Solution Structure of Calcium-saturated Cardiac Troponin C Bound to Cardiac Troponin I*

Cardiac troponin C (TnC) is composed of two globular domains connected by a flexible linker. In solution, linker flexibility results in an ill defined orientation of the two globular domains relative to one another. We have previously shown a decrease in linker flexibility in response to cardiac troponin I (cTnI) binding. To investigate the relative orientation of calcium-saturated TnC domains when bound to cTnI, 1 H- 15 N residual dipolar couplings were measured in two different alignment media. Similarity in alignment tensor orientation for the two TnC domains supports restriction of domain motion in the presence of cTnI. The relative spatial orientation of TnC domains bound to TnI was calculated from measured residual dipolar couplings and long-range distance restraints utilizing a rigid body molecular dynamics protocol. The relative domain orientation is such that hydrophobic pockets face each other, forming a latch to constrain separate helical segments of TnI. We have utilized this structure to successfully explain the observed functional consequences of linker region deletion mutants. Together, these studies suggest that, although linker plasticity is important, the ability of TnC to function in muscle contraction can be correlated with a preferred domain orientation and interdomain distance. Muscle contraction is controlled by Ca 2 (cid:1) gradients, with the troponin complex mediating the response to the Ca 2 (cid:1) influx. The troponin complex consists of three proteins, troponin T, troponin I (TnI), 1 and troponin as described (13) at a final molar ratio of (cid:2) 0.96. Typically, Ca 2 (cid:1) -satu- rated cTnC (cid:1) cTnI complexes (0.4–0.6 m M ) were made 4.5% in C 12 E 5 / n hexanol lamellar phase. NMR Spectroscopy— Experiments were carried out on a Varian Inova 800-MHz spectrometer. The temperature was set at 38 °C for the Ca 2 (cid:1) - saturated cTnC (cid:1) cTnI complex in 4.5% C 12 E 5 / n -hexanol and at 40 °C for the Ca 2 (cid:1) -saturated cTnC (cid:1) cTnI complex in Pf1 phage liquid-crystalline solvent. 1 H- 15 N correlation experiments utilized sensitivity-enhanced 1 H- 15 N heteronuclear single quantum coherence-based pulse sequences (14), and RDCs were measured using a gradient (cid:2) / (cid:1) -transverse relaxation optimized spectroscopy experiment (15). Typically, transverse relaxation optimized spectroscopy-based experiments were performed using spectral widths of 12,000 and 2500 Hz for 1 H( t 1 ) and 15 N( t 2 ) with corresponding acquisition times of 170 and 44 ms, respectively. Data were processed using a 90° shifted sine-squared apodization in F 1 ( 1 H) and F 2 ( 15 N). The time domain data in t 1 were extended using forward/ backward linear prediction to 220 points and then zero-filled to 1024 points. The time domain data in t 2 were zero-filled to 4096 points. Dipolar contributions to the 1 H- 15 N couplings were obtained from differences in couplings observed in partially oriented and non-oriented samples. described for full-length cTnC. A constrained/restrained simulated annealing algorithm was subsequently used to refine suitable low energy structures as described for full-length cTnC.

Cardiac troponin C (TnC) is composed of two globular domains connected by a flexible linker. In solution, linker flexibility results in an ill defined orientation of the two globular domains relative to one another. We have previously shown a decrease in linker flexibility in response to cardiac troponin I (cTnI) binding. To investigate the relative orientation of calcium-saturated TnC domains when bound to cTnI, 1 H-15 N residual dipolar couplings were measured in two different alignment media. Similarity in alignment tensor orientation for the two TnC domains supports restriction of domain motion in the presence of cTnI. The relative spatial orientation of TnC domains bound to TnI was calculated from measured residual dipolar couplings and long-range distance restraints utilizing a rigid body molecular dynamics protocol. The relative domain orientation is such that hydrophobic pockets face each other, forming a latch to constrain separate helical segments of TnI. We have utilized this structure to successfully explain the observed functional consequences of linker region deletion mutants. Together, these studies suggest that, although linker plasticity is important, the ability of TnC to function in muscle contraction can be correlated with a preferred domain orientation and interdomain distance.
