variant identified in a pediatric patient with sporadic dilated cardiomyopathy, and we determined a molecular mechanism. Trio whole-exome sequencing revealed a de novo missense variant in TNNC1 that encodes a p.I4M substitution in the N-terminal helix of cardiac troponin C (cTnC). Reconstitution of this human cTnC variant into permeabilized porcine cardiac muscle preparations significantly decreases the magnitude and rate of isometric force generation at physiological Ca2+-activation levels. Computational modeling suggests that this inhibitory effect can be explained by a decrease in the rates of cross-bridge attachment and detachment. For the first time, we show that cardiac troponin T (cTnT), in part through its intrinsically disordered C terminus, directly binds to WT cTnC, and we find that this cardiomyopathic variant displays tighter binding to cTnT. Steady-state fluorescence and NMR spectroscopy studies suggest that this variant propagates perturbations in cTnC structural dynamics to distal regions of the molecule. We propose that the intrinsically disordered C terminus of cTnT directly interacts with the regulatory N-domain of cTnC to allosterically modulate Ca2+ activation of force, perhaps by controlling the troponin I switching mechanism of striated muscle contraction. Alterations in cTnC–cTnT binding may compromise contractile performance and trigger pathological remodeling of the myocardium.

Aberrant regulation of myocardial force production represents an early biomechanical defect associated with sarcomeric cardiomyopathies, but the molecular mechanisms remain poorly defined. Here, we evaluated the pathogenicity of a previously unreported sarcomeric gene variant identified in a pediatric patient with sporadic dilated cardiomyopathy, and we determined a molecular mechanism. Trio whole-exome sequencing revealed a de novo missense variant in TNNC1 that encodes a p.I4M substitution in the N-terminal helix of cardiac troponin C (cTnC). Reconstitution of this human cTnC variant into permeabilized porcine cardiac muscle preparations significantly decreases the magnitude and rate of isometric force generation at physiological Ca 2؉ -activation levels. Computational modeling suggests that this inhibitory effect can be explained by a decrease in the rates of cross-bridge attachment and detachment. For the first time, we show that cardiac troponin T (cTnT), in part through its intrinsically disordered C terminus, directly binds to WT cTnC, and we find that this cardiomyopathic variant displays tighter binding to cTnT. Steady-state fluorescence and NMR spectroscopy studies suggest that this variant propagates perturbations in cTnC structural dynamics to distal regions of the molecule. We propose that the intrinsically disordered C terminus of cTnT directly interacts with the regulatory N-domain of cTnC to allosterically modulate Ca 2؉ activation of force, perhaps by controlling the troponin I switching mechanism of striated muscle contraction. Alterations in cTnC-cTnT binding may compromise contractile performance and trigger pathological remodeling of the myocardium.
Dilated cardiomyopathy (DCM) 2 is a leading cause of heart failure and the most frequent indication for heart transplantation in both children and adults (1)(2)(3). DCM is characterized clinically by ventricular wall thinning, chamber volume enlargement, and contractile (systolic) dysfunction in the absence of coronary artery disease (4,5). Primary DCM can result from genetic mutations in proteins related to diverse cellular functions, ranging from muscle contraction to RNA splicing and gene transcription (6,7). Although pathways that contribute to the DCM phenotype are complex and heterogeneous, sarcomeric protein mutations associated with DCM have been shown to alter the force-generating capacity of the myocardium by directly impacting myofilamentbased regulation of contraction (8).
Genetic variants account for a significant proportion of cardiomyopathies in both children and adults, which underscores the importance of fully evaluating sequence variations associated with cases of cardiomyopathy in humans (9,10). However, establishing the pathogenicity for sarcomere gene variants can be challenging in the absence of genetic segregation and linkage analyses (4). This is often the case for rare variants that arise de novo, which are often associated with lethality at early stages of development. Fortunately, structural and functional investigations into cardiomyopathy-associated variants can aid in establishing the potential genotype-phenotype relation in the absence of genetic linkage analysis, as well as provide mechanistic insights into physiological muscle regulation (11)(12)(13)(14)(15). Moreover, these studies bear the potential to unveil novel therapeutic targets to improve contractile performance in disease, and perhaps aging, by directly targeting the contractile machinery (16).
The sarcomere is a fundamental structural unit of striated muscle cells that is responsible for contractile function. This macromolecular complex is primarily composed of overlapping thick (myosin) and thin (actin, tropomyosin, and troponin) myofilaments that, seen in a cross-section, form a double-hexagonal lattice (8). Cardiac troponin (cTn) is a heterotrimeric complex that plays a central role in regulating muscle contraction by functioning as the Ca 2ϩ sensor in the contractile apparatus. It is composed of a Ca 2ϩ -binding subunit, cardiac/slow skeletal muscle troponin C (cTnC); an inhibitory subunit, cardiac troponin I (cTnI); and a tropomyosin-binding subunit, cardiac troponin T (cTnT) (17). cTn regulates contraction-relaxation through a series of conformational rearrangements and myofilament protein-protein interactions in response to cardiomyocyte intracellular Ca 2ϩ transients (18).
During the contractile (systolic) period of the cardiac cycle, elevated intracellular free Ca 2ϩ results in Ca 2ϩ binding to a single, low-affinity (ϳ10 5 M Ϫ1 ) regulatory site in cTnC (19). In this Ca 2ϩ -activated state, an interaction between cTnI and cTnC is stabilized, which shifts tropomyosin into a position that exposes myosin-binding sites on actin to allow crossbridge cycling and force generation (18). As free intracellular Ca 2ϩ levels fall during the relaxation phase (diastole), Ca 2ϩ dissociates from cTnC, and cTnI and tropomyosin are repositioned on actin filaments to inhibit strong myosin binding (18). Although it is well-established that cTnI-cTnC interactions are critical for cTn function, little is known about the potential functional significance of cTnT-cTnC interactions.
The TnT subunit has been viewed primarily as the subunit that anchors cTn to the thin filament via its highly-extended N terminus (20). However, there is evidence that cTnT might also have a regulatory role. Analysis of the crystal structure of the cTn core domain and results from small-angle neutronscattering experiments indicate putative interactions between cTnC and C-terminal region of cTnT (21,22). In addition, it was previously shown that skeletal muscle TnT is essential to confer Ca 2ϩ sensitivity of tension and ATPase activity independent of cTnT-cTnI interactions (23). Functional studies on a human cardiomyopathic variant that results in deletion of the last 14 C-terminal residues in cTnT have further highlighted the potential importance of cTnT in cardiac muscle regulation (24 -28). Taken together, these observations raise the possibility that cTnT directly interacts with cTnC for Ca 2ϩ -dependent control of myocardial force generation, alterations that may trigger contractile dysfunction and pathological remodeling of the heart.
In this report, we describe a previously unreported de novo, nonsynonymous variant (p.I4M) in the cTnC gene (TNNC1) that was identified in a pediatric patient with severe DCM. The objective of this study was to ascertain whether or not there is sufficient evidence to establish this variant as pathogenic and to elucidate an underlying molecular mechanism for its potential pathogenicity. By combining in silico, physiological, and biophysical approaches, we demonstrate this variant satisfies the American College of Medical Genetics and Genomics (ACMG) criteria for strong evidence of pathogenicity (29). We also identify a novel intersubunit interaction between cTnC and the C terminus of cTnT, which is disrupted by this cardiomyopathic variant. Our findings suggest that cTnC-cTnT interactions are potentially critical for modulation of contractility.

