Structure of Cardiac Muscle Troponin C Unexpectedly Reveals a Closed Regulatory Domain*

The regulation of cardiac muscle contraction must differ from that of skeletal muscles to effect different physiological and contractile properties. Cardiac troponin C (TnC), the key regulator of cardiac muscle contraction, possesses different functional and Ca2+-binding properties compared with skeletal TnC and features a Ca2+-binding site I, which is naturally inactive. The structure of cardiac TnC in the Ca2+-saturated state has been determined by nuclear magnetic resonance spectroscopy. The regulatory domain exists in a “closed” conformation even in the Ca2+-bound (the “on”) state, in contrast to all predicted models and differing significantly from the calcium-induced structure observed in skeletal TnC. This structure in the Ca2+-bound state, and its subsequent interaction with troponin I (TnI), are crucial in determining the specific regulatory mechanism for cardiac muscle contraction. Further, it will allow for an understanding of the action of calcium-sensitizing drugs, which bind to cardiac TnC and are known to enhance the ability of cardiac TnC to activate cardiac muscle contraction.

Transient increases in cytosolic Ca 2ϩ levels in the cardiac muscle cell must be recognized by the thin filament to regulate cardiac muscle contraction. This critical function is accomplished by cardiac TnC 1 (161 residues), a member of the EFhand family of Ca 2ϩ -binding proteins, which relays the Ca 2ϩ signal via a conformational change to the rest of the troponintropomyosin complex, and ultimately signals the activation of the myosin-actin ATPase reaction. Although the sequence of cardiac TnC is 70% identical to that of skeletal TnC, there are significant differences in the first 40 residues, the most crucial being the inactivation of Ca 2ϩ -binding site I due to an insertion (Val 28 ) and substitutions of key ligands relative to skeletal TnC (Leu 29 and Ala 31 in cardiac TnC instead of Asp 30 and Asp 32 in skeletal TnC) (1). Despite the many functional, binding, and modeling studies performed on cardiac TnC (2), the absence of direct structural data makes the Ca 2ϩ -induced conformational change in cardiac TnC unclear. The structures of TnC in the skeletal system, on the other hand, have been solved both in the 2-Ca 2ϩ (3,4) and 4-Ca 2ϩ states (5), showing TnC to be a dumbbell-shaped molecule with separate N-and C-terminal domains connected by a central linker. Upon Ca 2ϩ binding, the regulatory N-domain of skeletal TnC switches from a "closed" to an "open" conformation, thereby exposing a patch of hydrophobic residues, which is thought to interact with skeletal TnI (6). In this report, we show that, in contrast to predicted models (7)(8)(9), the analogous conformational change does not occur in cardiac TnC, and that this is the direct structural consequence of inactivating Ca 2ϩ -binding site I. In addition, a structural understanding of cardiac TnC has potential therapeutic value in the understanding of the mechanism of cardiac TnC-binding drugs known as "calcium-sensitizing drugs" (8,10).
For the purposes of this study, the two Cys residues at positions 35 and 84 of wild type cardiac TnC have been mutated to Ser residues. This prevents the formation of intra-and intermolecular disulfide bonds, which confer Ca 2ϩ -independent activity to cardiac TnC when assayed in skeletal muscle myofibrils (11). It has been shown that the conversion of these Cys residues to Ser residues has no effect on the ability of cardiac TnC to recover ATPase activity in TnC-extracted fast skeletal and cardiac myofibrils, and has little effect on Ca 2ϩ binding to site II of cardiac TnC (11). Thus, it is unlikely that the introduction of these two conservative mutations would result in gross conformational changes in the secondary or tertiary structure of cardiac TnC.