Muscle contraction is controlled by Ca 2ϩ gradients, with the troponin complex mediating the response to the Ca 2ϩ influx. The troponin complex consists of three proteins, troponin T, troponin I (TnI), 1 and troponin C (TnC). Troponin T interacts with tropomyosin and has been implicated in transmission of the contraction signal along the thin filament. TnI, the inhibitory subunit, has extensive contacts with TnC and actin and is responsible for inhibition of the actin-myosin interaction. The cardiac isoform of TnI (cTnI) differs from the skeletal isoform (sTnI) primarily by the presence of a 20 -32-residue-long extension at the amino terminus (1). This cardiac-specific extension contains two adjacent protein kinase A phosphorylation sites that play a role in modulating contraction in response to ␤-adrenergic stimulation.
TnC, the protein responsible for detection of the calcium signal and initiation of muscle contraction, consists of two globular domains separated by a flexible linker. In both skeletal (sTnC) and cardiac (cTnC) TnC (des-Met 1 ,Ala 2 , C35S), the C-terminal domain contains two high affinity Ca 2ϩ /Mg 2ϩ -binding sites (III and IV) and anchors TnC in the complex. The N-terminal domain also contains two EF-hand motifs that have high specificity, but a lower affinity for Ca 2ϩ (2). In the cardiac isoform, Ca 2ϩ -binding site I is inactivated by substitution of key metal-binding ligands and an amino acid insertion (3).
In the absence of bound Ca 2ϩ , the regulatory domain is in the inactive or "closed" conformation. Binding of Ca 2ϩ to the sTnC regulatory domain induces an opening of the regulatory domain and increased exposure of a hydrophobic pocket, the proposed binding site for the regulatory region of TnI (4). In the cardiac isoform, the presence of both Ca 2ϩ and the cTnI regulatory region appears necessary to stabilize a more "open" conformation with increased exposure of the hydrophobic pocket (4,5). The equilibrium between open and closed cardiac regulatory domain states can be modulated by protein kinase A phosphorylation of the cTnI-specific amino terminus or by binding of Ca 2ϩ -sensitizing agents (6 -9).
X-ray crystal structures of full-length TnC (10,11) are available along with high resolution solution structures of isolated N-and C-terminal domains (4,12). However, a structure of full-length Ca 2ϩ -saturated TnC bound to full-length TnI is currently unavailable. A complete description of the solution structure for Ca 2ϩ -saturated TnC bound to full-length TnI requires collection of both long-range distance and orientation restraints. In this study, orientation restraints were extracted from 1 H-15 N residual dipolar couplings (RDCs) measured by partially orienting Ca 2ϩ -saturated [ 2 H, 15 N]TnC⅐cTnI in two different alignment media. RDCs in combination with cTnC domain structures (4,12) and long-range distances (13)(14)(15) were used to elucidate the Ca 2ϩ -saturated cTnC structure in the cTnC⅐cTnI complex.
Alignment Media-Pf1 filamentous phage was prepared as described previously (16). Alignment of the medium was confirmed by observation of the quadrupole splitting of the 2 H 2 O signal. Typically, phage was titrated into the protein sample until a 2 H 2 O splitting of ϳ30 Hz was achieved.
