Structure of the C-domain of Human Cardiac Troponin C in Complex with the Ca2+ Sensitizing Drug EMD 57033*

Ca2+ binding to cardiac troponin C (cTnC) triggers contraction in heart muscle. In heart failure, myofilaments response to Ca2+ are often altered and compounds that sensitize the myofilaments to Ca2+possess therapeutic value in this syndrome. One of the most potent and selective Ca2+ sensitizers is the thiadiazinone derivative EMD 57033, which increases myocardial contractile function bothin vivo and in vitro and interacts with cTnCin vitro. We have determined the NMR structure of the 1:1 complex between Ca2+-saturated C-domain of human cTnC (cCTnC) and EMD 57033. Favorable hydrophobic interactions between the drug and the protein position EMD 57033 in the hydrophobic cleft of the protein. The drug molecule is orientated such that the chiral group of EMD 57033 fits deep in the hydrophobic pocket and makes several key contacts with the protein. This stereospecific interaction explains why the (−)-enantiomer of EMD 57033 is inactive. Titrations of the cCTnC·EMD 57033 complex with two regions of cardiac troponin I (cTnI34–71 and cTnI128–147) reveal that the drug does not share a common binding epitope with cTnI128–147 but is completely displaced by cTnI34–71. These results have important implications for elucidating the mechanism of the Ca2+ sensitizing effect of EMD 57033 in cardiac muscle contraction.

In order to function properly, heart muscle must respond efficiently to the transient increases in cytosolic Ca 2ϩ levels in the myocardial cell. In the syndrome of heart failure, there is strong evidence that the amount of Ca 2ϩ available for contraction is depressed as is the maximum force generating capability of the myofilaments (for reviews, see Refs. [1][2][3]. Treating this condition by elevating intracellular Ca 2ϩ has the potential threat of inducing arrhythmias. In contrast, the ability to sensitize cardiac muscle to Ca 2ϩ and increase cardiac contractility has the considerable advantage of increasing tension with little or no change in intracellular Ca 2ϩ and thus no increase in the energy required to release and transport Ca 2ϩ (for a review, see Ref. 1). A logical target for such Ca 2ϩ sensitizing drugs is cardiac troponin C (cTnC) 1 due to its role as the Ca 2ϩ binding receptor on the thin filament of cardiac muscle.
cTnC is a member of the EF-hand family of Ca 2ϩ -binding proteins. Its structure represents a dumbbell with the N-and C-domains connected by a flexible linker in solution (4). Both domains contain a core of hydrophobic residues. Once exposed, these hydrophobic residues are essential for the binding of cTnI to cTnC and transmitting the Ca 2ϩ signal to other proteins in the thin filaments, and ultimately signal the activation of the myosin-actin ATPase reaction (for reviews, see Refs. 5 and 6). Structural studies have shown that the apo N-domain of cTnC (cNTnC) adopts a "closed" conformation with most of its hydrophobic residues buried (7), like the apo N-domain of sTnC (sNTnC) (8,9). However, the binding of Ca 2ϩ has strikingly different structural consequences in cNTnC and sNTnC. In sTnC, the N-domain switches from a closed to an "open" conformation upon binding Ca 2ϩ (9), while the N-domain of cTnC remains in a closed state in the Ca 2ϩ bound state (7). Consequently, a large hydrophobic surface is exposed in the Ca 2ϩsaturated sNTnC, but not in the Ca 2ϩ -saturated cNTnC. This is mainly due to the fact that sNTnC contains two functional Ca 2ϩ -binding sites, while cNTnC contains only one (10).
The exposed hydrophobic surface on the Ca 2ϩ -saturated sNTnC has been shown as the sTnI-binding site (11)(12)(13)(14). Although Ca 2ϩ binding to cNTnC induces little structural changes, it sets the stage for cTnI binding. In the end, both cNTnC and sNTnC adopt similar conformations in binding their respective TnI regions. Specifically, sTnI 115-131 was found to bind to the hydrophobic cleft of Ca 2ϩ -saturated sNTnC (13) and the corresponding cTnI 147-163 also interacts with the hydrophobic cleft of Ca 2ϩ -saturated cNTnC and stabilizes the opening conformation of cNTnC (15). This region of TnI has been identified by many biological and biophysical studies to be the region responsible for binding to the regulatory domain of TnC and this interaction modulates the binding of the Nterminal and inhibitory regions of TnI to the C-domain of TnC (for reviews, see Refs. 16 and 17).