Clinical evaluation
The proband was a female neonate who presented with a heart murmur shortly after birth and was subsequently recommended for a follow-up. Medical records noted difficulty feeding and severe failure to thrive since birth. At 6 weeks of age, the proband underwent an extensive clinical work-up and was diagnosed with sporadic DCM. Complete blood count and comprehensive metabolic panel were reported to be normal. Enzymatic activity testing for Pompe disease was negative. On physical examination, severe hypotonia, global developmental delay, and nondysmorphia were noted. One parent is an Ashkenazi Jew and the other parent is Ashkenazi and Moroccan Jew. The proband had a negative family history for cardiomyopathy. Echocardiographic imaging revealed left ventricular hypertrabeculation and overt left ventricular and left atrial dilation (Fig. 1A). Electrocardiography indicated sinus rhythm with diagnostic criteria for biventricular hypertrophy, right atrial enlargement, and diffuse ST-T wave changes (Fig. 1B). Based on the clinical presentation, the proband and parents underwent genetic testing on the basis of a suspected genetic variant.

Molecular genetic analysis
Trio whole-exome sequencing revealed a previously unreported heterozygous variant of uncertain significance in exon 1 of TNNC1, in which a cytosine (C) to guanine (G) transversion at nucleotide position 12 was detected in the cDNA from the proband (c.12C3 G) (Fig. S1). This codon alteration (ATC 3 ATG) encodes for a nonsynonymous substitution, whereby isoleucine is replaced by methionine at position 4 in cTnC (cTnC-I4M). Both parents were genotype-negative for this variant ( Fig.  2A) and appeared to be clinically unaffected, suggesting that it arose de novo. No other genetic variants were identified in the sequencing. At 1 year old, the proband succumbed to acute viral infection with exacerbation of heart function ( Fig. 2A). This variant is absent in the Genome Aggregation database, suggesting it is not merely a benign polymorphism found in the general population of diverse racial and ethnic backgrounds (30). Furthermore, TNNC1 p.I4M was absent in the on-line database, ClinVar, at the time of this report (31). The I4M variant is located in the N-helix, which is positioned between the N-and C-domain in the tertiary structure of Ca 2ϩ -saturated cTnC (Fig. 2B). This affected residue, like most of the primary sequence of cTnC, is highly conserved across species (Fig. S2), suggesting that any protein-altering variant would not be tolerated. Indeed, 2 out of 4 in silico prediction programs revealed that TNNC1 p.I4M is predicted to be pathogenic (Table S1) (32)(33)(34)(35). Based on these collective findings, we carried out structural and functional studies to evaluate the putative pathogenicity and underlying mechanism of this variant.

Modulation of myocardial contraction by TnC-TnT interaction
There was also no significant difference in maximal restored tension between cTnC-WT (89.6 Ϯ 8.3%) and cTnC-I4M (88.0 Ϯ 10.0%), suggesting that the observed effects cannot be explained by differences in incorporation of exogenous cTnC-WT or cTnC-I4M ( Table 1).
The rate constant of isometric tension redevelopment (k TR ) provides insight into both actin-myosin cross-bridge kinetics and Ca 2ϩ -dependent dynamics of thin filament regulatory units. Therefore, we hypothesized that this Ca 2ϩ -desensitizing mutation would impact k TR . CMP reconstituted with cTnC-14M exhibited a slower rate of isometric tension redevelopment compared with cTnC-WT at all tested levels of Ca 2ϩ activation (Fig. 4, left). Noteworthy, this decrease in k TR occurred at both submaximal and maximal Ca 2ϩ activation (Fig. 4, left). The difference in the mean maximum k TR between cTnC-WT (7.0 Ϯ 0.54 s Ϫ1 ) and cTnC-I4M (4.64 Ϯ 0.4 s Ϫ1 ) was statistically significant ( Table 1), suggesting that this variant has an inhibitory effect on cross-bridge cycle kinetics in cardiac muscle, in addition to its influence on Ca 2ϩ sensitivity, an effect that is typically associated with changes in individual regulatory unit dynamics.
The 33% decrease in maximum k TR observed for cTnC-I4M could be explained by changes in the apparent cross-bridge attachment rate (f app ) and/or detachment rate (g app ), since maximal k TR ϰ f app ϩ g app . Therefore, we used a computational approach to estimate the best-fit values for f app , g app , and k off based on the three-state model of striated muscle contraction (36). We found 23 and 33% slower f app and g app , respectively, for cTnC-I4M compared with cTnC-WT (Fig. 4, right and Table  2). These data suggest that the slower k TR values observed for this variant derive from decreases in both cross-bridge attachment (f app ) and detachment (g app ) rates. Taken together, these results indicate that reconstitution of cTnC-I4M into CMP causes deficits in both the rate and magnitude of isometric tension generation at a given level of Ca 2ϩ activation.

