Solution Structure of Human Cardiac Troponin C in Complex with the Green Tea Polyphenol, (−)-Epigallocatechin 3-Gallate*

Heart muscle contraction is regulated by Ca2+ binding to the thin filament protein troponin C. In cardiovascular disease, the myofilament response to Ca2+ is often altered. Compounds that rectify this perturbation are of considerable interest as therapeutics. Plant flavonoids have been found to provide protection against a variety of human illnesses such as cancer, infection, and heart disease. (−)-Epigallocatechin gallate (EGCg), the prevalent flavonoid in green tea, modulates force generation in isolated guinea pig hearts (Hotta, Y., Huang, L., Muto, T., Yajima, M., Miyazeki, K., Ishikawa, N., Fukuzawa, Y., Wakida, Y., Tushima, H., Ando, H., and Nonogaki, T. (2006) Eur. J. Pharmacol. 552, 123–130) and in skinned cardiac muscle fibers (Liou, Y. M., Kuo, S. C., and Hsieh, S. R. (2008) Pflugers Arch. 456, 787–800; and Tadano, N., Yumoto, F., Tanokura, M., Ohtsuki, I., and Morimoto, S. (2005) Biophys. J. 88, 314a). In this study we describe the solution structure of the Ca2+-saturated C-terminal domain of troponin C in complex with EGCg. Moreover, we show that EGCg forms a ternary complex with the C-terminal domain of troponin C and the anchoring region of troponin I. The structural evidence indicates that the binding site of EGCg on the C-terminal domain of troponin C is in the hydrophobic pocket in the absence of troponin I, akin to EMD 57033. Based on chemical shift mapping, the binding of EGCg to the C-terminal domain of troponin C in the presence of troponin I may be to a new site formed by the troponin C·troponin I complex. This interaction of EGCg with the C-terminal domain of troponin C·troponin I complex has not been shown with other cardiotonic molecules and illustrates the potential mechanism by which EGCg modulates heart contraction.

States, ϳ1 in 3 deaths in 2004 were caused by CVD (1). In heart failure, the ability of the heart to distribute blood throughout the body is perturbed, and there is a growing interest to develop drugs that directly regulate the response of the myofilament to Ca 2ϩ . Regulation of muscle contraction is triggered by Ca 2ϩ binding to troponin. The troponin complex is situated at regular intervals along the thin filament, which is made up of two elongated polymers, f-actin and tropomyosin. The backbone of the thin filament is composed of actin molecules arranged in a double helix with tropomyosin wound around actin as a coiledcoil. Anchored at every seventh actin molecule is the heterotrimeric troponin complex, which consists of troponin C (TnC), troponin I (TnI), and troponin T (TnT). TnC is the Ca 2ϩ -binding subunit of troponin and has four EF-hand helix-loop-helix motifs. TnI is the inhibitory subunit of troponin. It regulates the actin-myosin cross-bridge formation by flipping between TnC and actin in a Ca 2ϩ -dependent manner. At low levels of cytosolic Ca 2ϩ , TnI is bound to actin, causing tropomyosin to sterically block the binding of the actomyosin cross-bridges. On the other hand, when Ca 2ϩ concentration is high, TnI translocates from actin to TnC inducing tropomyosin to change its orientation on actin so that the actin-myosin interaction may occur. The subunit TnT fetters the troponin complex to the thin filament by way of its association with TnI (for reviews on contraction see Refs. [2][3][4][5]. The large number of structural studies on troponin and the thin filament has helped gain insight into the molecular mechanism of muscle contraction. TnC is a dumbbell-shaped protein that consists of terminal domains connected by an elongated flexible linker, as shown by solution NMR (6). The overall folds of the terminal domains of skeletal TnC (sTnC) and cardiac TnC (cTnC) are very similar (7)(8)(9). The apo state of the N-domain of sTnC (sNTnC) and cTnC (cNTnC) reveals that the domain is in a "closed" conformation, such that the hydrophobic core of the protein is buried (8,10,11). In the skeletal system, sNTnC "opens" when two Ca 2ϩ ions bind (8,10,11). Alternatively, cNTnC contains only one functional Ca 2ϩ -binding site, and its global conformation does not change as significantly as in sNTnC (11). Nonetheless, Ca 2ϩ binding promotes the association of the switch region of cTnI (residues 147-163) with cNTnC. cTnI-(147-163) forms an ␣-helix when associated with cNTnC and has been elucidated by NMR in the solution structure of cNTnC⅐Ca 2ϩ ⅐cTnI-(147-163) (12) and by the x-ray crystallography structure of cTnC⅐3Ca 2ϩ ⅐cTnI⅐- ⅐cTnT-(183-288) (13). The interaction of cTnI-(147-163) with cNTnC⅐Ca 2ϩ is essential to draw the inhibitory (cTnI-(128 -147)) and C-terminal (cTnI-(163-210)) regions of cTnI away from actin. cTnI-(128 -147) is not visualized in the car-diac structure, probably due to disorder (13). In the skeletal crystal structure of sTnC⅐4Ca 2ϩ ⅐sTnI-(1-182)⅐sTnT-(156 -262), however, the inhibitory region of sTnI is visualized and makes electrostatic contacts with the central helix connecting the N-and C-terminal lobes of cTnC (14). The C-domain (CTnC) of both sTnC and cTnC has two functional binding sites for Ca 2ϩ and remains largely unstructured without Ca 2ϩ bound. The folding of this domain occurs in the presence of Ca 2ϩ (15,16). Throughout the relaxation-contraction cycle, cCTnC is Ca 2ϩ -saturated with both Ca 2ϩ -binding sites occupied (cCTnC⅐2Ca 2ϩ ) and is associated with the anchoring region of cTnI (cTnI- (34 -71)). The crystal structure of cTnC⅐3Ca 2ϩ ⅐cTnI⅐-(31-210)⅐cTnT-(183-288) shows cTnI-(34 -71) is ␣-helical when bound with cCTnC⅐2Ca 2ϩ (13). The interaction of cCTnC⅐2Ca 2ϩ with cTnI-(34 -71) is the primary site in which cTnC is tethered to the thin filament.
