HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Yang, E. Articles by Schulman, H. Search for Related Content PubMed PubMed Citation Articles by Yang, E. Articles by Schulman, H. Social Bookmarking                      What's this?

J Biol Chem, Vol. 274, Issue 37, 26199-26208, September 10, 1999

Structural Examination of Autoregulation of Multifunctional Calcium/Calmodulin-dependent Protein Kinase II^*

Eungyeong Yang and Howard Schulman^§

From the Department of Neurobiology, Stanford University School of Medicine, Stanford, California 94305-5125

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Regulation of Ca²⁺/calmodulin-dependent protein kinase II is likely based on an auto-inhibitory mechanism in which a segment of the kinase occupies the catalytic site in the absence of calmodulin. We analyze potential auto-inhibitory associations by employing charge reversal and hydrophobic-to-charged residue mutagenesis. We identify interacting amino acid pairs by using double mutants to test which modification in the catalytic domain complements a given change in the auto-inhibitory domain. Our studies identify the core pseudosubstrate sequence (residues 297-300) but reveal that distinct sequences centered about the autophosphorylation site at Thr-286 are involved in the critical auto-inhibitory interactions. Individual changes in any of the residues Arg-274, His-282, Arg-283, Lys-291, Arg-297, Phe-293, and Asn-294 in the auto-inhibitory domain or their interacting partners in the catalytic domain produces an enhanced affinity for calmodulin or generates a constitutively active enzyme. A structural model of Ca²⁺/calmodulin-dependent protein kinase II that incorporates these interactions shows that Thr-286 is oriented inwardly into a hydrophobic channel. The model explains why calmodulin must bind to the auto-inhibitory domain in order for Thr-286 in that domain to be phosphorylated and why introduction of phospho-Thr-286 produces the important Ca²⁺-independent state of the enzyme.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Multifunctional Ca²⁺/calmodulin-dependent protein kinase (CaM kinase)¹ II has received considerable attention because of its autoregulatory properties (reviewed in Refs. 1 and 2). Regulation of the kinase is likely based on an auto-inhibitory or pseudosubstrate mechanism in which a segment of the kinase occupies the catalytic site in the basal state. Ca²⁺/calmodulin activates the kinase by wrapping around its target sequence on the kinase, a site that overlaps the auto-inhibitory domain (3). The active enzyme not only phosphorylates exogenous substrates but also exhibits a prominent autophosphorylation of Thr-286 within the auto-inhibitory domain. Autophosphorylation is an intersubunit reaction occurring within each holoenzyme and requires that calmodulin activate one subunit serving as kinase, while a second calmodulin is bound to the subunit serving as substrate (4-6). Autophosphorylation traps bound calmodulin by greatly reducing its dissociation rate and prolonging the active state (7). Even after calmodulin dissociates, the autophosphorylated kinase remains partially active or autonomous of Ca²⁺/calmodulin (8-11). Phospho-Thr-286 may therefore be positioned to interfere with re-establishment of auto-inhibitory contacts. This autophosphorylation is critical for Ca²⁺ spike frequency-dependent activation of the enzyme (12) and enhances targeting of the kinase to synaptic sites (13-15) that may underlie the role of the kinase in regulation of synaptic strength. Mice defective in this autophosphorylation (-CaM kinase II Thr-286  Ala mutants) do not exhibit long term potentiation and are defective in spatial learning (16).

The prevailing hypothesis for kinase regulation is that the auto-inhibitory domain contains a pseudosubstrate that interacts directly with the catalytic site. The crystal structure of a related kinase, the monomeric CaM kinase I, indeed shows some elements of pseudosubstrate interactions with a structural motif resembling that of protein kinase inhibitor (PKI) complexed with cAMP-dependent protein kinase (PKA) (17). It contains a pseudosubstrate sequence with an Arg that mimics the basic residue at P(3), i.e. three residues amino-terminal to the phosphorylated residue of its substrates. In the absence of a crystal structure of CaM kinase II, attempts have been made to understand the molecular mechanisms of auto-inhibition. Previous studies have suggested that Thr-286 is unlikely to reside near the catalytic site because the small amount of autophosphorylation attained in the basal state modifies Thr-305 and Thr-306 but not Thr-286 (18-20). Furthermore, when Lys-300 is substituted with Ser, it also becomes autophosphorylated in the basal state (21), implying that if the kinase utilizes a pseudosubstrate it must be quite distant from Thr-286 (22).

Two conflicting computer models of CaM kinase II auto-inhibition have been presented. Based on the observation that a kinase truncated at amino acid 290 was constitutively active whereas a mutant truncated at 294 was inactive, a structural model was proposed using the coordinates of PKA complexed with PKI, with Lys-300 at P(3) (23). Subsequently, biochemical data implicating His-282 as pseudosubstrate of the ATP binding pocket (24, 25) have been used along with mutational results to build a molecular model with bisubstrate auto-inhibitory interactions of both the ATP and peptide binding sites (26). These models await further experimental verifications.

