Effects of Troponin I Phosphorylation on Conformational Exchange in the Regulatory Domain of Cardiac Troponin C*

Conformational exchange has been demonstrated within the regulatory domain of calcium-saturated cardiac troponin C when bound to the NH2-terminal domain of cardiac troponin I-(1–80), and cardiac troponin I-(1–80)DD, having serine residues 23 and 24 mutated to aspartate to mimic the phosphorylated form of the protein. Binding of cardiac troponin I-(1–80) decreases conformational exchange for residues 29, 32, and 34. Comparison of average transverse cross correlation rates show that both the NH2- and COOH-terminal domains of cardiac troponin C tumble with similar correlation times when bound to cardiac troponin I-(1–80). In contrast, the NH2- and COOH-terminal domains in free cardiac troponin C and cardiac troponin C bound cardiac troponin I-(1–80)DD tumble independently. These results suggest that the nonphosphorylated cardiac specific NH2 terminus of cardiac troponin I interacts with the NH2-terminal domain of cardiac troponin C.

Troponin C (TnC) 1 is the Ca 2ϩ binding component of the troponin complex that is required to confer Ca 2ϩ sensitivity on the actin-myosin interaction. The troponin complex consists of three proteins: troponin I (TnI), which inhibits the actomyosin Mg 2ϩ -ATPase; TnC, which removes TnI inhibition triggering the Mg 2ϩ -ATPase; and TnT, which makes primary proteinprotein contacts with tropomyosin. The cardiac isoforms of TnC and TnI differ significantly from skeletal isoforms. In the cardiac isoform of TnC, Ca 2ϩ binding site I, comprising residues 28 through 40, is naturally inactive (1). Residues 65 through 76 form the regulatory calcium binding site II. The cardiac isoform of TnI is unique in that it contains an additional NH 2 -terminal extension of approximately 32 residues. This extension contains two adjacent serine residues that can be phosphorylated by PKA. Phosphorylation has been demonstrated to modulate myofilament sensitivity to Ca 2ϩ by reducing Ca 2ϩ affinity for the NH 2 -terminal regulatory site of cTnC (2). The reduction in the sensitivity of myofilament force development to Ca 2ϩ induced by phosphorylation of cTnI can be mimicked by exchange of the native unphosphorylated cTnI with cTnI-DD, in which Ser 23 and Ser 24 have been mutated to Asp (3). Cardiac TnI-DD is also able to mimic effects of phosphorylation on the steadystate and pre-steady-state binding of cTnI to cTnC (4). Cardiac TnC and TnI are known to interact in an antiparallel manner such that the COOH-terminal domain of TnC interacts with the NH 2 -terminal domain of TnI (5).
Conformational changes induced by Ca 2ϩ binding to the NH 2 -terminal domain of skeletal TnC (sTnC) have been followed by both x-ray (6) and solution NMR (7). For the free skeletal TnC protein, Ca 2ϩ binding results in a conformational change from the "closed" form to an "open" form exposing a patch of hydrophobic residues for interaction with TnI (7). Surprisingly, the Ca 2ϩ -bound NH 2 -terminal domain of cTnC was found to maintain a closed conformation (8). The solution structure of the Ca 2ϩ saturated regulatory domain of human cTnC, cTnC-(1-91), in 9% trifluoroethanol, also reveals a "closed" conformation with little exposed hydrophobic surface (9). In addition, 15 N relaxation studies on human cTnC-(1-91) suggested conformational exchange for Val 28 , Thr 38 , Lys 39 , Ile 61 , and Val 64 . The authors suggest the origin of this conformational exchange results from exchange between the closed form and a low population of the open form. However, monomer-dimer exchange could not unequivocally be ruled out.
To elucidate the role of phosphorylation of the cardiac-specific amino terminus of cTnI on full-length cTnC and to unequivocally demonstrate the presence of conformational exchange in the regulatory domain of full-length cTnC, we have probed the dynamics of cTnC free and bound to cTnI-(1-80) and cTnI-(1-80)DD. We show that residues within inactive Ca 2ϩ binding site I undergo chemical exchange consistent with an equilibrium between closed and opened forms both in the presence and absence of the NH 2 -terminal cTnI domain. Based on chemical shift perturbation mapping and transverse relaxation rates, the cardiac-specific NH 2 terminus of cTnI-(1-80) appears to make additional interactions with the regulatory domain of cTnC, which results in slowing down the exchange rate in the defunct Ca 2ϩ binding site I as evidenced by the presence of 1 H-15 N cross-peaks representing two distinct conformations for Leu 29 , Glu 32 , and Gly 34 . These interactions were not observed in the free protein or in the complex with cTnI-(1-80)DD.

