Triple-Resonance Multidimensional NMR Study of Calmodulin Complexed with a Conformational Change in the Central Helix? the Binding Domain of Skeletal Muscle Myosin Light-Chain Kinase: Indication of

Heteronuclear 3D and 4D NMR experiments have been used to obtain 'H, 13C, and 15N backbone chemical shift assignments in Ca2+-loaded calmodulin complexed with a 26-residue synthetic peptide (M13) corresponding to the calmodulin-binding domain (residues 577-602) of rabbit skeletal muscle myosin light-chain kinase. Comparison of the chemical shift values with those observed in peptide-free calmodulin [Ikura, M., Kay, L. E., & Bax, A. (1990) Biochemistry 29,4659-46671 shows that binding of M13 peptide induces substantial chemical shift changes that are not localized in one particular region of the protein. The largest changes are found in the first helix of the Ca2+-binding site I (El 1-E14), the N-terminal portion of the central helix (M72-D78), and the second helix of the Ca2+-binding site IV (F141-M145). Analysis of backbone NOE connectivities indicates a change from a-helical to an extended conformation for residues 75-77 upon complexation with M13. This conformational change is supported by upfield changes in the Car and carbonyl chemical shifts of these residues relative to M13-free calmodulin and by hydrogen-exchange experiments that indicate that the amide protons of residues 75-82 are in fast exchange (kexch > 10 s-I at pH 7, 35 "C) with the solvent. No changes in secondary structure are observed for the first helix of site I or the C-terminal helix of site IV. Upon complexation with M13, a significant decrease in the amide exchange rate is observed for residues T110, L112, G113, and E l 14 at the end of the second helix of site 111. %e molecular recognition of intracellular enzymes and regulatory proteins by calmodulin (CaM)' plays a key role in coupling CaZ+ transients caused by a stimulus at the cell surface to events in the cytosol. The crystal structure of CaM shows an unusual dumbbell shape consisting of a long central helix and two globular homologous domains each containing two helix-loophelix calcium-binding motifs of the so-called "EF-hand" type (Babu et al., 1988; Kretsinger et al., 1986). Both domains have a hydrophobic patch that is considered capable of binding to several hydrophobic drugs that inhibit CaM activity. These patches presumably also form the binding site for the target protein. Binding studies with small naturally occurring biologically active peptides and synthetic analogues clearly indicate that CaM has a high affinity for positively charged amphiphilic a-helices (Anderson & Malencik, 1986; O'Neil & Degrado, 1990). The propensity to form an amphiphilic helix was also reported for the CaM-binding domains of several CaM-dependent enzymes, for which the amino acid sequence has been determined [for a review, see Blumenthal and Krebs (1988)l. The interaction between CaM and its target enzymes has been studied by a variety of chemical and physicochemical methods [see Klee (1988) for a review]. Small-angle X-ray scattering studies on the CaM complexes with mastoparan (Matsushima et al., 1989), with melittin (Kataoka et al., 1989), and with a 26 amino acid CaM-binding fragment of skeletal muscle myosin light-chain kinase (MLCK) known as M13 (Heidom et al., 1989) indicated that the two globular domains This work was supported by the AIDS directed anti-viral program of the Office of the Director of the National Institutes of Health. *To whom correspondence should be addressed. 3 Laboratory of Chemical Physics, NIDDK. 8 Laboratory of Biochemistry, NCI. in those complexes are much closer together than in the crystal structure of uncomplexed calmodulin, suggesting a disruption of the central-helix feature observed in the crystal structure. Jackson et al. (1986) showed that three of the lysine residues (K21, K75, and K148) are significantly protected from acetylation when CaM is complexed with MLCK. Manalan and Klee (1987) reported that calcineurin also protects K75 and K148 from acetylation. 