Unlike the apo N-terminal domain, the apo C-terminal do-main in both sTnC and cTnC possesses a largely unstructured state. Upon binding two Ca 2ϩ ions, this domain folds into a compact globular structure (18,19) and exhibits a similar but slightly less open conformation as that of the Ca 2ϩ -saturated sNTnC. Two regions of TnI have been shown to interact with the Ca 2ϩ -bound C-domain (12). These include the inhibitory region corresponding to sTnI [96][97][98][99][100][101][102][103][104][105][106][107][108][109][110][111][112][113][114][115] or cTnI 128 -147 , and the Nterminal region corresponding to sTnI  or cTnI 34 -71 . The inhibitory region is the critical functional region in the interaction of TnI with TnC and its movement from TnC to actintropomyosin is believed to be the major switch between muscle contraction and relaxation, while the N-terminal region plays primarily a structural role (12). NMR studies of the inhibitory peptides have yielded some structural information on the interaction of inhibitory region with TnC (20 -22), however, the exact binding sites for the inhibitory region on TnC has been under debate. The crystal structure of sTnC in complex with sTnI 1-47 has shown that sTnI 3-33 forms a long ␣-helix and binds to the hydrophobic groove of sCTnC (23). The corresponding region of cTnI  has also been shown to bind within the hydrophobic patch of cCTnC (18).
In view of the importance of the exposed hydrophobic surfaces on both domains of cTnC for the binding of cTnI, it is clear that one way to enhance the Ca 2ϩ sensitivity of cardiac muscle would be to stabilize the interaction of cTnC and cTnI by amplifying the hydrophobic cTnI-binding interface on cTnC. This can be accomplished by employing certain pharmacological agents that bind to the hydrophobic cleft but do not interfere with cTnC-cTnI interaction. Indeed, a variety of small hydrophobic compounds including the calmodulin antagonists bepridil, trifluoperazine, and calmidazolium have been shown to increase the Ca 2ϩ sensitivity of cardiac muscle preparations (for a review, see Ref. 1). However, most of these compounds are not good Ca 2ϩ sensitizers due to other undesired properties. This led to a search for new generations of Ca 2ϩ sensitizing compounds, among which is the thiadiazinone derivative EMD 57033 (24 -26). This drug has been found to increase the Ca 2ϩ sensitivity of both myofibrillar ATPase and force development by skinned muscle fibers (26,27). Recent in vivo studies have also demonstrated that EMD 57033 enhances cardiac contractile function without affecting Ca 2ϩ homeostasis (28 -32). The myofibrillar Ca 2ϩ sensitizing effects of EMD 57033 are remarkably stereo-specific. EMD 57033 is the (ϩ)-enantiomer of a racemate. The (Ϫ)-enantiomer, EMD 57439, exhibits no Ca 2ϩ sensitizing activity but acts as a pure phosphodiesterase III inhibitor (26,27). The molecular mechanism underlining the Ca 2ϩ sensitizing effects of EMD 57033 is not well understood. It is possible that the compound interacts directly with cTnC and increases its affinity for Ca 2ϩ . It is also possible that it exerts an effect on the interface of myosin-actin. An earlier fluorescence study has demonstrated that EMD 57033 interacts with isolated cTnC in a Ca 2ϩ -dependent and stereoselective manner (33). A NMR study has shown that EMD 57033 induces chemical shift changes of Met methyl groups in the C-domain of cTnC (34). Using NMR chemical shift mapping, we have demonstrated that the EMD 57033-binding site is in the hydrophobic pocket of the C-domain of cTnC and that this drug does not interfere with the binding of the inhibitory region of cTnI to cTnC (35). These findings suggest that EMD 57033 may exert its positive inotropic effect by binding to the structural domain of cTnC, modulating the interaction between cTnC and other thin filament proteins, rather than directly enhancing Ca 2ϩ binding to the regulatory domain of cTnC.
In order to delineate the molecular basis of the interaction between cTnC and EMD 57033, we have determined the NMR solution structure of the 1:1 complex between the Ca 2ϩ -satu-rated C-domain of human cTnC and EMD 57033. In the present structure, the drug molecule is oriented such that the chiral group of EMD 57033 fits deep in the hydrophobic pocket and makes several key contacts with the protein. This stereospecific interaction explains why the (Ϫ)-enantiomer of EMD 57033 is inactive. This structure provides a structural basis for the Ca 2ϩ sensitizing effect of EMD 57033. We also examined the possible competition of the drug with cTnI peptides and found that the drug does not share the common binding epitope with the inhibitory region of cTnI but is displaced completely by the N-terminal region of cTnI. These results have important implications in understanding the mechanism underlining Ca 2ϩ sensitizing effects of EMD 57033 in cardiac muscle contraction.