For NMR analysis, the protein was uniformly labeled with 13 C and/or 15 N by expression in Escherichia coli. Triple-resonance NMR experiments were used for assigning the resonances and subsequently to derive distance and dihedral angle restraints. 35 structures were then calculated using the simulated annealing protocol (12). Structural statistics for the 30 lowest energy structures (Table I) show that the N-and Cdomains are very well defined separately, with the central linker shown to be flexible by relaxation measurements. 2

MATERIALS AND METHODS
Sample Preparation-To produce high level expression of chicken cardiac TnC with the mutations C35S and C84S (denoted cTnC(A-Cys)), the expression plasmid pTnC(A-Cys)P L (11), which uses the P L promoter, was digested with NcoI and HindIII, and the small fragment containing the full amino acid coding region was ligated into the NcoI/ HindIII sites of the plasmid pET-23d (Novagen), which has a T 7 promoter. Plasmids were maintained, and cTnC(A-Cys) was expressed in * This work was supported by grants from the Medical Research Council of Canada, the Heart and Stroke Foundation of Canada, the National Institutes of Health, the Robert Welch Foundation, and the Alberta Heritage Foundation for Medical Research (to S. K. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The atomic coordinates and structure factors (code 1AJ4) have been deposited in the Protein Data Bank, Brookhaven National Laboratory, Upton, NY. Coordinates of the 30 calculated structures for the N-domain (code 2CTN) and the C-domain (code 3CTN) have also been deposited.
Structure Determination-All spectra were collected at 30°C on a Varian Unity Plus 500-MHz spectrometer or a Varian Unity Plus 600-MHz spectrometer, both equipped with pulsed field gradient accessories. The NMR samples contained 2-4 mM cTnC(A-Cys) in 16 mM CaCl 2 , 100 mM KCl, 10 mM imidazole at pH 6.7, in either 90% H 2 O, 10% D 2 O or 99.99% D 2 O. The high affinities of cardiac TnC for calcium (with dissociation constants on the order of 10 Ϫ6 M for the N-domain and 10 Ϫ8 M for the C-domain) ensure that the protein is Ca 2ϩ -saturated at 16 mM CaCl 2 (11).
Assignment of the main-chain NH, N, C␣, and C␤ resonances were made based on the three-dimensional experiments HNCACB and CB-CA(CO)NH (14). Three-dimensional 15 N-edited total correlation spectroscopy (TOCSY) (15) and three-dimensional HCCH-TOCSY (16) experiments were used to assign the chemical shifts for the side-chain atoms. A high resolution two-dimensional 1 H-1 H nuclear Overhauser enhancement spectroscopy (NOESY) experiment and assignments made by a previous study (17) were used to confirm the assignments for resonances of aromatic protons. Interproton distances were derived from three-dimensional 15 N-and 15 N/ 13 C-edited NOESY experiments (18), both collected at 50 ms of mixing time. In converting NOE intensities into distance restraints, the NOE intensities were calibrated for each residue using its intraresidue d N␣ (i, i) and sequential d ␣N (i Ϫ 1, i) NOEs as reference intensities (6). In cases where neither NOE was present, the upper bound distance was given the calibration factor corresponding to the loosest restraint. In all cases, a 40% error on NOE intensities was used. Based on the analysis of the crystal structures of skeletal TnC (19), 18 distance restraints to the Ca 2ϩ ion (6 at each of site II, III, and IV) of 2.0 -2.8 Å were assigned.
A three-dimensional HNHA experiment (20) was used to derive 3 J HNH␣ coupling constants. restraints were imposed if the 3 J HNH␣ values were greater than 8 Hz or less than 5.5 Hz. These restraints were given different errors depending on the region on the Karplus curve to which they corresponded, with a minimum error of Ϯ20°. restraints were assigned for the (i Ϫ 1) residue if the d N␣ (i, i) to d ␣N (i Ϫ 1, i) NOE intensity ratio was greater than 1.2 or less than 0.83, which were given loose restraints of -30 Ϯ 110°and 110 Ϯ 110°, respectively (28). Stereospecific assignments of the methyl protons of valines and leucines (20 out of 21 possible) were made based on the presence of singlets (pro-S) or doublets (pro-R) in a two-dimensional 13 C-1 H HSQC spectrum with the protein being expressed in 25% [ 13 C]glucose (21). H␤-methylene protons were stereospecifically assigned (32 out of 115 possible) by analyzing 3 J H␣H␤ values from the three-dimensional HACAHB experiment (22), 3 J NH␤ values from the three-dimensional HNHB experiment (23), and intraresidue NOE intensities. 