Interaction of calmodulin with its binding domain of rat cerebellar nitric oxide synthase. A multinuclear NMR study.

The intercellular messenger nitric oxide is produced through the action of nitric oxide synthases, a class of enzymes that is regulated by calcium-calmodulin (CaM). In this work, the interaction of CaM with a 23-amino-acid residue synthetic peptide, encompassing the CaM-binding domain of constitutive rat cerebellar nitric oxide synthase (cNOS), was investigated by various NMR methods. Cadmium-113 NMR studies showed that binding of the cNOS peptide increased the affinity of CaM for metal ions and induced interdomain cooperativity in metal ion binding as earlier observed for complexes of CaM with myosin light chain kinase (MLCK) peptides. By using specific isotopically labeled [13C]methyl-Met and selenomethionine-substituted CaM in two-dimensional proton-detected 13C and 77Se NMR studies, we obtained evidence for the involvement of the Met residues of CaM in the binding of the cNOS peptide. These residues form two hydrophobic surface areas on CaM, and they are also involved in the binding of other target proteins. A nitroxide spin-labeled version of the cNOS peptide caused broadening only for NMR resonances in the N-terminal half of CaM, showing that the peptide binds with a C to N orientation to the N- and C-terminal domains of CaM. pH titration experiments of CaM dimethylated with [13C]formaldehyde show that Lys-75 (and Lys-148) experience a large increase in pKa upon peptide binding; this indicates an unraveling of part of the helical linker region of CaM upon cNOS peptide binding. Taken together, our data show that the cNOS and MLCK peptides bind in a closely analogous fashion to CaM.

can regulate many different cellular events by interacting with almost 30 different target enzymes in a Ca 2ϩ -dependent manner (for a review, see Means et al., 1991;Vogel, 1994). The CaM-binding domains of the CaM targets are usually contained in a continuous stretch of about 20 amino acid residues. A large number of CaM-binding domains of various CaM targets have a tendency to form basic, amphiphilic ␣-helices, otherwise these peptides display very low amino acid sequence homology (O'Neil and DeGrado, 1990;James et al., 1995;Vogel and Zhang, 1995). The binding of two Ca 2ϩ ions to each domain of CaM induces significant conformational changes and exposes two Met-rich hydrophobic surfaces (Hiraoki and Vogel, 1987;Babu et al., 1988). The interaction of CaM with peptides encompassing the CaM-binding domains of various target enzymes occurs mainly via van der Waals interactions between these two hydrophobic surfaces and the hydrophobic face of the CaM-binding domain of its targets Meador et al., 1992Meador et al., , 1993. The Met residues in the hydrophobic surfaces play important roles by allowing CaM to bind to many targets in a sequence-independent manner (O'Neil and DeGrado, 1990;Vogel et al., 1990;Gellman, 1991;Vogel, 1994;Zhang and Vogel, 1994c;. The crystal structure of calcium-CaM reveals a dumbbell-shaped molecule, where the two globular domains of the protein are connected by a long, solvent-exposed ␣-helix (Babu et al., 1988). This central linker region is very flexible in solution Spera et al., 1991;Barbato et al., 1992), and its plasticity is one of the main reasons why CaM can interact effectively with such a wide spectrum of target proteins Meador et al., 1992Meador et al., , 1993. In complexes of CaM with its binding domains of MLCKs Roth et al., 1992;Meador et al., 1992) and CaM-dependent kinase II (Meador et al., 1993), the two globular domains of the protein essentially retain the same conformation, but in all cases the central helix of CaM unravels in a specific manner, allowing CaM to accommodate ␣-helical amphiphilic CaM-binding domains of different lengths.
