Drug Binding to Cardiac Troponin C*

Compounds that sensitize cardiac muscle to Ca2+ by intervening at the level of regulatory thin filament proteins would have potential therapeutic benefit in the treatment of myocardial infarctions. Two putative Ca2+sensitizers, EMD 57033 and levosimendan, are reported to bind to cardiac troponin C (cTnC). In this study, we use heteronuclear NMR techniques to study drug binding to [methyl- 13C]methionine-labeled cTnC when free or when complexed with cardiac troponin I (cTnI). In the absence of Ca2+, neither drug interacted with cTnC. In the presence of Ca2+, one molecule of EMD 57033 bound specifically to the C-terminal domain of free cTnC. NMR and equilibrium dialysis failed to demonstrate binding of levosimendan to free cTnC, and the presence of levosimendan had no apparent effect on the Ca2+ binding affinity of cTnC. Changes in the N-terminal methionine methyl chemical shifts in cTnC upon association with cTnI suggest that cTnI associates with the A-B helical interface and the N terminus of the central helix in cTnC. NMR experiments failed to show evidence of binding of levosimendan to the cTnC·cTnI complex. However, levosimendan covalently bound to a small percentage of free cTnC after prolonged incubation with the protein. These findings suggest that levosimendan exerts its positive inotropic effect by mechanisms that do not involve binding to cTnC.

Congestive heart failure results in desensitization of the myocardium to Ca 2ϩ as well as depressed cardiac contractility. A promising approach to the treatment of congestive heart failure is the development of positive inotropic Ca 2ϩ -sensitizing agents that are capable of increasing the sensitivity of myocardium to Ca 2ϩ , thereby compensating for low cardiac output. Ideal calcium-sensitizing compounds would enhance the maximal force of muscle contraction without increasing Ca 2ϩ ion flux into the cell or impairing relaxation in the failing myocardium. In addition, the compound should not inhibit phosphodiesterase activity (1) to minimize the possibility of heart arrhythmias and increased energy consumption in the heart (2).
Based on these criteria, cardiac troponin C (cTnC) 1 is an attractive target for potential Ca 2ϩ -sensitizing compounds.
Cardiac troponin C is the EF-hand Ca 2ϩ receptor in the thin filament of slow skeletal and cardiac striated muscle. The contraction relaxation cycle is regulated by the binding and release of Ca 2ϩ ions from the N-terminal regulatory site II of cTnC. Compounds that bind with high affinity and selectivity to cTnC and increase the affinity of Ca 2ϩ binding to site II could potentially increase the Ca 2ϩ sensitivity of myocardial contraction without altering Ca 2ϩ transients.
Knowledge of the structure and surface topology of cTnC is critical for an understanding of how existing Ca 2ϩ -sensitizing compounds interact with cTnC and to identify potential binding sites for new and more effective compounds. Previous structural models for cTnC based on known structures for Ca 2ϩbound fast skeletal troponin C (3,4) and calmodulin (5) predicted that the N-terminal regulatory domain would have an open conformation with a large contiguous hydrophobic surface. Ca 2ϩ -sensitizing compounds such as bepridil, levosimendan, pimobendan, and trifluoperazine were thought to bind to this predicted hydrophobic surface (6 -8). However, the recent NMR solution structures of intact Ca 2ϩ -bound cTnC and the apo and Ca 2ϩ -bound N-terminal domain of cTnC show the N-terminal regulatory domain to be partially closed in the presence of Ca 2ϩ (9,10). Hydrophobic surfaces in the N-terminal domain exist as discrete patches rather than a contiguous surface. Our recent study localized the binding site for bepridil and TFP in these N-terminal hydrophobic sites as well as in the C-terminal structural domain of cTnC (11).
Here we report studies on the binding of two newer generation Ca 2ϩ sensitizers, EMD 57033 and levosimendan, to cTnC. EMD 57033 is reported to interact with the C-terminal domain of cTnC (12), but its Ca 2ϩ -sensitizing effects are thought to result primarily by affecting the actin myosin interface (13). EMD 57033 has an isoform-specific effect on phosphodiesterase activity (14). The Ca 2ϩ -sensitizing effects of levosimendan are thought to result from direct binding to cTnC (12,15,16). Our data confirm that EMD 57033 binds to the C-terminal domain of cTnC but that levosimendan does not appear to bind to either free cTnC or the cTnC⅐cTnI binary complex. However, levosimendan can form covalent adducts with cTnC after prolonged exposure.