1 restraints of ϩ60°, 180°, or Ϫ60°(Ϯ60°) were also given if all of the above data were consistent with one 1 conformation. All spectra were processed with NMRPipe (24) and peak-picked with the program PIPP (25). Fig. 1 shows the solution structures of the N-and C-terminal domains, with the structural statistics provided in Table I. The overall solution structure of Ca 2ϩ -saturated cardiac TnC, like the solution structures of Ca 2ϩ -saturated skeletal TnC (5) and calmodulin-target peptide complex (26), resembles a dumbbell in shape, consisting of two separate domains connected by a flexible central linker (residues 86 -94 in cardiac TnC). However, despite the general structural similarities to homologous Ca 2ϩ -binding proteins, the regulatory N-domain of Ca 2ϩ -saturated cardiac TnC is significantly more compact than the Ndomain of Ca 2ϩ -saturated skeletal TnC (5, 6), exposing approximately 800 Å 2 less total accessible surface area (residues 5-84) than its skeletal counterpart (residues 7-85). In particular, the B-helix of defunct site I exists in the "closed" conformation, exhibiting an A-B interhelical angle of 142°(with the A-and B-helices corresponding to the two helices of the helixloop-helix motif in Ca 2ϩ -binding proteins; Table II). The closed conformation is evidenced by 21 NOE connectives observed between the A-and B-helices (Fig. 2), most of which would not be observed if the B-helix were in an "open" conformation as in skeletal TnC (Fig. 3A). A compact regulatory domain is also consistent with a previous cysteine-reactivity study on wild type cardiac TnC (27).

RESULTS
The difference in the conformation of the B-helix is reflected most clearly in the main chain conformation of residue Glu 40 in cardiac TnC (equivalent to Glu 41 in skeletal TnC). Glu 41 of skeletal TnC has been proposed to be pivotal in the mechanism of the coupling between Ca 2ϩ binding and the Ca 2ϩ -induced conformational change in Ca 2ϩ -binding proteins (28,29). Indeed, the apo form of skeletal TnC (3) features a kink in the B-helix at Glu 41 which straightens out upon Ca 2ϩ binding to sites I and II (28). In cardiac TnC, however, there exists a kink in the B-helix at Glu 40 , even in the Ca 2ϩ -bound state (Ca 2ϩ ions bound at sites II, III, and IV). The non-helical nature at Glu 40 is supported by a 3 J HNH␣ value of 7.8 Hz, an absence of an upfield-shift of its H␣ resonance (4.37 ppm), and an absence of a downfield shift in its C␣ resonance (56.8 ppm), all of which indicate non-helical conformations (30). On the other hand, both adjacent residues Lys 39 and Leu 41 exhibit 3 J HNH␣ values of less than 5.5 Hz, as well as appropriate shifts in their H␣ and C␣ resonances which indicate an ␣-helical conformation.
The above structural differences between cardiac and skeletal TnC can be explained by what is in fact the most striking functional difference between the two proteins: namely that site I in cardiac TnC is inactive due to an insertion and key substitutions of key ligands. In cardiac TnC, there is no Ca 2ϩ at site I to pull the Glu 40 side-chain carboxylate group over to contribute to the enthalpy necessary to overcome the entropic loss associated with exposing buried residues (which occurs in the "opening" of the B-and C-helices relative to helices N, A, and D). Thus, the B-helix remains closed due to favorable  b Note that calcium ions are not directly observed by NMR spectroscopy. No restraints involving the calcium ions were used in the initial stages of structure calculations, and were added only in the final stages of refinement (see "Materials and Methods").
c The final force constants were K NOE ϭ 50 kcal mol Ϫ1 and K dihedral ϭ 200 kcal mol Ϫ1 rad Ϫ2 . , core and allowed regions were as determined by the program PROCHECK (36). d There are no distance violations over 0.2 Å for the N-domain, and there is 1 distance violation over 0.2 Å for 30 structures for the Cdomain. a The parentheses indicate first the state of the N-domain (i.e., A-B and C-D helix-loop-helices), followed by the state of the C-domain (i.e., E-F and G-H helix-loop-helices). Note that in cardiac TnC(1 Ca 2ϩ /2 Ca 2ϩ ) as determined in the present study, defunct site I (i.e. A-B helix-loop-helix) is free of Ca 2ϩ , while sites II, III, and IV are Ca 2ϩbound. Protein Data Bank accession codes are: 1TNW for the NMR structure of skeletal TnC(2 Ca 2ϩ /2 Ca 2ϩ ), 5TNC for the crystal structure of skeletal TnC(apo/2 Ca 2ϩ ), and 4CLN for the crystal structure of calmodulin(2 Ca 2ϩ /2 Ca 2ϩ ). b A large angle defines a "closed" conformation, whereas a small angle defines an "open" conformation. The axis for an ␣-helix is defined by two points, the two points being the average coordinates of the first and last 11 backbone atoms of the ␣-helix.