EXPERIMENTAL PROCEDURES
Sample Preparation-The engineering of the expression vector for the cCTnC-(91-161) protein was as described in Chandra et al. (36). The expression and purification of 15 N-and 15 N/ 13 C-labeled protein in Escherichia coli follows the procedures as described previously for sNTnC (37), except a further purification step was done using a Superdex-75 column (Amersham Pharmacia Biotech) with a buffer containing 0.15 M NaCl and 50 mM Tris, pH 8.0. The handling and characterization of the stock solutions of EMD 57033 are as described in Li et al. (35). Two stock solutions (94 and 53 mM, respectively) of EMD 57033 in Me 2 SO-d 6 (Cambridge Isotopes Inc.) were prepared. The synthetic peptides, cTnI 128 -147 , acetyl-TQKIFDLRGKFKRPTLRRVR-amide, and cTnI 34 -71 , acetyl-AKKKSKISASRKLQLKTLLLQIAKQELEREA-EERRGEK-amide, were prepared as described for sTnI 96 -115 and sTnI 1-40 by Tripet et al. (12). Solid peptides were dissolved in double distilled water to make stock solutions. The concentrations were 60 mM for cTnI 128 -147 and 32 mM for cTnI 34 -71 , respectively, as determined by amino acid analysis. All NMR samples were 500 l in volume. The buffer conditions were 100 mM KCl, 10 mM imidazole, 0.2 mM 2,2dimethyl-2-silapentanesulfonic acid, and 0.01% NaN 3 in 90% H 2 O, 10% D 2 O, and the pH was 6.7. For structure determination, NMR samples contained 1-2 mM 15 N-cCTnC or 15 N/ 13 C-cCTnC saturated with ϳ10 mM CaCl 2 . The EMD 57033 was titrated to the Ca 2ϩ -saturated cCTnC until ratios reached 1:1. The NMR samples used for titrations with cTnI 128 -147 or cTnI 34 -71 were prepared in a similar manner as described above. The titrations of cCTnC by cTnI 128 -147 or cTnI 34 -71 follow the procedures as described previously (35).
Stability of EMD 57033 in Aqueous Solution-EMD 57033 is only marginally soluble in aqueous solutions. Once it is bound to cCTnC, the complex is soluble and is stable for 4 -5 days, which is long enough for a typical three-dimensional NMR experiment. Two-dimensional { 1 H, 15 N}-HSQC NMR spectra acquired before and after every threedimensional NMR experiment were compared and their identity was taken as an indication of sample stability throughout the three-dimensional NMR experiment. However, EMD 57033 tends to dissociate from the complex and precipitates out of the aqueous environment after 4 -5 days. This is reflected in the disappearance of the EMD 57033-induced chemical shift changes ( Fig. 1), which, however, can be reproduced upon the addition of more EMD 57033. This indicates that the protein is still intact and it is the drug that is disappearing from the solution. 15 N-Filtered DIPSI spectra of EMD 57033 in the cCTnC-EMD 57033 NMR sample detected no new resonances except those of intact EMD 57033, implying that the drug is not breaking down, but precipitating out of the solution. This conclusion was further supported by evidence from mass spectrometry measurements. 2 1 H NMR spectra were used to check any possible modification of EMD 57033 by chemicals present in the NMR buffer, such as imidazole and NaN 3 , which have been shown to interact with levosimendan (38), and the results show no sign of reaction between EMD 57033 and the chemical reagents.
NMR Spectroscopy-Most NMR data used in this study were collected at 30°C using Unity 600 MHz, Unity Inova 500 MHz, and Inova 800 MHz spectrometers. All three spectrometers are equipped with triple resonance probes and Z-pulsed field gradients (XYZ gradients for the 800 MHz). Unity 300 spectrometer was also used to collect the spectra of EMD 57033 in Me 2 SO-d 6 . Two-dimensional { 1 H, 15 N}-HSQC NMR spectra were acquired using the sensitivity enhanced gradient pulse scheme developed by Lewis E. Kay and co-workers (39,40). For cCTnC in the cCTnC⅐EMD 57033 complex, the chemical shift assignments of the backbone and side chain atoms and NOE interproton distance restraints were determined using the two-dimensional and three-dimensional NMR experiments described in Table I. For the bound EMD 57033, the proton chemical shifts of the drug were assigned by using two-dimensional 15 N/ 13 C-filtered NOESY (80 ms mixing time) and two-dimensional 15 N/ 13 C-filtered DIPSI experiments (43.2-ms spinlock time). The pulse sequences for these two experiments were based on Ogura et al. (41) with extensive modifications done in this lab (Leo Spyracopoulos, University of Alberta). One-dimensional 1 H and twodimensional DIPSI spectra of EMD 57033 in Me 2 SO-d 6 were obtained using the Unity 300 MHz spectrometer. The intermolecular proton distances were obtained using three-dimensional 15 N/ 13 C F 1 -filtered, F 3 -edited NOESY HSQC experiments employing linear frequency ramped broadband inversion pulses for 13 C (80-ms mixing time) (42).