Evidence for direct cTnT-cTnC binding
Recent studies suggest that the last 14 residues at the C terminus of cTnT are important for limiting thin-filament activation in the presence of Ca 2ϩ , but the mechanism is unknown. Therefore, we hypothesized that this region of cTnT directly interacts with cTnC to control Ca 2ϩ activation of the myofilament, the alterations in which could mechanistically explain the reduced myofilament Ca 2ϩ sensitivity and k TR value observed for cTnC-I4M.
To determine whether the C terminus of cTnT interacts with cTnC, we performed chemical cross-linking followed by MS analysis on reconstituted WT cTn in the presence of Ca 2ϩ . Prior to cross-linking, successful ternary complex formation was verified by native gel analysis (Fig. S3). Strikingly, we identified high-confidence cross-linked peptides corresponding to intersubunit interactions between Lys-282 of cTnT and Lys-6 in the N-helix of cTnC in the "light band" (Fig. 5A and Fig. S4). This finding is based on the corresponding mass spectrum in Fig. 5B, where a 4.0272-Da change in mass is characteristic of peptide cross-linking by BS3-d 0 /d 4 . Consistent with the crystal structure of the core domain of cTn, we also identified crosslinked peptides corresponding to intersubunit cTnT and cTnI interactions in the rigid I-T arm of the ternary complex in the "dark band" (Fig. 5A and Table S2).
Based on the chemical cross-linking results, we hypothesized that the C terminus of cTnT directly interacts with cTnC. To address this question, we titrated cTnC-WT or cTnC-I4M to a synthetic peptide (Fig. 6A) corresponding to the last 14 amino acids at the C terminus of cTnT and monitored intrinsic tryptophan (Trp-287) fluorescence using steady-state fluorescence spectroscopy. Titration of cTnC-WT or cTnC-I4M with the cTnT C-terminal peptide resulted in a concentration-dependent increase in tryptophan fluorescence and a blue-shifted emission wavelength maximum (Fig. 6, B and C). Normalized tryptophan fluorescence was plotted against free cTnC concentration to calculate K d values for cTnC-WT and cTnC-I4M (Fig. 6D). There was a statistically significant difference between the binding constants for cTnC-WT (K d ϭ 14.7 Ϯ 0.5 M) and cTnC-I4M (K d ϭ 10.9 Ϯ 0.4 M) ( Table 3). These  Table 2, and the dashed lines indicate the solutions from the sensitivity analysis whereby all four parameters were increased (upper traces) or decreased (lower traces) by 2%. Force values were normalized to the maximum force of cTnC-WT-reconstituted CMP. Data are represented as mean Ϯ S.D.

Table 1 Summary of contractile parameters for reconstituted CMP
The number of independent experiments for restored tension, pCa 50 , and n Hill was n ϭ 9 CMP for cTnC-WT and n ϭ 8 CMP for cTnC-I4M. For k TR, max , n ϭ 8 CMP for cTnC-WT and n ϭ 5 CMP for cTnC-I4M. Exclusion criteria for certain CMPs are described under "Experimental procedures." Data are represented as mean Ϯ S.D.

Modulation of myocardial contraction by TnC-TnT interaction
results show, for the first time, that the intrinsically disordered C terminus of cTnT interacts with cTnC-an interaction that is strengthened in the case of cTnC-I4M.
To determine whether full-length cTnT interacts directly with cTnC and to assess potential binding alterations for this cTnC variant, we used microscale thermophoresis (MST) to quantify cTnT-cTnC binding in the presence of Ca 2ϩ . The results indicate full-length cTnT directly binds to cTnC (Fig.  7A). Calculation of the equilibrium dissociation constant (K d ) yielded 13.2 Ϯ 1.6 M for cTnC-WT and 4.84 Ϯ 1.1 M for cTnC-I4M (Fig. 7A). Note that full-saturation was not feasible due to solubility issues of full-length cTnT at low ionic strength. These results indicate that full-length cTnT  Table 3 Affinity constants for cTnT peptide binding to cTnC Parameters were obtained by fitting the titration data in Fig. 6D

Modulation of myocardial contraction by TnC-TnT interaction
interacts with cTnC-WT and that the binding affinity is enhanced 2.7-fold for the cTnC-I4M variant.
Although we showed that isolated cTnC and cTnT subunits could form an interaction, we wondered whether this interaction was maintained in the presence of other major thin filament binding partners using a label-free approach. Therefore, we tested whether or not the cTnT-cTnC interaction also occurs in the context of F-actin and tropomyosin using a cosedimentation assay. We found that both cTnC-WT and -I4M associated with full-length cTnT, F-actin, and tropomyosin in the pellet upon ultra-centrifugation (Fig. 7B, leftmost and middle gels). To exclude the possibility that cTnC was forming an interaction with the thin filament proteins independent of cTnT, the assay was performed in the absence of cTnT, and we found that cTnC no longer associated with F-actin and tropomyosin in the pellet (Fig. 7B, rightmost gel). The co-sedimentation assay revealed that a statistically significant greater proportion of cTnC-I4M associated with cTnT-F-actin-tropomyosin in the pellet compared with cTnC-WT (Fig. 7C), further supporting the concept that this cTnC variant results in tighter binding to cTnT. These results suggest that cTnC interacts with cTnT in the setting of the thin myofilament apparatus.

Structure and backbone dynamics of cTnC
Next, we asked the question of whether perturbations in cTnC-I4M structure and/or dynamics could explain the observed effects on myofilament function and altered cTnT binding. Ca 2ϩ -induced hydrophobic exposure of cTnC is critical for activation of the thin filament and muscle contraction. To specifically examine the overall structural profiles of Ca 2ϩinduced hydrophobic exposure in cTnC, increasing concentrations of free Ca 2ϩ (pCa 8.0, 6.0, and 4.0) were titrated to cTnC-WT or cTnC-I4M with bis-ANS, a fluorescent probe sensitive to hydrophobic environments (Fig. 8, A and B). Examination of the bis-ANS fluorescence-emission curves revealed that both cTnC-WT and cTnC-I4M displayed identical fluorescence intensities at very low Ca 2ϩ (pCa 8.0), suggesting that there was no difference in hydrophobic exposure prior to Ca 2ϩ titrations (Fig. 8, A and B). Interestingly, at both sub-saturating (pCa 6.0) and saturating (pCa 4.0) Ca 2ϩ concentrations, the magnitude of the fluorescence intensity was lower for cTnC-I4M compared with cTnC-WT (Fig. 8B). These results indicate that Ca 2ϩ -induced hydrophobic exposure is markedly blunted as a consequence of this N-helix cTnC variant.
To acquire insight into potential alterations in cTnC-I4M structural dynamics, we designed experiments using solution-state nuclear magnetic resonance (NMR) spectroscopy. Because the 1 H- 15 N HSQC of cTnC-I4M revealed only modest changes in chemical shifts compared with cTnC-WT ( Fig.  9A and Fig. S5), we were able to successfully assign 1 H- 15 N correlations to this novel variant and track the chemical-shift changes induced by the I4M substitution ( Fig. 9A and Fig. S5) (37). Chemical-shift perturbation (CSP) analysis obtained by comparing cTnC-WT and cTnC-I4M 1 H- 15 N peak assignments identified protein segments mostly affected by the mutation (Fig. 9B). CSP outliers differing by Ն2 S.D. from the average across studied residues were mapped onto the cTnC atomic model (Fig. 9C). Perturbed residues were located at the N-helix (Asp-3), and on both the defunct site I (Gly-30, Asp-33, and Glu-40) and the Ca 2ϩ -binding EF-hand regulatory site II (Glu-66 and Asp-67). In addition to perturbations in the N-domain, the end of helix-G in the C-domain (Gly-140 and Lys-142) also experienced a different chemical environment compared with cTnC-WT. These results suggest that, whereas the overall tertiary structures are similar between cTnC-WT and cTnC-I4M, discrete regions are structurally impacted by this variant.
The diffusional properties measured by 15 N longitudinal (R 1 ) and transverse (R 2 ) relaxation times and the 1 H-15 N heteronuclear NOE provide information on the local dynamic changes on a fast (picosecond-nanosecond) timescale and also report, in a qualitative manner, slower (microsecond-millisecond) conformational exchange motions. The R 2 /R 1 plot obtained for cTnC-I4M revealed a cluster of residues close to the defunct Ca 2ϩ -binding loop presenting values above the average ϩ1 S.D. (4.95 Ϯ 0.93) (Fig. 10a). These data suggest a local conformational exchange at this region caused by the presence of I4M and that the N-and C-domains tumble independently with different rotational correlation times (Fig. 10a). Similar to the cTnC-WT, residues within the cTnC-I4M helix-D (Asp-73, Phe-74, Phe-77, and Val-79) also revealed increased R 2 /R 1 values (Fig. 10a) (37). Importantly, residues close to the defunct Ca 2ϩ -binding loop in cTnC-I4M display higher R 2 /R 1 values, which were not observed when the same experiments were carried out on cTnC-WT (37). Internal motions on a picosecondnanosecond timescale of 1 H- 15 N bonds can be assessed by NOE values. Except for residue Glu-32 that exhibits a higher flexibility of the H-N bond, no other changes in NOE values were observed between cTnC-WT and cTnC-I4M (Fig. 10b). Using the Lipari-Szabo approach, we also calculated the order parameter (S 2 ) and conformational exchange ratio (R ex ) for cTnC-I4M, which reflects movements of the backbone amide bonds in the picosecond-nanosecond and microsecondnanosecond time regimes, respectively. Residues throughout the regulatory N-domain (Met-8, Glu-10, Leu-29, Asp-33, Asn-50, and Phe-74), as well as the C-domain (Lys-117, Asp-109, and Lys-158), of cTnC-I4M show contributions from exchange between alternative conformations in the relatively slow

