A peptide model for calmodulin trapping by calcium/calmodulin-dependent protein kinase II.

Autophosphorylation of Ca2+/calmodulin-dependent protein kinase II (CaM-kinase) induces a more than 1000-fold increase in calmodulin (CaM)-binding affinity by dramatically decreasing the off-rate for CaM. In this report, we investigate the molecular mechanism for this phenomenon by comparing the rate of dissociation of a novel fluorescently labeled CaM from two synthetic peptides and from the phosphorylated and nonphosphorylated forms of a recombinant preparation of CaM-kinase. Dissociation of a complex of CaM and CKII(296-312), a peptide representing close to the minimum CaM-binding domain of the α subunit of CaM-kinase, exhibited a fast off-rate of 5.0 s−1. This was similar to the off-rate of 1.1 s−1 for the dissociation of CaM from the nonphosphorylated form of CaM-kinase. In contrast, dissociation of CaM from either autophosphorylated CaM-kinase or peptide CKII(290-314) was extremely slow with apparent off-rates of about 3-9 × 10−5 s−1. Along with information from the crystal structure of Ca2+/CaM bound to CKII(290-314) (Meador, W. E., Means, A. R., and Quiocho, F. A. (1993) Science 262, 1718-1721), our results suggest a model in which CaM-dependent autophosphorylation of CaM-kinase induces a conformational change in the region of the CaM-binding domain which allows the formation of additional stabilizing interactions with CaM. We predict that this involves amino acids 293-298 in CaM-kinase. The possible consequences of these observations on the reversibility of CaM trapping in native CaM-kinase are discussed.

gene transcription (for review, see Ref. 1). CaM-kinase is activated by binding to the Ca 2ϩ bound form of calmodulin, which dramatically increases the affinity of the enzyme for Mg 2ϩ / ATP, thus leading to substrate phosphorylation and phosphorylation of the enzyme itself (autophosphorylation).
CaM-kinase purified from rat forebrain is a holoenzyme composed of 8 -12 ␣ and ␤ subunits found in an approximate ratio of 3:1, respectively. The formation of holoenzymes is essential for rapid subunit autophosphorylation (2,3). Both inter-and intrasubunit phosphorylation occurs within a holoenzyme, both leading to distinct functional changes in enzyme behavior. Intersubunit autophosphorylation includes a site identified as Thr 286 in the ␣ subunit (2,3), while intrasubunit autophosphorylation occurs at Thr 305 and/or Thr 306 in the center of the CaM-binding domain (2,3). The functional consequences of CaM-kinase autophosphorylation are well studied and include generation of Ca 2ϩ /calmodulin-independent activity (4 -8) and inactivation and/or CaM insensitivity (9 -13). Recently, Meyer et al. (14) discovered that autophosphorylation of CaM-kinase at Thr 286 dramatically increases the affinity of the enzyme by decreasing the rate of dissociation of CaM by more than 3 orders of magnitude. This phenomenon is referred to as CaM trapping.
Two possible biological roles have been proposed for CaM trapping by CaM-kinase (14). The first involves an intrinsic effect on CaM-kinase activity as formalized by Hanson et al. (2). This model predicts that at a certain threshold, increased frequencies of Ca 2ϩ oscillations will produce the autophosphorylated form of CaM-kinase and thus transform the enzyme into the CaM-trapped state. Since CaM then remains bound to the enzyme for longer times, the probability of more subunits entering the trapped state is increased, sustaining a greater proportion of CaM-kinase subunits in the fully active state. A second possible biological role for CaM trapping would be to modulate the availability of free CaM in discrete subcellular locations. CaM-kinase is very abundant in brain, representing as much as 2% of total hippocampal protein (1), and it is likely that its subcellular distribution may be in excess of CaM in certain regions of the neuron where CaM-kinase is postulated to be particularly concentrated, such as postsynaptic densities (15,16). Conversion of CaM-kinase from a low to a high affinity CaM-binding protein has the potential to redistribute CaM in favor of binding to CaM-kinase during a period of high frequency Ca 2ϩ oscillations and thus decrease the activity of a variety of other CaM-dependent enzymes.