Our present study employs a modification of the method of Gibbs and Zoller (27) to identify the pseudosubstrate site and examine critical interactions in the auto-inhibitions of -CaM kinase II. We analyzed potential auto-inhibitory associations by employing charge reversal and hydrophobic-to-charged residue mutagenesis. We then identified interacting amino acid pairs by using double mutants to test which modification in the catalytic domain complemented a given change in the auto-inhibitory domain. The structure was tested by mutational investigation of hydrophobic and acidic residues in the catalytic domain that enabled us to identify the position of Thr-286 in the model and might explain important features of its autophosphorylation.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Materials and Chemicals-- T7 DNA Polymerase, Sequenase, sequencing reagents, and DNA modifying enzymes were purchased from U. S. Biochemical Corp. Mutagenesis kit was obtained from CLONTECH. Restriction enzymes were from Life Technologies, Inc. and New England Biolabs, Inc. [-³²P]ATP and [-³⁵S]dATP were from Amersham Pharmacia Biotech. Bovine brain calmodulin was purchased from Ocean Biologics (Edmonds, WA). Phosphocellulose paper (P-81) was obtained from Whatman, and nitrocellulose was from Schleicher & Schuell. Anti--CaM kinase monoclonal antibody, CB--2 (mouse ascites) was generated as described previously (28). CaM kinase substrates autocamtide-2 (AC-2; Lys-Lys-Ala-Leu-Arg-Arg-Gln-Glu-Thr-Val-Asp-Ala-Leu) and autocamtide-3 (AC-3; Lys-Lys-Ala-Leu-His-Arg-Gln-Glu-Thr-Val-Asp-Ala-Leu) were synthesized by Dr. Tim Mietzer and David King. The monoclonal antibody 12CA5 (mouse monoclonal, IgG2b) for the 9-amino acid epitope YPYDVPDYA from hemagglutinin HA1 was from Berkeley Antibody Co. (Richmond, CA). Oligonucleotides were synthesized by Ana-Gen Technologies, Inc. (Palo Alto, CA). Electrophoresis reagents were purchased from Bio-Rad. All other chemicals were of reagent grade or better.

Site-directed Mutagenesis-- Single-stranded M13--CaM kinase derived from M13 mp19 with an insert encoding the -subunit of rat CaM kinase or SR-CaM kinase was used for a template. Site-directed mutagenesis was performed by standard methods using single stranded M13--CaM kinase template as described (19) or by CLONTECH Transformer^TM site-directed mutagenesis protocol using a double-stranded SR-CaM kinase template. A hemagglutinin (HA) tag (GAPYPYDVPDYAGPGAQL) was inserted at the carboxyl terminus of the wild-type kinase as described previously (4), and mutant fragments were subcloned into the cDNA. M13-CaM kinase mutants were transferred into SR-CaM kinase as described (29). All mutations were identified by digestion with restriction enzymes and confirmed by DNA sequencing.

Expression of CaM Kinase-- COS-7 cells were transfected with calcium phosphate/DNA co-precipitate as described (29). Cells were harvested 78-90 h after transfection; cell pellets obtained by centrifugation were disrupted by sonication using a cup sonicator (Heat Systems-Ultrasonics) in a buffer containing 50 mM PIPES (pH 7.0), 1 mM EGTA, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 20 µg/ml leupeptin, 10 µg/ml pepstatin, and 2 mM dithiothreitol. The homogenates were then centrifuged at 12,000 × g for 15 min, and the supernatant was frozen at 80 °C. The quantity of wild-type and mutant kinases present in COS-7 cell lysates was determined by immunoblotting using CB--2 (mouse ascites). Immunoblot analysis demonstrated that all the mutant kinases were full-length and stably expressed at high levels.

Kinase Assays-- Kinase activity was measured using the CaM kinase-selective substrate AC-2 or AC-3. COS cell lysates, prepared as above, were used to determine wild-type and mutant kinase activity. Maximal activity was measured in reaction mixtures of 50 mM PIPES, pH 7.0, 10 mM MgCl₂, 0.1 mg/ml bovine serum albumin, 0.5 mM CaCl₂, 50 µM ATP containing [-³²P]ATP, 1 µM calmodulin, 20 µM AC-2 or AC-3 at 30 °C. The Ca²⁺/calmodulin-independent activity was determined in the presence of 0.5 mM EGTA in the absence of exogenous Ca²⁺ and calmodulin. Calmodulin activation properties of kinases were measured for 5-20-fold diluted cell lysates in the reaction mixture of 50 mM PIPES, pH 7.0, 10 mM MgCl₂, 0.1 mg/ml bovine serum albumin, 50 µM ATP containing [-³²P]ATP, 2 µM calmodulin, 20 µM AC-2, 0.1 mM EGTA with varying concentrations of Ca²⁺, where the free Ca²⁺ concentration determines the concentration of Ca²⁺/calmodulin (30). The quantitative changes in the calmodulin activation properties (relative K_CaM) of mutant kinases were assessed by comparing the ratios of activities at Ca²⁺ concentrations below that required for maximal activity as described previously (31, 32).