MATERIALS AND METHODS
Proteins-The single cysteine form of cTnC (cTnC35S) was used in the study (10,11). 15 N and 2 H labeling was accomplished using minimal medium containing 90% 2 H 2 O and 1 g/liter 15 NH 4 Cl. Cardiac troponin C was purified as described previously (5). Cardiac troponin I-(1-80) and cTnI-(1-80)DD were expressed as inclusion bodies purified in 8 M urea. 2 All proteins were judged to be homogeneous by SDS-polyacrylamide gel electrophoresis and staining with Coomassie Brilliant Blue.
NMR Spectroscopy-All experiments were carried out on Varian Inova 600 or 800 MHz spectrometers. Amide 1 H and 15 N resonances of cTnC were obtained by means of heteronuclear triple resonance NMR experiments. 3 To confirm the assignments for cTnC bound to cTnI-(1-80) and cTnI-(1-80)DD, 1 H-15 N NOESY-HSQC experiments with mixing times of 85 and 200 ms were performed at 800 MHz. The 15 N R 2 and dipolar CSA interference experiments were performed on Ca 2ϩ -saturated [ 2 H, 15 N]cTnC free and bound to either cTnI-(1-80) or cTnI-(1-80)DD. For all experiments, the spectral widths in t1 and t2 dimensions were 2.00 and 15.62 kHz, respectively. The 1 H carrier was set to the frequency of the water resonance at 4.77 ppm. The number of transients used in R 2 and I auto transverse cross-correlation experiments was 16, while I cross transverse cross-correlation experiments were performed with 48 transients. 15 N dipolar CSA interference experiments were performed as described by Kroenke et al. (12). Relaxation delays used for the transverse cross-correlation rate ( xy ) measurements were 0.0320, 0.0534, 0.0748, 0.0961, and 0.1068 s. The R 2 experiments were performed with recovery delays of 0.000, 0.032, 0.064, 0.096, 0.128, 0.160, 0.192, and 0.224 s as described previously (14). Double points were collected with recovery delays of 0.032, 0.128, and 0.192 s to estimate the uncertainties.
Data Processing-NMR data was processed using Felix 97.2 (MSI). Free induction decays were apodized with a skewed sine bell 90°function. Generally, peak heights were measured using Felix routines. For the transverse cross-correlation experiments I cross /I auto intensity ratios were normalized for the number of transients used in each experiment. Curve-fits for transverse relaxation and cross-correlation rates were performed using the CURVEFIT software (13).

RESULTS
To compare interactions of cTnI-(1-80) and cTnI-(1-80)DD with cTnC, 1 H and 15 N chemical shift differences of bound cTnC were analyzed. Chemical shift differences for amide 1 H and 15 N resonances observed between the two complexes are presented in Fig. 1. These chemical shift differences can be used to map to binding site of cTnI-(1-80) on the regulatory domain of cTnC. In addition, analysis of chemical shift differences between free cTnC and cTnC bound to cTnC⅐cTnI-(1-80) or cTnC⅐cTnI-(1-80)DD reveals that both proteins bind predominantly to the COOH-terminal domain of cTnC causing a conformational change, as was observed for cTnI-(33-80) binding to cTnC-(81-161). 4 Taken together, these results allow us to propose that both nonphosphorylated and phosphorylated forms of cTnI bind to the COOH-terminal domain of cTnC with the nonphosphorylated form making limited contacts with the NH 2 -terminal domain of cTnC. It is also apparent that binding of the nonphosphorylated NH 2 -terminal domain of cTnI causes only minor structure perturbations in the NH 2 -terminal domain of cTnC. 1 H-15 N correlation NMR spectroscopy was used to study the interactions of cTnC with TnI-(1-80) and TnI-(1-80)DD. Residues excluded from the analysis due to peak overlap, even at 800 MHz, in the 1 H-15 N correlation spectra of cTnC bound to cTnI-(1-80) or cTnI-(1-80)DD are listed in Fig. 2. In the cTnC⅐cTnI-(1-80)DD complex, cross-peaks for cTnC residues 38 and 39 were broadened beyond detection. Cross-peaks for residues 24, 28, 36, 37, 38, 39, and 40 were broadened beyond detection in the cTnC⅐cTnI-(1-80) complex.