Amplified DNA was purified and digested with NcoI and HindIII and subcloned into pET-23d (Novagen). The cTnI plasmid was transformed into E. coli strain Novablue (Novagen). Positive clones were confirmed by dideoxy DNA sequence analysis.
Generation of the truncated cTnI 33-211 (A-Cys) plasmid and subsequent purification of the protein followed the method used by Guo et al. (17), with minor modifications. The coding sequence corresponding to amino acids 33-211 of cTnI(A-Cys) was amplified by PCR using the 5Ј primer reported by Guo et al. (17) and a 3Ј primer, the T7 terminator primer (Novagen). The subsequent PCR product was cloned into pET-23d using the NcoI and HindIII sites. After confirmation of the sequence, the construct was transformed into BL21(DE3) for overexpression. Purification of cTnI  (A-Cys) protein was done essentially as reported previously for cTnI   (17).
NMR Sample Preparation-The bacterial expression of cTnC(A-Cys) and cTnC3 and their labeling with [methyl-13 C]Met and purification were reported previously (18 -20). The NMR samples varied in protein concentration from 0.1 to 1 mM in NMR buffer consisting of either 20 mM Tris-d ll , 200 mM KCl, and D 2 O (pH 7.0) or 10 mM HEPES, 100 mM KCl, and D 2 O (pH 7.0). Buffers used to prepare cTnC3 included five equivalents of dithiothreitol to reduce the Cys residues. In some experiments, the ionic strength or pH of the NMR samples was varied. Reconstitution of the binary complex with cTnC and cTnI was carried out by the method of Potter (21).
NMR Methods-All spectra, except for those in Fig. 3, were collected at 40°C on a Bruker AMX500 NMR spectrometer. HSQC (22) spectra were collected with 1024 complex data points in the t 2 domain and 72 increments in t 1 . The experiments took approximately 26 min to complete per titration point. The 1 H and 13 C spectra widths were 6012 and 600 Hz, respectively. 1 H and 13 C chemical shifts were reported relative to the H 2 O signal at 4.563 ppm and the [methyl-13 C]Met signal at 14.86 ppm, respectively. All spectra were processed using the FELIX software package (Biosym Technologies, Inc.).
Drug Solutions-Levosimendan is a light-sensitive drug; therefore, all stock solutions of the drug were prepared and stored in the dark. Stock solutions of levosimendan were prepared immediately before use in deuterated Me 2 SO or as described previously (8). After adding the drug to the protein, the sample pH was adjusted when necessary. The stock solution of EMD 57033 was prepared in deuterated DMSO immediately before use. Bepridil was prepared as reported previously (11).
Mass Spectroscopy-Matrix-assisted UV laser desorption mass spectroscopy spectra were acquired using a PerSeptive Voyager Elite Time of Flight mass spectrometer equipped with a delayed extraction and a nitrogen laser (337 mm). A saturated solution of (2,5)-dihydroxybenzoic acid in 0.1% trifluoroacetic acid in H 2 O was used as a matrix. 0.5 l of matrix was mixed with 0.5 l of sample, placed on the target, and dried at room temperature while protected from light. The sample was a 1:10 dilution in H 2 0 of the appropriate drug stock. The spectra in this study represent the summation of ϳ40 laser shots. In addition to the matrixassisted UV laser desorption mass spectroscopy analysis, a Kratos MS50 fast atom bombardment was used to analyze the levosimendan provided by Orion-Farmos. 3-Nitrobenzyl alcohol was used as a matrix.
