Solution structure and main chain dynamics of the regulatory domain (Residues 1-91) of human cardiac troponin C.

The three-dimensional structure of calcium-loaded regulatory, i.e. N-terminal, domain (1-91) of human cardiac troponin C (cNTnC) was determined by NMR in water/trifluoroethanol (91:9 v/v) solution. The single-calcium-loaded cardiac regulatory domain is in a "closed" conformation with comparatively little exposed hydrophobic surface. Difference distance matrices computed from the families of Ca2+-cNTnC, the apo and two-calcium forms of the skeletal TnC (sNTnC) structures reveal similar relative orientations for the N, A, and D helices. The B and C helices are closer to the NAD framework in Ca2+-cNTnC and in apo-sNTnC than in 2.Ca2+-sNTnC. However, there is an indication of a conformational exchange based on broad 15N resonances for several amino acids measured at several temperatures. A majority of the amides in the alpha-helices and in the calcium binding loop exhibit very fast motions with comparatively small amplitudes according to the Lipari-Szabo model. A few residues at the N and C termini are flexible. Data were recorded from nonlabeled and 15N-labeled samples, and backbone dynamics was investigated by 15N T1, T2, and heteronuclear nuclear Overhauser effect as well as by relaxation interference measurements.

Troponin C (TnC) 1 is the calcium-binding protein in the thin filament of skeletal and cardiac muscle and belongs to the EF-hand family of proteins. The structure of TnC closely resembles that of calmodulin, a ubiquitous calcium sensor, with an N-terminal and a C-terminal domain connected by a long central ␣-helix. Under physiological conditions the C-terminal domain of TnC is always occupied by Ca 2ϩ or Mg 2ϩ . In the N-terminal domain the low affinity sites of TnC bind calcium ions when they are released from the sarcoplasmic reticulum of skeletal or cardiac muscle myocytes. The conformational changes brought about by the calcium binding to the N-terminal domain of skeletal troponin C (sTnC) have been followed by NMR (1)(2)(3)(4). The data confirm the early hypothesis of an open and a closed conformation (5)(6)(7). According to the paradigm the Ca 2ϩ -induced conformational changes of the N-terminal domain are transmitted to the other components of the troponin complex and then to tropomyosin, triggering the muscle contraction (8 -12).
The sequence of the cardiac TnC (cTnC) is 70% identical to that of sTnC, whose structure in the calcium-saturated form has been determined by x-ray crystallography (7,13) and by NMR (14). However, while the N-terminal domain of sTnC contains two calcium-binding sites, the N-terminal domain of cTnC has only one intact Ca 2ϩ -binding site. The other site, often referred as a defunct site, does not bind calcium. The functional meaning of this difference between the cardiac and skeletal troponin C has not been explained in structural terms.
Even if there is no direct structural proof that the molecular mechanism of action of cTnC is the same as for sTnC, evidence has been given that also in cTnC the N-terminal domain is responsible for the regulation. It is assumed that the C-terminal domain plays primarily a structural role, since its calciumbinding sites are occupied even when the intracellular calcium concentration drops to micromolar values (15,16).
Recently, Sia et al. (17) working on cysteine (Cys-35 and Cys-84) to serine-mutated cTnC, showed that, in its calciumsaturated form, the N-terminal domain is in the closed conformation. This result implies that there is a much smaller conformational change to be expected upon calcium binding than in sTnC as proved by the very recent study by Spyracopoulos et al. (18). We aimed to confirm the existence of the closed form of the N-terminal domain for the human cTnC. Since the closed form of the N-terminal domain of cTnC was an unexpected result, we wanted to assess the structural integrity by dynamics measurements to get insight into a plausible conformational exchange between the closed and open states.