Modulation of myocardial contraction by TnC-TnT interaction
(microsecond-millisecond) timescale (Fig. 10, c and e). Lower than average S 2 values, which indicate increased flexibility, are seen in the Asp/Glu linker and hinge regions of cTnC-I4M (Fig.  10, d and f). Intriguingly, residues in the N-domain, a presumed region of cTnT C-terminal binding, show higher than average S 2 values, suggestive of reduced backbone flexibility in this region. Taken together, these results show that the N-helix I4M variant, in part through allosteric transmission, alters the structural dynamics of cTnC and can reasonably explain the enhanced cTnT-binding affinity.

Discussion
In this study, we identified a previously unreported cardiomyopathic variant as well as a novel molecular mechanism underlying its pathogenicity. By leveraging a broad array of experimental approaches to assess this variant, we found evidence for direct communication between the extreme C terminus of cTnT and the regulatory domain of cTnC, alterations that are expected to impact contractile force and potentially serve as a primary stimulus for pathological remodeling of the myocardium.
The first report to establish TNNC1 as a cardiomyopathic gene was published in 2004 (38). Since then, multiple distinct variants have been identified in patients with DCM, hypertrophic cardiomyopathy (HCM), and restrictive cardiomyopathy (39). The cTnC-I4M variant reported here should be added to the growing list of cardiomyopathic variants. Although genetic linkage and segregation analyses are undoubtedly important criteria for assigning pathogenicity to a variant, this information is often unavailable, especially for de novo variants that are often deleterious early in life (40). Furthermore, the vari-able expressivity and incomplete penetrance associated with sarcomeric cardiomyopathies can further complicate genetic analyses in cases where affected individuals reach reproductive age (4). Some investigators in the field have even questioned the relevance of studying de novo mutations. We argue that the clinical severity and frequent lethality often associated with rare TNNC1 variants only reinforces the importance of acquiring a better understanding of this essential gene. Indeed, investigating lethal cardiomyopathic variants provides insight into the regulatory mechanisms that are critical for sustaining contractile performance. As evidenced by previous reports and this specific case, it would appear that certain TNNC1 variants are incompatible with the normal human life span and therefore are not likely to be transmitted to offspring. Furthermore, the overall primary sequence of cTnC, including the affected residue in this report, is highly-conserved across species, suggesting an intolerance to mutations through the course of evolution (39).
Myofilament Ca 2ϩ sensitivity of steady-state isometric tension can be used as an index for contractility. We found that cTnC-I4M is associated with reduced Ca 2ϩ sensitivity of force, which is in alignment with the paradigm for sarcomeric cardiomyopathies in general and with DCM-associated variants in particular (41). Quantitatively, ϳ1.3-fold more Ca 2ϩ is required to achieve 50% of maximum isometric tension. Physiologically, the thin filament is never fully activated in the normal beating heart because cytosolic free Ca 2ϩ concentration varies throughout the cardiac cycle within the approximate range of pCa 7 to 6 (ϳ0.1 to 1 M) (42). Therefore, expression of cTnC-I4M in the heart is expected to depress myocardial contractility in vivo, despite no significant reduction in maximal restored tension for cTnC-I4M-reconstituted CMP. Thus, the

Modulation of myocardial contraction by TnC-TnT interaction
importance of myofilament Ca 2ϩ sensitivity should not be underestimated. Modulation of myofilament Ca 2ϩ sensitivity by engineered mutants or pharmacological treatment has previously been shown to ameliorate, and sometimes prevent, pathological remodeling associated with sarcomeric cardiomyopathies (43)(44)(45).
The observed inhibitory effect of cTnC-I4M on Ca 2ϩ activation of k TR (Fig. 4) suggests that expression of this N-helix variant in the myocardium would impair contractile function. Indeed, it has been previously suggested, and borne out by experimental evidence, that k TR reflects processes that limit the rate and magnitude of isovolumetric pressure rise in the cardiac ventricles (46,47).
We observed a Ca 2ϩ dependence of k TR across physiological activation, consistent with previous observations in cardiac muscle (47)(48)(49)(50). Computational modeling suggested that the slower k TR value displayed by cTnC-I4M can be explained by the combination of slower rates of cross-bridge transition from the nonforce-bearing state to the strong, force-generating state (f app ), as well as the reverse process (g app ) because, to a first approximation, maximal k TR ϭ f app ϩ g app (51, 52). Our mod-eling further suggests that the decrease in Ca 2ϩ sensitivity ( Fig.  3 and Table 1) could be due to faster deactivation of thin filament regulatory units (Table 2); while this is the most likely scenario, in the absence of experimental measurements on the cTnC-I4M variant, we cannot rule out an effect on the kinetics of Ca 2ϩ binding and regulatory unit activation.
According to Brenner's two-state model of cross-bridge cycling, the fraction of cross-bridges in strong, force-generating states and thus isometric tension are both proportional to f app / (f app ϩ g app ) (51); this proportionality is valid to a first approximation in the three-state model at saturating levels of activating Ca 2ϩ (48,50,53). The model-derived values of f app and g app ( Table 1) are consistent with the lack of a significant difference in maximum isometric force between cTnC-WT and cTnC-I4M (Figs. 3 and 4 and Table 1) because we were able to use this information to constrain the parameter space examined in the modeling. Thus, in the absence of an indirect effect on Ca 2ϩhandling in the living cardiomyocyte, the cTnC-I4M variant would be expected to reduce force production at a given level of intracellular free Ca 2ϩ due to a smaller number of strong, forcegenerating cross-bridges that results from lower Ca 2ϩ sensitiv-