The HA-tagged kinase in transfected cell extracts was assayed on Falcon microtiter plates (12). The plate wells were precoated with HA-antibody (10 µg/ml in 20 mM NaHCO₃, pH 9.6) for at least 4 h at 4 °C, followed by washing with PIPES-buffered saline (50 mM PIPES, pH 7.0, 150 mM NaCl) containing 0.1% BSA and 0.1% Tween 20. COS-7 cell extracts diluted in 20-40 volumes of PIPES-buffered saline containing 0.1% BSA, 1 mM EGTA, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 20 µg/ml leupeptin, 10 µg/ml pepstatin, and 2 mM dithiothreitol were incubated in the precoated wells for 2 h at 4 °C. Unbound proteins were then removed by extensive washing with PIPES-buffered saline containing 0.1% BSA and 0.1% Tween 20.

Calmodulin activation properties were measured using a Ca²⁺/EGTA buffer system as described above. Autophosphorylation reaction was performed in 50 mM PIPES, pH 7.0, 10 mM MgCl₂, 0.1 mg/ml bovine serum albumin, 0.3 mM CaCl₂, 250 µM ATP, 0.5 µM calmodulin, for 15 s at 30 °C. The reaction was stopped by addition of EGTA (3 mM final concentration) and EDTA (16 mM final concentration) on ice. Aliquots (5 µl) of the autophosphorylated kinases were immediately used to measure kinase activity in reaction mixtures (50 µl total) containing 50 mM PIPES, pH 7.0, 10 mM MgCl₂, 0.1 mg/ml bovine serum albumin, 20 µM AC-2, either 1 mM CaCl₂ and 0.5 µM calmodulin or 1 mM EGTA, and ATP containing [-³²P]ATP at a final concentration of 50 µM at 30 °C.

Molecular Modeling-- The coordinates of CaM kinase I crystal structure were kindly provided by Dr. Angus Nairn (17). The sequences of the catalytic domain of CaM kinase II, PKA, and CaM kinase I were aligned according to Goldberg et al. (17). Within the auto-inhibitory domain of CaM kinase II (residues 272-302), the sequence was aligned to those of CaM kinase I and PKI, preserving the position of the substrate recognition residue at P(3). The alignment was then adjusted, conserving the overall secondary structure and positioning of the residues known to be critical in binding ATP and in the catalytic site. An initial hybrid model was constructed using the coordinates from CaM kinase I and the closed form of PKA. Individual interactions obtained by mutational analyses were then incorporated by several trials of coordinate assignment and homology modeling. The modeling was carried out with Look^TM and Insight II version 95.0 on a Silicon Graphics Indigo workstation. In each step of the modeling process, the validity of the tertiary structure was assessed using the three-dimensional profile approach (33).

	RESULTS

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Identification of the Core "Pseudosubstrate" Segment-- The auto-inhibitory domain contains a number of basic residues that could anchor a core pseudosubstrate sequence in the peptide-binding site of the kinase, including Arg-283 within the autophosphorylation site (RQET). In PKA, which also prefers a basic residue at P(3) of its substrates, this basic residue interacts with Glu-127 in the crystal structure (34); therefore, we began our analysis of pseudosubstrate interactions in CaM kinase II by identification of the acidic residue(s) homologous to Glu-127 of PKA. A trial molecular model built based on kinase sequence alignments (17, 35) assisted in identifying a number of charged residues in the catalytic domain of CaM kinase II that may interact with substrates and with the auto-inhibitory domain (Fig. 1A). We mutated Glu-96 and Glu-139 to basic residues to determine whether this would increase the K_m for substrates, since these residues appeared to correspond to Glu-127 and Glu-170 of PKA that interact with basic residues at P(3) and P(2) of PKA substrates, respectively (Fig. 1A). The homologous residues are involved in auto-inhibitory interactions in MLCK as well (36, 37). Mutants E96K and E139R displayed K_m values of ~57 µM and ~61 µM for the synthetic peptide AC-2, respectively, compared with ~4 µM for the wild-type kinase. The increased K_m values suggested that these two acidic residues are likely to participate in substrate binding (Fig. 1B, top).

View larger version (14K): [in this window] [in a new window]   Fig. 1.   Summary of mutations and interactions in the catalytic and auto-inhibitory domains. A, schematic representation of site-directed mutants in the catalytic domain. The residues in -CaM kinase II catalytic domain predicted to interact with its inhibitory domain or with substrates are shown in bold, along with some adjacent residues. The specific amino acid substitutions are indicated below the sequence and the residue numbers are indicated above. B, summary of interactions between substrate autocamtide-3 and -CaM kinase II. The sequence of AC-3 is displayed, with some key residues in closed circles. Their interacting amino acid residues in the catalytic domain of the kinase are shown above (hydrophobic interactions) or below (electrostatic interactions) the sequence. C, schematic representation of site-directed mutants in the regulatory domains. The amino acid sequence from Ser-272 to Lys-300 within the autoregulatory domain is shown in single-letter code, including the inhibitory and part of the calmodulin-binding subdomains. Some residue numbers are provided as a guide and the specific amino acid substitutions are indicated below the sequence of the regulatory region. Double mutants are shown below the single mutants with one underline. D, summary of auto-inhibitory interactions in -CaM kinase II. The sequence of the auto-inhibitory domain is displayed, with important residues in closed circles. Their interacting amino acid residues in the catalytic domain are indicated above (hydrophobic interactions) or below (electrostatic interactions). Hydrophobic interactions are indicated with a solid arrow, electrostatic interactions with a dotted line, and acidic residues near Thr-286 that interfere with phospho-Thr-286 with a dashed line.