NMR Spectroscopy-Experiments were carried out on a Varian Inova 800-MHz spectrometer. The temperature was set at 38°C for the Ca 2ϩsaturated cTnC⅐cTnI complex in 4.5% C 12 E 5 /n-hexanol and at 40°C for the Ca 2ϩ -saturated cTnC⅐cTnI complex in Pf1 phage liquid-crystalline solvent. 1 H-15 N correlation experiments utilized sensitivity-enhanced 1 H-15 N heteronuclear single quantum coherence-based pulse sequences (14), and RDCs were measured using a gradient ␣/␤-transverse relaxation optimized spectroscopy experiment (15). Typically, transverse relaxation optimized spectroscopy-based experiments were performed using spectral widths of 12,000 and 2500 Hz for 1 H(t 1 ) and 15 N(t 2 ) with corresponding acquisition times of 170 and 44 ms, respectively. Data were processed using a 90°shifted sine-squared apodization in F 1 ( 1 H) and F 2 ( 15 N). The time domain data in t 1 were extended using forward/ backward linear prediction to 220 points and then zero-filled to 1024 points. The time domain data in t 2 were zero-filled to 4096 points. Dipolar contributions to the 1 H-15 N couplings were obtained from differences in couplings observed in partially oriented and non-oriented samples.
Calculation of cTnC Domain Orientation-Starting coordinates for the cTnC solution structure determination were created by merging coordinates for the solution structure of Ca 2ϩ -saturated cTnC-(1-89) bound to cTnI-(147-163) (Protein Data Bank code 1MXL) (4) with coordinates for the solution structure of Ca 2ϩ -saturated cTnC-(81-161) bound to cTnI-(33-80) (Protein Data Bank code 1FI5) (12), which had been further refined. 2 To ensure sampling of all possible domain orientations, seven starting structures having extended conformations with different N-and C-terminal domain orientations were generated. Residues in the linker region (positions 84 -92) were treated as rigid bodies separated by the known distance between consecutive residues assuming planar and trans-peptide bonds.
Rigid body minimization was used to evaluate relative domain orientation by simultaneously minimizing the N-and C-terminal domains to the experimentally measured RDCs recorded in two different alignment media. A long-range fluorescence resonance energy transfer-derived restraint of 48 Å from the thiol proton of Cys 35 to Tb 3ϩ bound at site III (17) and a radius of gyration restraint of 23.9 Ϯ 0.5 Å derived from neutron scattering (18) were used to tether the two domains together. Calculations were carried out in XPLOR (19), modified to incorporate refinement against RDC restraints (20) using rigid body minimization and constrained/restrained annealing protocols (21). Structures with the lowest energy were selected for further refinement. Constrained/restrained simulated annealing was performed in several stages to further refine the domain-oriented conformation of Ca 2ϩsaturated cTnC bound to cTnI. A mock crystal symmetry molecule was used to generate artificial crystal symmetry restraints, preventing rotation of cTnC domains relative to each other during the remainder of the refinement process.
Initially, the linker region was refined using peptide bond distance restraints and Ramachandran-based dihedral restraints. The C-terminal domain of cTnC was then subjected to simulated annealing using the RDCs together with published nuclear Overhauser effect and dihedral restraints collected for the cTnC-(81-161)⅐cTnI-(33-80) complex (12). The resulting structure was then passed through this procedure again, except that the C-terminal domain was treated as a rigid body, whereas the N-terminal domain was subjected to constrained/restrained simulated annealing using the residual RDCs together with published nuclear Overhauser effect and dihedral restraints for the N-terminal domain in the cTnC-(1-89)⅐cTnI-(147-163) complex (4). The atomic coordinates have been deposited in the Protein Data Bank (code 1LA0).
Definition of the Rotation Axis-Relative cTnC domain orientation was described in terms of three consecutive rotations defined as bend, azimuth, and twist. The bend is defined as the angle between the D-helix (residues 73-83) and E-helix (residues 95-105). The azimuth of the bend was calculated relative to the orientation of sTnI-  bound to the C-terminal domain of sTnC in the crystal structure of sTnC⅐sTnI-(1-47) (Protein Data Bank code 1A2X) (11). The cross-product of unit vectors along the bend (E-helix) and azimuth axis defines the twist axis, rotation about which completes the transformation. The azimuth axis was fully orthogonalized using the cross-product between the bend and the twist axis as the final azimuth axis.
