A Direct Test of the Reductionist Approach to Structural Studies of Calmodulin Activity

Ca2+-saturated calmodulin (CaM) directly associates with and activates CaM-dependent protein kinase I (CaMKI) through interactions with a short sequence in its regulatory domain. Using heteronuclear NMR13C-15N-1H correlation experiments, the backbone assignments were determined for CaM bound to a peptide (CaMKIp) corresponding to the CaM-binding sequence of CaMKI. A comparison of chemical shifts for free CaM with those of the CaM·CaMKIp complex indicate large differences throughout the CaM sequence. Using NMR techniques optimized for large proteins, backbone resonance assignments were also determined for CaM bound to the intact CaMKI enzyme. NMR spectra of CaM bound to either the CaMKI enzyme or peptide are virtually identical, indicating that calmodulin is structurally indistinguishable when complexed to the intact kinase or the peptide CaM-binding domain. Chemical shifts of CaM bound to a peptide (smMLCKp) corresponding to the calmodulin-binding domain of smooth muscle myosin light chain kinase are also compared with the CaM·CaMKI complexes. Chemical shifts can differentiate one complex from another, as well as bound versus free states of CaM. In this context, the observed similarity between CaM·CaMKI enzyme and peptide complexes is striking, indicating that the peptide is an excellent mimetic for interaction of calmodulin with the CaMKI enzyme.

Calmodulin is the prototypic transducer of the Ca 2ϩ ion second messenger, present in all eukaryotes where it modulates numerous intracellular proteins, including a number of Ser/Thr protein kinases (1,2). An elevation of intracellular Ca 2ϩ concentrations results in saturation of calmodulin by calcium, inducing conformational changes in both N-and Cdomains (3)(4)(5)(6). In this activated state, Ca 2ϩ -saturated calmodulin (CaM) 1 binds to target enzymes, modulating their activity. The Ca 2ϩ -dependent association is well studied, yet a mechanistic description of CaM function is confounded by the apparently diverse target sequences and numerous activities that are regulated by this protein (1,(7)(8)(9).
The structural paradigm of CaM signal transduction has been formed from a number of crystallographic and NMR studies, typically in the context of CaM binding to biochemically inactive peptide fragments from larger proteins (1,9). This has been necessitated by the fact that the smallest intact enzymes that are regulated in CaM-dependent pathways are generally too large for standard NMR experiments and have been apparently resistant to crystallization efforts, forcing reliance on studies of the interaction of CaM with peptide fragments of calmodulin-binding domains. Indeed, peptide models are often used in many different types of studies of calmodulin-target protein interactions (2).
Commonly, CaM binds a single amphiphilic peptide, which forms an ␣-helix upon being sequestered from solvent via collapse of the two Ca 2ϩ -binding domains of CaM around the peptide (10 -15). Recently, structures of several large CaM⅐protein complexes have been determined (16 -18), each suggesting that extrahelical interactions are available to CaM. In a study of a small conductance Ca 2ϩ -activated K ϩ channel (SK2 channel) by Schumacher et al. (16), two molecules of CaM each span three helical segments in the tetrameric CaM⅐SK2 channel peptide complex. Similarly, Larsson et al. (17) found a multimeric complex between CaM and the transcription factor SEF2-1/E2-2. CaM has also been crystallized in association with a domain of Edema factor (EF) exotoxin from Bacillus anthracis (18). Remarkably, however, it is CaM that is engulfed by the EF domain in the ternary complex, with CaM bound as a new subdomain in the EF protein via "a molecular full nelson" (19). These examples indicate that CaM may be capable of numerous modes of interaction, further supporting a model of a conformationally adaptive protein. These studies suggest that CaM may utilize substantially larger surfaces in activating target enzymes than is indicated by CaM⅐peptide complexes. Thus a means for verifying the relevance of using a CaMpeptide mimetic in studies of the interaction between CaM and intact enzymes is needed.
CaM-dependent protein kinase I (CaMKI) is part of a CaMdependent kinase signaling cascade, comprised of CaMKK, CaMKI, and CaMKIV (20 -22). A crystal structure of CaMKI(320) has provided insights into the structural basis for CaM-mediated activation of these and related enzymes (23). * This work was supported by National Institutes of Health (NIH) Grants DK39806 and GM50402 and by equipment grants from the NIH and the Army Research Office. 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. Many kinases are maintained in an autoinhibited state via interactions of the C-terminal regulatory domain with the catalytic core (24,25). In CaMKI, the core catalytic domain (1-298) is flanked by a regulatory domain (299 -320) that is comprised of overlapping pseudosubstrate and CaM-binding sequences. The regulatory domain inhibits enzyme activity through contacts of the pseudosubstrate sequence and the substrate-binding site and through contacts between the CaMbinding sequence and ATP-binding sites, thus providing an autoinhibitory mechanism (24,25). The interaction between CaM and the CaM-binding domain sequesters the regulatory domain, releasing pseudosubstrate inhibition, and resulting in activation of the kinase.