In light of the importance of the Ca 2ϩ -dependent cTnI-cTnC interaction in the signaling of muscle contraction, the design of drugs that modulate this interaction would be useful in the treatment of heart disease. Compounds that treat CVD through modulation of the activity of cTnC are called Ca 2ϩ sensitizers or desensitizers, depending on whether they positively or negatively influence its function. These drugs are safer than other currently prescribed medicines that alter the cytosolic Ca 2ϩ homeostasis (such as milrinone and dobutamine), which may cause arrhythmia or death with prolonged usage.
The potential therapeutic advantage of Ca 2ϩ (de)sensitizers has led to the development of a number of compounds that target cTnC. Compounds have been identified that elicit their activity through binding either cNTnC or cCTnC. Levosimendan and pimobendan are examples of molecules that increase heart muscle contractility through binding to cNTnC. Conversely, the molecule W7 decreases contractility via its interaction with cNTnC. For recent reviews on the molecular mechanism of these compounds and others see Refs. 17-19. The discovery of small molecules that bind to cCTnC to elicit their Ca 2ϩ -sensitizing effects suggests that cCTnC is also a suitable target for the development of therapeutics. The Ca 2ϩ sensitizer, EMD 57033, is approved for the treatment for heart failure in dogs and binds to cCTnC⅐2Ca 2ϩ (20). In the NMR structure of cCTnC⅐2Ca 2ϩ ⅐ EMD 57033, EMD 57033 is associated in the hydrophobic cavity of cCTnC⅐2Ca 2ϩ (21). The interaction of EMD 57033 with cCTnC is stereospecific for the (ϩ)-enantiomer and explains why the (Ϫ)-enantiomer is inactive (22). Because EMD 57033 has been shown to bind cCTnC⅐2Ca 2ϩ concurrently with cTnI-(128 -147) but not with cTnI-(34 -71) (23), one postulate is that EMD 57033 acts as a Ca 2ϩ sensitizer by weakening the interaction of cTnI-(34 -71) with cCTnC⅐2Ca 2ϩ , thus increasing the propensity of cTnI-(128 -147) to bind cCTnC⅐2Ca 2ϩ in vivo. The dilated cardiomyopathy (DCM) mutation, G159D, of cCTnC has renewed interest in the role of the C-lobe for regulation in contraction. The mutation has been identified to decrease the sensitivity of the thin filament to Ca 2ϩ (24). The source of the DCM phenotype of G159D might come from the modulation of the interaction of cCTnC⅐2Ca 2ϩ with cTnI-(34 -71) (25).
Green tea (Camellia sinensis) is one of the most widely consumed beverages in the world, and several epidemiological studies have linked the consumption of tea with a decrease in CVD (26,27). (Ϫ)-Epigallocatechin gallate (EGCg) is a polyphenol that exists abundantly in unfermented teas and has been identified as a modulator of heart contraction through its interaction with cTnC (28 -30). Here we use NMR spectroscopy to elucidate the three-dimensional structure of the cCTnC⅐ 2Ca 2ϩ ⅐EGCg complex. The solution structure reveals that EGCg binds at the hydrophobic core of cCTnC inducing a small structural "opening." We also use two-dimensional NMR spectroscopy to monitor the binding of EGCg to cCTnC⅐2Ca 2ϩ and cCTnC⅐2Ca 2ϩ ⅐cTnI- (34 -71). Because EGCg and cTnI-(34 -71) can bind cCTnC concurrently, the inotropic effect of EGCg may stem from its modulation of the cTnI-(34 -71)-cCTnC⅐2Ca 2ϩ interaction. The solution structure of cCTnC⅐ 2Ca 2ϩ ⅐EGCg provides insight into the mechanism in which EGCg might influence heart contraction. These results taken with previous research on the Ca 2ϩ sensitizer EMD 57033 and the DCM mutation G159D bring into question the dogma that cNTnC is the exclusive site for regulation of contraction in cTnC.
Stability of EGCg in Aqueous Solution-Certain steps had to be taken when working with EGCg, because it is rapidly oxidized in aqueous solution. For most of the experiments, TCEP was used to keep EGCg reduced for the duration of the longer three-dimensional experiments (4 -5 days). TCEP has been shown to be an adequate reducer of ascorbic acid (33). For some of the shorter experiments, the sample was simply flushed with Ar 2 gas prior to the addition of EGCg. Titrations with EGCg were all initially done with no TCEP present and then repeated with TCEP. This confirmed that TCEP had no influence on the interaction of EGCg with cCTnC (data not shown). To monitor sample stability 1 H, 15 N HSQC spectra were acquired before and after each long three-dimensional experiment. In cases where degradation of EGCg was witnessed, the solution first began to change from clear to a brownish hue, and eventually, precipitate started to form at the bottom of the NMR tube. In this stage the 1 H, 15 N HSQC spectrum revealed a slight recession of the amide correlation peaks toward the unbound chemical shifts of cCTnC. In addition to the visual cues of EGCg degradation, one-dimensional 1 H NMR spectra were used to monitor the transition of EGCg from its reduced form to the oxidized state. When EGCg was oxidized, additional peaks began to appear in the spectrum, and peaks representative of the reduced form of EGCg decreased in intensity (data not shown). Fig. 2 shows the chemical structure and one-dimensional 1 H NMR spectrum of 1.1 mM EGCg in DMSO-d 6 .