Modeling-To model the structural consequences of progressive deletions within the cTnC linker region, the structure of cTnC bound to cTnI (Protein Data Bank code 1LA0) was used as a template. Two 7-residue deletion models were created by removal of amino acids corresponding to residues 90 -96 and 86 -92 from cTnC. The remaining linker residues were treated as rigid bodies separated by the known distance between consecutive residues assuming planar and transpeptide bonds. Nine-and 12-amino acid deletions corresponding to residues 88 -96 and 85-96 of cTnC, respectively, were also constructed. To provide additional linker flexibility in the 12-amino acid deletion mutant, residues 81-83 were treated as rigid bodies in the rigid body minimization procedure.
For each linker region deletion mutant constructed, rigid body minimization was used to determine cTnC domain orientation and position as described for full-length cTnC. A constrained/restrained simulated annealing algorithm was subsequently used to refine suitable low energy structures as described for full-length cTnC.

RESULTS
Dipolar Couplings-Amide 1 H and 15 N resonance assignments for Ca 2ϩ -saturated [ 2 H, 15 N]cTnC⅐cTnI in Pf1 filamentous phage or C 12 E 5 /n-hexanol were transferred from previously published assignments for Ca 2ϩ -saturated cTnC bound to cTnI (8). Comparison of Ca 2ϩ -saturated cTnC chemical shifts in oriented and non-oriented cTnC⅐cTnI complexes showed minimal chemical shift perturbations due to the presence of the alignment media (data not shown). The absence of substantial chemical shift changes suggests that the cTnI-bound structure of cTnC is unaffected by addition of the alignment media. In the Pf1 filamentous phage liquid-crystalline medium, a total of 36 and 29 RDCs were measured in the N-and C-terminal domains of cTnC bound to cTnI, respectively. In the lamellar phase produced with C 12 E 5 /n-hexanol, 23 and 30 RDCs were measured in the N-and C-terminal domains of cTnC, respectively. In each alignment medium, a combination of resonance overlap, signal/noise, and line broadening prevented measurement of additional 1 H-15 N RDCs. The magnitude of the principal axis and the rhombicity of the alignment tensors were estimated from the distribution of the dipolar couplings (22). Values of Ϫ9.5 Hz for the magnitude of the principal axis of the alignment tensor and 30% rhombicity were estimated from RDCs measured in C 12 E 5 /n-hexanol lamellar phase. Similarly, values of Ϫ6.5 Hz and 25% for the magnitude and rhombicity, respectively, were estimated from RDCs measured in the Pf1 filamentous phage medium.
Estimated alignment tensor parameters were used with experimentally measured RDCs to determine the relative cTnC domain orientation in the presence of cTnI. To avoid the multiple minima associated with using RDCs for solution structure determination (23), an initial model was constructed for cTnC bound to cTnI from the domain solution structures using Ca 2ϩsaturated cTnC-(1-89)⅐cTnI-(147-163) (Protein Data Bank code 1MXL) (4) for the N-terminal domain and Ca 2ϩ -saturated cTnC-(81-161)⅐cTnI-(33-80) (Protein Data Bank code 1FI5) (12) for the C-terminal domain. The RDC data, collected in both the C 12 E 5 /n-hexanol and Pf1 filamentous phage liquid-crystalline media, fit the cTnC N-terminal domain structure with an R-factor of 26%. Similarly, the RDC data collected in the C 12 E 5 / n-hexanol and Pf1 filamentous phage liquid-crystalline media fit the cTnC C-terminal domain structure with R-factors of 26 and 31%, respectively. The reasonable fit of experimentally measured RDCs in intact cTnC bound to cTnI with solution structures of the individual cTnC domains justified use of the individual domain structures as starting points in the structure determination of cTnC bound to cTnI.