Drug Binding and Ca 2ϩ Titration-Dialysis was carried out at 25°C using two-chamber equilibrium dialysis units from Sialmed, Inc. Each chamber had a 0.1 ml capacity. Prepared dialysis membranes (Life Technologies, Inc.) had a molecular mass cutoff of 12,000 -14,000 daltons. Protein was added to one chamber at a concentration of about 0.2 mM to approximate the concentration used in NMR experiments. Protein was prepared in buffers of 20 mM Tris, pH 7.0, and 200 mM KCl or 20 mM HEPES, pH 7.0, and 100 mM KCl. Drug was added to the opposing chamber at concentrations between 0.2 and 0.4 mM in the corresponding buffer. Sufficient CaCl 2 was added to saturate all three Ca 2ϩ binding sites on the protein. After equilibration (typically 16 h), aliquots of samples from both chambers were diluted at least 50-fold and assayed spectrophotometrically at 406 nm for levosimendan (⑀ ϭ 93,285 cm Ϫ1 M Ϫ1 ) and 296 nm for bepridil (⑀ ϭ 3,544 cm Ϫ1 M Ϫ1 ). After dilution, and at these wavelengths, the protein contributed insignificantly to total absorbance. The concentration of cTnC(A-Cys) in the samples was determined by the method of Bradford (23) and bicincho-ninic acid assay (24).
A competitive dye binding assay to determine the relative Ca 2ϩ binding properties of cTnC in the presence and absence of levosimendan was accomplished essentially as described previously by Linse et al. (25). Apo cTnC(A-Cys) was prepared by treatment with 10 mM EGTA followed by desalting into 3 mM Tris, pH 7.5, and 100 mM KCl. Samples used for titration contained 30 M apo troponin C(A-Cys), 30 M 5,5Ј-dibromoBAPTA, 3 mM Tris, pH 7.5, and 100 mM KCl. One ml of the protein solution without Ca 2ϩ was titrated with 3 l aliquots of the same solution made 2 mM in CaCl 2 , and absorbance was monitored at 264 nm. Under these conditions, the concentrations of BAPTA and protein did not change during the titration, and there was no dilution effect on absorbance. The experimental data were fit to a Ca 2ϩ binding isotherm as described by Linse et al. (25). The basal Ca 2ϩ in the buffer/proteins sample was determined to be 7 M.
Gel Filtration and HPLC Analysis-NMR samples were analyzed by both gel filtration and HPLC chromatography to detect possible covalent binding of levosimendan to cTnC. Samples were run on a Bio-Gel P-6 column equilibrated in 20 mM Tris, 150 mM KCl, and 10 mM EDTA (pH 7.5) to remove free drug. Eluted protein was concentrated and stored in 6 M urea, 20 mM Tris, 150 mM KCl, and 10 mM EDTA (pH 8.0) overnight and then loaded onto a Bio-Gel P-6 column equilibrated in 3 M urea, 20 mM Tris-Cl, and 150 mM KCl (pH 7.5). These samples were analyzed by HPLC after exchanging the sample buffer by dialysis to remove the urea. The HPLC chromatographic system consisted of a Waters model 501 pump, a Waters model 680 automated gradient controller, and an Alltech Macrosphere TM GPC 60 7 m (250 ϫ 4.6 mm, length ϫ inner diameter) HPLC column. A Waters Lambda TM model 481 liquid chromatography spectrophotometer was used to monitor the elution of levosimendan at 406 nm. The buffer consisted of 20 mM Tris and 150 mM KCl (pH 7.5). In addition, the samples were run on 15% SDS-polyacrylamide gel electrophoresis urea gels.

Effect of EMD 57033 on the Met Methyl Chemical Shifts of
Free cTnC-Using equilibrium dialysis and tyrosine fluorescence, Pan and Johnson (12) have shown that the thiadiazinone derivative EMD 57033 binds selectively to the C-terminal domain of cTnC (12). This contrasts with bepridil and TFP, which, as we (11) have shown, bind to both the N-and Cterminal domains of cTnC. To confirm the selectivity of EMD 57033 and as a further test of our methodology for detecting drug binding to cTnC, we determined the effect of EMD 57033 on the methionyl methyl chemical shifts of cTnC in the absence and presence of Ca 2ϩ . cTnC(A-Cys), which has both endogenous Cys residues converted to Ser and was chosen for these studies because the solution structure of this cTnC derivative is known (9), and it has been used in previous drug binding studies (11). Fig. 1A shows that EMD 57033 did not bind to [methyl-13 C]Met-labeled cTnC(A-Cys) in the absence of Ca 2ϩ , as evidenced by the absence of drug-dependent changes in 1 H-13 C HSQC spectra. In the presence of Ca 2ϩ (Fig. 1B), significant drug-induced changes in the positions of 1 H-13 C correlations were observed for Met 157 and Met 120 located in the C-terminal domain of cTnC. Drug-induced changes in the chemical shifts of Met 157 and Met 120 were maximal at a drug:protein ratio of 1:1. Chemical shifts for N-terminal Met residues were unaffected by EMD 50733, even at a 2:1 drug:protein ratio (data not shown). These data agree with the results Pan and Johnson (12) and demonstrate that EMD 57033 binds selectively to the C-terminal domain of cTnC in the presence of Ca 2ϩ . The data also validate the use of the Met methyl chemical shifts as location-specific markers to monitor the binding of structurally distinct drugs to cTnC.