Data Processing and Peak Calibration-All two-dimensional and three-dimensional NMR data was processed using NMRPipe (43), and all one-dimensional NMR data were processed using VNMR (Varian Associates). The spectra were analyzed using NMRView (44). For cCTnC in the complex, intramolecular distance restraints obtained from the NOESY experiments were calibrated according to Gagné et al. (45). Intramolecular proton distances for EMD 57033 in the complex were calibrated based on NOEs corresponding to known distances (neighboring protons on aromatic ring are separated by 2.48 Å). Intermolecular NOEs obtained from the 15 N/ 13 C-filtered/edited experiment were categorized as either strong (1.8 -3.0 Å), medium (1.8 to 4.0 Å), or weak (1.8 to 5.5 Å). Dihedral angle restraints were derived from data obtained from HNHA, HNHB, and NOESY-HSQC experiments according to Sia et al. (4).
Structural Calculations-Using an initial set of intramolecular NOE restraints for cCTnC, 100 structures of cCTnC without EMD 57033 were calculated starting from an extended conformation. The calculations were done using simulated annealing protocol implemented in X-PLOR (46) with 10,000 high-temperature steps (time step of 30 ps) and 6000 cooling steps (time step of 30 ps). Approximately 50% of the initial structures converged. These structures were used as templates for further rounds of refinements. Dihedral angle restraints and 12 artificial distance restraints from chelating oxygens to the two Ca 2ϩ ions were added at later stages of the refinement process. The structure of the cCTnC⅐EMD 57033 complex were calculated starting from the extended conformations of cCTnC and EMD 57033 using simulated annealing protocol with the same conditions as above. The calculations were carried out using the distance and dihedral restraints for cCTnC, 12 distance restraints to Ca 2ϩ ions, 14 intermolecular distance restraints between cCTnC and EMD 57033, as well as 8 intramolecular distance restraints for the bound EMD 57033 molecule (see Table II). The final family of solution structures presented in this article consists of 30 of the lowest energy structures.
Coordinates-The coordinates for the structure have been deposited in the RCSB Protein Data Bank (1IH0). Chemical shifts assignments for cCTnC and EMD 57033 have been deposited in the BioMagResBank (4994).

RESULTS AND DISCUSSION
Structure of the cCTnC⅐EMD 57033 Complex-We have shown clearly that EMD 57033 forms a 1:1 complex (K D ϭ ϳ8 M) with intact cTnC and the binding site resides in the Cdomain (35). For the purpose of this study, we have made a complex between an isolated Ca 2ϩ -saturated C-domain (residues 91-161) of human cTnC and EMD 57033. This reduces the NMR spectral overlap and facilitates the structure determination process. The two-dimensional { 1 H, 15 N}-HSQC NMR spectrum of the Ca 2ϩ -saturated cCTnC, shown in Fig. 1, indicates that it is a well structured domain characterized by the dispersion of amide proton signals. Titration of EMD 57033 induces progressive shifts of the cross-peaks. All the chemical shift changes fall into the fast exchange limit on the NMR time scale. The linear movement of the cross-peaks indicates that EMD 57033 binding to cCTnC occurs with a 1:1 stoichiometry. At the end of titration, a stable 1:1 cCTnC⅐EMD 57033 complex (K D ϭ ϳ10 M) was formed.
The two-dimensional { 1 H, 15 N}-HSQC NMR spectrum ( Fig. 1) of cCTnC in the cCTnC⅐EMD 57033 complex was highly resolved, which allowed the chemical shifts of the backbone and the side chain atoms to be readily assigned using 15 N-or 15 N/ 13 C-labeled protein. Distance restraints for cCTnC in the cCTnC⅐EMD 57033 complex were obtained by analyzing threedimensional 15 N-or 15 N/ 13 C-NOESY experiments. Dihedral angle restraints for cCTnC in the cCTnC⅐EMD 57033 complex were obtained from three-dimensional HNHA and HNHB experiments. The NMR experiments performed are summarized in Table I.