Titrations of EGCg to cTnC, cCTnC, and cCTnC⅐cTnI-(34 -71) Monitored by NMR Spectroscopy-All NMR samples contained 500 l of aqueous NMR buffer (see under "Sample Preparation"). The protein concentration was determined by amino acid analysis and tyrosine absorption at 280 nm. Protein and EGCg concentrations were corrected for the dilution factor. The stock solutions of EGCg were prepared in DMSO-d 6 or in an aqueous TCEP buffer. TCEP was used to keep EGCg reduced during the titration. The pH was adjusted with NaOH when necessary. EGCg concentration was determined by weight and by comparing EGCg peak heights or integrals with the internal standard 2,2-dimethyl-2-silapentane-5-sulfonate sodium salt in the one-dimensional 1 H NMR spectrum. EGCg was titrated into a 0. 45  NMR Spectroscopy-Most NMR experiments were collected on either a Varian Inova 500-MHz or a Unity 600-MHz spectrometer. All data were collected at 30°C. Both spectrometers have triple resonance HCN probes and z-pulsed field gradients. The titration of EGCg to 80 M 15 N-cCTnC was done at 25°C on a Varian Inova 800-MHz equipped with a cryogenic probe. The supplemental Table 1 lists the experiments acquired for this work.
Data Processing and Peak Calibration-All two-and threedimensional NMR data were processed with NMRPipe (34). One-dimensional NMR spectra to assign EGCg in DMSO-d 6 and address EGCg degradation were processed using VNMRJ (Varian Inc.) Assignment of chemical shifts was carried out in NMRView (35), and backbone assignments were aided with the software package SmartNotebook (36). Intramolecular NOEs of EGCg measured from two-dimensional NOESY acquired in D 2 O were categorized as strong (1.8 -4.0 Å), medium (1.8 -5.0 Å), and weak (1.8 -6.0 Å). Intermolecular NOEs were categorized as strong (1.8 -4.0Å), medium (1.8 -5.0 Å), and weak (1.8 -6.0 Å). Intramolecular NOEs of cCTnC were calibrated automatically with the CYANA standard procedure, with upper bounds set to 6 Å. The chemical shifts that were assigned in NMRView were converted to CYANA nomenclature for NOE calibration. After the CYANA refinement, the final restraints were converted to XPLOR-NIH nomenclature. Because CYANA calculates distances differently than XPLOR-NIH, NOE restraints were loosened by 1 Å.
Generation of EGCg Structure File for Structure Refinement-The structure file of EGCg was generated for XPLOR-NIH by the PRODRG2 webserver (37). This on-line resource converts the chemical structure of a small ligand into PDB format. Following PDB conversion, XPLO2D was used to generate XPLOR-NIH compatible structure files of EGCg.
Structural Calculation-Structures were initially generated using the program CYANA 2.1 (38 -40). Unambiguous restraints were assigned manually and were forced to keep their assignments during the first four runs of CYANA calculations, after which they were open for automatic assignment with the "NOEassign" command of CYANA (41). Distance restraints were calibrated with CYANA standard procedure using upper limits of 6 Å. Dihedral angle restraints from TALOS (42) were used as well as 12 distance restraints from x-ray crystallographic data of chelating oxygen atoms to the two Ca 2ϩ ions. CYANA was used to calculate 100 structures, of which the 30 conformers with the lowest target function were used to further refine the structure. The 30 conformers were averaged in X-PLOR-NIH and used as a template structure in the simulated annealing protocol, with 10,000 high temperature steps and 6000 cooling steps. After the structure of cCTnC was well defined, the binary cCTnC⅐2Ca 2ϩ ⅐EGCg structure was solved in a similar manner, starting with an extended conformation of cCTnC. The calculations contained 10 intramolecular EGCg NOE restraints and 18 intermolecular NOE restraints. The final ensemble discussed in this study is represented by the 30 lowest energy structures of the 100 calculated (see Table 1 for statistics). The final refined ensemble has been deposited in the Protein Data Bank with the accession code of 2kdh.

RESULTS
Two-dimensional 1 H, 15 N HSQC and 1 H, 13 C HSQC NMR experiments were used to monitor the binding of EGCg to cTnC⅐3Ca 2ϩ , cCTnC⅐2Ca 2ϩ , and cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71). The HSQC experiment correlates backbone or side chain 1 H with 15 N or 13 C nuclei, so that each cross-peak in the spectrum represents an individual 1 H-15 N (or 1 H-13 C) from the 13 C, 15 N-labeled protein. Typically, when a ligand binds a labeled protein, the chemical shifts of individual cross-peaks in the spectrum move as a function of the ligand concentration. The changes in chemical shifts can be quantified to derive ligand stoichiometry and affinity, as well as be used to approximate the ligand-binding site on the protein, an approach commonly referred to as chemical shift mapping.
Effect of EGCg on cTnC⅐3Ca 2ϩ -EGCg was titrated into cTnC⅐3Ca 2ϩ to assess its primary binding site. The backbone of EGCg-free cTnC⅐3Ca 2ϩ was assigned by the three-dimensional CBCACONNH and HNCACB NMR experiments, and well resolved amide resonances in the 1 H, 15 N HSQC spectrum were followed throughout the EGCg titration. EGCg perturbed amide resonances of both the C-and N-terminal domains; however, it induced the largest chemical shift changes in the C-terminal domain (see supplemental Fig. 1). We posit that the amide resonances that are shifted greatly in the C-terminal domain represent direct binding of EGCg, and the smaller shifts in the N-terminal domain of cTnC are from "communication" between the two domains. This indicates that the principal binding site for EGCg is in the C-terminal domain of cTnC. It has also been shown by fluorescence spectroscopy (28) and NMR spectroscopy (30) that EGCg targets cCTnC. A global fitting approach using the program xcrvfit was used as described previously (43) to determine the dissociation constant (K D ) that best fit Reaction 1, cTnC⅐3Ca 2ϩ ϩ EGCg 7 cTnC⅐3Ca 2ϩ ⅐EGCg REACTION 1 The binding curves that were globally fit are shown in supplemental Fig. 1B. The binding of EGCg to cTnC⅐3Ca 2ϩ was fit to a 1:1 stoichiometry, with a best fit K D of 1.12 mM. The K D value was also calculated by averaging the normalized individual chemical shifts as a function of the ligand to protein ratios and fitting using xcrvfit. This approach yielded a K D of 1.1 Ϯ 0.12 mM. The observation that EGCg targets the C-terminal domain of cTnC focused our subsequent structural analysis to cCTnC.