Effect of Levosimendan on Met Methyl Chemical Shifts of Free cTnC-Initial experiments showed that Met methyl chemical shifts of [methyl-13 C]Met-labeled cTnC(A-Cys) were unaffected by an equimolar concentration of levosimendan in the presence or absence of Ca 2ϩ (data not shown). Thus, the effects of higher concentrations of drug were determined. Fig. 2A shows that a 4-fold molar excess of levosimendan has generalized and minor effects on the methionyl methyl chemical shifts of cTnC(A-Cys) in the absence of Ca 2ϩ . In the presence of Ca 2ϩ (Fig. 2B), only Met 120 and Met 157 exhibit significant chemical shift changes in the proton dimension upon the addition of 2-fold and 4-fold molar excesses of levosimendan. Observed changes for Met 81 at high drug:protein ratios are restricted to the carbon dimension. However, the interpretation of these changes with respect to specific drug binding to cTnC must be made cautiously because they are observed only at drug:protein ratios of greater than 1:1. Binding of bepridil and TFP to cTnC(A-Cys) induced significant spectral changes at a drug: protein ratio of 0.5:1 (11). Even the fluorescent dye 4,4Ј-dianilino-1,1Ј-binaphthyl-5,5Ј-disulfonic acid, which binds to hydrophobic surfaces, induces large changes in the HSQC spectra of [methyl-13 C]Met-labeled cTnC(A-Cys) (data not shown).
We next used isotope-filtered NMR techniques in an attempt to identify NOEs between aromatic protons in levosimendan and Met methyl protons in cTnC(A-Cys) that was metabolically labeled with [methyl-13 C]Met and Phe-(D) 8 . No NOEs were detected (data not shown). In contrast, similar experiments identified strong NOEs between bepridil and cTnC(A-Cys) (11).
cTnC(A-Cys) encodes Ser rather than Cys at positions 35 and 84. It is possible that these mutations affect the ability of levosimendan to bind to cTnC. To test this, we labeled cTnC3, which contains Cys 35 and Cys 84 , with [methyl-13 C]Met and monitored Met methyl chemical shifts upon the addition of levosimendan in the presence of Ca 2ϩ as shown in Fig. 3. The Met methyl chemical shifts for cTnC3 in the presence of Ca 2ϩ are slightly different than those for cTnC(A-Cys) but are identical to those reported previously for cTnC3 (19). The addition of levosimendan to cTnC3 at a probe:protein ratio of 1.6:1 had Direct Binding Studies-The results in Figs. 2 and 3 and the absence of NOEs between cTnC and levosimendan suggest either that the drug does not bind to cTnC or that it binds with very low affinity and preferentially to the C-terminal domain. Both of these interpretations are not consistent with the Ca 2ϩsensitizing effects of levosimendan, which is proposed to act via binding to cTnC (8,15) and to be effective at concentrations as low as 0.03-1 M levosimendan (26). An alternative explanation for the data is that levosimendan binds to cTnC but does not greatly affect the Met methyl chemical shifts. To address this, we used equilibrium dialysis to determine direct binding of levosimendan to Ca 2ϩ -saturated cTnC. The antianginal agent bepridil (27) was used as a positive control because it is known to bind to cTnC (7,11). Protein and drug concentrations were selected to ensure that even low affinity binding would be detected. Fig. 4A shows the results of equilibrium dialysis to detect the binding of bepridil to cTnC(A-Cys). The absorbance spectrum clearly shows significant binding of bepridil to cTnC(A-Cys), as evidenced by an accumulation of drug in the chamber containing protein. The apparent dissociation constant for binding of bepridil to cTnC(A-Cys) was approximately 20 M, which agrees well with values of 10 -20 M that were reported using NMR and fluorescence techniques (7,28). Fig. 4B shows the results of equilibrium dialysis to detect the binding of levosimendan to cTnC. It is clear that levosimendan does not bind significantly to cTnC because chambers with and without protein contained equal concentrations of levosimendan. Equilibrium dialysis experiments performed at pH 7.5, pH 7.0, and pH 6.8 and at different ionic strengths all showed no evidence of binding levosimendan to cTnC.