'H and '13Cd NMR studies showed that when calmodulin binds target peptides, substantial chemical shift changes occur in the spectra of both domains (Klevit et al., 1985; Seeholzer et al., 1986; Linse et al., 1986; Ikura et al., 1989). However, no direct evidence has been obtained regarding the exact location where changes in the structure of calmodulin take place upon binding of target peptide. The most detailed NMR analysis of the complex of CaM with a MLCK fragment (Seeholzer & Wand, 1989) showed that despite substantial chemical shift changes in the two short antiparallel 6-sheet regions of the two globular domains, the secondary structure of these regions was preserved. Using novel heteronuclear triple-resonance ( lH,13C,lSN) three-dimensional NMR techniques (Kay et al., 1990b), we recently were able to make complete backbone 'H, 13C, and I5N assignments of CaM (M, -16.7 kDa). This protein was I Abbreviation: CaM, Calmodulin, HCACO, a-proton to a-carbon to carbonyl correlation; HCACON, a-proton to a-carbon to carbonyl to nitrogen correlation; HCA(CO)N, a-proton to a-carbon (via carbonyl) to nitrogen correlation; H(CA)NHN, a-proton (via a-carbon) to nitrogen to amide proton correlation; HMQC, heteronuclear multiple-quantum correlation; HNCO, amide proton to nitrogen to carbonyl correlation; HOHAHA, homonuclear Hartmann-Hahn; MLCK, myosin light-chain kinase; M13, a 26-residue fragment of the CaM-binding domain of MLCK comprising residues 577-602; NMR, nuclear magnetic resonance; TPPI, time-proportional phase incrementation; 2D, two dimensional; 3D, three dimensional; 4D, four dimensional. This article not subject to US. Copyright. Published 1991 by the American Chemical Society NMR of Calmodulin-MLCK Peptide Complex the first example for which backbone assignments were obtained with this approach (Ikura et al., 1990a). Here we report results of this new approach for studying the CaM-M13 complex. The larger molecular weight of the CaM-M13 complex (19.7 kDa) and the concomittant increase in line width forced us to use a slightly modified procedure from that described previously. In this new procedure the only experimental step that relied on poorly resolved lH-lH J coupling, the HOHAHA-HMQC experiment (Marion et al., 1989b), is replaced by a triple-resonance technique that relies entirely on single-bond heteronuclear couplings for correlating HN and H a chemical shifts (Kay et al., 1991a). In addition, a fourdimensional (4D) NMR experiment is used to facilitate automated assignment for spectra of the complexity as that obtained for CaM-M13. The present paper briefly describes the improved triple-resonance approach and reports the 'H, 13C, and ISN CaM backbone assignments for the CaM-M13 complex. A chemical shift comparison is made with values found for M13-free Ca2+-ligated CaM. In addition, NOES are reported for amide protons in the central helix part of CaM where substantial chemical shift changes occur, and these NOES are compared with patterns observed for uncomplexed CaM. EXPERIMENTAL PROCEDURES Sample Preparation. Drosophila CaM was over expressed by using the PAS vector in Escherichia coli (strain AR58) and was purified as reported previously (Ikura et al., 1990a,b). Samples used for the NOE and hydrogen-exchange experiments contained uniformly "N-labeled (-95%) CaM in H20 solution; all other experiments were carried out on 15Nand 13C-labeled (-95%) CaM, in either H 2 0 or D20 solution. Chemically synthesized HPLC-purified peptide M 13 (KRRWKKNFIAVSAANRFKKISSSGAL) (Peptide Technologies Corp. Washington, D.C.) was used without further purification. NMR samples were prepared according to the procedure outlined by Seeholzer and Wand (1989): 12 mg of decalcified CaM was dissolved in 4.5 mL of H 2 0 solution containing 0.01 M KCI and 0.68 mM CaC12, and the pH was adjusted to 6.8. The concentration of CaM (0.16 mM) in this solution was determined by UV spectroscopy with an extinction coefficient (&,) of 0.945 in the presence of Ca2+. M13 stock solution (0.18 mM) was added slowly to the CaM solution until the concentration ratio of M13 and CaM reached 1: 1. No precipitation was observed under these conditions. The solution was concentrated without freezing by using a speed vacuum concentrator to a total volume of 0.43 mL after which 0.02 mL of D 2 0 was added for a deuterium lock, and the pH was readjusted to 6.8. In the final NMR sample, the concentration of the complex was 1.5 mM. The D20 sample was prepared as described above, except that D 2 0 was used as solvent, with 7 mg of calmodulin, resulting in a final sample concentration of 1 .O mM. NMR Spectroscopy. With the exception of the 15N 3D NOESY -HMQC experiment, all NMR experiments