Effect of EGCg on cCTnC⅐2Ca 2ϩ -The assigned 1 H, 15 N HSQC and 1 H, 13 C-HSQC NMR spectrum of cCTnC⅐2Ca 2ϩ was used to observe the binding of EGCg to cCTnC⅐ 2Ca 2ϩ (Fig. 1, a and b). The 1 H, 15 N HSQC spectrum of cCTnC⅐2Ca 2ϩ has been previously assigned (21,44), and the 1 H, 15 N HSQC and 1 H, 13 C HSQC spectra of the cCTnC⅐2Ca 2ϩ ⅐EGCg complex were assigned using three-dimensional CBCACONNH and HNCACB NMR spectra. Because the resonance perturbations of cCTnC are in fast exchange, the chemical shifts could be easily followed throughout the titration. The chemical shift of each assigned amide peak was recorded for every titration point, and a total chemical shift was normalized for each resonance. A global fitting approach using the program xcrvfit was used to determine the K D value that best fit Reaction 2, cCTnC⅐2Ca 2ϩ ϩ EGCg 7 cCTnC⅐2Ca 2ϩ ⅐EGCg REACTION 2 The binding curves that were globally fit are shown in Fig.  1c. The binding of EGCg to cCTnC⅐2Ca 2ϩ was fit to a 1:1 stoichiometry, with a best fit K D of 1.10 mM. The K D value was also calculated by averaging the normalized individual chemical shifts as a function of the ligand to protein ratios and fitting using xcrvfit. This approach gave a K D of 1.09 Ϯ 0.08 mM. Given EGCg binding to the C-terminal domain of cTnC⅐3Ca 2ϩ and to cCTnC⅐2Ca 2ϩ induced effectively identical chemical shift changes and dissociation constants, we concluded that EGCg binds both the C-terminal domain of cTnC⅐3Ca 2ϩ and cCTnC⅐2Ca 2ϩ in a similar fashion.
The dissociation constant was significantly higher than previous groups reported for EGCg (28). To investigate whether FIGURE 1. Titration of cCTnC⅐2Ca 2؉ with EGCg. Two-dimensional 1 H, 15 N HSQC (a) and 1 H, 13 C-HSQC (b) spectra arising from backbone and side chain amide groups (a) and side chain methyl groups (b) are overlaid for a series of EGCg additions. Each titration point represents the titration points described under "Experimental Procedures." The titration was made into 13 C, 15 N-labeled cCTnC⅐2Ca 2ϩ , and both the 1 H, 15 N HSQC and 1 H, 13 C-HSQC spectra were acquired at each titration point. Assignments of some of the cross-peaks are labeled. The multiple contours (F) represent the initial point in the titration, with no EGCg added, and each open contour (E) represents a specific point in the titration for a given residue. b, red contours represent cross-peaks with negative intensity, a feature of the constant time 1 H, 13 C-HSQC experiment. The direction the peaks shift is indicated with arrows, for example see Gly 125 . c, curves represent the amide resonances belonging to residues affected by ligand binding, as shown in a. The curves were fit as a function of normalized total chemical shift perturbation versus [EGCg] total /[cCTnC⅐2Ca 2ϩ ] total . The total chemical shift changes were calculated in hertz as follows: ⌬␦ ϭ ((⌬␦ 1 H) 2 ϩ (⌬␦ 15 N) 2 ) 1/2 . Because hertz is used instead of parts/million, a correction factor of 1/5 for the 15 N dimension is not used. d, chemical shift mapping on the structure of cCTnC⅐2Ca 2ϩ . The ribbon representation of cCTnC⅐2Ca 2ϩ is shown in yellow, and residues that were perturbed greater than the mean chemical shift change for all backbone amide resonances of cCTnC are colored in red.
there were concentration-dependent effects of EGCg, such as aromatic stacking, we lowered the protein concentration to 80 M and repeated the titration. The data gave a best fit K D of 385 M, which indicates that EGCg is undergoing a competing equilibrium in aqueous solution. Therefore, the observed or apparent K D value is an upper limit of the actual dissociation constant. The chemical shift changes were identical in both the low and high concentration titrations, suggesting that the location of the binding site of EGCg is independent on cCTnC concentration.
A number of cCTnC⅐2Ca 2ϩ 1 H, 15 N and 1 H, 13 C cross-peaks underwent significant chemical shift perturbations during the EGCg titration. Chemical shift mapping of the amide peaks was used to identify specific regions of the protein that underwent large perturbations induced from EGCg binding to localize the binding site of EGCg (Fig. 1d). The amide resonances of residues on the loop connecting helices F and G (Thr 124 -Thr 129 ) underwent the largest changes in chemical shifts, suggesting a close proximity of the ligand to their backbone and side chain nuclei or a large change of conformation or dynamics in the loop. This may reflect a change in the positions of helices F and G relative to one another upon formation of the EGCg⅐cCTnC complex. Other residues of cCTnC that had significant backbone chemical shift perturbations were residues Met 120 , Leu 121 , Gln 122 , and Ala 123 of helix F, residues Leu 136 , Gly 140 , and Asp 141 of helix G, and residues Met 157 , Lys 158 , Gly 159 , Val 160 , and Glu 161 of helix H. Amide and methyl resonances of residues on the ␤-sheet (Tyr 111 , Ile 112 , Asp 113 , Arg 147 , Ile 148 , and Asp 149 ) did not undergo any significant chemical shift changes. This suggests that the binding of EGCg is near the opening of the hydrophobic cleft of cCTnC, rather than deep within the pocket. To obtain a more detailed knowledge of the interaction of cCTnC with EGCg, the solution structure of cCTnC⅐2Ca 2ϩ ⅐EGCg was determined by NMR spectroscopy.