Potential Effect of Levosimendan on Ca 2ϩ Binding to cT-nC(A-Cys)-The potential effect of levosimendan on the Ca 2ϩ binding affinity of cTnC was determined using Ca 2ϩ -sensitive dyes. Fig. 5 shows the absorbance of 5,5Ј-dibromoBAPTA when titrated with Ca 2ϩ in the presence of cTnC(A-Cys) and in the presence or absence of levosimendan. Binding of Ca 2ϩ to 5,5Ј-dibromoBAPTA causes a decrease in absorbance at 262 nm. Competition between 5,5Ј-dibromoBAPTA and cTnC for binding Ca 2ϩ allows determination of the capacity and macromolecular binding constants for Ca 2ϩ binding sites in cTnC. The line represents a fit of experimental data in the absence of levosimendan using the method of Linse et al. (25). It is clear from Fig. 5 that the presence of levosimendan had no effect on the Ca 2ϩ binding properties of cTnC. The fractional change in absorbance was essentially identical for all level of Ca 2ϩ in the presence and absence of levosimendan. 33-211 -Together, the data in Figs. 2-5 demonstrate that levosimendan at concentrations of less than 1 mM does not bind to free Ca 2ϩ -bound cTnC. The modest chemical shift changes seen in Fig. 2 at high concentrations of levosimendan (4 mM) may be due to nonspecific or very weak binding. However, it is possible that levosimendan binds to cTnC only when it is bound to cTnI. To address this possibility, we monitored drug-induced chemical shift changes when cTnC(A-Cys) was bound to cTnI. For these experiments, we generated mouse cTnI(A-Cys) 33 were converted to Ser. A derivative of mouse cTnI in which the first 33 amino acids had been deleted was shown previously by Guo et al. (17) to be functional and has been used for structural studies (29,30). We elected to convert both Cys residues to Ser to prevent formation of intramolecular disulfide bonds in native cTnI. 2 A derivative of cTnI in which both Cys residues were converted to Ser was shown to be functional (31). Table I lists the chemical shifts for the 10 methionyl methyl 1 H-13 C correlations cTnC(A-Cys) when free and when associated with cTnI(A-Cys)  . All Met methyl chemical shifts in the C-terminal domain of cTnC(A-Cys), except Met 137 , were significantly affected by cTnI(A-Cys)  . The absolute chemical shift values for Met residues in the C-terminal domain of cTnC(A-Cys) in the presence or absence of cTnI(A-Cys)  are in excellent agreement with those of Krudy et al. (29), who used cTnC(C35S) and cTnI  . The chemical shifts of Met 85 , Met 80 , and Met 60 in the N-terminal domain are not affected by cTnI(A-Cys)  . Met 45 and Met 81 show a small effect, whereas Met 47 shows a significant change in the 1 H dimension of 0.09 ppm. These data differ from those of Krudy et al. (29),who reported that only methionyl methyl 1 H chemical shifts of Met 81 and Met 85 in the N-terminal domain of cTnC(C35S) were altered upon association with cTnI  . Fig. 6 shows the effect of levosimendan on the cTnC(A-Cys)⅐cTnI(A-Cys)  complex in the presence of Ca 2ϩ . The 1 H-13 C correlations attributed to Met groups located in the N-terminal domain (Fig. 6A) are shown separately from those in the C-terminal domain (Fig.6B). Solid contour lines indicate cross-peaks in the absence of drug, whereas dashed contour lines indicate cross-peaks upon the addition of 1 equivalent of drug. The lack of change in the chemical shifts indicates that levosimendan does not bind to the cTnC⅐cTnI complex.