In this work, we assay to what extent complexes of CaM with CaMKI(320) and with CaMKIp are similar at the level of the backbone structure of CaM. Studies comparing interactions of CaM with intact CaMKI versus a synthetic CaMKIp peptide (residues 299 -320) show largely similar spectroscopic changes, and a similar responsiveness to mutations (26,27), providing support for the use of peptides as models of CaM interactions. Here, the extraordinary sensitivity of the NMR chemical shift is used to report on the subtle structural changes that occur throughout the entire calmodulin molecule upon interaction with a target peptide or protein. The observed chemical shifts show that CaM⅐CaMKIp and CaM⅐CaMKI(320) are virtually identical in terms of backbone structure.

Protein Purification and Sample Preparation-Recombinant chicken
CaM was prepared as described previously (28). For overexpression of 13 C, 15 N-labeled CaM, cells were grown on minimal medium containing 0.1% 15 NH 4 Cl with 0.2% (U-13 C 6 -99%) D-glucose (Isotec). For overexpression of [ 2 H, 15  The DNA encoding CaMKI(320) was subcloned from a previously described fusion protein expression system (23). The CaMKI(320) DNA was inserted after a hexahistidine N-terminal leader coding sequence in the pET-15b vector (Novagen) utilizing NdeI and BamHI restriction sites, using standard methods. Overexpression of recombinant CaMKI(320) in Escherichia coli BL21(DE3) cells was similar to that described previously (23), with all growth performed on minimal medium. After 4-h isopropyl ␤-D-thiogalactopyranoside induction, cells were harvested, lysed by sonication, and clarified by centrifugation. The His-tagged kinase was isolated on a nickel-nitrilotriacetic acid His-bind column (Novagen) and used without cleavage of the His-tag. CaM and CaMKI(320) preparations are typically Ͼ98% pure as judged by SDS-PAGE.
The CaMKIp peptide sequence, Ac-AKSKWKQAFNATAGGRN-MRKLQ-NH 2 , was produced by solid state synthesis (PerSeptive Biosystems Pioneer Synthesizer) using the Fmoc (N-(9-fluorenyl)methoxycarbonyl)/t-butyl protection strategy (29) at 0.1-mmol scale. Single coupling cycles with OPfp/HOBt chemistry were used for all amino acids. The N terminus was manually acetylated. Crude peptide was deprotected, precipitated, and washed with cold ether, followed by dissolution in H 2 O and reversed-phase C 18 high performance liquid chromatography purification. Purity of CaMKIp peptide was Ͼ98% as judged by matrix-assisted laser desorption ionization time-of-flight mass spectroscopy.
The complex between CaM and either CaMKIp or CaMKI(320) was formed under dilute solution conditions, typically below 200 M protein, by titration of CaM into a 15-25% molar excess of kinase or peptide to ensure that CaM is maintained entirely in the bound state. Buffer conditions were 100 mM KCl, 20 mM bis-Tris, 20 mM CaCl 2 , 1 mM dithiothreitol (in [ 2 H, 15 N]CaM⅐CaMKI(320) sample only), 0.02% NaN 3 , pH of 6.5, and 7% D 2 O.
NMR Spectroscopy-NMR spectra were recorded on a 750 MHz ( 1 H) Varian Inova spectrometer using a Varian 5-mm triple resonance probe (z-axis pulse field gradients) at 303 K. Proton chemical shifts were referenced to an external standard of 2,2-dimethyl-2-silapentane-5-sulfonate in D 2 O. 13 C and 15 N chemical shifts were determined via the ratio method (30). For all experiments, the 1 H carrier frequency was set to the water resonance. Complete backbone (H N , N, C a , C b , and CЈ) resonance assignments of [ 13 C, 15

CaM Binding to CaMKI Enzyme and Peptide-Previous
studies have shown the interaction of CaM with CaMKI protein and peptides requires Ca 2ϩ , involves sequestration of Trp 303 from solvent, and binding of CaM to a helical peptide (23,24,26,27). Backbone resonance assignments of Ca 2ϩ -saturated [ 13 C, 15 N]CaM bound to CaMKIp peptide (Fig. 1) were completed using standard triple-resonance assignment strategies. Briefly, using three-dimensional amide-detected HNCACB, CBCA(CO)NH, HNCO, and HN(CA)CO spectra, backbone assignments were determined for all CaM residues except for the first two residues. Titration of labeled CaM with CaMKIp indicates a 1:1 stoichiometry that is in slow exchange (Fig. 2) consistent with high affinity binding. The residue-specific chemical shift differences between CaM⅐CaMKIp and free CaM are suggestive of substantial structural changes upon peptide association (Fig. 3A), consistent with chemical shift changes and structural rearrangements observed in formation of other CaM⅐peptide complexes (10 -15, 42).