Structure of cCTnC⅐2Ca 2ϩ ⅐EGCg-The solution structure of cCTnC⅐2Ca 2ϩ bound to EGCg was determined using the NMR experiments listed in supplemental Table 1 (see references for  experiments therein). Dihedral angle restraints were calculated from chemical shifts with the program TALOS (42). The chemical shifts corresponding to backbone atoms of cCTnC in the cCTnC⅐2Ca 2ϩ ⅐EGCg complex were assigned using the two-dimensional 1 H, 15 N HSQC and the three-dimensional CBCA-CONNH and HNCACB NMR spectra. The two-dimensional 1 H, 13 C-HSQC and the three-dimensional HCCH-TOCSY, CCONH, and HCCONH NMR spectra were acquired to assign side chain resonances. The 15 N-edited HNHA and HNHB experiments were acquired to assign H␣ and H␤ resonances. Distance restraints for cCTnC were determined using 13 C-NOESY HSQC and 15 N-NOESY HSQC NMR spectra. Resonances for aromatic residues of cCTnC were assigned using the two-dimensional NOESY NMR spectrum in D 2 O.
The one-dimensional 1 H spectrum of free EGCg was first assigned in DMSO-d 6 ; the 1 H spectrum of free EGCg in D 2 O has been previously assigned (45), and the 1 H chemical shifts are virtually identical in either solvent. Following assignment of free EGCg, the 1 H chemical shifts were assigned for cCTnC⅐2Ca 2ϩ -bound EGCg using the two-dimensional NOESY NMR spectrum in D 2 O (Fig. 2). 13 C/ 15 N-edited/fil-tered experiments (two-dimensional and three-dimensional) were run to assign intermolecular NOEs between EGCg and cCTnC to identify the binding site and orientation of EGCg when bound to cCTnC⅐2Ca 2ϩ . 12 Ca 2ϩ distance restraints from crystallographic data were incorporated into the structure determination as described previously (21). There were a total of 1053 structural restraints used in the structure determination, including 899 intramolecular cCTnC distance restraints, 18 intermolecular distance restraints, 10 intramolecular EGCg distance restraints, 12 Ca 2ϩ distance restraints, and 114 dihedral restraints. Table 1 contains a list of the structural statistics for cCTnC⅐2Ca 2ϩ ⅐EGCg.
Plots of the intermolecular NOEs between EGCg and cCTnC⅐2Ca 2ϩ are shown in Fig. 3a. To assign the intermolecular NOEs, both the three-dimensional and two-dimensional versions of the 13 C/ 15 N-edited/filtered experiment were acquired. The two-dimensional experiment was acquired because it provides a better signal-to-noise ratio than the three-dimensional experiment, and the 13 C-edited three-dimensional experiment was run to confirm the two-dimensional assignments. The NOE restraints observed between EGCg and cCTnC⅐2Ca 2ϩ are between a number of hydrophobic residues that line the hydrophobic pocket. Residues that have NOEs to EGCg include Met 157 and Val 160 on the terminal end of helix H, Leu 121 and Met 120 on helix F, and Leu 136 on helix G, all of which point toward the hydrophobic pocket of cCTnC. The tetrahydropyran ring of EGCg contains four hydrogen atoms (H01, H02, H15, and H33), all of which make significant contacts to cCTnC. The two hydrogen atoms on the benzenediol ring, on the other hand, do not make any NOE contacts to cCTnC, hence its loosely defined orientation. Fig. 3b depicts the specific observed NOEs between cCTnC residues and EGCg hydrogen atoms. There were no intermolecular NOEs observed between EGCg and the hydrophobic residues of the ␤-sheet (Ile 148 or Ile 112 ), indicating that EGCg is not buried deep in the hydrophobic pocket as is EMD 57033. Also, no NOEs were observed between EGCg and any residues along the loop between helices F and G, even though these residues had the greatest chemical shift perturbation (Fig. 1). The large perturbations of the loop residues could result from a proximity to the aromatic polyphenol rings of EGCg or from a change in the conformation or dynamics of the loop upon binding EGCg.
The ensemble of the 30 lowest energy structures of EGCg in complex with cCTnC⅐2Ca 2ϩ is depicted in Fig. 4, a and b. The ensemble of EGCg is shown in Fig. 4c, and the lowest energy structure of cCTnC⅐2Ca 2ϩ ⅐EGCg is shown in schematic representation in Fig. 4d. The overall fold of cCTnC is similar to that which has been previously described for cCTnC. There are four well defined helices as follows: helices E, F, G, and H. The two Ca 2ϩ -binding sites (sites III and IV) are between helices E and F and G and H. There is a short twisted anti-parallel ␤-sheet that joins the two EF-hands. The r.m.s.d. of cCTnC for the backbone atoms of the well defined residues is 0.80 Ϯ 0.14 Å. The well defined regions (r.m.s.d. Ͻ 1 Å) involve residues 94 -122, 132-143, and 145-157. These regions include residues from helices E, F, G, and H as well as those of the Ca 2ϩ -binding loops and anti-parallel ␤-sheet. The N and C termini as well as the interhelical F-G loop of cCTnC⅐2Ca 2ϩ had fewer structural restraints, and hence had a larger r.m.s.d. in the ensemble. Given the rotational freedom of EGCg around the galloyl (ring BЈ) and pyrogallol (ring B) moieties, there was significant mobility in the ensemble of EGCg (Fig. 4c). In addition to the rotational freedom of the trihydroxyphenyl rings, the pairs of hydrogen atoms on the rings (H39/H41 and H12/H4) are chemically equivalent, and so it was not possible to differentiate NOE contacts within the pairs. Nonetheless, the 10 intramolecular NOEs of the bound EGCg and the 18 intermolecular NOEs between EGCg and cCTnC⅐2Ca 2ϩ positioned the three functional moieties of EGCg with reasonable precision. The galloyl moiety is positioned near the loop connecting helix F and helix G. The large chemical shift perturbations of the F-G loop residues (Fig. 1) could in part result from a propinquity to the gal-FIGURE 2. Assignment of EGCg. a, chemical structure of EGCg. The benzenediol is labeled as ring A, the pyrogallol ring as B, the galloyl moiety as BЈ, and ring C is the tetrahydropyran moiety. The hydrogen atoms attached to carbon atoms are also labeled. b, assigned one-dimensional 1 H NMR spectrum of EGCg in DMSO-d 6 . c, a few strip plots from the two-dimensional NOESY with resonances assigned that belong to EGCg in complex with cCTnC⅐2Ca 2ϩ . The data were acquired in D 2 O as to remove amide signals that predominate this region of the two-dimensional NOESY spectrum in H 2 O. Details of the experiment are outlined in Table 1.