Nitric oxide is a small gaseous messenger molecule which can regulate diverse physiological processes, such as smooth muscle relaxation, platelet inhibition, neurotransmission, immune regulation, and blood vessel dilation (for reviews, see Nathan, 1992;Feldman et al., 1993;Bredt and Snyder, 1994;Nathan and Xie, 1994). The cellular level of nitric oxide is regulated by the action of nitric oxide synthases (NOS). There are two distinct isoforms of NOS, a constitutively expressed brain and endothelial NOS (Bredt and Snyder, 1990;Bredt et al., 1991;Pollack et al., 1991;Sessa et al., 1992) and a cytokineinduced isoform from macrophages (Stuehr et al., 1991;Xie et al., 1992). The constitutively expressed NOS (cNOS) are activated by CaM in a Ca 2ϩ -dependent manner; the inducible NOS are not further activated by exogenous Ca 2ϩ -CaM, since they contain CaM as an integral, tightly bound subunit (Nathan and Xie, 1994). Synthetic peptides encompassing the putative CaMbinding domain of rat brain cNOS (Bredt et al., 1991) have been shown to bind to CaM in a Ca 2ϩ -dependent manner with a 1:1 stoichiometry and a K D of Ϸ2 nM (Vorherr et al., 1993;. From CD and isotope-edited FTIR studies, it is known that the cNOS peptide binds to CaM in an ␣-helical conformation . However, further details about the mode of binding and the location of the cNOS peptide binding site on CaM are lacking at present. In the present work, we have used 1D and 2D 1 H, 13 C, 77 Se, and 113 Cd NMR spectroscopy to study the interaction of the cNOS peptide with CaM. The involvement of the functionally important Met residues of CaM for binding the cNOS peptide was studied by 1 H, 13 C HMQC spectroscopy and 1 H, 77 Se HMBC spectroscopy. From the pK a titration behavior of Lys-75, we have obtained information about the unraveling of the central helix of CaM upon peptide binding . Cadmium-113 NMR spectra were acquired to study the effect of peptide binding on the metal-ion binding sites (Vogel and Forsén, 1987;Ikura et al., 1989). Finally, a paramagnetic nitroxide spin-labeled cNOS peptide was prepared to map the orientation of the cNOS peptide in its complex with CaM.

MATERIALS AND METHODS
All chemicals (analytical grade) were obtained commercially. The nitroxide spin labeling reagent 4-(2-bromoacetamido)-2,2,6,6-tetramethylpiperidine-N-oxyl was purchased from Sigma. The 23-residue cNOS peptide KRRAIGFKKLAEAVKFSAKLMGQ was obtained and handled as described previously . Preparations of wild-type CaM, CaM mutants, isotopically labeled 13 CH 3 -Met-CaM, and natural abundance 77 Se-Met-incorporated CaM have been described in detail in our earlier reports , 1994c. Reductive methylation of the Lys residues in CaM with 13 C-labeled formaldehyde was performed according to published procedures; this provides a convenient means of measuring individual pK a values in CaM . The CaM⅐peptide complexes used in this study were prepared in essentially the same manner as the complex of CaM with the CaM-binding domain of MLCK .
Spin Labeling of the cNOS Peptide-The spin-labeled cNOS peptide in which the single Met residue was specifically reacted was prepared following published procedures (Lundblad and Noyes, 1984;Musci et al., 1988). Briefly, the cNOS peptide was incubated with a 10-fold excess of 4-(2-bromoacetamido)-2,2,6,6-tetramethylpiperidine-N-oxyl for 24 h at room temperature in 0.2 M acetate buffer, pH 3.0. Under these conditions this reagent will react specifically with the sulfur atom of the single Met in the peptide (Musci et al., 1988). The labeled cNOS peptide (yield 50%) was separated from the unreacted cNOS peptide by reverse-phase high performance liquid chromatography; only a single peak was observed for the modified peptide indicating that no side reactions have taken place.
NMR Experiments-All 1 H, 13 C and 77 Se NMR spectra were acquired on a Bruker AMX500 NMR spectrometer equipped with a 5-mm broadband inverse detection probe at 25°C unless otherwise indicated. A total of 128 scans were collected for the 1D 1 H NMR spectra with a sweep width of 5000 Hz. 1 H, 13 C HMQC spectra were recorded in the phase-sensitive mode (Bax et al., 1983). The two-dimensional natural abundance 1 H, 77 Se HMBC spectrum of the SeMet-CaM⅐cNOS peptide complex was recorded in the magnitude mode (Zhang and Vogel, 1994c). All NMR spectra were processed on a Bruker X32 computer using commercially available Bruker UXNMR software.