Studies using synthetic peptides (17)(18)(19)(20)(21) as well as the crystal structure of CaM bound to a synthetic CaM-kinase peptide (22) provide information concerning the structural basis for enzyme regulation and CaM binding. Amino acids 281-301 in CaM-kinase encompass two inhibitory domains, one (surrounding Thr 286 ) that binds to the ATP binding site and another (surrounding Lys-291) that binds to the substrate binding site, thus inhibiting the enzyme in the absence of Ca 2ϩ /CaM. The minimum CaM-binding domain was defined as amino acids 296 -309 of the ␣ subunit (19). However, the crystal structure of Ca 2ϩ /CaM bound to a synthetic peptide spanning amino acids 290 -314 of CaM-kinase (22) shows that amino acid residues 293-310 adopt an ␣-helical structure and form stabilizing contacts with Ca 2ϩ /CaM. This suggests that the minimum CaMbinding domain (residues 296 -309) does not encompass all amino acids with the potential to interact with CaM.
Proteolysis studies showed that autophosphorylation of Thr 286 induces a conformational change in CaM-kinase that leads to exposure of previously cryptic proteolytic sites (23,24). This observation, coupled with the peptide studies described above, lead us to hypothesize that autophosphorylation of Thr 286 exposes residues on the enzyme that make stabilizing contacts with CaM leading to the CaM-trapped state. If this is true, then phosphorylation may only be required for high affinity binding of CaM within the context of native CaM-kinase, and synthetic peptides delineating different regions of the CaM-binding domain could potentially represent the untrapped and trapped forms of the enzyme. In this report, we identify two synthetic peptides that appear to kinetically mimic the untrapped and trapped forms of CaM binding to CaMkinase, neither of which include the autophosphorylation sequence surrounding Thr 286 .

EXPERIMENTAL PROCEDURES
Expression, Purification, and Mutagenesis of Calmodulin-A bacterial expression plasmid for CaM which uses the phage P L promoter was generated by subcloning a BglII/NcoI fragment from pOTSNcoI2 (25) into the plasmid pCaM23N (26) which had been digested with BamHI/NcoI. Introduction of a single Cys residue by conversion of Lys at amino acid 75 to Cys to produce CaM(C75) was accomplished by polymerase chain reaction splicing by overlap extension (27). The presence of the desired mutation and the absence of polymerase chain reaction artifacts was confirmed by DNA sequencing. Cells expressing CaM were grown and lysed, and a soluble fraction was prepared as described previously (28). After dialysis against 50 mM Tris, pH 7.5, 0.2 mM EGTA, the soluble fraction was applied to a Macro-Prep Q 2.5 ϫ 30-cm column (Bio-Rad), washed with the same buffer, and then eluted with a linear 0 -500 mM KCl gradient. The appropriate fractions were pooled, made 1 mM in CaCl 2 , and loaded on a phenyl-Sepharose fast flow high-sub 2.5 ϫ 30-cm column (Pharmacia Biotech Inc.). The column was washed with 50 mM Tris, pH 7.5, and 0.2 mM CaCl 2 and then eluted with 50 mM Tris, pH 7.5, and 0.2 mM EGTA. If necessary, CaM was further purified by anion exchange chromatography using a semiprep HPLC DEAE-5PW column (Waters).
Labeling of CaM(C75) with IAEDANS-IAEDANS (Molecular Probes) was dissolved in N,NЈ-dimethyl formamide at concentration of 100 mM and stored at Ϫ20°C. CaM(C75) at a concentration of 2-2.5 mg/ml was made 5 mM in dithiothreitol and then desalted into 50 mM MOPS, pH 7.5, and 0.5 mM CaCl 2 with or without 6 M urea. IAEDANS was added in 2-fold molar excess over protein and allowed to react overnight at room temperature, followed by exhaustive dialysis against 50 mM MOPS, pH 7.0. Final protein concentrations were determined using the BCA assay (Pierce) with CaM as a standard, and the amount of bound probe was determined by absorbance spectroscopy using a molar extinction coefficient at 336 nm of 5700. Probe-to-protein ratios were 0.9 and 1.0 for proteins labeled in the absence and presence of urea, respectively.