Modulation of myocardial contraction by TnC-TnT interaction
ity in combination with slower kinetics of tension generation. Whereas haploinsufficiency has been observed for other cardiomyopathic variants in the sarcomere, it cannot explain the disruption in myofilament function for this cTnC variant; both cTnC-WT and cTnC-I4M were able to restore comparable levels of tension upon incorporation into CMP, supporting the concept that cTnC-I4M acts as a dominant-negative mutation in the contractile apparatus.
In contrast to the cTnC-I4M variant, we previously reported that a Ca 2ϩ -sensitizing mutation in the N-helix of cTnC (cTnC-A8V) increases k TR at all tested levels of activation, which could be explained by an increase in both f app and g app (48,54). Hence, these two N-helix variants, which are associated with divergent pathways of cardiac remodeling (hypertrophy versus dilation), appear to have opposing effects on k TR . In addition to reinforcing the notion that sarcomeric cardiomyopathies are mutation-specific rather gene-specific, these observations raise the possibility that the N-helix confers specific regulatory properties to cTnC to allow for the Ca 2ϩ -dependent control of myocardial force production.
To understand how a variant in the N-helix of cTnC could influence molecular processes downstream of Ca 2ϩ binding (i.e. cross-bridge cycle kinetics), we investigated the possibility of intersubunit communication between cTnT and cTnC. The cTnT subunit was a logical target because it directly interacts with the other key thin-filament regulatory proteins, actin and tropomyosin (18). In fact, results from in vitro motility assays have suggested that TnT may have a direct role in controlling k TR (55). This study is the first to quantify direct binding between the cardiac isoforms of TnT and TnC and show that this interaction is altered by a DCM-related variant in cTnC.
In the context of a molecular mechanism, cTnC-cTnT binding might impact contractile performance by shifting the distribution of thin filament equilibrium states (56,57). The premise of this supposition is primarily derived from a series of reports by Chalovich and co-workers (25-27), in which they suggest that the C-terminal region of cTnT has a dual Ca 2ϩ -dependent regulatory role. Their findings indicate that the C-terminal residues of cTnT are important for both the formation of the inactive B-state at low levels of Ca 2ϩ and limiting the active M-state at saturating Ca 2ϩ concentrations. This cTnT region is particularly relevant because it is associated with genetic cardiomyopathy. A single missense mutation (R286H) as well as C-terminal deletions (⌬14 or ⌬21) in cTnT are linked to HCM in humans (24, 58). Furthermore, preliminary results from our laboratory have suggested that this C-terminal cTnT peptide has an inhibitory effect on force generation in permeabilized cardiac muscle fiber preparations. 3 Thus, one possibility is that the increased binding affinity of the cTnT C terminus for cTnC-I4M might shift the equilibrium in favor of stabilizing an inactive, nonforce-generating state of regulated actin. If true, this could reasonably explain the markedly slower rate (k TR ) of transition between the nonforce-generating (C-state) and the force-generating (M-state) states that we observed for this variant upon reconstitution into permeabilized CMP (Fig. 4).
Our chemical cross-linking results suggest that the C terminus of cTnT is in proximity (ϳ1 nm) to the N-helix of cTnC. It is therefore not surprising that the cTnC I4M variant impacted the binding affinity for both full-length and C-terminal peptide of cTnT. Recent molecular dynamics simulations performed on cTn in the Ca 2ϩ -saturated state have suggested that the C terminus of cTnT most frequently samples a conformation that is situated in a cleft between the two lobes of cTnC, simultaneously interacting with the inhibitory peptide and N-terminal regions of cTnI (59). In this conformation, it would appear that the cTnT C terminus could influence intrasubunit cTnI contacts (28). Interestingly, it has been shown that mutation of conserved acidic residues to alanine in the regulatory N-domain of skeletal muscle TnC simultaneously reduced TnT binding and disrupted the Ca 2ϩ -induced structural rearrangements of troponin I (60). Considering the abundance of basic residues at the C terminus of cTnT, this region is expected to electrostatically interact with this patch of acidic residues in cTnC. Although it has been recognized that the cardiac thin filament represents an extensive allosteric system, the molecular mechanisms are not fully understood (61)(62)(63). Our findings in this report raise an intriguing possibility that the C terminus of cTnT allosterically controls thin filament activation via modulation of the cTnI-cTnC switching mechanism.
It is reasonable to expect that the C terminus of cTnT would have a different binding partner at low (diastolic) levels of Ca 2ϩ . The conformational plasticity of intrinsically disordered regions allows for interactions with a diverse set of binding partners in the cell (64,65). We could find no evidence for cTnC-cTnT binding in the absence of Ca 2ϩ bound to the regulatory site of cTnC. 3 This is consistent with previous findings from binding studies with the skeletal muscle isoforms of TnT and TnC, which were shown to be Ca 2ϩ -dependent (23,66,67). Moreover, hydrogen-deuterium exchange experiments performed on cTn revealed that the C terminus of cTnT is stabilized by Ca 2ϩ , suggestive of the formation of inter-or intrasubunit unit contacts in cTn (68). One possibility is that the C terminus of cTnT binds to actin or tropomyosin in the absence of Ca 2ϩ to stabilize the nonforce-generating B-state of the thin filament (25). Whereas much of this discussion is merely speculation, one certainty is that further investigations are needed to unravel the complexity of myofilament protein-protein interactions underlying modulation of contractile regulation in the myocardium.
Time-dependent fluctuations in atomic coordinates of proteins (i.e. protein dynamics) have emerged as a critical determinant of function (69). The observed alterations in cTnT-cTnC binding led us to ask the question of whether this cardiomyopathic variant impacted the structural dynamics of cTnC. Our group and others have previously reported on the importance of cTnC dynamics for its regulatory function (37, 70 -72). The NMR relaxation results revealed remote molecular dynamics changes in the cTnC molecule caused by the I4M substitution, because the residues distant from the mutation site showed different NMR relaxation rates when compared with the same residues in WT cTnC (Fig. 10). Our bis-ANS fluorescence studies revealed similar hydrophobic exposure profiles in the apo (minus Ca 2ϩ ) form. However, upon titration with Ca 2ϩ , hydro-phobic exposure was markedly blunted for cTnC-I4M (Fig. 8), suggesting that this variant might stabilize the closed conformation or destabilize the open conformation of cTnC. One possibility is that introduction of a methionine at a solvent-exposed position in the N-helix of cTnC could render the cTnC molecule vulnerable to oxidation-induced conformational destabilization (73). Our CSP analysis by NMR suggested that key acidic residues in cTnC-I4M experienced a different chemical environment compared with cTnC-WT. We found that the regulatory N-domain of cTnC, the expected region of cTnC C-terminal binding, experienced relatively slower backbone dynamics as a direct result of this N-helix variant. These results offer a potential atomic-level explanation for the enhanced binding affinity between cTnC-I4M and the C terminus of cTnT.
Although we addressed this study mechanistically from a biophysical perspective, it should be acknowledged that DCM is a complex, heterogeneous disorder that is influenced by additional factors (e.g. modifier genes, epigenetics, and environment); myofilament dysregulation represents just one conceivable explanation for the genotype-phenotype relation (4,74). In some instances, sarcomeric gene variants associated with cardiomyopathy have been suggested to render a host susceptible to bacterial or viral infections and further exacerbate cardiac dysfunction (75). Intriguingly, immunofluorescence studies have indicated nuclear localization of cTnC, raising the possibility that pathogenic cTnC variants might also disrupt cellular processes beyond contractile regulation (76). With this complexity in mind, it is not surprising that unequivocally establishing a genotype-phenotype relation for cardiomyopathic variants remains an enormous challenge, particularly in the case of de novo variants associated with DCM. While this study focused on an isolated rare variant, we surmise that other cardiomyopathic variants in the contractile apparatus might alter cTnC-cTnT-based regulation of myocardial contractility.