To further investigate interactions between the regulatory domain of cTnC and the NH 2 -terminal domain of cTnI, 15 N transverse relaxation and 1 H-15 N diploar/ 15 N CSA cross-corre-lation rates were measured at 800 MHz on [ 2 H, 15 N]cTnC free and bound to cTnI-(1-80) and cTnI-(1-80)DD. Perdeuteration offered an opportunity to obtain precise relaxation parameters and enhanced spectral resolution. Data analysis was performed as described under "Materials and Methods." The experimental values for R 2 obtained at 800 MHz on cTnC⅐cTnI-(1-80), cTnC⅐cTnI-(1-80)DD, and free cTnC are depicted in Fig. 2. Not surprisingly, low R 2 values for flexible NH 2 and COOH termini of cTnC free and bound are observed. It is also apparent that Met 85 through Thr 93 in the linker region of cTnC free and cTnI bound exhibit reduced R 2 values. In addition, Ala 30 through Ser 35 in the inactive calcium binding site I of cTnC bound to cTnI-(1-80) or cTnI-(1-80)DD exhibit reduced R 2 values. This fact is indicative of increased mobility in these regions.
Elevated R 2 relaxation rates are found for Val 28 , Ile 36 , and Ser 37 in both free cTnC and cTnC bound to cTnI-(1-80)DD (Fig.  2B). R 2 relaxation rates could not be obtained for these residues in the cTnC⅐cTnI-(1-80) complex due to extreme broadening of 1 H-15 N cross-peaks in HSQC spectra recorded for the R 2 measurements. This broadening may best be explained by the possibility that these residues are engaged in slow conformational exchange.
To prove the existence of conformational exchange in the inactive Ca 2ϩ binding site I of cTnC, measurements of 1 H-15 N dipolar/ 15 N CSA transverse ( xy ) cross-correlation rates were carried out. These cross-correlation rates do not depend on chemical exchange effects and are directly proportional to the generalized order parameters (15). Therefore, the xy values are sensitive only to internal and overall motions (12, 16). The transverse cross-correlation rates are also presented in Fig. 2 In general, the transverse cross-correlation rates repeat the same trend as R 2 values. However, for residues Val 28 , Ile 36 , and Ser 37 the cross-correlation values are not increased like the R 2 relaxation rates in this region. Thus, the results of these experiments prove that residues adjacent to the calcium binding site I undergo conformational exchange in free cTnC and when bound to cTnI-(1-80)DD. Conformational exchange in Val 28 , Ile 36 , and Ser 37 is slower when cTnC is bound to cTnI-(1-80), since these residues have broadened beyond detection. In support of this, multiple resonances for Leu 29 , Glu 32 , and Gly 34 of cTnC were observed in the complex with cTnI-(1-80) (Fig. 3). Careful analysis of NOESY-HSQC spectra revealed that both cross-peaks for each residue have NOEs to the same amide protons. In the case of Leu 29 and Gly 34 , mutual sequential NOEs are observed to Gly 30 (Fig. 3A) and Asp 33 (Fig. 3B), respectively. These results allowed us to suggest that the new cross-peaks represent a different conformation for the defunct Ca 2ϩ binding site I of cTnC. The resonance for Glu 32 also appears to consist of multiple peaks, but H N -H N NOEs for Glu 32 could not be observed in the recorded NOESY-HSQC spectra. Thus, we are unable to unequivocally assign both cross-peaks to Glu 32 . Cross-peaks for Leu 29 , Glu 32 , and Gly 34 are well resolved, and it is unlikely that the new peaks belong to some other unassigned amide protons experiencing extremely large chemical shift perturbations and having identical i to i ϩ 1 NOEs. It seems possible that cTnI-(1-80) binds to the NH 2 -terminal domain of cTnC and slows down chemical exchange in defunct Ca 2ϩ binding site I, presumably changing the dynamic equilibrium between open and closed conformations.
On average, the R 2 values for the COOH-terminal domain of the free protein are smaller than those values for the NH 2terminal domain (Fig. 2C). This can be attributed to the fact that the NH 2 -terminal domain of cTnC is slightly larger than the COOH-terminal domain. The average R 2 value for the COOH-terminal domain, comprising residues 104 through 135, was 11.0 Ϯ 1.9 s Ϫ1 , while the average R 2 value for the NH 2 terminus, comprising residues 14 through 23 and residues 41 through 80, was 14.6 Ϯ 1.6 s Ϫ1 . The linker region, the NH 2 and COOH termini, and inactive calcium binding site I were excluded from this analysis on the basis of either increased mobility or possible conformational exchange in these regions.