Cloning and Expression of Recombinant Human Cardiac Troponin C N-terminal Fragment-
The coding sequence of human cTnC was cloned by using the reverse transcription-polymerase chain reaction technique and human heart poly(A) ϩ RNA as a template and primers (20). The amplified DNA fragment was digested, purified, and subcloned to the pGEM3-vector (Promega). For protein expression the subcloned insert was isolated and ligated to the glutathione S-transferase fusion protein vector pGEX-2T (Pharmacia PL-Biochemicals) (19). The bacterial expression for production of human cTnC glutathione S-transferase fusion protein was carried out in Escherichia coli DH5 ␣-cells. The cells were grown at 37°C overnight in minimum medium according to Jansson et al. (20), using ammonium chloride instead of ammonium sulfate. The culture was diluted 1:25 in minimum medium containing 15 Nlabeled ammonium chloride (1 g/liter) and grown at 37°C to the middle of the growth phase, prior to induction with isopropyl ␤-D-thiogalactopyranoside (0.5 mM) for 4 h. The cells were harvested by centrifugation, and an aliquot of the collected cells was analyzed for estimation of the * The work was supported by the Academy of Finland. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Purification of Recombinant Human 15 N-Labeled cTnC N-terminal
Fragment-The collected bacterial paste of the 5-liter culture was diluted in 50 ml of 16 mM Na 2 HPO 4 , 4 mM NaH 2 PO 4 , 150 mM NaCl, 1% Triton-100, pH 7.3, buffer containing 0.2 mM phenylmethanesulfonyl fluoride and 25 l of bentzonuclease. The cells were disrupted by sonication on ice. The suspension was clarified by centrifugation and used as the starting material for purification on the glutathione Sepharose (10 ml; Pharmacia) affinity column. The recombinant human cTnC was further purified by HPLC anion exchange chromatography (Mono-Q HR 5/5) as described by Pollesello et al. (19). The free calcium was removed using Chelex-100 affinity chromatography (30 ml; Bio-Rad) equilibrated in water/ammonia solution, pH 8.0, and the eluted purified protein was freeze-dried. The amount of protein was estimated according to Bradford (21). The purity of the purified human cTnC fragment was analyzed by reversed phase HPLC (C 1 TSK TMS 250 column, 0.46 ϫ 4 cm) (19), and by SDS-polyacrylamide gel electrophoresis followed by Coomassie Brilliant Blue staining (22). The reversedphase chromatography-purified protein peak was further analyzed by matrix-assisted laser desporption isonization time-of-flight mass spectrometry. NMR Spectroscopy-Samples of the freeze-dried N-terminal fragment of cTnC (10 mg) were dissolved in 500 l of a solution containing 7.7 mM dithiotreitol in H 2 O. Dithiothreitol was added to prevent potential dimerization of cNTnC due to a possible disulfide formation between Cys-35 or Cys-84. The pH was adjusted at 6.0 with the addition of microliters of NaOH 0.5 M. Thereafter, CaCl 2 was added to a final concentration of 3.9 mM. Finally, D 2 O and perdeuterated trifluoroethanol (TFE) were added in order to obtain 750 l of a 1.3 mM protein solution in 82% H 2 O, 9% D 2 O, 9% perdeuterated TFE, at pH 6.0 Ϯ 0.1 (not corrected for the deuteron effect). 15 N-Labeled sample (1.3 mM) was prepared to 300 l in a Shigemi microcell.
Structure Generation-Spin systems were assigned from the through bond correlation spectra and sequence specific assignments were deduced from the NOEs between the adjacent spin systems. The spectral overlap was partly unraveled by 15 N editing and exploiting temperature dependence of HN resonances. Distance restraints were extracted from the homonuclear NOE data by fitting a second-order polynomial to integrated cross-peak volumes (I) of the NOE series. Additional distance restraints were extracted from NOEs observable in the 15 N-edited spectra according to intensity. The intramethylene and helical NH i -NH iϩ1 NOEs served for the calibration. Distances were given ϩ20 and Ϫ50% uncertainties. There was no objectives to tighten the distance limits because of a likely presence of conformational exchange and spin diffusion in the 15 N-edited NOE spectrum. When a distance could not be extracted from the build-up curve, owing to a overlap, a poor signal-tonoise ratio or disturbances, the distance was restrained to be at most 5.0 Å. The upper bounds were extended by 1.0 Å for each pseudo atom in methyl groups and 1.5 Å for pseudo atoms in aromatic rings. Backbone dihedral angles characterized by small J NH␣ , measured from COSY spectra, were restrained to the helical conformation (Ϯ30 degrees) on the basis of the Karplus relation and NOEs (25). Dihedral angles characterized by intermediate J NH␣ were not constrained. The temperature dependence of HN ␦ NH (T) was measured from two-dimensional spectra, ␣ proton chemical shifts (␦ HA ) and J NH␣ served to recognize secondary structures. A residue was constrained by NH i -CO iϩ4 hydrogen bond (Ϯ0.2 Å) provided that at least for three subsequent residues ␦ NH (T) Ͻ 0.005 ppm/°C and ␦ HA were smaller by 0.1 ppm than the corresponding random coil values and J NH␣ Ͻ 6 Hz. The calcium binding site was constrained by six distances between the calcium-ion and carbons in the coordinating carboxyl groups (2.1-3.9 Å). The short ␤-sheet between the calcium binding loop and the pseudo site was constrained by two hydrogen bonds.