The proton NMR chemical shifts assignments and intramolecular distance restraints for the bound EMD 57033 required the collection of 15 N/ 13 C filtered two-dimensional DIPSI and two-dimensional NOESY experiments using unlabeled drug bound to 15 15 N}-HSQC NMR spectra arising from the backbone NH and side chain NH 2 groups of cTnC are superimposed for various EMD 57033 additions. Each titration point represents the addition of 1 l of 94 mM EMD 57033 stock solution to a NMR sample containing 0.92 mM 15 N-labeled cCTnC. Assignments of the crosspeaks are indicated. Cross-peaks corresponding to free cCTnC with no drug added are shown as multiple contours. Cross-peaks from subsequent spectra are shown as single contours. The arrow indicates the direction of the movement of cross-peak for Gly 140 with increasing EMD 57033 concentrations. dimensional 1 H and two-dimensional DIPSI NMR spectra of free EMD 57033 in Me 2 SO-d 6 were also analyzed to aid the assignments. Intermolecular NOE correlations between cCTnC and EMD 57033 were determined from three-dimensional 15 N/ 13 C-filtered, 15 N/ 13 C-edited NOESY data. These experiments were optimized 3 for detecting NOEs between protons attached to 14 N/ 12 C and protons attached to 15 N/ 13 C, thereby editing out all the intramolecular NOEs. A total of 1000 experimental distance restraints (approximately 14 restraints per residue) including 974 for cCTnC, 8 for EMD 57033, 14 intermolecular contacts between cCTnC and EMD 57033, 61 dihedral restraints for cCTnC, and 12 restraints to Ca 2ϩ were used to calculate the high resolution structure of the cCTnC⅐EMD 57033 complex. Fig Fig. 3A), the tetrahydroquinolinyl (B and C rings in Fig. 3A), and the dimethoxybenzoyl (D ring in Fig. 3A) moieties, respectively. The three functional groups have rigid conforma-tions, however, the bonds that connect the thiadiazinone-tetrahydroquinolinyl units and the tetrahydroquinolinyl-dimethoxybenzoyl units can rotate freely (see Fig. 3A). The drug molecule is completely assigned and the chemical shifts for all the protons are labeled in Fig. 3A. Eight intramolecular NOEs of the bound drug were observed and shown in Fig. 3, B and C. The relative orientations of the three moieties in the cCTnC⅐EMD 57033 complex are determined by both the intramolecular NOEs within the drug molecule and the intermolecular NOEs between cCTnC and the drug. In the ensemble of structures ( Fig. 2A), the interplanetary angles for the thiadiazinone-tetrahydroquinolinyl and the tetrahydroquinolinyl-dimethoxybenzoyl units are 44°Ϯ 21°and 69°Ϯ 14°, respectively. When the heavy atoms of EMD 57033 superimpose onto the average structure in the cCTnC⅐EMD complex, the average r.m.s.d. is 0.29 Ϯ 0.07 Å. This r.m.s.d. increases to 1.15 Å when the backbone atoms of residues 95-158 of cCTnC in the ensemble of solution structures for the cCTnC⅐EMD 57033 complex are superimposed onto the average cCTnC structure. As a result, the stereochemical quality of the drug molecule assembly is slightly lower than the structures of cCTnC in the cCTnC⅐EMD 57033 complex. This is due in part to the loosely imposed intermolecular restraints between cCTnC and EMD 57033 because of a lack of calibration for those NOEs (see "Experimental Procedures").
Binding Interface between EMD 57033 and cCTnC in the cCTnC⅐EMD 57033 Complex-Strip plots taken from the threedimensional 15 N/ 13 C F 1 -filtered, F 3 -edited spectrum of the cCTnC⅐EMD 57033 complex are shown in Fig. 4A. In this spectrum, only NOEs arising from the drug protons (attached to 12 C) and terminating on protein protons (attached to 13 C) are observed. For example, the H7# methyl protons from the thiadiazinone unit show strong NOE contacts with the protons attached to the methyl groups of Leu 117 located in the F-helix (Fig. 4B, c) and of Ile 112 and Ile 148 located in the ␤-sheet of cCTnC (Fig. 4B, a); the H9, H12, and H13 protons from the B ring (Fig. 3A) show NOE contacts with the protons attached to the methyl groups of Leu 136 located in the G-helix (Fig. 4B, b); and the two H17 protons from the C ring (Fig. 3A) show NOE contacts with the protons attached to the methyl groups of 3 L. Spyracopoulos, unpublished data. Met 157 located in the H-helix (Fig. 4B, d). These experimental intermolecular restraints serve to orient the drug molecule in the cCTnC⅐EMD 57033 complex. As a result, the thiadiazinone group is located at the bottom of the cCTnC pocket in a small hydrophobic cavity (Fig. 2C). It seems that the strong contacts between the chiral methyl group and three residues of cCTnC (Ile 112 , Leu 117 , and Ile 148 ) serve as the primary anchoring site for this drug molecule (see below for biological implications). The dimethoxybenzoyl group protrudes toward outside of the hydrophobic pocket (Fig. 2C). Although most interactions between cCTnC and EMD 57033 are hydrophobic in nature, the fact that the chiral methyl (H7#) group delves deep into the hydrophobic pocket of cCTnC while the two methoxy moieties, despite their hydrophobic nature, stick out to indicate the interaction is not nonspecific. Another important interaction is between the methyl group of Met 157 located in the C-terminal G-helix and H17# protons from ring C (Fig. 3A) of EMD 57033. This contact serves mainly to hold the plane formed by rings B and C of the drug in an orientation that is almost perpendicular to the plane formed by ring A, and to allow the residues on helix G (Leu 136 ) and at the end of helix H (Met 157 ) to interact with the same proton (H9) on ring B. The contacts between the protons (H12 and H13) on ring B and the methyl groups of Leu 117 and Leu 136 help the drug bound to the protein firmly. Leu 