Expression and Purification of the ␣ Subunit of CaM-kinase-The cDNA encoding the rat ␣ subunit of CaM-kinase lacking all of the 5Ј untranslated sequence was inserted into the BakPak-9 baculovirus expression vector (Clonetech). Recombinant virus was produced in monolayers of Sf21 cells exactly as described by the manufacturer. Cloned virus stocks were used to produce enzyme by infecting Sf21 cells at a multiplicity of infection of between 2-10. After 72 h, the cells were harvested, washed once with phosphate-buffered saline, and stored at Ϫ80°C. CaM-kinase was purified from the cells utilizing a protocol adapted from Brickey et al. (29). Cells were lysed in 10 volumes of 40 mM HEPES, pH 7.5, 5% betaine, 1 mM dithiothreitol, 1 mM EDTA, 1 mM EGTA, 5 mg/liter soybean trypsin inhibitor, and 20 mg/liter leupeptin with 10 strokes of a glass/Teflon homogenizer and clarified by centrifugation at 25,000 ϫ g for 30 min and then at 100,000 ϫ g for 1 h at 4°C.
The proteins were precipitated from the supernatant with ammonium sulfate (50% final), resuspended in 40 mM HEPES, pH 7.5, with 2 mM CaCl 2 and 10% glycerol and protease inhibitors and applied to a CaM-Sepharose column. The specifically bound proteins were eluted with 40 mM HEPES, pH 7.5, containing 2 mM EGTA, 0.6 M NaCl, 10% glycerol and protease inhibitors, again ammonium sulfate-precipitated (35% final), and the pellets following centrifugation were solubilized in 40 mM HEPES, pH 7.5, 10% ethylene glycol, 0.1% Tween 40, 0.2 M NaCl, and protease inhibitors. The solubilized proteins were layered onto a 5-24% linear sucrose gradient made in the same buffer and centrifuged for 15 h at 225,000 ϫ g at 4°C. Gradients were fractionated and the peak of enzyme activity was determined as described previously (30). Finally, the enzyme was concentrated from solution by binding to phosphocellulose resin, and the kinase eluted by the addition of 0.5 M NaCl was dialyzed into storage buffer (10 mM HEPES, pH 7.4, 0.1 mM EGTA, 100 mM KCl, and 50% glycerol) and frozen as aliquots at Ϫ80°C.
Synthesis and Purification of Peptides-The peptide CKII(290 -314) was a generous gift from Dr. Florante Quiocho and William Meador (Baylor College of Medicine). Peptide CKII(296 -312) was synthesized by the Analytical Chemistry Center at the University of Texas Medical School at Houston. All peptides were purified by high performance liquid chromatography and analyzed by mass spectroscopy. Stopped-flow measurements were accomplished using an Applied Photophysics Ltd. (Leatherhead, UK) Model SF.17 MV sequential stopped-flow spectrofluorimeter with a dead time of 1.6 ms. Excitation was at 340 nm with 5-nm slit widths, and emitted light was collected using a 390-nm cut-off filter. Samples were excited using a 150-watt xenon lamp. Off-rates were determined by rapid mixing (17 l/ms) of equal volumes of solutions from two syringes. Typically, one syringe contained the standard buffer with CaM(C75) IAE (0.6 M) and sufficient peptide or enzyme to achieve maximal fluorescence intensity. The second syringe contained excess unlabeled CaM (30 M) in the same buffer. All data were fit to a single exponential function. All fluorescence measurements were made at room temperature.

Characterization of Mutated and Labeled Calmodulin-
The cDNA for chicken calmodulin was mutated to introduce a single Cys residue for Lys at amino acid 75. A comparison of the crystal and NMR solution structures of free CaM (31) and CaM bound to a peptide derived from CaM-kinase II (22) indicated that the side chain of Lys 75 does not directly participate in peptide binding but would undergo a change in its environment upon formation of the peptide⅐CaM complex. Presumably, a probe bound to Cys at position 75 in CaM would not drastically influence the interaction of CaM with a target peptide or protein but would exhibit a change in fluorescence characteristics upon binding ligands. Fig. 1 shows that there is no discernible difference in the abilities of bovine brain CaM, recombinant CaM synthesized in bacteria, and CaM(C75) IAE to activate CaM-kinase.