If Glu-96 and Glu-139 also function in pseudosubstrate interactions with the inhibitory domain, then their mutation to basic residues should reduce such interactions and thereby facilitate activation by calmodulin. Ca²⁺/calmodulin dependence was determined by performing assays in which the concentration of free Ca²⁺ was varied with a Ca²⁺/EGTA buffer at a fixed high calmodulin concentration (31). E96K did, in fact, have an enhanced affinity for calmodulin (2.6-fold decrease in K_CaM), whereas the calmodulin binding affinity of E139R was not significantly better than wild-type. Glu-96, but not Glu-139, may therefore interact with a basic residue P(3) in the pseudosubstrate; both may contribute equally to interactions with actual substrates.

We employed the strategy of Gibbs and Zoller (27) to find the critical basic residue at P(3) in the pseudosubstrate P(3) that interacts with Glu-96. The general idea is that reversing the charge of this basic residue would decrease K_CaM, just like E96R, and that combining the two charge reversals in a double mutant would restore some of the original electrostatic interactions and shift the Ca²⁺/calmodulin dependence curve toward that of wild-type. We mutated a number of potential P(3) residues at the carboxyl terminus of the auto-inhibitory domain and assayed their activation properties. Some were further checked by expression as HA-tagged kinase constructs. This facilitated their purification from COS cell extracts by adsorption to monoclonal HA-antibody immobilized on 96-well plates in which assays could also be performed (see "Experimental Procedures"). The most likely residue in the inhibitory domain to pair with Glu-96 in the catalytic site and support auto-inhibition is Arg-297. HA-R297E exhibited a Ca²⁺ activation curve that was shifted to higher affinity (relative K_CaM of 0.77) (Fig. 2; Table I). The shift for Arg-297 was small, and we reasoned that since it interacts with calmodulin in the crystal structure (3), HA-R297E may exhibit the counteracting effects of a weakened auto-inhibitory interaction with Glu-96 (a decrease in relative K_CaM) and reduced binding to calmodulin (an increase in relative K_CaM). This was confirmed in a double mutant containing both HA-E96R and HA-R297E. The leftward shift in the calmodulin activation curve of HA-E96R (relative K_CaM of 0.18) was indeed reversed when combined with HA-R297E in a double mutant, with a relative K_CaM of 3.31 (Fig. 2; Table I). The inhibitory interaction appeared to be reestablished in HA-E96K/R297E; the remaining effect of reduced binding to calmodulin due to HA-R297E shifted the Ca²⁺ activation curve to the right of wild-type (Fig. 2). Arg-297 may therefore mimic the basic residue at P(3) of substrates, as both interact with Glu-96. No such complementation was seen when either mutation was paired with other charged reversal mutants, supporting the specificity and validity of this approach. For example, the combination mutant of untagged E96K (with decreased K_CaM of 0.38) and R296E (with increased K_CaM of 10.28) displayed an intermediate K_CaM value (7.65). This is what would be expected for non-interacting pairs in which the Ca²⁺ activation curve of a double mutant would simply be the average of the curves for the two individual mutants.

View larger version (16K): [in this window] [in a new window]   Fig. 2.   Pseudosubstrate interaction between Glu-96 and Arg-297. HA-tagged -CaM kinase II mutants were expressed in COS-7, partially purified via immobilized HA antibody, and normalized kinase activity assayed at varying concentrations of Ca2+ in the presence of 0.1 mM EGTA, as described under "Experimental Procedures." The final concentration of AC-2 in the assay was 0.2 mM. The data were normalized to the maximal activity. Symbols represent a mean value of at least four independent assays from two independent transfections, each performed in duplicate. The data are presented without error bars for clarity. , HA-wild-type; , HA-R297E; , HA-E96K; , HA-E96K/R297E.

Identification of Inhibitory Domain Residues Critical for Autoregulation-- Previous studies on CaM kinase II using synthetic peptides (24) and on MLCK using mutagenesis (36, 37) implicated charged residues in autoregulation. We therefore employed site-directed mutagenesis to reverse the charge of amino acids in the auto-inhibitory domain; Glu was substituted for positively charged residues, Lys for negatively charged residues, and Lys for His (Fig. 1C). In addition, since hydrophobic residues participate in inhibitory associations of PKA/PKI and CaM kinase I (17, 34), Phe-293 and Asn-294 were mutated to acidic residues (Fig. 1C). We reasoned that there would be numerous contacts contributing to auto-inhibition so that disruption of individual contacts would facilitate activation by calmodulin, detected as a reduced K_CaM, but would not sufficiently displace the entire auto-inhibitory domain to produce a constitutively active kinase.