Cadmium-113 NMR experiments were performed on a Bruker AM400 widebore NMR spectrometer, which was equipped with a 10-mm broadband probe. All spectra were recorded at a frequency of 88.75 MHz. Isotopically enriched 113 CdO (94.8%) was purchased from MSD Isotopes. A 100 mM stock solution of 113 Cd(ClO 4 ) 2 (pH Ϸ 1 in D 2 O) was used for the preparation and titration of the samples. All 113 Cd NMR samples were prepared in 25% (v/v) D 2 O, 100 mM KCl, 20 mM Tris, pH 7.5 Ϯ 0.05. For the 113 Cd 4 2ϩ -CaM sample titrated with cNOS peptide, microliter quantities of a concentrated peptide solution were added into the sample. The apo-CaM-cNOS peptide mixture was prepared by gradually adding cNOS peptide to a diluted apo-CaM solution until a 1:1 ratio was reached. The mixture was then lyophilized and redissolved in 2.0 ml of buffer. The 113 Cd NMR acquisition parameters are as follows: a 45°flip angle, acquisition time 0.14 s, relaxation delay 0.5 or 5 s, sweep width 30 kHz, 8 K data points, and 70,000 scans. The temperature of all experiments was 298 K. The total acquisition time for a 0.5 s delay experiment is 12 h, 40 min. Each FID was zero-filled once and an exponential line broadening of 30 Hz was applied prior to Fourier transformation. All 113 Cd spectra are referenced to external 100 mM 113 Cd (ClO 4 ) 2 in D 2 O.

Binding of the cNOS Peptide Induces Total Cooperativity in Metal Ion Binding-Cadmium-113 NMR titration experiments
showed that both domains of CaM are affected simultaneously by the binding of the cNOS peptide. It should be noted that only two signals are detected for 113 Cd 4 2ϩ -CaM in the absence of the peptide; these represent the two cations bound in the two C-terminal sites of CaM; those in the N-terminal part are not detected due to an exchange process that broadens both resonances beyond detection . Fig. 1 shows the spectra of 113 Cd 4 2ϩ -CaM titrated with the cNOS peptide; upon addition of the peptide a total of four resonances is detected, as observed earlier with other target-bound CaMs (see Table I). The spectra reveal that the cNOS peptide binds to 113 Cd 4 2ϩ -CaM in slow exchange on the 113 Cd NMR time scale. All of the four new resonances increase simultaneously in intensity with the concomitant disappearance of the two signals from 113 Cd 4 2ϩ -CaM. These four signals can be tentatively assigned to the four Cd 2ϩ ions bound to the four calcium-binding sites in the CaM-cNOS peptide complex, by comparison with values obtained for other complexes (see Table I). No minor signals are detectable in the spectra, suggesting that the peptide binds in only one orientation. Upon integration of the peaks, we found that the 1:1 complex only equals to 2 equivalents of Cd 2ϩ , as does the 113 Cd 4 2ϩ ⅐CaM complex. It is quite common in 113 Cd NMR that spectra are not fully relaxed if the relaxation delay used is short (Coleman, 1993). Therefore we acquired spectra with a 5 s delay for both 113 Cd 4 2ϩ -CaM and the 113 Cd 4 2ϩ ⅐CaM⅐ cNOS peptide complex at the same protein concentrations. The integration data obtained for these samples showed that the protein-peptide complex equals 3.7 equivalents of 113 Cd 2ϩ , which is close to the expected 4 equivalents; the integrated intensity of 113 Cd 4 2ϩ -CaM remains 2.0 equivalents indicating a significant change in the T 1 relaxation rate upon the binding of the peptide to the protein. Similar integration results have also been obtained for 113 Cd 4 2ϩ -CaM titrated with the skeletal mus- cle MLCK peptide (see Fig. 2).