Effect of Levosimendan on Met Methyl Chemical Shifts in cTnC(A-Cys) when Bound to cTnI(A-Cys)
Evidence for a Covalently-bound Drug-Protein Complex-During the course of our experiments, we observed that incubation of [methly-13 C]Met-labeled cTnC(A-Cys) with levosimendan at 40°C for 24 h resulted in the splitting of a number of 1 H-13 C correlations. This phenomenon is illustrated for apo cTnC(A-Cys) in Fig. 7. No splitting was seen for the C-terminal Met residues. However, splitting is observed for Met 45 , Met 47 , and Met 85 (see boxes in Fig. 7). The more intense of the split peaks, which corresponds to the untreated protein, represents about 85% of the total protein. This observation suggests the presence of more than one conformational state for the Nterminal domain of cTnC that is induced by prolonged exposure of the protein to levosimendan.
Conformational heterogeneity can result from relatively high affinity binding of a ligand such that it is in slow exchange between the bound and free form on the NMR time scale. Such splitting of the C-terminal Met resonances in cTnC(C35S) was observed when the high-affinity C-terminal Ca 2ϩ -binding sites are partially filled with Ca 2ϩ . However, Figs. 2-6 show that levosimendan does not bind to cTnC. Thus, we reasoned that the split resonances could be due to a slow covalent attachment of levosimendan to cTnC, likely in the N-terminal domain. This conclusion was supported by the observation that a distinct yellow color characteristic of levosimendan co-migrated with cTnC(A-Cys) when NMR samples were analyzed by SDS-polyacrylamide gel electrophoresis.
We used analytical HPLC gel filtration to investigate what appeared to be levosimendan covalently bound to cTnC(A-Cys). Elution profiles of various samples were monitored at the absorbance maxima for levosimendan (406 nm). Fig. 8A shows the chromatographic profile of cTnC(A-Cys) that had not been exposed to levosimendan. No peak is observed because the protein does not absorb light at 406 nm. The arrow in Fig. 8A shows the elution position of cTnC(A-Cys) detected at 280 nm. Fig. 8B shows the elution profile of levosimendan. Fig. 8C shows the elution profile of an NMR sample of cTnC(A-Cys) that had been exposed to levosimendan for an extended period of time. Fig. 8D shows the elution profile of the NMR sample after removal of free drug by desalting in the presence of 6 M urea and EGTA. Comparison of these elution profiles demonstrates that a chromophore, which absorbs at 406 nm, remains associated with cTnC(A-Cys) after extended exposure to levosimendan and subsequent removal of free drug under denaturing conditions. This data, as well as the data in  SDS-polyacrylamide gel electrophoresis analysis of levosimendan-treated protein strongly suggest the covalent binding of the drug to cTnC(A-Cys).
Mass Spectra of Levosimendan-Both fast atom bombardment and matrix-assisted UV laser desorption mass spectroscopy were used to determine whether breakdown products or derivatives of levosimendan, which could covalently bind to protein, accumulate in organic or aqueous solutions. Levosimendan was first analyzed immediately after preparation in Me 2 SO as shown in Fig. 9A. The peak at m/z 281.0 agrees well with the expected mass of levosimendan. The peak at m/z 550.5 is not identified but was seen in all spectra to varying extents. Levosimendan was stable in Me 2 SO because no change in the mass species was apparent after several days of storage (data not shown).
Levosimendan was next prepared in 0.5 M NaOH as described by Pollesello et al. (8) and analyzed immediately (Fig.  9B) or 72 h (Fig. 9C) after dilution into water. The major peak at m/z 302.9 corresponds well to the mass of the monosodium form of levosimendan. After 72 h in aqueous solution, a number of higher mass peaks appear (Fig. 9C). Peaks at m/z 561.0, m/z 583.2, and m/z 601.1 correspond well to the mass of (2 M), (2 M ϩ Na 2ϩ ), and (2 M ϩ 2Na 2ϩ ) forms of levosimendan. These data suggest that levosimendan can dimerize in solution. A time-dependent accumulation of higher molecular mass species was also seen when levosimendan was prepared in Tris or HEPES buffers (data not shown). DISCUSSION The primary purpose of the present study was to identify binding sites for levosimendan on cTnC. As part of this overall study, we also report the identification of binding sites for EMD 57033 on cTnC.