Mutations in the pseudosubstrate region (R296E, R297E, K298E, and K300E) did not, in fact, produce constitutive activity (Table II). Paradoxically, mutant residues far from the pseudosubstrate (positions 297-300) appeared more critical for auto-inhibition; R274E, H282K, and R283E were 4-15% Ca²⁺/calmodulin-independent (Table II). In addition to these, T286D, which places a charged residue at the autophosphorylation site, has previously been shown to produce a constitutively active kinase. Hydrophobic-to-acidic residue mutants in the region between Thr-286 and the pseudosubstrate also showed relatively high constitutive activities; Ca²⁺-independent activity was 15-18% for N294D. Although the single mutant F293E did not alter Ca²⁺ independence, it could be shown to participate in auto-inhibition by use of a double mutant, F293E/N294D, which was 38-40% Ca²⁺-independent (Table II). The extent of Ca²⁺/calmodulin-independent activity resulting from mutagenesis was not due to differences in K_m for ATP, as all such mutants used in this study exhibited K_m values ranging from 6 to 14 µM ATP, well below the 50 µM used in our standard assays.

We determined K_CaM values in order to identify the important residues for kinase auto-inhibition among the mutant enzymes that showed no constitutive activity. K291E showed a decrease in relative K_CaM to 0.46 (Table II), likely due to a decrease in the potency of the auto-inhibitory domain. In contrast, E285K, R296E, R298E, and K300E required significantly more Ca²⁺ for half-maximal activation than the wild-type kinase, which resulted in increases in relative K_CaM values (Table II). The much larger K_CaM values obtained for the R296E, R298E, and K300E mutants are consistent with their interactions with calmodulin (3). D288K, K292E, and R297E, on the other hand, exhibited no significant changes in calmodulin binding affinity (Table II).

Identification of Catalytic Domain Residues Critical for Autoregulation-- Mutagenic studies on MLCK have suggested that the active site is kept inactive in the basal state by numerous electrostatic interactions between an auto-inhibitory domain on the surface of the enzyme and the catalytic core of the enzyme (36). We therefore mutated the corresponding residues in -CaM kinase II; we chose Glu-99, Asp-100, Arg-104, Glu-105, Asp-111, and Glu-236 (Fig. 1A) for mutagenesis based on the sequence alignment with MLCK (38). Charge reversal mutations in Glu-105 and Glu-236 showed ~3-fold decreases in relative K_CaM (Table III), which suggests their importance for inhibition, along with Glu-96 described above. By contrast, E99K, E99K/D100K, R104E, and D111R had relative K_CaM values similar to those for the wild-type kinase.

We next selected Phe-98, Lys-148, Ile-205, Asp-238, and Glu-243 for mutagenesis based on our temporary homology model of CaM kinase II that incorporated the electrostatic interaction between Glu-96 and Arg-297 established above. A hydrophobic-to-basic residue mutant, I205K, exhibited a significantly increased calmodulin binding affinity (Table III). Other mutations induced Ca²⁺-independent activity, suggesting that they may make critical interactions for positioning of the auto-inhibitory domain. Charge reversal mutants K148E, D238R, and E243R had significant constitutive activity (Table III) but bound AC-2 with normal affinity (data not shown). F98K had 7.9-12.3% constitutive activity and a 10-fold decrease in K_m for AC-2 (42 µM). Based on our computer model, Phe-98 possibly interacts with Leu at the P(5) position in AC-2 as well as with hydrophobic residues near the corresponding P(5) position in the inhibitory domain (Fig. 1B). In summary, the residues whose mutations caused autonomous activity or increased calmodulin binding affinity are likely to participate in the inhibitory segment recognition. These include Glu-96, Phe-98, Glu-105, Lys-148, Ile-205, Glu-236, Glu-238, and Glu-243.

Identification of Paired Interactions of Residues Carboxyl-terminal to Thr-286-- Having identified a number of residues critical for auto-inhibition, we next used the molecular model to begin the process of matching amino acids in the auto-inhibitory domain that participate in paired electrostatic or hydrophobic interactions with critical residues in the catalytic domain. Since K291E displayed a decrease in relative K_CaM (Tables II), the possibility of the interaction between Lys-291 and Glu-105 was examined using a double mutant. Whereas the individual mutations of the putative pair, HA-E105K and HA-K291E, shifted Ca²⁺ activation curves to the left, a double mutant HA-E105K/K291E corrected the shift and exhibited a similar Ca²⁺ activation curve as the wild-type kinase with relative K_CaM of 1.03 (Fig. 3; Table I). This implies that the electrostatic interaction has been reestablished in the double mutant, suggesting the interaction of Glu-105 with Lys-291. In further support of the specificity of this method, double mutants in which other residues were paired with E105K or K291E did not correct the individual shifts. For example, combination mutants E105K/R297E and E96K/K291E had 3-5% Ca²⁺-independent activities, implying cumulative disruption of electrostatic interactions rather than re-establishment of the original interacting pairs (Table IV). Other paired combinations with E105K or K291E similarly failed to produce an enzyme with the calmodulin activation properties seen in wild-type or in E105K/K291E.

View larger version (15K): [in this window] [in a new window]   Fig. 3.   Interactions between Glu-105 and Lys-291. HA-tagged -CaM kinase II mutants were expressed in COS-7 cells, and kinase activity was measured in cell lysates with the normalized amount of the enzyme at varying concentrations of Ca2+ in the presence of 0.1 mM EGTA, as described under "Experimental Procedures." The final concentration of AC-2 in the assay was 20 µM. The data were normalized to the maximal activity. Symbols represent a mean value of at least four independent assays from two independent transfections, each performed in duplicate. The data are presented without error bars for clarity. , HA-wild-type; , HA-K291E; , HA-E105K; , HA-E105K/K291E.