121 located on the C-terminal of F-helix makes only weak NOE contacts with aromatic protons on both ring B (H12) and ring D (H20, H23, and H24) of EMD 57033, which also play a role in binding the drug to cCTnC. Overall, the protons from EMD 57033 make 14 NOE contacts with protons of the methyl groups from six hydrophobic residues (Ile 112 , Leu 117 , Leu 121 , Leu 136 , Ile 148 , and Met 157 ) of cCTnC and these interactions keep the drug molecule fit snuggly in the hydrophobic cavity of Ca 2ϩ -saturated cCTnC (Fig. 2C). The total "exposed" nonpolar accessible surface area for residues 95-158 of cCTnC, excluding the drug, in the cCTnC-EMD 57033 complex is ϳ 2689 Å 2 , as measured by GRASP (47)  overall fold of cCTnC in the cCTnC⅐EMD 57033 complex is similar to the structure of the C-domain of intact cTnC determined by solution NMR spectroscopy (4), with a r.m.s.d. of ϳ 1.7 Å for backbone overlay (Fig. 5A). There are several EMD 57033-induced structural changes. Among the most notable ones, the cCTnC in the cCTnC⅐EMD 57033 complex is more open, i.e. the F/G helix unit is farther away from the E/H helix unit, than the free C-domain of cTnC. This is mostly due to the strong interactions of the drug with residues Leu 117 and Leu 121 located on the F helix, which pulls helix F away from helix E and toward helix G. This is quantified by a decrease in the E/F interhelical angles (from 115 Ϯ 4°for the C-domain in cTnC to 92 Ϯ 4°for cCTnC in the cCTnC⅐EMD 57033 complex). The interhelical angle between G and H also showed a slight decrease (from 121 Ϯ 4°for the C-domain in cTnC to 113 Ϯ 7°for cCTnC in the cCTnC⅐EMD 57033 complex) (see Table III). The slight opening of cCTnC and movement of the F/G helix unit away from the E/H helix unit was also observed by the binding of the cTnI 33-80 peptide to cCTnC (18). It seems that these movements are necessary for the C-domain of cTnC to provide a binding site for cTnI or a drug molecule. The degree of EMD 57033-induced structural opening of cCTnC is also measured by an increase of ϳ20 Å 2 of exposed apolar surface area for cCTnC in the cCTnC⅐EMD 57033 complex, compared with the free C-domain of cTnC. Although EMD 57033 induced little G/H interhelical angle changes, the multiple contacts between Leu 136 of the G helix with several ring protons of EMD 57033 resulted in a more flexible G helix in the present structure than the C-domain in the cTnC structure (r.m.s.d. of 0.36 Å for cCTnC in the cCTnC⅐EMD 57033 versus 0.24 Å for the Cdomain in cTnC). Another significant EMD-induced change on cCTnC is the unwinding of residues 156 -161 at the end of helix H. This last helix has always been well defined to the Cterminal end in many EF-hand domains such as sNTnC (9,48), cNTnC (7), and CaM domains (49). However, in this complex, helix H ends at residue 156, the last portion of the C-terminal residues (157-161) forms an extended structure and appears to be more flexible than usual. This can be attributed to the interactions between EMD 57033 and residue Met 157 of cCTnC. Interestingly, the same extended conformation at the end of the H helix was observed in the C-domain of sTnC bound to the N-terminal region of sTnI (sTnI 1-47 ) (23). This suggests that the nature of the interactions between EMD 57033 and cCTnC may be similar to that between sTnI 1-47 and sTnC. An overlay (Fig. 5B) of the cCTnC in the cCTnC⅐EMD complex and sCTnC in the sCTnC⅐sTnI 1-47 complex shows that the backbones of cCTnC and sCTnC adopt very similar conformations (r.m.s.d. ϭ 1.3 Å). It also shows that the binding of EMD 57033 to cCTnC possesses similar features as the binding of sTnI  to sCTnC. In fact, residues Ile 110 , Leu 115 , Phe 119 , Leu 134 , Ile 146 , and Met 155 of sCTnC, corresponding to the six EMD 57033contacting residues in cCTnC, are all involved in the interaction with sTnI 1-47 (23). It is interesting to note that binding of sTnI 1-47 to sCTnC does not greatly perturb the fold of this domain from a peptide-free conformation. In the free state, sCTnC exhibits E/F and G/H interhelical angles of ϳ105°and ϳ115°, respectively (Table III). These angles do not change very much upon sTnI 1-47 binding to sCTnC. On the other hand, both EMD 57033 and cTnI  induce interhelical angle changes in cCTnC and the bound cCTnC exhibits a more open conformation. This is analogous to the ligand induced structural changes in the N-domain of sTnC and cTnC. Ca 2ϩ binding switches sNTnC from a closed to open conformation (9) and the open sNTnC is ready to bind sTnI 115-131 with no need to open further (13). On the other hand, Ca 2ϩ binding to cNTnC induces little structural changes but sets the stage for cTnI binding and cNTnC undergoes a large closed to open transition upon binding cTnI 147-163 (15) or Bepridil (50). Thus, the interaction of both domains of cTnC with cTnI would require more free energy than those of sTnC with sTnI, because cTnI has to overcome the energy barrier of opening the domains of cTnC (see discussions in McKay et al. (51)). This energy cost may be compensated by employing pharmacological agents, which bind to both domains of cTnC and thereby help to stabilize the open conformation of cTnC. The ideal situation would be that the interaction between the drug and cTnC occurs without interfering with cTnI binding to cTnC. The analysis presented herein raises issues about the effect of EMD 57033 on the interaction of cCTnC and cTnI. Since two regions of cTnI (cTnI 128 -147 and cTnI 34 -71 ) were identified to interact with the C-terminal domain of cTnC (12), we examined the binding of these two synthetic peptides (cTnI 128 -147 and cTnI 34 -71 ) to the cCTnC⅐EMD 57033 complex as discussed below.