loyl trihydroxyphenyl ring. The pyrogallol ring rests near the C terminus of cCTnC, which explains the large chemical shift perturbations of helix H residues Met 157 -Glu 161 during titration with EGCg (Fig. 1). The fused tetrahydropyran (ring C) and benzenediol (ring A) rings lie near the surface of helix E that faces the hydrophobic core of cCTnC⅐2Ca 2ϩ .
Comparison of Solution Structures cCTnC⅐2Ca 2ϩ and cCTnC⅐2Ca 2ϩ ⅐EGCg-The solution structure of cCTnC⅐ 2Ca 2ϩ ⅐EGCg adopts a similar fold to other structures determined for cCTnC (6,13,21). The differences in the structures determined for cCTnC are all primarily the result of varying degrees of "openness," in which the helices are spread out from the hydrophobic core of the protein. The degree in which cCTnC is open is described by the inter-helical angles between helices E and F and between helices G and H. The closer the inter-helical angles are to 90°, the more open the structure is. Inter-helical angles were calculated using the program interhlx (K. Yap, University of Toronto).
When the structure of cCTnC⅐2Ca 2ϩ ⅐EGCg was overlaid with the NMR structure of cCTnC⅐2Ca 2ϩ (6), a minor perturbation in the structure was observed. Between the two structures, an r.m.s.d. of 1.53 Å for backbone atoms of helices E-H was observed (Fig. 5, a and b). Notable differences between the structures include helix E, helix G, and helix H that are positioned slightly away from the core of the protein when compared with cCTnC⅐2Ca 2ϩ . The E-F inter-helical angle of cCTnC⅐2Ca 2ϩ is 112°, and the G-H angle is 117°. The E-F interhelical angle of cCTnC⅐2Ca 2ϩ ⅐EGCg is 105°, and the G-H interhelical angle is 113°, revealing that the overall outcome of EGCg binding is a more open conformation of cCTnC.
The solution structure of cCTnC⅐2Ca 2ϩ bound to EMD 57033 has also been solved (21). When cCTnC⅐2Ca 2ϩ ⅐EGCg was overlaid with cCTnC⅐2Ca 2ϩ ⅐EMD 57033, an r.m.s.d. for the backbone atoms for the helices is 1.84 Å (Fig. 5, c and d). The inter-helical angles of the E-F and the G-H helices of cCTnC⅐2Ca 2ϩ ⅐EMD 57033 are 96 and 118°, respectively. The largest difference between the structures is in the positions of the G and F helices. In the EMD 57033-bound structure, the G helix is shifted nearer to the core of cCTnC (as in the unbound form of cCTnC⅐2Ca 2ϩ ). In contrast, helix F is further from the core of cCTnC in the EMD 57033-bound structure when compared with the EGCg-bound structure of cCTnC⅐2Ca 2ϩ . The location of the two ligands is similar, with both drugs binding the core of the protein; however, EMD 57033 is buried deep within cCTnC, and EGCg remains near the surface of the opening of cCTnC. The methyl on the thiadiazinone ring of EMD 57033 makes several NOE contacts with Ile 148 and Ile 112 of the ␤-sheet. In the case of EGCg, the ring protons all make contacts exclusively to hydrophobic residues that line the surface of cCTnC. The region of the crystal structure of the cardiac troponin complex (13) corresponding to cCTnC⅐ 2Ca 2ϩ ⅐cTnI-(34 -71) was also overlaid with cCTnC⅐2Ca 2ϩ ⅐EGCg and gives an r.m.s.d. of 1.45 Å for the backbone atoms of the helix residues (Fig. 5, e and f). In accordance with the good agreement in r.m.s.d., the inter-helical angles of cCTnC in the cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) complex also resemble the cCTnC in the cCTnC⅐2Ca 2ϩ ⅐EGCg complex; the E-F inter-helical angle is 100°, and the G-H inter-helical angle is 114°. Therefore, EGCg induces a conformational change in cCTnC most closely akin to that caused by cTnI- (34 -71). The helices of cCTnC bound to cTnI-(34 - . Intermolecular NOEs between EGCg and cCTnC⅐2Ca 2؉ . a, series of strip plots assigned from the two-dimensional 13 C-edited/filtered NOESY NMR experiment. The 1 H resonances that correspond to EGCg are labeled on the right side of the spectra, and the 1 H that correspond to 15 N, 13 C-labeled cCTnC⅐2Ca 2ϩ are labeled at the top of the strips plots. The circled peak is an artifact from the intermolecular NOE experiment. b, schematic depiction of several of the NOE contacts assigned in a. EGCg is shown in stick representation with carbon atoms colored in purple, oxygen atoms color in red, and hydrogen atoms colored in white. cCTnC is depicted in schematic representation with the residues involved in making NOEs to EGCg shown in stick representation. Carbon atoms for cCTnC are colored in gray, sulfur atoms in yellow, and hydrogen atoms in white. The dotted lines indicate contacts measured by the 13 C-edited/filtered NOESY NMR experiment. 71) all align well with the cCTnC⅐EGCg complex, suggesting an analogous opening of cCTnC. Helix G is further from the core of the protein in the EGCg-bound structure when compared with the cTnI-(34 -71)-bound form of cCTnC⅐2Ca 2ϩ . EGCg and the backbone of cTnI-(34 -71) occupy the same surface of cCTnC. To test how this steric clash affects EGCg binding, we titrated EGCg into the cTnI-(34 -71) saturated cCTnC⅐2Ca 2ϩ complex.