We extended this strategy for matching interacting pairs to hydrophobic residues in the carboxyl-terminal half of the auto-inhibitory domain. Since introduction of charged residues in place of hydrophobic ones in both the auto-inhibitory (Phe-293 and Asn-294; Table II) and catalytic (Phe-98 and Ile-205; Table III) domains altered activation properties, we combined mutants in these two domains and tested for interacting residues. The high constitutive activity of the single mutant N294D was greatly suppressed in double mutants N294D/I205K or N294D/F98K (Table V), consistent with establishment of electrostatic interactions between mutated residues in place of the hydrophobic interactions in wild-type enzyme. In addition, combination of F293E with F98K partially abolished the Ca²⁺-independence of F98K (Table V), suggesting that Phe-98 may be positioned close to both Asn-294 and Phe-293 in the inactive kinase. The role of Phe-293 in auto-inhibition was revealed in F293E/N294D, which had a much higher constitutive activity than the two individual mutants combined, in fact the most constitutively active mutant we have found. This hydrophobic pair of residues may interact with Ile-205 in the catalytic domain since the constitutive activity of F293E/N294D is suppressed when combined with I205K in the triple mutant, F293E/N294D/I205K (Table V). It should be noted that our approach using hydrophobic-to-charged residue mutations might not be applicable to address the specific residue pairs of the association, since hydrophobic interactions can occur in a closer distance than electrostatic interactions. However, the data support participation of Phe-293 and Asn-294 with Phe-98 and Ile-205 in hydrophobic auto-inhibitory associations (Fig. 1D).

Identification of Paired Interactions of Residues Amino-terminal to Thr-286-- Three opposite substitutions (R274E, H282K, and R283E) in the region amino-terminal to Thr-286 caused partial but significant Ca²⁺-independent activities (Table II). To identify their electrostatic pairs in the kinase, double mutants in the catalytic domain and in the auto-inhibitory domain were generated and their constitutive activities were analyzed. In some cases, the paired combination of individual mutants resulted in cancellation of constitutive activity; E243R/R274E, K148E/H282K, and D238R/R283E (Table VI) exhibited no significant Ca²⁺-independent activities. Further, these double mutants showed Ca²⁺ activation properties similar to those for the wild-type kinase (Table IV). By contrast, double mutants with mismatched combinations had higher constitutive activities than individual mutants. The results are consistent with pairing of Glu-243 with Arg-274 at the amino-terminal end of the auto-inhibitory domain and of Lys-148 with His-282 and Asp-238 with Arg-283 near the important autophosphorylation site (Fig. 1D).

Modeling of CaM Kinase II-- The results from the mutational analyses were complemented by construction of a molecular model. We originally attempted to model CaM kinase II from the coordinates of CaM kinase I, but this led to a trial structure in which incorporation of data about interacting pairs from the mutational analysis was unsuccessful, particularly in the loop region amino-terminal to Thr-286. The displacement of the D helix by this region differs in the structures of CaM kinase I and PKA/PKI (17). We therefore made a hybrid model based on both CaM kinase I and PKA in which the sequence of the D helix of CaM kinase II was aligned with that of PKA rather than CaM kinase I. Additionally, two regions with high conformational disorder, the loop between 8 and EF (activation loop) and the one between 3 and C in the ATP binding domain, were also assigned from the PKA coordinates. The structures of CaM kinase I and the closed form of PKA were then optimally superimposed, and hybrid modeling was performed.

An initial hybrid model containing the auto-inhibitory domain was built by aligning residues 272-302 of CaM kinase II with the inhibitory domain of CaM kinase I, preserving the P(3) sites. Biochemical data of ionic and hydrophobic interactions identified above and summarized in Fig. 1D were then sequentially incorporated into the initial hybrid model. First, Lys-148 was assigned the coordinates of Pro-313 in PKA and then hybrid modeling was performed. Since the resulting hybrid model did not contain the interaction between Arg-297 and Glu-96, Arg-297 was aligned with P(3) of PKI and the modeling was carried out. To introduce the interactions between Arg-283 and Asp-238, and between Arg-274 and Glu-243, coordinates in the neighboring regions of Arg-283 and Arg-274 were assigned using the Generate Loops program. Among 10 possible loops, the ones containing the interactions were selected. The association between Lys-291 and Glu-105 was incorporated by first modeling of CaM kinase I in which its Val-108 was "mutated" to Glu, followed by homology modeling of CaM kinase II based on the sequence alignment of Glu-105 in CaM kinase II with the mutated Glu-108 in CaM kinase I. Hydrophobic interactions were refined by aligning Phe-293, Asn-294, Phe-98, and Ile-205 in CaM kinase II with Asn-297, Phe-298, and their interacting residues in CaM kinase I, and performing homology modeling.