Effects of Different Peptides and CaM-kinase on Fluorescence of CaM(C75) IAE --Two synthetic peptides were used in the current study ( Fig. 2A). A crystal structure of the longer peptide, CKII(290 -314), bound to CaM was reported by Meador et al. (22). The underlined residues in CKII(290 -314) are those which were resolved in the crystal structure. The shorter peptide, CKII(296 -312), is based on the minimal CaM-binding peptide (19). The C-terminal extension of Gly-Cys on CKII(296 -312) was included to provide a flexible site for covalent attachment of various sulfhydryl-specific probes for future studies. Fig. 2B shows the steady-state emission spectra for CaM(C75) IAE in the absence (ϪCa 2ϩ ) and presence (ϩCa 2ϩ ) of calcium. Ca 2ϩ binding produced a 20 -30% increase in the fluorescence intensity and a slight shift of the emission maxima to a shorter wavelength. Addition of CKII(296 -312) to Ca 2ϩ / CaM resulted in an additional, larger increase in fluorescence intensity and a further blue shift in emission maxima. The emission maxima of 465 nm in the presence of the peptide was selected for subsequent experiments. Fig. 3 compares the effects of sequential additions of Ca 2ϩ and either CaM-kinase peptides or CaM-kinase on the fluorescence intensity from CaM(C75) IAE . A typical increase in fluorescence intensity is detected when Ca 2ϩ is added to CaM(C75) IAE . Addition of CKII(296 -312) produces an additional increase in fluorescence similar to that shown in Fig. 2. A more robust increase in fluorescence intensity is produced by addition of CKII(290 -314). This difference in fluorescence intensities induced by the two peptides suggests distinct conformations of CaM in the respective complexes. Peptide-induced increases in fluorescence intensity are inhibited by more than 97% if 50-fold excess unlabeled CaM is added prior to the addition of CKII(290 -314). This would be anticipated if the affinities of CaM and CaM(C75) IAE for peptides are similar as suggested by Fig. 1.
Addition of CaM-kinase (3.5 g) also produces an increase in fluorescence intensity, although the absolute increase is smaller relative to either of the peptides. This increase in fluorescence intensity produced by the addition of CaM-kinase is similar to that reported by Meyer et al. (14). They further showed that the increase in fluorescence intensity was paralleled by an increase in fluorescence anisotropy. Addition of 1 mM Mg 2ϩ /ATP to the cuvette to allow autophosphorylation of the enzyme produced a second and larger increase in fluorescence, which is not seen when 1 mM Mg 2ϩ is added to the cuvette (data not shown). The observed increase in fluorescence intensity is likely due to a combination of nucleotide binding and autophosphorylation of the enzyme and reached a stable plateau fluorescence increase within the few seconds required to open the fluorimeter, mix the sample manually, close the fluorimeter, and resume data collection. Similar to the two peptides, CaM appears to adopt distinct conformations when bound to the autophosphorylated and nonautophosphorylated enzyme. Interestingly, the final fluorescence intensity is equivalent for CaM(C75) IAE bound to either CKII(290 -314) or to autophosphorylated CaM-kinase. However, in interpreting this data, it is important to note that the similarities attained in intensity changes between the peptide⅐CaM(C75) IAE complex and enzyme⅐CaM(C75) IAE complex do not necessarily indicate that they adopt the same conformations.  10 (N(logCaM50ϪlogCaM)) ), where A is the activity at a given CaM concentration, A max is the maximal activity, N is the Hill coefficient, and CaM 50 is the concentration of CaM which yields 50% of maximal activity. The insets indicate the log CaM 50 and Hill coefficients for each protein with percent errors derived from the computer fit.

FIG. 2. Peptide sequences and characterization of changes in emission spectra of CaM(C75) IAE by binding Ca 2؉ and peptide.
A, sequence of the two synthetic peptides used in the current study. In B, CaM(C75) IAE (1 g, 0.06 M final concentration) was added to 1 ml of 25 mM MOPS, pH 7.0, 150 mM KCl, 0.1 mM EGTA, and 0.1 mg/ml BSA and an emission spectra obtained between 400 and 600 nm (ϪCa 2ϩ ). CaCl 2 (0.5 mM final concentration) was added and a second scan was obtained (ϩCa 2ϩ ). Finally, an excess of CKII(296 -312) was added (5 M final concentration) and a third scan was taken (ϩCa 2ϩ /ϩCKII(296 -312)). The samples were excited at 345 nm at room temperature. The excitation and emission slit widths were 2 and 2 nm, respectively. Each trace shown is the average of three scans.