were carried out at 36 OC on a Bruker AM-500 spectrometer, equipped with a triple-resonance probe head optimized for 'H detection, and home-built hardware (Kay et al., 1990b) to minimize overhead time and to generate the frequencies for the third channel. The pulse sequences used for the tripleresonance 3D and 4D experiments have been described previously (Ikura et al., 1990a; Kay et al., 1990b, 1991a,b). The HNCO and HNCA 3D spectra, recorded in H20, result from matrices comprising 32 complex X 64 complex X 1024 real data points. The spectral widths used in F1 and F3 were 1000 and 8064 Hz, respectively, with corresponding Biochemistry, Vol. 30, No. 22, 1991 5499 acquisition times of 32.0 and 63.5 ms. The F3 carrier was set at the water resonance position to ensure that zero-frequency artifacts fall outside the spectral region of interest. For HNCO, the F2 (C') spectral width was 1388.9 Hz (acquisition time 46.1 ms). For HNCA, the F2 (Ca) spectral width was 4310 Hz (acquisition time 14.8 ms). For the HCACO and HCA(C0)N 3D experiments, recorded in D20 solution, the acquired data matrices comprised 32 complex X 64 complex X 512 real data points. The spectral widths in F1 (Ca) and F3 (Ha) were 2994 and 5000 Hz, respectively, with corresponding acquisition times of 10.7 and 51.2 ms. For the HCACO experiment, the F2 (C') spectral width was 1389 Hz (acquisition time 46.1 ms) and 1000 Hz (acquisition time 32 ms) for the F2 (lSN) dimension of the HCA(C0)N experiment. The H(CA)NHN spectrum, recorded in HzO, resulted from a 64 complex X 32 complex X 1024 real data matrix with spectral widths of 2180 Hz (FI, Ha), 1000 Hz (F2, lSN), and 8250 Hz (F3, HN), and acquisition times of 29.4,32, and 62 ms, respectively. The total measuring time needed for recording all four 3D triple-resonance experiments was about 9 days. In the t l and t2 dimensions of all 3D experiments, complex data were acquired in a States-TPPI manner (Marion et al., 1989a). For the HNCO, HNCA, and H(CA)NHN, the length of the 15N time domain data was extended to 64 complex points by using linear prediction. For HCACO, the t2 domain (C') was extended to 128 complex points in the same manner. In addition, zero filling was employed in all dimensions prior to Fourier transformation. For the HNCO, HNCA, and H(CA)NHN experiments, the right half of the spectrum in the F3 dimension (upfield of H20), where there is no signal, was discarded after F3 Fourier transformation. The absorptive part of the final processed data matrix comprised 256 X 128 X 512 points for all 3D spectra. The HCACON 4D spectrum was recorded as described previously (Kay et al., 1991b). The size of the acquired data matrix was 32 complex X 8 complex X 8 complex X 512 real. The t2 and t3 data were extended to 16 complex points by using linear prediction. Zero filling was used in all dimensions. The spectral widths were 2994, 1388.9, 1000, and 5000 Hz in the F, , F2, F3, and F4 dimensions, respectively. Spin system identification was made for a substantial number of amino acids with the HCCH-COSY technique (Kay et al., 1990a), which correlates vicinal protons via a three-step 'H 13C, 13C 13C, 13C 'H magnetization transfer. The spectrum, recorded in D20, resulted from a 64 complex X 32 complex X 512 real data matrix, with acquisition times of 25.6 ms (Fl, lH), 10.7 ms (F2, 13C), and 51.2 ms (F3, lH). Linear prediction was used in the Fl dimension to double the length of the time domain data, and zero filling was used in all dimensions. A 3D lSN-separated 3D NOESY-HMQC experiment was carried out at 36 OC on a Bruker AM-600 spectrometer, with a pulse sequence that does not require presaturation of the H 2 0 resonance (Messerle et al., 1989). The size of the acquired data matrix was 128 complex X 16 complex X 512 real, with acquisition times of 21.3 ms (Fl, IH), 13.3 ms (F2, lSN), and 56.3 ms (F3, IH). The t2 time domain was extended to twice its original length by using linear prediction, and zero filling was used in all three dimensions prior to Fourier transformation. The 3D and 4D data were processed on a Sun Sparc-l workstation and on a Sun4 using simple in-house routines for Fourier transformation in the F2 dimension (Kay et al., 1989) and linear prediction together with the commercially available software package NMRZ (New Methods Research, Inc., Syr5500 Biochemistry, Vol. 30, No. 22, 1991 Ikura et al.