The final structural model of CaM kinase II is shown in Fig. 4, which includes only residues 3-300 since no information about interacting pairs is available for the rest of the regulatory domain of the kinase. Although the model differs markedly in positioning of the auto-inhibitory domain from that of a previous model (26), it nicely incorporates the paired interactions identified by mutagenesis in a structure with the auto-inhibitory domain oriented similarly to that of CaM kinase I and PKA/PKI. A sufficient number of interacting pairs were implicated by mutagenesis to place the entire length of the auto-inhibitory domain on the surface of the kinase (Fig. 4). Its amino-terminal end begins at the bottom of the large lobe, across a hydrophobic channel, and finally making several electrostatic interactions across the sites that bind peptide, and ending with the beginning of the calmodulin-binding sequence.

View larger version (100K): [in this window] [in a new window]   Fig. 4.   Molecular model of monomeric -CaM kinase II (residues 3-300). The catalytic domain (residues 3-271, blue color) was built by hybrid modeling based on the coordinates of CaM kinase I and PKA. Residues involved in electrostatic interactions (Glu-96, Glu-105, Lys-148, Asp-238, and Glu-243; gray color) and those involved in hydrophobic interactions (Phe-98 and Ile-205; dark blue color) are highlighted (yellow letter). Residues 272-300, which correspond to the auto-inhibitory region, are shown in ribbon: Arg-297-Lys-300, green; Phe-293-Arg-296, purple; Thr-286, yellow; others, red. Residues that interact with the catalytic domain are indicated by white letters.

Molecular Basis for Ca²⁺-independent Activity of Autophosphorylated Kinase-- In a molecular model constructed as above, Thr-286 appears to interact with hydrophobic amino acid residues in the catalytic domain. If correct, then substitution of Thr-286 with a basic residue should be as disruptive as the acidic substitution (T286D) that is thought to mimic the phospho-Thr-286 in the autophosphorylated kinase (39, 40). We tested this prediction of the model by mutating Thr-286 to an acidic (Asp), basic (Lys), or hydrophobic (Leu) residue and assaying for Ca²⁺-independent activity. Introduction of a basic residue did, in fact, generate constitutive activity, similar to the ~10% Ca²⁺-independent activity of T286D, whereas T286L exhibited minimal constitutive activity (Fig. 5). The findings with T286D and T286L have been reported before (39, 40) and are reexamined here for comparison. These results support positioning of Thr-286 in the hydrophobic environment.

View larger version (15K): [in this window] [in a new window]   Fig. 5.   Effects of Thr-286 mutations. The activity of kinases in transfected COS cell extracts with the normalized enzyme level was determined in the presence of Ca2+/calmodulin (solid bars) or EGTA (hatched bars). The constitutive activity is indicated above the bars.

Autophosphorylation modifies Thr-286 with a bulky anionic phosphate moiety that cannot be reintroduced near the hydrophobic channel due to unfavorable energetics and steric constraint, making it particularly effective at disrupting the auto-inhibitory domain. We reasoned that if a cationic residue was introduced into the hydrophobic channel, then repulsion of phospho-Thr-286 and the resultant Ca²⁺-independent activity would be reduced relative to that seen with autophosphorylated wild-type. We selected three residues in the hydrophobic channel of the catalytic domain located close to Thr-286 in the model for further examination. Val-208 and Trp-237 were replaced with Lys, and Val-240 with Arg. Since a charge is introduced into the hydrophobic environment, this mutation should cause disruption of the inhibitory interaction, leading to some constitutive activity. Indeed, V208K, W237K, and V240R showed 7-12% Ca²⁺-independent activities (Fig. 6A). Despite this basal constitutive activity, however, we were able to observe a reduction in Ca²⁺-independent activity generated by autophosphorylation. Following autophosphorylation, V208K and W237K exhibited only ~30% Ca²⁺-independent activity, corresponding to an approximately 20% increase due to autophosphorylation, compared with a 50% increase seen with wild-type. Autophosphorylation increased Ca²⁺-independent activity of V240R by 50%, similar to the increase in wild-type (Fig. 6A). These results support a model in which phospho-Thr-286 is repulsed by a hydrophobic environment that includes Val-208 and Trp-237 in the catalytic domain, since their replacement with basic residues appears to enable attractive interactions with phospho-Thr-286 (Fig. 1D).

View larger version (38K): [in this window] [in a new window]   Fig. 6.   Effects of mutations in the catalytic domain near Thr-286 on Ca2+-dependent autophosphorylation. Transfected cell extracts with the normalized enzyme level were incubated at 30 °C for 15 s in the presence of Ca2+/calmodulin (hatched bars, + autophosphorylation) as described under "Experimental Procedures." Reactions were terminated by addition of excess EDTA and EGTA. Control reactions (solid bars,  autophosphorylation) received excess EDTA and EGTA before addition of the cell extracts and were kept on ice. Aliquots were then immediately assayed for total and Ca2+-independent activities. The Ca2+-independent activity is represented as a percentage of the total Ca2+-dependent activity for hydrophobic-to-basic residue mutations (A), hydrophobic-to-acidic residue mutations (B), and charge reversal mutations (C).