with excess unlabeled CaM. The fluorescence intensity after full exchange should approach the value for Ca 2ϩ -bound CaM(C75) IAE . Fig. 4A shows the kinetics of dissociation of CaM(C75) IAE from CKII(296 -312) or CaM-kinase which is not autophosphorylated. The data fit well to a single exponential with rates of dissociation of CaM from the peptide and dephosphorylated enzyme of 5.0 s Ϫ1 and 1.1 s Ϫ1 , respectively. Control experiments in which the CaM(C75) IAE ⅐peptide or CaM(C75) IA -E⅐enzyme complexes were mixed with buffer without excess CaM showed no decrease in fluorescence (data not shown). The presence of 1 mM dithiothreitol to reduce potential disulfide bonded dimers of CKII(296 -312) had no effect on the rate of CaM exchange. Thus, the off-rate observed for CKII(296 -312) appears similar to the low affinity (untrapped) state of CaM-kinase. Fig. 4B shows that the rates of dissociation of CaM from either the autophosphorylated form of CaM-kinase or CKII-(290 -314) were extremely slow and only partial exchange was observed after several hours. We utilized autophosphorylation conditions similar to those described my Meyer et al. (14), who demonstrated the CaM trapping effect was due specifically to autophosphorylation of Thr 286 and not to binding of nucleotides. Exchange reactions involving the CaM(C75) IAE ⅐CKII-(290 -314) complex were about 50% complete after 15 h. If the final fluorescence intensity after full exchange is defined as that observed when CKII(290 -314) is added to CaM(C75) IAE in the presence of excess CaM (Fig. 4B, dotted line), then the estimated rates of dissociation of CaM from CKII(290 -314) or autophosphorylated enzyme are 3 ϫ 10 Ϫ5 s Ϫ1 and 9 ϫ 10 Ϫ5 s Ϫ1 , respectively. Loss of enzyme activity at room temperature precludes meaningful interpretation of prolonged incubations with the holoenzyme. Additionally, because these experiments were performed at room temperature in the sustained presence of Mg 2ϩ /ATP, it is likely that CaM-kinase underwent autophosphorylation at sites in addition to Thr 286 . We do not know how these subsequent autophosphorylation steps may influence the observed off-rates for CaM dissociation. Nevertheless, it is clear that CKII(290 -314) kinetically mimics the trapped form of CaM binding to CaM-kinase. DISCUSSION We have shown that peptides which span different amino acids in the CaM-binding domain of CaM-kinase have greatly different rates of dissociation from CaM. CaM binding to peptide CKII(296 -312) kinetically mimics binding to unphosphorylated CaM-kinase, while peptide CKII(290 -314) mimics binding to the phosphorylated/trapped state of the enzyme. We  (30 M) in the same buffer without peptide or enzyme at room temperature. Excitation was at 345 nm and emission was monitored using a 390-nm cut-off filter. Each curve represents the average of 5-6 exchange reactions. The solid lines and indicated rate constants were derived by fitting the experimental data to the single exponential equation F ϭ (F initial ϫ e Ϫkt ) ϩ F final . B, time course for CaM dissociation from CKII(290 -314) (E) or from the phospho-form of CaM-kinase (q). The experimental design was exactly the same as described in the legend to Fig. 3, except that after addition of Ca 2ϩ and peptide or CaM-kinase, 50-fold excess CaM was added to the cuvette, mixed manually, and data collected for the indicated times. A control to determine the extent of photobleaching was performed using CKII(290 -314) without excess CaM (Ç). Excitation was at 345 nm and emission was monitored at 465 nm. Excitation and emission slit widths were 1 and 10 nm, respectively. The dotted line indicates the fluorescence intensity observed when 50-fold excess CaM is added prior to the addition of CKII(290 -314). The indicated rate constants are estimates derived by fitting the experimental data to a single exponential equation in which the fluorescence intensity at maximal exchange is defined as the value indicated by the dotted line.
conclude that autophosphorylation of Thr 286 in CaM-kinase is necessary for high affinity binding of CaM only within the context of the enzyme. Phosphorylation-dependent conformational changes in the CaM-binding region may allow formation of additional interactions with CaM that lead to a dramatic decrease in the rate of dissociation of CaM from CaM-kinase.