Clinical and molecular genetics
Physicians obtained family history and performed physical examination of the proband. Trio Whole-Exome Sequencing was performed at Centogene (Centogene AG, Germany). Written informed consent was obtained from the parents of the proband.

Molecular visualization and protein sequence alignment
The cTnC-I4M variant was visualized by UCSF Chimera using Protein Data Bank code 1J1E (22,77). Multiple protein sequence alignments were performed with Clustal Omega, 1.2.4 (EMBL-EMI).

Mutagenesis, cloning, protein expression, and purification
QuikChange (Agilent) PCR site-directed mutagenesis was used to generate the cTnC mutants from pET-3d/11d constructs. Plasmid DNA sequences were verified by sequencing. Recombinant human cTnC, cTnT, and cTnI were cloned, expressed, and purified as described previously (41,78). Both cTnT and cTnI were stored in aliquots at Ϫ80°C, and cTnC-WT and -I4M were lyophilized and stored at 4°C. For NMR spectroscopy experiments, the cTnC-I4M construct was transformed into competent Escherichia coli cells (strain BL21-DE3) and inoculated for growth in [ 15 N]ammonium chloride (NLM-467-PK, Cambridge Isotope Laboratories, Inc.)-enriched minimal media as described previously (37). Native porcine cardiac tropomyosin was purified from left ventricular tissue, and filamentous actin was prepared from rabbit skeletal muscle acetone powder as described previously (41). SDS-PAGE was used to assess protein purity and potential degradation.

Protein quantification
NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA) was used to quantify protein concentrations (absorbance at 280 nm) for peptide-binding experiments. Extinction coefficients used for cTnC and a 14-amino acid C-terminal cTnT peptide were 4080 and 5960 cm Ϫ1 M Ϫ1 , respectively. Pierce Coomassie Plus/Bradford assay kit (Thermo Fisher Scientific TM , catalogue no. 23236) was used for all other experiments with BSA as the standard.

Formation of troponin complexes
Recombinant human cardiac troponin complexes were assembled following standard procedures in our laboratory (15,79). Briefly, troponin complexes were formed by first dialyzing the individual subunits (Ն1.0 mg/ml) for 4 h in 4 liters of troponin complex buffer (3 M urea, 1 M KCl, 10 mM MOPS, 1 mM DTT, pH 7.0) and then twice against 4 liters of troponin complex buffer without urea for 4 h each. After a 1-h incubation on ice, recombinant cTnT, cTnI, and cTnC-WT or -I4M subunits were combined at 1.3:1.3:1 ratio into dialysis tubing and sequentially dialyzed against 4 liters of troponin complex buffer containing 1 mM MgCl 2 with decreasing salt concentrations (0.7, 0.5, 0.3, and 0.1 M KCl) under constant stirring at 4°C for 4 h each (15,79). Precipitants were removed by centrifugation at 19,000 rpm in a JA-20 (Beckman Coulter) rotor for 15 min at 4°C. Complexation was confirmed by native PAGE (4% stacking, 8% resolving).

Peptide synthesis
The C-terminal peptide (acetyl-Ser-Lys-Thr-Arg-Gly-Lys-Ala-Lys-Val-Thr-Gly-Arg-Trp-Lys-amide), corresponding to the last 14 amino acid residues in human cTnT (Uniprot ID P45379 -6), was purchased from Genscript USA (Piscataway, NJ) and supplied in lyophilized form at 99.1% purity and stored at Ϫ20°C. Peptide sequence and purity were assessed by MS and analytical HPLC.

Preparation of buffered Ca 2؉ solutions for CMP experiments
Calculations for buffered pCa (Ϫlog[Ca 2ϩ ] free ) solutions were carried out using the computer program, pCa calculator

Modulation of myocardial contraction by TnC-TnT interaction
perature (ϳ21°C), as described previously (48). Experiments performed at different temperatures (as indicated in the figure legends) and involving dilutions (CaCl 2 titrations) were accounted for in the calculations.

Cardiac muscle preparations
Fresh porcine hearts were procured from a local abattoir and transported in an ice-cold, O 2 -sparged buffer containing the following: 140 mM NaCl, 4 mM KCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 1.8 mM NaH 2 PO 4 , 5.5 mM D-glucose, 50 mM HEPES, pH 7.4. Strips of papillary muscle were dissected from the myocardium on ice and permeabilized for 4 h in pCa 8.0 relaxing solution containing 1% (v/v) Triton X-100 at 4°C. The CMP were further dissected to an average length of ϳ1 mm, clipped on each end with aluminum T-clips, and stored in pCa 8.0 relaxing solution containing 52% (v/v) glycerol at Ϫ20°C until used for experiments, as described previously (11).