packing forces with the A-helix and D-helix (Figs. 1A and 3A;  Table II, C-D interhelical angle). These observations demonstrate that the inability of Glu 40 to coordinate Ca 2ϩ results in a more compact conformation for the B-helix, and possibly the C-helix, than is observed in skeletal TnC (Fig. 3A); in effect, Ca 2ϩ binding to sites I and II of skeletal TnC locks open the whole regulatory domain, whereas Ca 2ϩ binding to site II of cardiac TnC only partially opens up the regulatory domain. (This discussion assumes that the structure of the apo form of the regulatory domain of cardiac TnC is similar to that of skeletal TnC, as has been recently demonstrated. 3 ) The model of Glu 40 acting as a pivot for the N-domain is further supported by a recent structural study of a skeletal TnC mutant in which Glu 41 is replaced by Ala 41 , such that residue 41 can no longer coordinate the Ca 2ϩ ion present at site I (29). In the Ca 2ϩ -saturated state of this protein, the single substitution results in a kink at Ala 41 and a closed conformation for the B-helix, similar to what is seen in Ca 2ϩ -saturated cardiac TnC.
The structure of defunct site I shows that Leu 29 , which comes just after the insertion at Val 28 , forms an extra half-turn at the end of the A-helix, as evidenced by d ␣N (i, i ϩ 3) and d ␣␤ (i, i ϩ 3) NOE connectives from Ile 26 to Leu 29 . Site I, being Ca 2ϩ -free, is not as well defined as the rest of the molecule (root mean square deviation of 0.78 Å for backbone atoms of residues 30 -33), and is shown to be more flexible than the rest of the regulatory domain by relaxation measurements.
The structural C-domain of cardiac TnC (Figs. 1B and 3B) is predictably similar to those in skeletal TnC and calmodulin, although the interhelical angles of the two EF-hands in the C-domain indicate that this domain is in fact slightly more compact in cardiac TnC than in its counterparts (10 -20°more closed in the E-F and G-H interhelical angles; see Table II). Overall, the backbone atoms of residues 95-157 of cardiac TnC superimpose within 1.9 Å with their equivalent residues (96 -158) in the NMR structure of skeletal TnC with Ca 2ϩ -saturated N-domain, and 1.3 Å with the same region in the crystal structure of skeletal TnC with apo N-domain. DISCUSSION We have shown for the first time the three-dimensional structure of Ca 2ϩ -saturated cardiac TnC, which reveals an unexpected compact regulatory domain as a direct consequence of an inactive Ca 2ϩ -binding site I. These results provide a structural precedent for a Ca 2ϩ -binding regulatory protein in which one of the two sites in the paired set of EF-hands is inactive (for example, some invertebrate TnCs also have this feature; Ref. 31)). This unique structural feature sets cardiac TnC apart from other "calcium sensor" EF-hand Ca 2ϩ -binding proteins such as skeletal TnC and calmodulin, as well as "calcium buffer" EF-hand proteins such as parvalbumin and calbindin. The compact regulatory domain is a surprising result because it violates the general rule with Ca 2ϩ -binding proteins that a small conformational change accompanies Ca 2ϩ binding in buffering proteins, and that a large conformational change accompanies Ca 2ϩ binding in regulatory proteins such as cardiac TnC (32). In particular, it is believed that in general the action of Ca 2ϩ binding in calcium sensor proteins is to induce an exposure of a large hydrophobic surface, allowing the protein to interact with targets to accomplish regulatory functions, whereas the capture of Ca 2ϩ ions by calcium buffer proteins is accompanied by only minor conformational changes.