Structure Determination-Rigid body minimization was used to determine the relative domain orientation by simultaneously minimizing the N-and C-terminal domains to the experimentally measured RDCs recorded in both alignment media. In addition, the overall shape of cTnC was constrained using long-range distance restraints obtained from the radius of gyration derived from neutron scattering data (18) and an interdomain fluorescence resonance energy transfer distance (17). The inclusion of RDC data from two different alignment media does not completely eliminate the degeneracy in determining relative domain orientation. From seven initial starting structures, two different ensembles were obtained. In each ensemble, both sets of RDCs were simultaneously satisfied. In one ensemble, the linker region was significantly distorted, resulting in unfavorable backbone conformations as a consequence of simultaneously satisfying long-range distance restraints. Thus, structures within this ensemble were not further refined.
The remaining ensemble of cTnC domain-oriented structures was further refined, with a mock crystal symmetry molecule to prevent domain rotation, using a constrained/restrained simulated annealing algorithm (24). This permitted refinement of individual domain structures and minimization of the linker region against experimentally determined RDCs without significantly altering domain orientation. The cTnI-bound cTnC structure was refined to R-factors of 15 and 22% for RDCs collected in the lamellar phase and Pf1 filamentous phage, respectively (Fig. 1). The average domain-oriented solution structure for Ca 2ϩ -saturated cTnC bound to cTnI is shown in Fig. 2. Fig. 2 depicts a structure in which the hydrophobic pockets in each domain face each other and act as a latch to constrain separate regions of the cTnI molecule. The quality of the average cTnC domain-oriented structure bound to cTnI was further examined using PROCHECK (25). Residues in both domains show good covalent geometries with well defined secondary structures, having 97% of residues in the most favorable and allowed regions of the Ramachandran plot.
Comparison with Existing TnC Structures-The structure of cTnC-(81-161) bound to cTnI-(33-80) (cTnC-C) (Protein Data Bank code 1FI5) (12) is available with a peptide corresponding to residues 33-80 of cTnI. Superposition of this structure with the C-terminal domain of cTnC in the intact cTnC⅐cTnI complex yielded a backbone root mean square deviation of only 2.2 Å. Comparison of the structure of cTnC-(1-89) bound to cTnI-(147-163) (Protein Data Bank code 1MXL) (4) with the Nterminal domain in cTnC⅐cTnI yielded a backbone root mean square deviation of only 2.1 Å. The similarity of the cTnC-C and cTnC-N structures to their corresponding domains in the intact complex establishes the utility of using isolated domain structures as models for understanding the atomic details of divalent cation and TnI peptide interactions. Unfortunately, isolated domain studies cannot provide information on domain orientation, long-range protein-protein interactions, or the interdomain linker.
We also compared amide 1 H and 15 N chemical shifts for cTnC in the intact cTnC⅐cTnI complex with those for cTnC-C (12) and cTnC-N (26). Differences in amide 1 H chemical shifts are a sensitive probe of backbone conformational changes. The average chemical shift difference between cTnC-C and the C-terminal domain of cTnC in the intact cTnC⅐cTnI complex was 0.02 Ϯ 0.01 ppm. The chemical shift variation between cTnC-N and the N-terminal domain of cTnC in the cTnC⅐cTnI complex was 0.12 Ϯ 0.01 ppm. Although larger chemical shift differences were observed between N-terminal domain structures, these were modest and generally smaller than the chemical shift changes that occur upon binding cTnI-(148 -163) and Ca 2ϩ to cTnC (26). Furthermore, the magnitude and direction of Nterminal domain chemical shift differences are consistent with minor perturbations of the closed-to-open regulatory domain equilibria. Because most cTnC N-and C-terminal domain amides showed the same resonances in the intact cTnC⅐cTnI complex as observed in the individual isolated domains, the contact area between cTnC and cTnI must not be much larger than those observed in the cTnC(1-89)⅐cTnI(147-163) peptide (Protein Data Bank code 1FI5) (12) and the cTnC-N⅐cTnI peptide (Protein Data Bank code 1MXL) (4) structures. In addition,

FIG. 1. Experimentally determined versus calculated RDCs for the N-terminal domain (छ) and C-terminal domain (؋) of Ca 2؉saturated cTnC bound to cTnI obtained in nonionic liquid-crystalline media consisting of C 12 E 5 /n-hexanol (A) and Pf1 filamentous phage (B).