Muscle mechanics
Extraction-reconstitution experiments on CMP were performed as described previously (11,13). Briefly, CMP were mounted between a force transducer (model 403A, Aurora Scientific Inc.) and a high-speed servomotor (model 322C, Aurora Scientific Inc.) in pCa 8.0. Sarcomere length (SL) was set to 2.1 m by He-Ne laser diffraction. Endogenous cTnC was extracted by incubation of a CMP with extraction buffer (5.0 mM CDTA, 25 mM Tris, pH 8.4). Drops of 55 M cTnC-WT or -I4M were applied directly to the cTnC-extracted CMP. Measurements and analysis of myofilament Ca 2ϩ sensitivity of steady-state isometric tension generation and k TR were performed as described previously with minor modifications (48). The pCa-force relation was fitted with a two-parameter, sigmoid Hill equation (48). The ends of the CMP were fixed with 1% glutaraldehyde (w/v) to minimize end compliance (81). In conditions where force was below 15% of the maximal force, the CMP were excluded for k TR analysis. Extraction and reconstitution of cTnC, as well as recording of initial maximal tension, were carried out at 21°C to prevent CMP rundown. However, the temperature of the chambers was switched to 30°C immediately prior to Ca 2ϩ -dependent activation (71). Mathematical modeling of tension kinetics was carried out in the MatLab environment as described previously (48).

bis-ANS fluorescence
Equimolar concentrations (2 M) of bis-ANS (Sigma D4162) and cTnC-WT or -I4M were combined in fluorescence buffer (120 mM MOPS, 1.25 mM MgCl 2 , 90 mM KCl, 2 mM EGTA, 1 mM fresh DTT, pH 7.0) to a final volume of 2 ml in quartz cuvettes. Incremental (l) amounts of 1 M CaCl 2 were added to achieve the desired pCa based on the calculations described above. bis-ANS was excited at 380 nm, and emission maxima were monitored at ϳ440 -600 nm. Spectra were acquired on a four-cell Jasco FP-8300 spectrofluorometer connected to a compact recirculating cooler (Julabo) set to 21°C. Acquisition parameters were set to 5.0-nm slit widths, a constant photomultiplier (PMT) voltage of 340 V, and automated stirring (600 rpm).

Intrinsic tryptophan fluorescence
The 14-amino acid C-terminal cTnT peptide, cTnC-WT and cTnC-I4M, was suspended and dialyzed in 25 mM HEPES, 60 mM NaCl, 3 mM MgCl 2 , 0.5 mM CaCl 2 , 2 mM ␤-mercaptoethanol (␤ME), pH 7.0, prior to titrations. Incremental (l) amounts, from 0 to 50 M, of cTnC-WT-or -I4M were titrated into 5 M peptide. To account for potential contribution of fluorescence intensity changes upon titration of cTnC, cTnC-WT or -I4M was titrated into buffer without peptide and subtracted from the fluorescence intensity values for each condition. Intrinsic tryptophan was excited at 295 nm, and the wavelength emission maximum was monitored within the range 300 -400 nm. Spectra were acquired on a four-cell Jasco FP-8300 spectrofluorometer connected to a compact recirculating cooler (Julabo) set to 21°C. Acquisition parameters were set to 5.0 nm slit widths, a constant PMT voltage of 300 V, and automated stirring (600 rpm). Results were fitted using Origin-

Solution-state NMR spectroscopy
Solution-state NMR experiments on 15

Modulation of myocardial contraction by TnC-TnT interaction
equipped with a 5-mm trip resonance TXI probe at the National Center of Nuclear Magnetic Resonance, Jiri Jonas (Rio de Janeiro, Brazil) or Bruker Avance III 700 MHz equipped with a cryoprobe at the Department of Chemistry and Biochemistry, Florida State University (Tallahassee, FL). We performed the 1 H- 15 N assignment of cTnC-I4M using the following procedures: (i) cross-peak transferring from our previous cTnC-WT assignment (37) to the -I4M 1 H- 15 N HSQC spectrum, and (ii) peak-by-peak evaluation to decide whether transferred crosspeaks belong to its closest neighbor peak. The inclusion criteria to assign was naked eye inspection of the surrounding area and clear evidence that a selected cross-peak exclusively belongs to the closest neighbor. We did not assign few peaks at crowded areas because dualities arose (37). CSP analysis was performed as described previously using Equation 2, where ⌬␦ H and ⌬␦ N represent the change in chemical shifts for 1 H and 15 N, respectively, between cTnC-WT and -I4M. All spectra were processed using Topspin 3.2, and chemical-shift values were measured using the software CCPN Analysis 2.4.1. The diffusion properties (R 1 , R 2 , and NOE) of I4M were obtained as described previously (37). 15 N R 1 and 15 N R 2 relaxation rates were measured from spectra with different relaxation delays: T 1 ϭ 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.7, 0.9, and 1.1 s for R 1 and T 2 ϭ 0.01696, 0.03392, 0.05088, 0.06784, 0.0848, 0.10176, 0.13568, and 0.1696 s for R 2 . The errors in the peak intensities were calculated from the standard deviation of background noise levels in relation to each residue peak intensities (82). The heteronuclear NOEs were determined from the ratio of peak intensities with and without the saturation of the amide protons. NOE spectra was acquired using recycle relaxation delays of 6 s between the scans. Errors in HetNOE values were calculated from the peak intensities and noise levels in the reference and saturated spectra. Rotation correlation times, c , of the N-and C-terminal domains were calculated separately from R 2 /R 1 rates assuming isotropic tumbling and using Equation 3, c ϭ 1 where 15N is the Larmor frequency of 15 N. For these calculations, all residues showing a higher contribution of local dynamics on picosecond-nanosecond timescales, characterized by HetNOE rates lower than 0.6, and those showing contribution of conformational exchange, characterized by R 2 /R 1 ratio values above the average value plus the standard deviation, were excluded. Relaxation rates R 1 , R 2 , and HetNOE were also analyzed with the Lipari-Szabo formalism (83,84) using the software Tensor 2.0 (85). For this analysis, the N-and the C-terminal domains were analyzed separately because they showed very different values of rotational correlational times. The maximum theoretical value (HetNOE max ) for cTnC was estimated using the software HYDRONMR (86). HYDRONMR simulates NMR relaxation data by calculating anisotropic rotational diffusion and by modeling protein structure (PDB code 1J1E) with a bead model. The following parameters used in the calculation were: temperature ϭ 298 K; solvent viscosity ϭ 0.01 poise; magnetic field ϭ 21.15 tesla; gyromagnetic ratio of 1 H ϭ 2.675 ϫ 10 4 rad s Ϫ1 G Ϫ1 ; gyromagnetic ratio of 15 N ϭ Ϫ2.713 ϫ 10 3 rad s Ϫ1 G Ϫ1 ; internuclear H-N distance ϭ 1.02 Å; and 15 N chemical-shift anisotropy ϭ Ϫ160 ppm.