Competition of EMD 57033 with cTnI 128 -147 and cTnI 34 -71 -Previously, we have shown that both EMD 57033 and cTnI 128 -147 bind with a 1:1 stoichiometry to cTnC but do not compete for the same binding sites on cTnC (35). EMD 57033 binding is not affected by cTnI 128 -147 nor does the drug affect cTnI 128 -147 binding. In the end, a stable ternary cTnC⅐EMD 57033⅐cTnI 128 -147 complex is formed. In the present work, sim- ilar titrations were performed on cCTnC and similar conclusions were obtained (data not shown). Both EMD 57033 and cTnI 128 -147 can bind to cCTnC simultaneously and the binding affinities (K D ϭ ϳ10 M for EMD 57033 and K D ϭ ϳ100 M for cTnI 128 -147 ) are similar to those determined from binding to intact cTnC (35). However, unlike cTnI 128 -147 , the N-terminal region of cTnI, cTnI 34 -71 , can displace EMD 57033 completely from cCTnC. When the 38-residue cTnI 34 -71 peptide is titrated to the Ca 2ϩ -saturated cCTnC, it binds tightly (K D Յ 1 M) to form a stable cCTnC⅐cTnI 34 -71 complex. Fig. 6A shows a superimposition of the two-dimensional { 1 H, 15 N}-HSQC NMR spectra of cCTnC and the cCTnC⅐cTnI 34 -71 complex. Similar to sTnI 1-40 binding to sCTnC (19), cTnI 34 -71 binding to cCTnC occurs with slow exchange kinetics on the NMR time scale. Thus, as the titration progresses, the resonance peaks corresponding to cCTnC becomes less intense while those corresponding to cCTnC-cTnI 34 -71 grow. When the cTnI 34-71 :cCTnC ratio reaches 1:1, all cross-peaks corresponding to cCTnC have completely disappeared while those corresponding to cCTnC-cTnI 34 -71 attain maximum intensity. When cTnI 34 -71 is titrated to the cCTnC⅐EMD 57033 complex, EMD 57033 is displaced completely by cTnI 34 -71 and a stable cCTnC⅐cTnI 34 -71 complex forms (K D Յ 1 M). The binding behavior of cTnI 34 -71 to the cCTnC⅐EMD complex is very similar to that of cTnI 34 -71 to cCTnC (Fig. 6A) and is illustrated in Fig. 6B, which shows a superimposition of the two-dimensional { 1 H, 15 N}-HSQC NMR spectra of cCTnC-EMD 57033 and cCTnC-cTnI 34 -71 . The only difference between Fig. 6, A and B, is the starting spectra. The former started with the spectrum of cCTnC and the latter started with the spectrum of cCTnC-EMD 57033. The end spectra are identical and represent that of the cCTnC⅐cTnI 34 -71 complex. Thus, cTnI 34 -71 associates tightly with cCTnC regardless of the presence of EMD 57033. It is necessary to put these results into perspective with respect to the interaction of cTnC and cTnI. The noncompetitive binding sites of cTnI 128 -147 and EMD 57033 on cTnC (35) or cCTnC (present data) suggest that the inhibitory region of cTnI does not block the binding of EMD 57033 to cTnC, and vice versa. This conclusion is important in terms of addressing the Ca 2ϩ sensitizing role of EMD 57033 because the inhibitory region of cTnI constitutes a major switch between muscle contraction/relaxation by moving between cTnC and actin-tropomyosin (52). A good Ca 2ϩ sensitizer would not interfere with the inhibitory function of cTnI. A pertinent question is whether this is also the case with intact cTnI, especially considering the present clear-cut results that EMD 57033 cannot compete with cTnI 34 -71 for cCTnC. The very tight association of sTnI 1-40 and the sCTnC have been shown by several groups. In addition to the sTnC-sTnI 1-47 crystal structure (23), early functional studies of this region by Ngai and Hodges (53) have shown that sTnI  can effectively compete with sTnI or sTnI 96 -115 inhibitory peptide for sTnC and a recent NMR study has shown that sTnI 1-40 binds strongly to sCTnC with a K D of ϳ2 M and can displace sTnI 96 -115 completely (19). This raises questions regarding how the inhibitory region binds to TnC to release inhibition if the N-terminal region of TnI is always present. In order to rationalize these results, two models have been proposed for the interaction of TnC and TnI. One suggests that these two regions of TnI share overlapping binding sites on the C-domain of TnC, which are alternatively occupied by either one or the other depending on the interactions between the N-domain of TnC and the C-domain of TnI or the C-domain of TnT (12). The second model proposes that the N-terminal region of TnI always binds to the C-domain of TnC, regardless of the Ca 2ϩ -dependent interactions between the N-domain of TnC and the C-domain of TnI (23,54,55)  C-domain) of TnC in a Ca 2ϩ -dependent