Calculated dipolar shifts were determined by the tensor fitting procedure using the domain-oriented solution structure of cTnC bound to cTnI. For each domain, solvent-exposed hydrophobic surface area was calculated and compared with that observed in the individual domain structures. The surface-accessible nonpolar area was determined utilizing the NMR-Refine module of Insight II (MSI). The calculated accessible nonpolar area for the N-terminal domain (residues 4 -83) is 2614 Å 2 for intact cTnC⅐cTnI and 2631 Å 2 for cTnC-N (Protein Data Bank code 1MXL). For the C-terminal domain (residues 93-159), the calculated accessible nonpolar area is 2280 Å 2 for intact cTnC⅐cTnI and 2320 Å 2 for cTnC-C (Protein Data Bank code 1FI5). Similarity in accessible nonpolar surface area further shows that individual cTnC domain conformations in cTnC⅐cTnI are similar to the aforementioned structures of the isolated N-and C-terminal domains bound to their respective cTnI binding peptides. DISCUSSION Uncertainty in domain orientation is a persistent problem in structure determination of multidomain proteins. This is especially true in the multidomain EF-hand protein family, exemplified by TnC and calmodulin (CaM), due to the presence of a long flexible linker connecting individual domains. Analysis of heteronuclear relaxation data established that binding of cTnI decreases conformational fluctuations within the cTnC interdomain linker (27). The observed decrease in linker flexibility indicates that, in the presence of cTnI, cTnC interdomain motion is restricted. In addition, paramagnetic effects on [methyl-13 C]Met residues in cTnC from a nitroxide spin label attached at Cys 84 suggested that binding of cTnI decreases interdomain motion, resulting in a more extended cTnC structure (28). In the absence of TnI, the interdomain linker is considerably more flexible, resulting in an ill defined orientation between the two domains, with both domains aligning in the magnetic field to differing degrees.
Restriction of interdomain motion in the presence of bound cTnI suggested that RDCs measured in liquid-crystalline media could be used to determine the average domain orientation of Ca 2ϩ -saturated cTnC bound to cTnI. With experimentally measured RDCs in two different alignment media and previously published long-range distance restraints, we determined the relative domain orientation of Ca 2ϩ -saturated cTnC bound to cTnI. The RDC-determined solution structure for cTnC, having both hydrophobic pockets highlighted, with cTnI residues 33-80 interacting with the C-terminal domain hydrophobic pocket and cTnI residues 147-163 interacting with the Nterminal domain hydrophobic pocket is shown in Fig. 2. This structure represents the first full-length TnC solution structure defining the relationship between globular domains in the presence of full-length cTnI.
The relative domain orientation is such that hydrophobic pockets within each domain are positioned to "latch on" to different helical segments of cTnI. The relatively good fit of RDCs measured in Ca 2ϩ -saturated cTnC bound to cTnI with available domain structures (4,12) together with the modest differences in chemical shifts between the intact complex and isolated domains support the assumption that high resolution TnC domain structures provide biologically relevant models for understanding atomic details of domain-cation and domaintarget interactions.
Validation of domain structures permitted the use of longrange distance and orientation constraints to investigate how individual domains are functionally assembled to control muscle contraction. In addition, these studies suggest that cTnC may interact with relatively small regions of cTnI, being structurally unaffected by the remaining portions of the protein. A similar finding for CaM binding to CaM kinase I and peptide complexes has recently been reported (29). In that study, chemical shift comparisons were used to show that the CaM⅐CaM kinase I peptide complex is a nearly perfect mimic for the interaction of CaM with the intact enzyme. Taken together, these studies suggest that both CaM and TnC interact with target proteins by recognizing and binding short EF-hand recognition motifs.