Structures were generated by distance geometry followed by restrained simulated annealing. The structures were iteratively refined. At each iteration, distance restraints corresponding to degenerate NOEs were included provided that it was possible to exclude all but one alternative based on clearly longer distances. The procedure was finished when no more restraints could be imposed. The assignment and structure calculations were carried out by Felix and InsightII 95 software. A final set of structures was analyzed for RMSD, energy, backbone dihedrals, exposed surface, distance violations by PROCHECK and InsightII programs. Difference distance matrices (DDM) were calculated with a program modified to compute DDMs for families of structures.
The relaxation and heteronuclear NOE data were analyzed in terms of the commonly used Lipari-Szabo model-free spectral density (26) with the programs by Palmer (27). The overall correlation time m was estimated from R 2 /R 1 ratios (28). The cross-correlation rate between the dipole dipole and chemical shift anisotropy (29) was measured by a modified HSQC 1 H-15 N correlation experiment with a relaxation period prior to the 15 N evolution time, as described by Tjandra et al. (24). During the relaxation period (2⌬) the two 15 N doublets relax with rates proportional to the square of the sum and difference of double dipole and chemical shift assay interactions. The in-and anti-phase components are subsequently chosen separately for the detection yielding two spectra. The cross-correlation rate () is obtained from the ratio of the signal intensities (I A and I B ) of the two spectra by ϭ (1/2⌬)tanh Ϫ1 (I A /I B ) (24). and C helices. The defunct site is between the A and B helices and the calcium binding site between the C and D helices. The two sites are spatially close, and there are interstrand hydrogen bonds between the backbone amide protons and carbonyl oxygens (residues 36 and 72). The spatially adjacent N, A, and D helices are approximately orthogonal to each other, and they form a compact structural unit, the NAD-unit (18).

Structure-The
The structural statistics are given in Table I. The overall fold is defined by 224 long range NOEs. The backbone heavy atom RMSD is 1.1 Ϯ 0.2 Å. The helices and the calcium binding site are well defined. Three residues at each termini, the defunct site, and the central loop have larger RMSD (Ϸ2 Å). The all heavy atom RMSD is 1.8 Ϯ 0.2 Å. The extensive spectral overlap in the aliphatic region prevented us from deriving stereospecific assignments for side chain methylene groups. The RMSD per residue reflects approximately the distribution of the number of restraints (on the average 19) per residue.
The orientation of the A and B helices is based on the observation of 20 NOEs. These NOEs are primarily between aliphatic protons. For example, we find an NOE between Phe-20 and Met-45, which are in the middle of the A helix and the B helix, respectively. In an open conformation these distances would be clearly longer, and NOEs except in the vicinity of the loop would be beyond detection. However, the uncertainty in the side chain to side chain distance bounds involving particularly the pseudo atoms of methyl groups do not allow us to determine the interhelical angles precisely. Due to the overlap of in the two-dimensional homonuclear spectra, few NOEs between the A and B as well as between the C and D helices were not assigned. The loop between the A and B helices, i.e. the defunct site, is quite irregular and not particularly well defined. We identify 17 NOEs from the defunct loop. In the central loop between the B and C helices the prolines 52 and 54 are both in the trans configurations. We did not find any evidence for cis-trans isomerism.