EMD 57033 is structurally distinct from bepridil and TFP. It has positive inotropic effects on cardiac myocytes (13) but is believed to act at the actin myosin interface (13). It is reported to bind selectively to the C-terminal domain of cTnC with a K d of approximately 40 M (12) but does not increase the Ca 2ϩ affinity of the regulatory site II (13). Our interest in EMD 57033 was to determine whether it can induce domain-specific changes in the Met methyl chemical shifts that would be consistent with selective binding to the C-terminal domain of cTnC. The data show a single binding site for EMD 57033 in the C-terminal domain of cTnC. Chemical shifts of Met methyl groups in the N-terminal domain of cTnC are unaffected, even at high concentrations of EMD 57033. It is likely that binding of EMD 57033 to the C-terminal domain of cTnC will be inhibited by cTnI because both bepridil and TFP are displaced from their C-terminal binding sites when cTnC binds to cTnI. 3 This would be consistent with the hypothesis put forth by Solaro et al. (13) that EMD 57033 acts at the actin/myosin interface downstream of cTnC.
All data lead to the conclusion that levosimendan does not bind to cTnC or the cTnC⅐cTnI complex. This conclusion is in contrast to previous reports. Haikala et al. (15) showed that the retention time of levosimendan on a cTnC affinity column was increased in the presence of Ca 2ϩ or Mg 2ϩ . From this it was concluded that levosimendan binds to both the N-and C-terminal domains of cTnC. However, significant effects on retention time were observed only at concentrations of Ca 2ϩ greater than 1-3 mM, and no maximal retention time appeared to be reached even at 30 mM Ca 2ϩ . Solution phase equilibrium dialysis shown in Fig. 4 failed to show direct binding of levosimendan to cTnC at a variety of pH values and ionic strengths, whereas bepridil binds to cTnC under similar conditions. Pollesello et al. (8) studied an N-terminal fragment of cTnC using homonuclear 1 H NMR, and reported NOEs between levosimendan and Met 81 , Met 85 , and Phe 77 in cTnC. However, these NOEs were weak, and the sequence-specific assignment of Met methyl groups was tentative. Our experiments used full-length cTnC(A-Cys) and cTnC3, coupled with high-resolution heteronuclear NMR and unambiguous assignments of the Met methyl groups derived from mutagenesis (19) and the solution structure (9) of full-length cTnC(A-Cys). Our data show that the 1 H chemical shifts of N-terminal Met residues in free cTnC(A-Cys), free cTnC3, and the cTnC(A-Cys)⅐cTnI(A-Cys) binary complex are unaffected by all concentrations of levosimendan tested. Moreover, no NOEs were observed between free cTnC(A-Cys) and levosimendan. The chemical shifts for Met 157 and Met 120 in free cTnC are affected by levosimendan, but the observed changes are significant only at drug: protein ratios of greater than 1:1, which equates to millimolar concentrations of the drug, and this effect is not seen for the binary complex. These data, coupled with equilibrium dialysis experiments shown in Fig. 4, support the conclusion that levosimendan does not bind to the N-terminal regulatory domain of cTnC. Potential interactions with the C-terminal domain that affect the chemical shifts of Met 157 and Met 120 may be due to a nonspecific or very weak association that is prevented when cTnI binds to cTnC.
Pollesello et al. (8) reported that levosimendan increased the Ca 2ϩ binding affinity of site II in cTnC as evidenced by Ca 2ϩdependent changes in fluorescence from dansylated cTnC in the presence and absence of drug. We have not directly addressed this issue; however, we see no evidence that levosimendan alters the Ca 2ϩ binding properties of cTnC(A-Cys) using a competitive dye assay in which the protein is not covalently modified.
Our data show that levosimendan, or a dansylated, can covalently link to cTnC with apparent specificity for the N-terminal domain, as evidenced by the splitting of resonances for N-terminal Met residues after prolonged incubation with the drug. NMR data sets were collected in 20 -30 min to minimize the extent of modification. Covalent modification of cTnC by levosimendan provides a plausible mechanism to account for the levosimendan-dependent effects observed by Pollesello et al. (8).
Our demonstration that levosimendan does not bind to cTnC does not question the positive inotropic effects of this compound that have been reported by different groups (26,32), it simply questions the mechanism by which this effect is achieved. Several groups have reported that the positive inotropic effects of levosimendan result, at least in part, from inhibition of phosphodiesterase activity and accumulation of cAMP (26,(33)(34)(35). Our data would be consistent with this, or with other mechanisms of action.