Studies of "intact" CaMKI have utilized a truncated version of the enzyme (1-320) that results from proteolytic cleavage during expression and purification of the full-length enzyme (23). This truncated form is fully autoinhibited, includes the complete consensus CaM-binding sequence (9), displays CaMdependent activation (24), and is the same form of the protein as was utilized in crystallization studies (23). A standard 15 N HSQC spectrum of the CaM⅐CaMKI(320) complex showed significant line-broadening of CaM resonances (data not shown), requiring the use of NMR strategies suitable for studies of large (Ͼ35 kDa) proteins. The TROSY-HSQC (38) spectrum of [ 2 H, 15 N]CaM⅐CaMKI(320) is virtually identical to that of the CaM⅐CaMKIp complex (Fig. 1), allowing complete cross-assignment of the CaM⅐CaMKI(320) spectrum by reference to the assigned spectrum of the CaM⅐CaMKIp complex. The chemical shift differences between these two are on the order of the precision of individual chemical shifts, clearly denoting that CaM adopts virtually identical structures in association with either the CaMKI enzyme or peptide (Fig. 3D). Substantial line-broadening is observed in HSQC spectra of the CaM⅐CaMKI(320) complex, which indicates the hydrodynamic properties of CaM are influenced by association with the comparatively large CaMKI enzyme. However, from the perspective of backbone chemical shifts, there is no difference in the backbone structure of CaM when bound to the 35-kDa enzyme or when attached to the 22-amino acid peptide model.
Comparison of CaM⅐CaMKI with CaM⅐smMLCKp-How unique is CaMKI sequence in promoting changes in the structure of CaM? Is the similarity between CaM⅐CaMKIp and CaM⅐CaMKI(320) expected for all CaM⅐enzyme complexes? To address these issues, we compared the effect of 1 H and 15 N chemical shift changes in CaM upon binding two different recognition sequences, CaMKIp and a distantly homologous CaM-binding peptide from smooth muscle myosin light chain kinase (smMLCKp). Both interactions are known to require Ca 2ϩ and are known to interact with similar sites on CaM (12,26,27). A comparison of CaM⅐CaMKIp relative to free CaM (Fig. 3A) with chemical shift changes in CaM⅐smMLCKp relative to free CaM (Fig. 3B) indicates that binding to either peptide produces effects of significant magnitude throughout CaM but with a quite different distribution. A comparison of the crystal structures of free CaM and CaM⅐smMLCKp shows the change in backbone conformation is small within either N-terminal (0.43-Å r.m.s. deviation) or C-terminal domain (0.55-Å r.m.s. deviation). 2 Despite small changes in structure, substantial chemical shift changes are induced upon peptide binding; CaM chemical shifts are extremely sensitive to the small conformational changes that take place upon complexation. A comparison of amide chemical shifts from CaM⅐CaMKIp and CaM⅐smMLCKp shows substantial effects over the entire sequence of CaM (Fig. 3C), with larger chemical shift differences for residues in the central helix, indicating that CaM can differentiate between the substrates. In this context, the lack of differences in the chemical shifts of CaM in the CaM⅐CaMKIp and CaM⅐CaMKI(320) enzyme complexes are even more striking.
In light of the intrinsic sensitivity of backbone chemical shifts to perturbations in the structure of CaM, the equivalence of 1 H and 15 N chemical shifts of CaM⅐CaMKIp and those of CaM⅐CaMKI(320) indicate no differences in the structure of CaM (Fig. 3D) bound to these two substrates. While our results do not presently address changes induced in the structure of the intact CaMKI enzyme upon CaM association, the data are consistent with a model whereby interactions between CaM and CaMKI are limited to the CaM recognition sequence in the regulatory domain. While this is likely also true for some CaMregulated proteins, there is evidence from other systems that the mechanism of CaM activation can be more complex than could be extrapolated from peptide binding studies (18,43,44  However, comparisons of binding energies of CaM to either peptides or enzymes may not take into account possible interactions between the regulatory domain and the catalytic domain that must be disrupted upon CaM binding. Here we have demonstrated the feasibility of directly examining this issue in large protein complexes using TROSY-based NMR methods. Furthermore, the work presented here strengthens the model for CaM-dependent activation of an autoinhibited CaMKI enzyme, relieving allosteric inhibition by sequestering the regulatory domain, preventing its interaction with both the substrate and ATP binding domains. In this activated state, CaM is a silent partner unaware of the presence of an active enzyme tethered to the CaM recognition sequence.