Effect of EGCg on cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71)-The interaction of EGCg with cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) was monitored using the 1 H, 15 N HSQC and 1 H, 13 C-HSQC NMR spectra acquired at a series of EGCg-cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) ratios (Fig. 6, a and b). The total chemical shift was normalized for a number of resonances, and a global fitting approach using xcrvfit was used to determine the dissociation constant that best fit Reaction 3, cCTnC⅐2Ca 2ϩ ⅐cTnI-͑34-71͒ ϩ EGCg 7 cCTnC⅐2Ca 2ϩ ⅐cTnI-͑34-71͒⅐EGCg REACTION 3 The binding curves that were globally fit are shown in Fig. 6c. The binding of EGCg to cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) was fit to a 1:1 stoichiometry, with a best fit K D of 1.8 mM, about 1.5-fold weaker than the K D for EGCg binding to cCTnC⅐2Ca 2ϩ calculated at the higher concentration. The K D value was also calculated by averaging the normalized individual chemical shifts as a function of the ligand to protein ratios and fitting using xcrvfit. This approach gave a K D of 1.64 Ϯ 0.24 mM.
The perturbation of 1 H, 15 N cross-peaks was less than that for the binding of EGCg to cCTnC⅐2Ca 2ϩ alone, and the perturbed residues correspond to several isolated regions of cCTnC. This made chemical shift mapping difficult to interpret, forestalling the localization of the binding surface of EGCg. 1 H, 13 C-HSQC NMR spectra of cCTnC⅐2Ca 2ϩ in the cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) complex were also acquired during the EGCg titration to hone in on the binding surface on cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) complex. The methyl region of the 1 H, 13 C-HSQC NMR spectra is shown in Fig. 6b. Two or three regions of the protein experienced large chemical shift changes in the 1 H, 13 C-HSQC NMR spectra as follows: near the E-H helix interface, along the F helix, and E-F loop.
The amide resonances that were most perturbed upon EGCg titration are mapped onto the structure of cCTnC⅐2Ca 2ϩ ⅐cTnI-  (34 -71) (Fig. 6d) from the core troponin structure (13). It seems that EGCg binds in the proximity of helix F. Further evidence of EGCg binding near helix F is given when the methyl region of the 1 H, 13 C-HSQC spectrum was monitored during the titration of cCTnC⅐Ca 2ϩ ⅐cTnI-(34 -71) with EGCg. The terminal methyls of the F helix residues, Leu 114 and Ile 119 , underwent large chemical shift perturbations (Fig. 6b). It may be that EGCg binds to the interface between cTnI and cCTnC near helix F or simply to the side of the protein near the F-G loop. There were also perturbations of amide resonances near the N terminus of helix H toward the ␤-sheet and of residues on the ␤-sheet. These perturbations may be caused by direct contact with EGCg or from a conformational change in the structure of cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) necessary to lodge EGCg.

DISCUSSION
Common treatment schemes of heart failure modify levels of cytosolic Ca 2ϩ . This provides immediate improvement in heart function, but it can lead to serious side effects if used for an extended period of time. Drugs that alter the Ca 2ϩ sensitivity of the thin filament, rather than the cytosolic Ca 2ϩ concentration, provide a safer alternative. There are compounds that increase or decrease the sensitivity of the thin filament through interacting specifically with troponin. An increase in Ca 2ϩ sensitivity would be beneficial for the treatment of heart failure, whereas the use of Ca 2ϩ desensitizers may provide protection against the development of hypertrophic cardiomyopathy (HCM). HCM is identified by an enlargement of the heart muscle and a decrease in chamber volume of the ventricles. Patients with HCM often suffer from shortness of breath and angina but may also eventually experience heart failure, arrhythmia, and sudden death. The treatment of HCM has been traditionally pursued with the use of negative inotropes that block neurohormones, target pathological load on the heart, or block calcium channels (for reviews on HCM and therapies, see Refs. 46 -49). The use of Ca 2ϩ desensitizers would provide another treatment option, because Ca 2ϩ desensitizers do not disrupt the cytosolic Ca 2ϩ homeostasis or hormone levels. A compound that has demonstrated the ability to inhibit cardiac muscle activation is W7. In skinned rabbit psoas fibers, W7 was shown to inhibit the striated muscle activation (50). Silver et al. (51) also indicated that W7 inhibited ATPase activity and proposed this deactivation occurred through interaction with cTnC. NMR has been utilized to show that W7 binds both the C-and N-terminal domains of cTnC in the absence of cTnI (43); however, in the presence of cTnI-(34 -71) and cTnI-(128 -163), W7 associates exclusively in the N-terminal domain of cTnC (52). Similarly, EGCg has been identified by the preliminary study of Tadano et al. (30) and by the recent work of Liou et al. (28) to reduce the Ca 2ϩ sensitivity of myofibrillar ATPase activity in cardiac myofibrils. Contrary to W7, however, we illustrate by NMR that EGCg binds to the C-domain of cTnC preferentially. This observation has also been shown by fluorometry and NMR spectroscopy (28,30).
Structural biology has supplemented the understanding of cardiac muscle contraction and has revealed interesting therapeutic opportunities with cTnC as the primary target. In this study we used NMR spectroscopy to define the molecular details of the interaction between EGCg and cCTnC. Analogous to other ligands, we found that EGCg binds in the hydrophobic cavity of cCTnC⅐2Ca 2ϩ . In addition to the EGCg-cCTnC interaction, we found that EGCg also binds to the cCTnC⅐Ca 2ϩ ⅐cTnI-(34 -71) complex. Both of these observations have also been described by others, and it is thought that it is these interactions with the C-domain of cTnC that is responsible for the activity of EGCg (28,30). Two-dimensional HSQC NMR spectra were acquired, and chemical shift changes of 15 N, 13 C-labeled cCTnC were followed during the EGCg titration into solutions containing cCTnC⅐2Ca 2ϩ or cCTnC⅐2Ca 2ϩ ⅐ cTnI-(34 -71). Because two-dimensional HSQC NMR spectroscopy provides information regarding the chemical environment surrounding individual nuclei in the protein, we were able to identify specific residues that were affected by EGCg. This branded residues that are proximal to EGCg in the protein⅐ligand complex or that experience conformational changes upon ligand binding. The 1 H, 13 C-HSQC NMR experiment was utilized to elucidate nearby residues via side chain resonance perturbations, and thus gain insight into the binding location of EGCg to cCTnC.