Regions of cTnI outside the known peptide binding sequences have little effect on cTnC, consistent with disordered or flexible regions of polypeptide separating functional cTnI domains. Model peptide studies utilizing cTnI residues 129 -149, containing the inhibitory region, suggest that the inhibitory region may not interact directly with cTnC (30). This finding is consistent with flexible linkers allowing TnI domains to move relative to one another.
To allow quantitative comparison of domain orientations between intact cTnC bound to cTnI with other Ca 2ϩ -binding protein⅐target complexes, the relative domain orientations are described by three parameters: bend, azimuth, and twist. When classified by the bend between the D-and E-helices, available TnC and CaM structures fall into two general families. The first family, as typified by the two Ca 2ϩ -loaded x-ray structure of TnC (Protein Data Bank code 1TOP), exhibits a slight bend along a particular azimuth with little or no twist. This structure has a helical linker and adopts an extended conformation with the hydrophobic pockets facing in opposition. The second family shows a larger bend with a portion of FIG. 3. Surface representations of EF-hand protein-target interactions. A, the solution structure of cTnC bound to cTnI. cTnC was found to have bend, azimuth, and twist angles of 70°, 30°, and Ϫ29°, respectively. B, the x-ray structure of sTnC bound to sTnI-(1-47) (11). The orientation of sTnC domains is defined by bend, azimuth, and twist angles of 93°, Ϫ162°, and 25°, respectively. C, the crystal structure of four Ca 2ϩ -loaded sTnC (36). The orientation of sTnC domains is defined by bend, azimuth, and twist angles of 12°, 94°, and Ϫ3°, respectively. The N-and C-terminal domains of TnC are shown in gold and red, respectively. TnI peptides are shown in magenta and blue. Superposition of TnC C-terminal domains was used to provide similar orientations. D, the NMR structure of Ca 2ϩ -bound CaM complexed with the myosin light chain kinase peptide (37). The orientation of CaM domains is defined by bend, azimuth, and twist angles of 111°, Ϫ95°, and Ϫ68°, respectively. The N-and C-terminal domains of CaM are shown in gold and red, respectively. The CaM-bound myosin light chain kinase peptide is shown in blue. Superposition of the CaM C-terminal domain and the TnC C-terminal domain was used to provide similar orientations. the D/E-helical linker unwound (Fig. 3). The cTnI-bound cTnC structure falls into this second family, having a 70°bend in the linker region. Similarly, the crystal structure of sTnC bound to sTnI-(1-47) has a 93°bend in the linker region (11). The azimuths for TnC bound to full-length TnI and TnI-(1-47) are 30°and Ϫ162°, respectively, producing different relative orientations of the hydrophobic binding pockets (Fig. 3). Differences in the disposition of the hydrophobic pockets between the solution structure of cTnC bound to cTnI and the crystal structure of sTnC bound to cTnI-(1-47) (11) may result from the influence of full-length cTnI or as a consequence of crystalpacking forces.
In the solution structure of cTnC, each Ca 2ϩ -saturated TnC domain is positioned to act as a latch constraining the separate helical regions of cTnI. The binding of Ca 2ϩ and the regulatory region of cTnI to the N-terminal domain of cTnC is generally believed to be the key interaction in a series of conformational rearrangements necessary for relieving inhibition of the actomyosin ATPase and initiation of muscle contraction. Assuming a largely structural role for the cTnC C-terminal domain-cTnI N-terminal domain interaction, the role of the linker region may be to optimally position the N-terminal domain for interaction with the regulatory region of cTnI. The removal of Ca 2ϩ from the regulatory domain would result in the unlatching of the cTnI regulatory region. Such a model suggests an essential role for linker plasticity in the ability of TnC to regulate muscle contraction in a Ca 2ϩ -dependent manner. Comparison of domain orientation in the absence of Ca 2ϩ is not possible because cTnC would only be anchored to the N-terminal domain of cTnI via C-terminal domain interactions.