Dynamics-The rate constants of the 15 N longitudinal relaxation (R 1 ) are nearly constant (1.5 Ϯ 0.2 s Ϫ1 ) over the entire sequence. Only for Gly-91, the R 1 is substantially smaller. The variation in the rate constant of the transverse relaxation (R 2 ) is larger. For the helices and the calcium binding site the R 2 values are 10 Ϯ 2 s Ϫ1 and toward the N-and C termini the R 2 values drop for a few residues. Furthermore, for residues Val-28, Ser-37, Thr-38, Lys-39, Glu-40, Ile-61, and Val-64 the transverse relaxation proceeds so fast that the rate constants (Ͼ15 s Ϫ1 ) could not be determined reliably. The values of R 1 for residues 37-40 were smaller than the corresponding R 2 values but owing to the off-resonance effects (30) quantitative comparisons were not made. The 15 N line widths for these residues measured from the HSQC spectrum recorded at 30°C are indeed larger than on the average (Fig. 2). Whereas at increased temperature (40 and 50°C), the line widths are only 1-2 Hz wider than on the average. At lower temperatures (10 and 20°C), also the lines for residues Ile-36, Leu-41 and Gly-42 become broader than the average (Fig. 3). The rate of the 15 N cross-correlation () between the dipole and chemical shift anisotropy was insensitive to the conformational exchange. The comparison of with R 2 reveals residues experiencing fast transverse relaxation (Fig. 3). For the residues (Val-28, Thr-38, Lys-39, Ile-61, and Val-64) with very fast transverse relaxation, no values of were obtained due to the insensitivity of the cross-correlation experiment.
The heteronuclear Overhauser enhancement was nearly constant and close to the maximum value (ϩ0.82) (28) for the helices and the calcium-binding site but decreased for a few residues at the N and C termini. Heteronuclear NOEs could not be determined reliably for residues Ser-37 and Thr-38 due to the line broadening.
The principle components of the inertia tensor computed for the cNTnC structure were (0.81:0.85:1), which implies that the overall rotational diffusion can be regarded approximately isotropic. Furthermore, the ratio of R 2 /R 1 for the helices did not vary significantly, which also implies an approximately isotropic rotational diffusion characterized by m ϭ 7.7 ns. Therefore we considered the analysis of relaxation data in terms of the commonly used Lipari-Szabo model-free spectral density (26) reasonable. First the simple model was used with the order parameter (S 2 ) as the only free parameter. Subsequently, the time constant e for the fast internal motion was allowed to vary as well. In this way a statistically good fit was obtained for the majority of the residues. For the residues in the helices and in the calcium binding site S 2 is approximately 0.9. The extended model in which the internal motion is divided into two components (31) resulted for the residues at the N and C termini (2-8 and 86 -91) in a statistically good fit. At the N and C termini S 2 drops rapidly (Fig. 4). For the remaining residues  20,23,25,28,30,[37][38][39][40]48, 60, and 63-64, the conformational exchange contribution (R ex ) was included in the analysis. For the residues 28, 37-40, 61, and 64, reliable values for S 2 were not obtained due to the fast loss of transverse coherence.

DISCUSSION
For comparisons to skeletal troponin C, the angle between helices A and B in cNTnC (135°) is approximately the same as that of apo-sNTnC (138°) but very different from that of 2⅐Ca 2ϩ -sNTnC (81°) (17). The angle between the C and D helices (130°) is more similar to the corresponding angle in apo-sNTnC (145°) than to the angle in 2⅐Ca 2ϩ -sNTnC (78°). Owing to the comparatively large distance bounds of the few NOEs in the calcium binding site the angle between the C and D helices is subject to uncertainty which does not allow us to prove unambiguously differences between cNTnC and sNTnC. The relative orientations of the N, A, and D helices of cNTnC are similar to the corresponding orientations in both apo-and 2⅐Ca 2ϩ -sNTnC. In this NAD frame of reference the orientation of the B and C helices of cNTnC resemble more those of the apo-form than the two-calcium form of sNTnC. This is obvious from the DDMs computed from the families of structures in order to obtain unbiased comparisons (Fig. 5). In cNTnC the B helix is closer to (approximately 10 Å) the N and D helices than in 2⅐Ca 2ϩ -sNTnC. On the other hand the differences between cNTnC and apo-sNTnC are small; however, the cNTnC form appears slightly more compact (Fig. 5B). The residue Val-28 not in the sNTnC sequence was excluded from the comparisons. For the sake of completeness a DDM was computed between apo-and 2⅐Ca 2ϩ -sNTnC (Fig. 5C). The result is quite similar to the DDM between cNTnC and 2⅐Ca 2ϩ -sNTnC. These comparisons imply only minor conformational changes to be expected upon the calcium binding as already shown by Spyracopoulos et al. (18). This may indeed be a fundamental property of cTnC, which enables the fast switching between apo and holo states necessary for the heart function. On the other hand, so far no high resolution structure has been obtained from a complex comprising the key parts of cardiac TnC and troponin I. The association of cTnC with troponin I may affect the degree of opening of cTnC upon calcium binding (18).