The solution structure of cCTnC⅐2Ca 2ϩ ⅐EGCg was determined to unravel the mode of action of EGCg. It was found that EGCg binds to the hydrophobic pocket of cCTnC⅐2Ca 2ϩ as do other ligands of cCTnC, such as the anchoring region of cTnI (cTnI-(34 -71)) and the cardiotonic drug (EMD 57033) (13,21). Unlike cTnI-(34 -71) and EMD 57033, EGCg binds closer to the surface of the hydrophobic pocket rather than deep within the core of cCTnC. EGCg was shown to open the core of cCTnC⅐2Ca 2ϩ in a similar manner as cTnI-(34 -71) does. It was also seen that EGCg occupies the same binding site as cTnI-(34 -71) (Fig. 5, e and f), which might suggest the mode of action of EGCg. To address the possibility that EGCg may compete with cTnI-(34 -71) for binding to cCTnC, the interaction of EGCg with cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) was measured.
We found that EGCg induced chemical shift perturbations of cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71); however, the overall magnitude of the perturbations, when compared with cCTnC⅐2Ca 2ϩ , appear smaller, and the affinity of EGCg for the complex is decreased. Possible reasons for the lessened affinity of EGCg for cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) are EGCg and cTnI-(34 -71) compete for the same binding site on cCTnC⅐2Ca 2ϩ or there is a new binding site for EGCg in the cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) complex. The chemical shift changes induced by EGCg on the cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) complex do not indicate a dissociation of cTnI-(34 -71) from cCTnC⅐2Ca 2ϩ , but rather suggest the formation of a ternary complex, cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71)⅐EGCg. The smaller chemical shift changes suggest less of a structural perturbation of the troponin C-I complex than of cCTnC.
The affinity of EGCg for cTnC has been measured by intrinsic tyrosine fluorescence quenching to be 3-4 M for EGCg to cTnC (28). At the higher concentrations typically used for NMR spectroscopy, we found that EGCg bound to cTnC with an K D of 1.1 Ϯ 0.12 mM, to cCTnC with a K D of 1.09 Ϯ 0.08 mM, and to cCTnC⅐cTnI-(34 -71) with a K D of 1.64 Ϯ 0.24 mM. It has been shown that aromatic stacking of EGCg occurs in aqueous solution (53,54), and this additional equilibrium would confound accurate K D determination. Prompted by these reports, we repeated the titration of EGCg into cCTnC⅐2Ca 2ϩ at a lower concentration. Our results show that as we decrease the concentration of cCTnC, the apparent K D was decreased as well (from 1.09 mM to 385 M). This enhanced affinity supports the notion of EGCg stacking in vitro, and it explains the weaker affinity we measured when compared with fluorescence spectroscopy.
The perturbation of the cCTnC-cTnI-(34 -71) interaction by EGCg could weaken the anchoring of cCTnC to the thin filament and thus decrease the sensitivity of the thin filament for Ca 2ϩ . There are compounds that have been identified to bind the N-terminal domain of cTnC and function by modulating the interaction of cNTnC and cTnI. Levosimendan is expected to work by increasing the affinity of cTnI-(147-163) for cNTnC, thus increasing the Ca 2ϩ sensitivity of the thin filament through an indirect mechanism (55). There has been renewed interest in the role of the so-called structural domain of cTnC (cCTnC) in regulation of contraction. EMD 57033 is a drug that has been shown to interact exclusively with cCTnC. EMD 57033 may act as a Ca 2ϩ sensitizer by modulating the interaction of cCTnC and cTnI- (34 -71). Small compounds that bind to and inhibit or strengthen the interaction of cCTnC for cTnI-(34 -71) may have a pronounced effect on contraction rate or force and Ca 2ϩ sensitivity.
With the aid of this solution structure it might be possible to design new agents using EGCg as a lead compound. The strategic methylation of some of the hydroxyl groups on the polyphenol rings, for example, could potentially increase the potency of EGCg. In fact, it has been shown that catechin and epicatechin are O-methylated in rat small intestine (56), and the in vivo mechanism of EGCg on thin filament activity may include these substituted metabolites. The concept of chemical modification of natural products to improve their effectiveness is common, and there are many examples of pharmaceuticals currently prescribed that are derived from natural products (for a review see Ref. 57). The potential role of EGCg as a Ca 2ϩ desensitizer is particularly interesting in regards to treatment for HCM, because there is evidence that reactive oxygen species may be one of the causes of cardiac hypertrophy (49). Because EGCg is a known scavenger of radicals, it may help treat and/or prevent HCM by sequestering reactive oxygen species as well as by inhibiting ATPase activity.
The data presented in this work provide evidence to support the notion that cTnC is one of the primary targets for EGCg in the myofilament, the effective binding site is in the hydrophobic pocket of cCTnC, and the binding induces an opening of the domain. We describe the interaction of EGCg with the cCTnC⅐2Ca 2ϩ ⅐cTnI-(34 -71) complex, indicating a possible mechanism in which EGCg modulates contraction. EGCg may compete with cTnI-(34 -71) and weaken the anchoring of cTnC to the thin filament. This has been postulated for the DCM mutation G159D. The G159D mutation is in the cCTnC and has been shown to weaken the affinity of cTnI-(34 -71) for cCTnC (25). EGCg may work in a similar manner, protecting the heart from the development of hypertrophy.