Phosphorylation-dependent conformational changes in CaMkinase have been demonstrated previously by differential proteolysis. Kwiatkowski and King (23) showed that a site close to Thr 286 becomes accessible to proteolysis only when Thr 286 is autophosphorylated, indicating exposure of this area. Subsequently, Yamagata et al. (24) showed that limited chymotrypsin treatment cleaved CaM-kinase at Phe 293 in the absence of phosphorylation and at Ile 271 when autophosphorylated. Together, these results indicate that the region just N-terminal to the CaM-binding domain of CaM-kinase undergoes a conformational change of sufficient magnitude to expose the region to proteolytic attack following autophosphorylation of Thr 286 . Because the peptide mimics the autophosphorylated enzyme, we predict that this conformational change allows new contacts to form between CaM and CaM-kinase following autophosphorylation leading to the high affinity CaM-binding state.
Our results suggest that the crystal structure resolved for the complex of CaM and CKII(290 -314) represents high affinity binding of CaM to CaM-kinase. In this structure, only amino acids 293-310 form stable interactions with CaM (see underlined residues in Fig. 2A) and are seen as an ␣-helical rod that is enveloped by CaM (22). Amino acids 290 -292 and 311-314 are not resolved in the crystal structure, presumably because they do not interact with CaM and are free to adopt random conformations. Thus, amino acids included in the CaMbinding region as defined by the crystal structure, but which are absent in the minimal CaM-binding peptide, are Phe 293 , Asn 294 , and Ala 295 . It seems most probable that these residues are responsible for the difference in dissociation rates of the two peptides from CaM.
If one assumes similar rates of association of CaM from CKII(296 -312) and CKII(290 -314), as is seen for binding CaM to phospho-and dephospho-CaM-kinase (14), then the difference in free energy between complexes of CaM and the two peptides would be about 7 kcal/mol. This difference in free energy could easily be achieved by the formation of several hydrogen bonds or salt bridges coupled with hydrophobic interactions. Meador et al. (22) showed that Phe 293 and Ala 295 are positioned to form hydrophobic interactions with residues in the C-terminal half of CaM. They also noted hydrogen bonds and salt bridges between amino acids 296 Arg-Arg-Lys 298 of CKII(290 -314) and Glu residues in both the N-terminal and C-terminal halves of CaM. It is possible that these salt bridges cannot form unless Thr 286 is phosphorylated and Phe 293 and Ala 295 interact with CaM.
The results described in the present study also suggest an important potential limitation on the reversibility of the CaM trapping process. Once CaM-kinase is autophosphorylated and converted to the CaM-trapped state, dephosphorylation may not completely return the off-rate for CaM to the basal state. This hypothesis is based on the observation that the synthetic peptide CKII(290 -314) kinetically mimics the trapped state and has no associated phosphate group. Once CaM is trapped, autophosphorylation of Thr 286 may no longer be essential. This is not meant to indicate that dephosphorylation of Thr 286 would have no effect on CaM-binding affinity. A reasonable prediction would be that an intermediate affinity may be obtained between the fully trapped (autophosphorylated) and untrapped (dephosphorylated) form of the enzyme. In either case, once established in the trapped conformation, the dominant mech-anism for CaM dissociation from the enzyme would be to decrease Ca 2ϩ concentrations. Future experiments will evaluate whether dephosphorylation of autophosphorylated CaM-kinase causes alterations in CaM off-rate of the enzyme in the presence of maintained Ca 2ϩ levels.
In summary, the data presented here, together with an analysis of the crystal structure of CaM bound to CKII(290 -314) and previous studies using peptides and limited proteolysis, provide a model that can be tested for the way in which CaMkinase transitions from the low to high affinity CaM binding, which involves the following events: 1) CaM binds to two adjacent subunits in CaM-kinase holoenzymes, which likely involves amino acids 296 -310 of each subunit; 2) intersubunit autophosphorylation of Thr 286 occurs following Mg 2ϩ /ATP binding; 3) a conformational change is induced in the CaMbinding region; and 4) stabilizing contacts between CaM and CaM-kinase are established, which may include hydrophobic interactions involving Phe 293 and Ala 295 and salt bridges involving 296 Arg-Arg-Lys 298 .