Qualitative comparisons with the structures of the avian (17) and human (18) regulatory domains of cardiac TnC, reported just recently, reveal a high degree of similarity. Apparently the C35S and C84S mutations, used to prevent possible disulfide formation, in the avian cTnC study are immaterial for the overall structure. Nevertheless, we notice that NOEs between Ala-23 and Val-44 from the A and B helices, respectively, reported for the avian cTnC are absent in our NOE spectra. This could be explained by a small difference in the twist of the B helix with respect to the A helix between the human and avian cTnC. Furthermore, we notice some differences in the chemical shifts of certain NHs between our data and the data by Spyracopoulos et al. (18). This is partly due to the slightly longer (1-91 amino acids) sequence used in the present study compared with the protein (1-89 amino acids) studied by Spyracopoulos et al. (18) (Fig. 1 caption). In addition NH chemical shifts depend on the concentration of TFE but we do not observe noticeable structural consequences. According to an earlier study on sTnC (33) the use of 15% TFE should prevent the dimerization involving the N-terminal domain. The lack of apparent variation in R 2 /R 1 for the approximately globular tertiary structure of the cNTnC excludes the presence of specific dimers, but it may be that 9% of TFE used in the present study does not completely prevent a transient unspecified dimerization. Indeed our estimate of the overall correlation time ( m ϭ 7.7 ns) based on R 2 /R 1 (28) appears somewhat high in comparison with values obtained for proteins of similar sizes. This leaves the possibility for the monomer-dimer exchange, which leads to an erroneous estimate for the m and consequently also to a small systematic error in the values of S 2 assuming the isotropic model. The fast dynamics itself is unlikely to be affected by the dimerization (34). The monomerdimer exchange is indistinguishable from a conformational exchange (34). In the present study it appears more likely that the comparatively fast loss of the transverse coherence for residues 28, 37-40, 61, and 64 is caused by the conformational exchange rather than by the monomer-dimer exchange. These residues do not appear to form a contact surface.
Evenäs et al. (35) have studied a mutated C-terminal half of calmodulin, E140Q, and observed a conformational equilibrium between the open and closed forms with a population ratio close to 1:1. They observed the most pronounced effects on resonances from residues in the calcium binding loops. The strongest line broadening observed in the present work is for residues in the defunct calcium binding loop 37-40 and hydrophobic residues Val-28, Ile-61, and Val-64. Coincidentally, Val-64 is also the "hinge" about which the end of the C helix rotates relative to the NAD unit in sTnC (36) as well as Glu-40 is the "hinge" about which the B helix reorients with respect to the NAD unit in cTnC (18). Therefore, it is interesting to consider the possibility of an exchange between the closed form and a low population of the open state as an explanation for the line broadening. In particular because prior to the results by Sia et al. (17) a large conformational change similar to the one observed for sTnC was generally expected for cTnC. It was shown, for example, that upon a calcium titration the chemical shift of the methyl of Val-28 changes much (␦ CH3 ϭ 0.16). In retrospect this significant change is probably explained by small movements of the nearby phenyls, because the structures of apo-and Ca 2ϩ -cNTnC are so similar (18). In the case of the mutant calmodulin two conformers were distinguished from a double set of mutually excluding NOEs. We find no clear evidence for a similar situation, which at least implies that the determined closed conformation of cNTnC has a population above 80 -90%. A small population of the minor conformation could still cause the rapid transverse relaxation provided that the chemical shifts for the two conformations would differ significantly for instance due to ring current effects. A study of the structures of the apo and holo states of cNTnC in the presence of the binding peptide from TnI, now in progress, should settle if the presently known calcium-loaded closed conformation is truly the biologically relevant conformation of the regulatory domain of cardiac troponin C.
The time-limiting step in muscle relaxation is most likely the release of calcium from the N-terminal half of troponin C. For free troponin C and calmodulin the off-rate of the Ca 2ϩ ions from the regulatory sites is about 500 s Ϫ1 (3). However, in the presence of a binding peptide or drug this rate is dramatically reduced. For skeletal muscle with sTnC and smooth muscle with calmodulin, even a 100-fold reduction is acceptable, but not so for heart muscle. For the heart muscle to relax to about 90% in half a second the calcium off-rate has to be about 7 s Ϫ1 . Maybe the cardiac TnC has developed to its present sequence with the defunct site to cope with the strict timing requirements, which can be more easily met with a somewhat weaker calcium binding.