When cTnC is bound to cTnI-(1-80)DD the average R 2 value (20.0 Ϯ 1.9 s Ϫ1 ) for the COOH-terminal domain is larger than the average R 2 value for the NH 2 -terminal domain (16.0 Ϯ 1.8 s Ϫ1 ) as shown in Fig. 2B. This difference could be explained by the fact that the NH 2 terminus of cTnI binds to the COOH terminus of cTnC (5). However, for the cTnC⅐cTnI-(1-80) complex both the NH 2 -and COOH-terminal domains have similar average R 2 values, (30.3 Ϯ 2.9 s Ϫ1 ) and (31.6 Ϯ 2.1 s Ϫ1 ), respectively ( Fig. 2A). A possible explanation of this phenomenon is that the NH 2 terminus of the nonphosphorylated form of cTnI also makes contacts with the NH 2 -terminal domain of cTnC.
The two domains of cTnC which are connected by a flexible linker tumble independently, with different rotational correlation times, in both the free protein and in a complex with cTnI-(1-80)DD. This results in different 15 N R 2 values for the two domains of cTnC. Binding of cTnI-(1-80) to both the NH 2and COOH-terminal domains of cTnC reduces the possibility of independent tumbling of the two domains, which results in more uniform rotational correlation times across the molecule and subsequently in similar 15 N transverse relaxation rates. Interaction of cTnI-(1-80) with both domains of Ca 2ϩ -saturated cTnC does not preclude flexibility in the linker region as is observed in the complex ( Fig. 2A). DISCUSSION The solution structure for cardiac troponin C unexpectedly revealed a closed conformation for the regulatory domain of cardiac TnC (8). In this regard, the Ca 2ϩ -saturated cardiac isoform of TnC differs from the Ca 2ϩ -saturated skeletal isoform, which is in the open state (7). The authors suggest that the functional form of Ca 2ϩ -saturated cTnC could be the closed form. However, based on relaxation studies performed on the isolated NH 2 -terminal domain of human cTnC, it has been suggested that the regulatory domain exists in two conformations: open and closed (9).
The results of our experiments demonstrate that chemical exchange does occur in the defunct Ca 2ϩ binding site I in both intact free cTnC and cTnC bound to the NH 2  Ca 2ϩ binding site I. Taken together, our studies suggest that when cTnI is phosphorylated, the regulatory domain of cTnC is destabilized, resulting in a decrease in Ca 2ϩ affinity. This would explain the enhanced relaxation observed when cTnI is phosphorylated by PKA in response to ␤-adrenergic stimulation of the heart (17). Binding of the nonphosphorylated form of cTnI does not prevent chemical exchange in inactive calcium binding site I of cTnC. We propose, based on chemical shift mapping, R 2 values, and the presence of multiple cross-peaks for certain residues in defunct Ca 2ϩ binding site I that nonphosphorylated cTnI interacts with both domains of cTnC and decreases chemical exchange presumably changing the dynamic equilibrium between open and closed conformations (8,18). The fact that the transverse relaxation rates for the NH 2 terminus of cTnC are not increased upon binding cTnI-(1-80)DD provides strong evidence that the phosphorylated mimetic of cTnI binds only to the COOH terminus of cTnC. This interpretation is in agreement with the fact that there are no significant amide proton and amide nitrogen chemical shift differences in the NH 2 -terminal domain of cTnC upon binding of cTnI-(1-80)DD. The crystal structure of sTnC complexed with a fragment of sTnI that corresponds to cardiac TnI-(33-80) clearly shows that sTnI interacts with both the NH 2 and COOH termini of TnC (19). Specifically, the A-helix of sTnC makes several contacts with the NH 2 -terminal portion of sTnI-(1-47) (19). We have shown recently that the solution structure of cTnC-(81-161) bound to cTnI-(33-80) is similar to that of sTnC bound to sTnI-(1-47). General agreement between sTnC and cTnC bound TnI structures suggest a common binding motif for the Ca 2ϩ /Mg 2ϩ -dependent interaction site in the TnI⅐TnC complex. It now appears that in the absence of phosphorylation, the Ca 2ϩ /Mg 2ϩ -dependent interaction site of cTnI makes additional contacts with the regulatory domain of cTnC similar to those interactions observed in the crystal structure for sTnC⅐sTnI-(1-47) (19). Data presented here are all consistent with this hypothesis. It is intriguing to speculate that loss of these interactions are responsible for a decrease in calcium affinity in Ca 2ϩ binding site II of cTnC when cTnI is phosphorylated. However, the exact sequence of events in this process is not understood and requires further investigation.
This study provides the first structural and dynamic insight on the mechanism by which PKA phosphorylation of cTnI, in response to ␤-adrenergic stimulation of the myocyte, produces a decrease in the calcium affinity of the regulatory site in cTnC. The proposed mechanism utilizes the unique cardiac isoform differences found in both TnC and TnI. Knowledge of this mechanism provides the basis for molecular approaches aimed at modifying cardiac muscle contraction.