Thus, it has long been believed that the mechanism for the activation of cardiac TnC involves the exposure of a large hydrophobic patch upon Ca 2ϩ binding as observed for other calcium sensors. In fact, cardiac TnC models based on the conformational changes observed in skeletal TnC have been widely used to interpret the functional, Ca 2ϩ -binding and drugbinding properties of cardiac TnC (7)(8)(9), despite the unique inactive Ca 2ϩ -binding site I in cardiac TnC. The present results show that the hydrophobic exposure in the Ca 2ϩ -saturated regulatory domain of cardiac TnC (Fig. 4B) is dramatically reduced compared with that of skeletal TnC (Fig. 4A) as well as a previous widely used model of cardiac TnC (9) (Fig. 4C).
The substantially reduced hydrophobic surface of Ca 2ϩ -saturated cardiac TnC has important implications for the association of cardiac TnI with cardiac TnC. In particular, given that in both the cardiac and skeletal systems the Ca 2ϩ -dependent binding of TnI involves the N-domain of TnC (33), and that residues 5-84 of Ca 2ϩ -saturated cardiac TnC expose less total and hydrophobic surface area than residues 7-85 of Ca 2ϩsaturated skeletal TnC (Fig. 4), it is possible that the mode of interaction between TnI and TnC in cardiac muscle is in fact different from that in skeletal muscle. A smaller surface of interaction between 3Ca⅐cardiac TnC-cardiac TnI as compared with 4Ca⅐skeletal TnC-skeletal TnI would explain the finding that the free energy (⌬G) of Ca 2ϩ binding to the TnC-TnI complex is 4 times smaller in cardiac than it is in skeletal muscle (34). It may also be that the hydrophobic or electrostatic force dominates more in one isoform than in the other in the binding of TnI to TnC. If indeed the interaction between cardiac TnC and cardiac TnI involves less surface contact than that between skeletal TnC and skeletal TnI, cardiac TnC would be a more dynamic calcium sensor than its skeletal counterpart. In fact, a recent study has shown that the Ca 2ϩ off rate measured for site II in cardiac TnC is about 3-fold faster than observed for the N-terminal signaling domain of calmodulin or skeletal troponin C. 4 On the other hand, as an alternative to the above proposal, it is possible that the Ca 2ϩ -dependent binding of TnI forces open the regulatory domain of cardiac TnC, with the end result being that the cardiac TnI-TnC complex binds in a similar fashion to skeletal TnI-TnC (here cardiac TnC may adopt a structure similar to that in Fig. 4C). Such a model would be consistent with the finding that the chemical environment of Met 81 , which is mostly buried in this structure (accessible surface area of 14 Å 2 ), changes upon the binding of cardiac TnI (33). This may also imply that cardiac TnC opens up to different degrees in response to events of muscle contraction such as TnI phosphorylation. At present, there is no compelling evidence to either favor or discount either model for cardiac TnI-TnC binding.
Cardiac TnC is a potential target in therapy for patients with acute myocardial infarctions and subsequently congestive heart failure, where the diseased myocardium is "desensitized" to increases in cytosolic Ca 2ϩ levels. A novel group of positive inotropic agents known as "calcium sensitizers" (10) is known to increase the affinity of cardiac TnC for Ca 2ϩ , possibly by binding to a hydrophobic patch in the N-domain of cardiac TnC (8). The exposed hydrophobic patches in Ca 2ϩ -saturated cardiac TnC can now be identified (Fig. 4B). Although several residues (e.g. Phe 77 , Met 81 , and Met 85 ) have been implicated in earlier studies as possible binding sites for these drugs (8), the proposed modes of binding must now be re-evaluated since most of these residues lie on the side of the D-helix facing the B-helix, and therefore are more buried by the B-helix than previously suspected (B-D interhelical distance of 12 Å for cardiac TnC versus 18 Å for skeletal TnC). In addition, significant surface topology differences between the cardiac TnC model and the solution structure (Fig. 4) warrants for a reinterpretation of most of the previous drug binding studies performed based on the now-disproved model (8). Thus, the solution structure of cardiac TnC presented here will allow for the accurate modeling of the binding of calcium-sensitizing drugs to cardiac TnC, in addition to revealing the structural basis for the regulation of cardiac versus skeletal muscle contraction.