Co-sedimentation assay
Co-sedimentation assays were performed following previously described methods with minor modifications (67). Prior to running the assay, troponin subunits were centrifuged to remove precipitates, and the supernatants were run on a gel to confirm solubility. Tropomyosin (2.86 M), actin (20 M), and cTnT (2.86 M) were combined in a final volume of 100 l of reaction buffer containing 25 mM HEPES, 60 mM NaCl, 3 mM MgCl 2 , 0.5 mM CaCl 2 , 2 mM ␤ME, pH 7.0. Next, cTnC-WT or cTnC-I4M was added to a final concentration of 40 M. The samples were then gently mixed, incubated on ice for 30 min, and centrifuged at 54,000 rpm in a fixed-angle TLA-100 rotor (Beckman Coulter) for 10 min at 4°C. The protein-containing pellets were washed twice with reaction buffer and subsequently resuspended in 100 l of 1ϫ Laemmli buffer, boiled, and resolved by SDS-PAGE on a 15% gel. As a control, the experiment was repeated exactly as described but with omission of cTnT from the reactions. Proteins were visualized by Coomassie staining and imaged on an Odyssey IR system (LI-COR Biosciences). Image Studio Lite (LI-COR Biosciences) was used for densitometric quantification.

Chemical cross-linking-MS
WT cTn was subjected to BS3 chemical cross-linking followed by MS analysis using a recently published protocol (87). A 25 mM stock of equimolar BS3-d 0 and BS3-d 4 (bis(sulfosuccinimidyl) 2,2,7,7-suberate, Thermo Fisher Scientific) was freshly prepared in double-distilled water. A series of incubation times and varying concentrations of BS3-d 0 /d 4 were tested to optimize cross-linking conditions. In a final reaction volume of 20 l, cross-linking was initiated by addition of equimolar BS3-d 0 /d 4 (0.25 mM, final concentration) to 1 mg/ml cTn in 25 mM HEPES, 60 mM NaCl, 3 mM MgCl 2 , 0.5 mM CaCl 2 , 2 mM ␤ME, pH 7.0. The reaction was incubated in a MixMate (Eppendorf) at 350 rpm for 1 h at room temperature and quenched by addition of glycine, pH 8.0 (100 mM, final concentration). The uncross-linked and cross-linked samples were prepared for SDS-PAGE and run on a 4 -15% Mini-PROTEAN TGX gel (Bio-Rad).

In-gel protein digestion
The band corresponding to the cross-linked troponin complex band was cut from gel and digested with trypsin. Briefly, the gel band was de-stained with a de-staining buffer (1:1 H 2 O/ acetonitrile with 50 mM ammonium bicarbonate), cut into ϳ2-mm pieces, shrunk with acetonitrile, and dried in Speed-Vac. Then, the gel pieces were rehydrated with a digestion buffer (10% aqueous acetonitrile with 50 mM ammonium bicarbonate) and reduced with 0.5 mM tris(2-carboxyethyl)phosphine (Sigma, catalogue no. C4706) at 37°C for 10 min. After cooling to room temperature, 1 mM iodoacetamide (Sigma, catalogue no. I1149) was added to the mixture and vortexed for 10

Modulation of myocardial contraction by TnC-TnT interaction
min. Then 0.005 g/l trypsin (Thermo Fisher Scientific, catalogue no. 90058) was added, and the mixture was incubated overnight at 37°C. The supernatant was collected. 0.5% formic acid was added to the residual gel pieces; the mixture was incubated at 37°C for 15 min, and the supernatant was collected. The residual gel pieces were dried after adding acetonitrile, and incubation was at 37°C for 10 min. The combined supernatant was dried in SpeedVac (Thermo Fisher Scientific).

Nano-LC nLC/MS 2
The dried tryptic peptide mixture was reconstituted in 0.1% aqueous formic acid and separated by nano-LC (nLC). An Easy Nano LC II system (Thermo Fisher Scientific) equipped with a 75 m ϫ 10-cm C18AQ analytical column (catalogue no. SC003, Thermo Fisher Scientific) and a 100 m ϫ 2-cm trap column (easy column, catalogue no. SC001, Thermo Fisher Scientific) was used. A 3-h linear gradient from 1 to 35% B was performed with a flow rate of 300 nl/min (mobile phase A: 99.9% H 2 O and 0.1% formic acid; mobile phase B: 99.9% acetonitrile and 0.1% formic acid). Eluate was ionized online with 2 kV spray voltage and detected by a Velos LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific). The precursor ions were detected with a mass resolution of 60 K (at m/z of 800 Da) and Automatic Gain Control (AGC) of 1e6 in the Orbitrap. Centroid data-dependent MS 2 was carried out on the top 10 most abundant precursor ions with AGC of 5e4 with collisioninduced dissociation (2.0 m/z isolation window and 35 normalized collision energy) in the LTQ.

Data analysis
An open-access software ProteoWizard MSConvert (Prot-eoWizard version 3.0.10158) (88) was used to convert the original .raw file to .mzXML file. The converted file was then analyzed by another open-access software StavroX (StavroX version 3.6.6) to identify cross-linked peptides (89). The following parameters were used in StavroX: protein sequence FASTA database containing the three human cardiac troponin complex proteins (cTnC, cTnI, and cTnT); up to three trypsin missed cleavages; static modification of cysteine to carbamidomethyl cysteine; variable modification of methionine by oxidation; precursor precision of Ͻ3 ppm; fragment ion precision of Ͻ0.8 Da; signal/noise ratio above 2; false discovery rate cutoff of Ͻ5.0%. Software-generated results were then manually checked for correct assignments of those cross-linked peptide MS 2 spectra. For every confirmed cross-linked peptide, the extracted ion chromatogram and MS of the corresponding precursor ions were checked for presence of peptide doublets, peptides crosslinked by BS3-d 0 /d 4 . The cross-linked peptide doublets should have a separation of 4.0256 Da (the mass difference with 4 H and 4 D). A list of selected peptide sequences identified, precursor charge, m/z for each assignment, peptide identification scores, and LC retention time can be found in Table S3. All cross-linked peptides identified and annotated MS/MS spectra can be found in Fig. S6.

Statistical analyses
Statistical analyses were performed using SigmaPlot version 12.0 or OriginPro software (Northampton, MA). Experimental results are reported as mean Ϯ S.D. as indicated in each figure legend. Student's t-test (paired or unpaired, indicated in figure legends) was used to determine statistical significance (p Ͻ 0.05) between control and experimental groups. Sample size (N) for a given condition is reported in each respective figure legend.