The functional and structural significance of the interdomain linker in TnC is poorly understood. Deletion of 7 residues was found not to impact regulatory function (31). In a similar study, systematic deletion of 3-12 linker residues showed that up to 7 residues could be deleted with little change in maximal force development (31,32). However, with further deletions, inhibition increased until at 12 deletions the inhibition was complete (31,32). In an attempt to provide a structural context to the observed functional consequences of linker region deletions, we have examined the influence of deleting 7, 9, and 12 residues from the linker region on cTnC length and domain orientation. Two different 7-amino acid deletions in the linker region have been reported (31,32). Seven-residue deletions in the sTnC linker domain, corresponding to residues 90 -96 and 86 -92 in recombinant cTnC, were found to have little effect on maximal force development or the ability of sTnC to activate actomyosin ATPase (31)(32)(33). Computationally generated mutants, created by deletion of appropriate linker residues, were used as input structures in rigid body minimization and constrained/restrained simulated annealing protocols.
Computationally generated deletion mutants were evaluated for their ability to simultaneously satisfy both experimentally measured RDCs and long-range distance restraints. Acceptable low energy structures, satisfying all experimental constraints, were obtained for both computationally generated mutant cTnC proteins with 7-residue linker deletions. No significant differences in domain orientations were found between the 7-residue linker domain deletion mutants and wild-type cTnC.
Next, a computationally generated cTnC mutant with a 9-residue linker deletion, corresponding to residues 88 -96, was subjected to rigid body minimization and constrained/restrained simulated annealing. Although structures satisfying the experimental constraints were obtained, significant strain within the linker domain was introduced. The strain was evidenced by the appearance of distorted covalent geometry in the linker region, including elongated covalent bonds and bond angles that deviate significantly from the ideal values. The PROCHECK G-factor (25), a composite measure of covalent geometry, was used to evaluate the consequences of each deletion. The G-factor for the solution structure of cTnC bound to cTnI is Ϫ0.21. For both 7-residue deletions, the G-factor remained essentially unchanged at Ϫ0.22. However, for the 9-residue deletion, the G-factor decreased to Ϫ0.63. This value is below the Ϫ0.5 threshold, indicating the development of significant deviations from ideal covalent geometry. Although TnC can accommodate a 9-residue linker deletion and not violate experimental constraints, the increase in intermolecular strain required to adapt a wild-type conformation is likely responsible for the significant decrease in maximal force generated (31). Finally, orientation and distance restraints were inconsistent with computationally generated TnC structures having 12 linker residues deleted. Contraction was completely inhibited by TnC containing a 12-residue linker deletion (31), in agreement with our modeling studies. Thus, the solution structure for cTnC bound to cTnI can successfully explain the observed functional consequences of linker region deletions. Although plasticity within the linker region may play an important role, the ability of TnC to function in muscle regulation can be correlated with a preferred domain orientation and interdomain distance. Thus, the linker region also functions as a spacer to maintain the N-and C-terminal domains at an appropriate distance that likely correlates with the separation of TnI target peptides 33-80 and 147-163 in cTnI. Recent efforts to model the disposition of TnI relative to TnC have been hindered by the absence of structural information on TnI and long-range distance restraints (34,35). In addition, the crystal structure of free TnC (10, 36) was assumed to provide an accurate model for the solution structure of TnC in the complex and was used to define the overall length and domain orientation. The experimentally determined solution structure for cTnC bound to TnI demonstrates that the domain orientation differs from that of the free protein in the crystal (Fig. 3). Although initial events in muscle contraction are well characterized at the domain level, propagation of the signal remains poorly understood largely due to the uncertainty of interdomain position. Our structure provides a scaffold for further studies aimed at mapping interactions within the troponin complex responsible for the initiation of muscle contraction.