We also performed a titration of the apo-CaM-cNOS peptide mixture with 113 Cd(ClO 4 ) 2 . Under these conditions, four 113 Cd NMR peaks increased simultaneously in intensity until a 4:1 ratio of metal ions to protein was reached (Fig. 3). The chemical shifts obtained in this experiment were consistent with those in Table I. The parallel increase of the four signals shows that CaM possesses total positive cooperativity in metal ion binding in the presence of the cNOS peptide. This is not evident in the absence of the peptide, but it can be induced by the binding of MLCK peptide, as well as other targets Ikura et al., 1989;Vogel, 1994). The line widths of the various 113 Cd resonances are listed in Table II. The cNOS Peptide Binds to the Met-rich Hydrophobic Surfaces of CaM-The two hydrophobic surfaces of CaM that are generally involved in the binding of the CaM-binding domains of target proteins are very rich in Met residues (Babu et al., 1988;O'Neil and DeGrado, 1990). Therefore, we studied the interaction between CaM and the cNOS peptide by heteronuclear NMR spectroscopic techniques using 13 CH 3 -Met isotopically labeled CaM (Siivari et al., 1995). Fig. 4, A and B, show the 1 H, 13 C HMQC spectra of the methyl groups of the Met residues from Ca 2ϩ ⅐CaM and the CaM⅐cNOS peptide complex. The assignment of the Met resonances were obtained by using individual Met 3 Leu mutants of the protein (for description see Siivari et al., 1995). With the exception of Met-76 which is part of the central linker of CaM, all the Met residues in the hydrophobic patches undergo significant chemical shift changes upon binding the cNOS peptide (see Table  III). Because of the sensitivity of the 77 Se chemical shift to changes in local environment (Luthra and Odom, 1986), the effects of complex formation on the Met residues in CaM were also studied by 77 Se NMR using CaM with Se-Met incorporated for Met (Zhang and Vogel, 1994c). Again, the binding of the cNOS peptide causes drastic chemical shift changes for the resonances of the 8 Se-Met residues from the two hydrophobic patches (see Table III). These results demonstrate that the two Met-rich hydrophobic regions of CaM are directly involved in the binding of the cNOS peptide.
The Conformation of the Central Helix of Ca 2ϩ -CaM Changes upon Binding to the cNOS Peptide-The possibility of conformational changes in the central helix/linker region of CaM upon binding of the cNOS peptide was evaluated by determining the pK a changes of Lys-75. It has been shown that the unraveling of the helical part of this region upon complex formation  is accompanied by a drastic increase in the pK a of this residue . Fig.  5 shows the 1 H, 13 C HMQC spectra of reductively dimethylated CaM and its K75R and K148Q mutants complexed with the cNOS peptide. From these data, it is immediately apparent that the missing resonances belong to Lys-75 and Lys-148, respectively. pH titration of the 13 C-methylated-CaM⅐cNOS   a C-terminal C III , C IV resonances (calcium binding sites III and IV) and N-terminal N A , N B resonances (sites I and II are not distinguished) have been assigned in these studies by using tryptic and thrombic proteolytic fragments of CaM. b NO, not observable due to exchange broadening (see Forsén et al., 1986).
peptide complex allowed us to obtain the pK a values of the individual Lys residues of CaM (see Table IV). Fig. 6 shows the pH titration curves of Lys-75 in Ca 2ϩ ⅐CaM and in the CaM⅐cNOS peptide complex. Lys-75 displays a significant pK a change from 9.29 in CaM to 9.94 in the complex. Similar to what was observed for the CaM⅐MLCK peptide complex, Lys-148 also shows a large pK a change upon complex formation with the cNOS peptide, whereas the other Lys residues only   Siivari et al. (1995). c Data were acquired at 30°C. d Data were acquired at 55°C. e Data obtained from Zhang and Vogel (1994c).

FIG. 5. 1 H, 13 C HMQC NMR spectra of 13 C-dimethylated wild type calmodulin (A) and the mutant K75R (B) and K148Q (C).
The procedure for isotopic dimethylation of CaM was described in detail elsewhere . experience minor changes.