manner. The later model does not adequately explain the experimental data (20, 35, 56 -59), supporting a binding site for the inhibitory region of TnI primarily in the C-domain of TnC, especially when only isolated C-domains of TnC were used in the study (19,57). Based on our titration data of cTnI 128 -147 binding to cTnC (35) and cCTnC (present work) and the competitive binding of sTnI  and sTnI 96 -115 to sCTnC (19), we suggest that cTnI 128 -147 binds primarily to the C-domain of cTnC and in order for this binding to occur, the interaction between cTnI 34 -71 and cCTnC has to be weakened. This can be accomplished by the mechanisms proposed for model 1 (12). Since cTnC does not act alone, and the contractile proteins work in a highly organized and cooperative manner in muscle contraction, it is possible that cCTnC in myofilaments may have a lower affinity for cTnI 34 -71 than it does in isolation. Although EMD 57033 alone is too small to compete with the extensive contacts between the long ␣-helix of cTnI 34 -71 and cCTnC, the fact that EMD 57033 interacts with many of the same residues on cCTnC as sTnI 1-47 on sCTnC suggests that the drug may play a role in disrupting and therefore weakening the interaction of cTnI 34 -71 with cCTnC in the myofilaments, and consequently, in enhancing the binding of the inhibitory region of TnI to TnC. Since this interaction is Ca 2ϩ -dependent, the apparent Ca 2ϩ sensitivity of the contractile system can be modulated by EMD 57033.
Implications in the Ca 2ϩ Sensitizing Effect of EMD 57033 in Cardiac Muscle Contraction-The number of patients suffering from congestive heart failure is rising accompanying the aging of baby boomers. Ca 2ϩ sensitizers have been proposed as a treatment for this common disease. These agents increase myocardial contractility by generating more force for a given amount of cytosolic free Ca 2ϩ . This allows an achievement of positive inotropic effect more economically as compared with other positive inotropic drugs that exert effect by simply enhancing Ca 2ϩ influx into myocytes and therefore add intracellular Ca 2ϩ overload. Numerous studies have documented the Ca 2ϩ sensitizing effects of this class of agents under in vivo and in vitro conditions (for a review, see Ref. 1). Among which, EMD 57033 is one of the most potent and selective Ca 2ϩ sensitizers available. Earlier physiological studies have shown that EMD 57033 exerts detectable positive inotropic effects on isolated cardiac myocytes at concentrations as low as 1 M (27), and in skinned cardiac muscle fibers, less than 10 M EMD 57033 induced significant Ca 2ϩ sensitivity of force development (26). In a recent study, 0.3-1 M EMD 57033 is shown to have positive inotropic effects on both normal and failing cardiac myocytes (32). The cCTnC⅐EMD 57033 complex structure presented in this work provided a structural basis for the understanding of the mechanism underlining the Ca 2ϩ sensitizing effect of EMD 57033 in cardiac muscle contraction. In the present structure of the cCTnC⅐EMD 57033 complex, the drug molecule is orientated such that the chiral group of EMD 57033 fits deep in the hydrophobic pocket and makes several key contacts with the protein. This stereospecific interaction explains why the (Ϫ)-enantiomer of EMD 57033 is inactive (27). Since the methyl group attached to the chiral carbon of EMD 57033 makes extensive contacts with methyl groups of Ile 112 , Leu 117 , and Ile 148 , these contacts may be weakened or lost if the stereospecificity of the chiral carbon is changed. This is especially true for the interactions between residues on the short ␤-sheet and the drug. These interactions would be eliminated if the chirality of the drug has reversed, not even the rotation of ring A can completely restore all the contacts between the drug and these residues. Interestingly, the (Ϫ)-enantiomer of EMD 57033, EMD 57439 is also capable of stimulating muscle con-traction, but through a different mechanism. EMD 57439 is a potent phosphodiesterase III inhibitor, and as such, is capable of producing an increase in cAMP level inside the cell. This will result in the activation of protein kinase A and ultimately, an increase in Ca 2ϩ concentration in cardiac muscle cells (26,28). The fact that EMD 57439 has no Ca 2ϩ sensitizing activity and EMD 57033 is only a weak phosphodiesterase III inhibitor points to the sensitivity of protein toward stereospecificity of ligands.