The C-terminal Region of the cNOS Peptide Binds to the N-terminal Hydrophobic Surface of CaM-Introduction of nitroxide spin labels covalently attached to ligands provides a useful means of determining the binding orientation of ligands on proteins (see for example de Jong et al., 1988;Penington and Rule, 1992;Folkers et al., 1993;Girvin and Fillingame, 1995). We have utilized a Met nitroxide spin-labeled cNOS peptide in order to map out the orientation of the CaM-bound cNOS peptide with respect to the two domains of the protein. Fig. 7A shows the 1 H, 13 C HMQC spectrum of the 13 CH 3 -Met-labeled CaM complexed with the nitroxide spin-labeled cNOS peptide. Comparing Figs. 7A and 4B, it is clear that the four Met resonances from the C-terminal domains of CaM do not show any chemical shift and line width changes, whereas the Met resonances from the N-terminal domain undergo significant chemical shift as well as line width changes (the contour levels for the three visible N-terminal resonances were about half of those in the C-terminal domain). Most noticeably, a resonance tentatively assigned to Met-72 was broadened beyond detection by the spin label in the cNOS peptide. The chemical shift changes observed for Met-36, Met-51, and Met-71 result from the conformational perturbation induced by the relatively large hydrophobic spin labeling reagent attached to the cNOS peptide, as these persist following the reduction of the spin label (see below). In order to ascertain that the broadening of the resonances of the N-terminal Met residues in Fig. 7A results from the paramagnetic effect of the spin label, we have added ascorbic acid, which can abstract the single electron from the nitroxide (Kosen et al., 1986). Fig. 7B provides the spectrum of the complex recorded 24 h after the addition of ascorbic acid. The reappearance of the Met-72 and the increased intensity for the other N-terminal Met resonances demonstrates that the broadening effect results from the close proximity of the spin label (in panel 7B, the contour levels for Met-36, Met-51, and Met-71 are identical to those in the C-terminal domain). Hence, the C-terminal end of the cNOS peptide binds to the N-terminal hydrophobic surface of the protein.

DISCUSSION
The synthetic cNOS peptide used in this study has earlier been shown to contain the complete CaM-binding domain of rat brain cNOS (Vorherr et al., 1993;. NMR studies of the isolated cNOS peptide have also shown that it can adopt a nascent ␣-helical structure in aqueous solution that it forms a stable ␣-helix in aqueous trifluoroethanol (Zhang and Vogel, 1994;Vogel and Zhang, 1995) and that the ␣-helix has a typical amphiphilic structure as seen in many other CaM-binding peptides (O'Neil and DeGrado, 1990;James et al., 1995). Using CD and isotope-edited FTIR spectroscopy, we have demonstrated that the cNOS peptide also adopts an ␣-helical structure in its complex with CaM . Here, we have used 1 H, 13 C, 77 Se and 113 Cd NMR spectroscopic techniques to further characterize the details of the interaction between the cNOS peptide and CaM. It is clear from our NMR data that the cNOS peptide binds to both domains of calcium-CaM in slow exchange on the NMR time scale(s), since one observes distinct sets of resonances for the complex and free CaM. These data also show that the cNOS peptide forms a stable 1:1 complex with CaM, as the spectral changes in the titrations are completed once the amount of the cNOS peptide reaches this molar ratio.
Cadmium-113 NMR provides a useful method for the study of various metalloproteins including a range of calcium-binding proteins (for reviews see Vogel and Forsén, 1987;Coleman, 1993). For example, the binding of metal ions to CaM, its proteolytic fragments, and its complex with drugs and modeltarget CaM-binding domain peptides have been studied in this fashion (Andersson et al., 1983;Thulin et al., 1984;Linse et al., 1986;Ikura et al., 1989;Ohki et al., 1993). Here we have performed titration studies of the binding of the cNOS peptide to CaM by 113 Cd NMR. The four 113 Cd resonances in the CaM⅐cNOS peptide complexes could be tentatively assigned to the four Ca 2ϩ -binding sites of CaM, by comparison to earlier studies (see Table I). When the cNOS apo-  . CaM mixture was titrated with cadmium, four 113 Cd signals simultaneously increased in intensity, which showed that cooperativity in metal ion binding is induced between the N and C domains of CaM in the presence of the peptide. This is likely what will happen in vivo because CaM always exists in the presence of CaM-binding proteins in the living cell (Vogel, 1994). Integration of the 113 Cd NMR spectra indicated that the 113 Cd T 1 relaxation time in the complex is longer than in 113 Cd 4 2ϩ -CaM. Also the T 2 relaxation time for the complex appears to be longer than that in Cd 4 2ϩ -CaM from the obvious line width narrowing (⌬ 1/2 ) in the complex (Table II); however, it is important to realize that chemical exchange processes, rather than T 2 relaxation generally play a determining role for the line width of protein-bound Cd 2ϩ ions (Kördel et al., 1992;Aramini et al., 1995). Because chemical shift anisotropy contributes significantly to the 113 Cd relaxation of similar proteinbound ions at 9.4 Tesla (Kördel et al., 1992;Aramini et al., 1995), we can calculate the T 2 (and line width) values for Cd 4 2ϩ ⅐CaM and the Cd 4 2ϩ ⅐CaM⅐MLCK or cNOS complexes, assuming correlation times of 8.5 and 10 ns, respectively, at 25°C and using a chemical shift anisotropy value ⌬ ϭ 91 parts/ million, as determined for the homologous protein calbindin (Kördel et al., 1992). If we assume that the 113 Cd relaxation can be attributed completely to chemical shift anisotropy and dipole-dipole relaxation, we can calculate line widths of 0.87 and 1.02 Hz for CaM and CaM⅐peptide complexes, respectively. These values are obviously smaller than the line widths measured in our NMR spectra, suggesting that some chemical exchange processes are involved. Because the line widths of the free Cd 2ϩ signals did not change significantly upon peptide addition (Table II), we do not think that exchange between protein-bound and free Cd 2ϩ is involved. A possible explanation is that some fluctuations exist in the Ca 2ϩ -binding sites corresponding to slightly different Cd 2ϩ chemical shifts (Vogel and Forsén, 1987). Be that as it may, our 113 Cd NMR are consistent with the notion that the CaM⅐cNOS peptide complex, as well as the CaM⅐MLCK peptide complex, provide a more constrained Ca 2ϩ -binding environment than Ca 2ϩ -CaM.
The binding of CaM with its target peptides takes place mainly via van der Waals interactions between the Met-rich hydrophobic surfaces of CaM and the hydrophobic face of the CaM-target peptides Meador et al., 1992Meador et al., , 1993. In particular, the polarizability of the Met sulfur atoms can contribute significantly to the binding affinity (Gellman, 1991). Selectively 13 CH 3 -Met isotopically labeled and Se-Met containing CaM were used to study the involvement of the Met residues in the interaction with the cNOS peptide. The dramatic chemical shift changes observed for all of the Met and Se-Met residues in the two hydrophobic surfaces of CaM upon complex formation clearly indicate that the cNOS peptide also interacts with CaM through hydrophobic interactions. An analysis of the chemical shift changes of the Met 1 H, 13 C, and 77 Se resonances upon binding of the MLCK and cNOS peptides (see Table III) show that the changes induced by MLCK and cNOS are not identical. This behavior is expected, as the different amino acid sequence of these two target peptides would give rise to distinct bound conformations for the Met side chains of CaM.
The helical parts of the central linker region of CaM unravel further upon complex formation Meador et al., 1992Meador et al., , 1993. It is anticipated that the linker region of CaM will also unravel when CaM forms a complex with the cNOS peptide. We have shown earlier that the binding of a target peptide to the hydrophobic surfaces of CaM will displace the side chain of Lys-75 from the N-terminal hydrophobic surface; this process may induce the unraveling of the central ␣-helix (Vogel, 1994). This displacement causes a significant increase in its pK a . Here we have observed that the pK a of Lys-75 changes from 9.29 to 9.94 when CaM forms a complex with the cNOS peptide, and this increase is comparable to what was seen for MLCK , see also Table IV). This suggests that the helical parts of the linker region of CaM also undergo a conformational change upon the formation of the complex with the cNOS peptide. The pK a of Lys-148 also increases significantly when CaM binds to the cNOS peptide, and this again resembles the results with the MLCK peptide . By analogy to Lys-75, it is believed to be caused by a displacement of the Lys-148 side chain from an orientation on top of the C-terminal hydrophobic domain.
In principle, the cNOS peptide can bind to the protein in two orientations. Using a specific nitroxide spin-labeled cNOS peptide derivative, we have shown that the C-terminal part of the cNOS peptide binds to the N-terminal hydrophobic surface of CaM; consequently, the N-terminal portion of the peptide binds to the C-terminal domain of CaM. This orientation resembles the complexes of CaM with the CaM-binding domains of MLCK Meador et al., 1992), CaM-kinase II (Meador et al., 1993), and adenylate cyclase (Craescu et al., 1995). From a sequence alignment of the cNOS peptide with the CaMbinding domains of MLCK, we find that two anchoring hydrophobic residues are found with a spacing of 14 residues, analogous to MLCK, but distinct from CaM kinase II. Moreover, isotope-edited difference FTIR studies have shown that the two domains of CaM retain their secondary structure upon binding the cNOS peptide, as well as the MLCK peptide . Therefore, our data establish that the cNOS peptide binds to CaM in an analogous fashion as the MLCK peptide.