If the interpretation above is correct, then introduction of an anionic residue into the hydrophobic channel should increase repulsion of phospho-Thr-286 and the resultant Ca²⁺-independent activity should be increased relative to wild-type. Indeed, we found that hydrophobic-to-acidic residue mutations V208E and W237D displayed a basal Ca²⁺-independent activity of 20% and 12%, respectively, and developed higher Ca²⁺-independent activity than the wild-type kinase after autophosphorylation (Fig. 6B). In contrast, V240D had no basal constitutive activity and autophosphorylation induced similar Ca²⁺ independence as for the wild-type kinase (Fig. 6B). The data with both cationic and anionic substitutions suggest that Val-208 and Trp-237, but not Val-240, are near Thr-286 and may provide a hydrophobic environment that repulses phospho-Thr-286.

The model also indicated that several acidic residues (Glu-236, Glu-109, and Asp-111) were near the hydrophobic channel and could further contribute to the repulsion of phospho-Thr-286. We tested this by mutating the three acidic residues to basic ones in order to see whether the basic residues would provide an electrostatic attraction to phospho-Thr-286 and reduce the autonomy of the autophosphorylated state. In fact, autophosphorylation generated lower Ca²⁺-independent activities in E236K, E109R, and D111R than the wild-type kinase (Fig. 6C). If Glu-285 is near these acidic residues, then converting it to a basic residue should increase auto-inhibitory interactions and reduce the effect of autophosphorylation. After autophosphorylation E285K showed 19% Ca²⁺-independence compared with ~50% for the autophosphorylated wild-type kinase (Fig. 6C). Further, a double mutant E236K/E285K showed only 13% Ca²⁺-independent activity after autophosphorylation compared with 9% constitutive activity without autophosphorylation (Fig. 6C). These suggest that not only the hydrophobic environment but also negative charges near Thr-286 might participate in producing autonomous activity after autophosphorylation.

	DISCUSSION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

This report attempts the first experimental mapping of the ionic and hydrophobic interactions involved in the intrasteric auto-inhibition of CaM kinase II. Our approach takes advantage of known crystal structures of CaM kinase I and PKA and of the sequence and topological similarities in protein kinase catalytic domains to facilitate development of a model for CaM kinase II, whose crystal structure has not been solved. The catalytic domain of protein kinases consists of two lobes joined by a hinge polypeptide (41) with a catalytic site at their interface and the relative movements of the lobes upon displacement of an auto-inhibitory segment probably serving to regulate the kinase (42). We employed the approach of charge reversal and hydrophobic-to-charge mutagenesis with the assumptions that there are no intersubunit inhibitory interactions and that amino acid substitutions weaken or disrupt their normal interactions without radically altering the conformation of the core enzyme.

We identified hydrophobic interactions and five pairs of electrostatic interactions (Fig. 1D) by using complementation of two individually disruptive mutants with the aid of homology modeling (Fig. 4). The model structure contains the auto-inhibitory domain with some -helical (residues 283-292), and more extended -helical or loop-like regions (residues 293-300 and residues amino-terminal to residue 283). The amino acids important in the inhibitory association, His-282, Arg-283, Lys-291, Phe-293, Asn-294, and Arg-297, are spatially aligned on the internal face of this inhibitory strand. In accordance with this secondary structure, the synthetic peptide corresponding to residues 281-302 has been shown by circular dichroism to contain 57% -helix (24). Furthermore, constitutive activity is generated when Cys-289, which appears to be situated on the outer face of the -helical region in our model, is substituted with a proline residue, likely due to disruption of the helix (5).

Ca²⁺/calmodulin activates CaM kinase II by increasing affinity for ATP as well as peptide substrates (20, 24, 43), suggesting that the auto-inhibitory domain suppresses binding of both substrates in the basal state. A peptide inhibitor containing residues 281-309 was shown to be competitive with respect to ATP (25) and to protect the ATP binding site from chemical modification by phenylglyoxal (44). Such an ATP-competitive mechanism is not observed in CaM kinase I (45). His-282 has been specifically implicated in the ATP-directed mechanism of inhibition; substitution or protonation of His-282 decreases the inhibitory potency of the CaM kinase II peptide up to 50-fold, and changes the nature of the kinetics from competitive to noncompetitive with ATP (24). This finding was incorporated into a provocative model in which His-282 directly blocks ATP binding by occupying the adenosine-binding motif of the catalytic site (26). Our data delineating paired interactions along the length of the auto-inhibitory domain are inconsistent with such positioning of His-282 in the ATP-binding site. However, it is tempting to speculate that a cascade of interactions starting with His-282 and Lys-148 in the loop segment (Fig. 4) regulates the interface between the small and large lobes and provides a unique ATP-directed auto-inhibitory mechanism.

One of the most striking features of the modeled CaM kinase II (Fig. 4) supported by mutational studies (Figs. 5 and. 6) is that Thr-286 fits on the internal face of the helical inhibitory domain in an environment containing both hydrophobic and acidic residues far from the catalytic site. Goldberg et al. (17) suggested a hydrophobic environment near Thr-286 based on the corresponding structure in CaM kinase I. Our findings provide important insights into the molecular basis for the mechanism of autophosphorylation and for the autonomous activity that is generated by it. First, the Thr-286 region is not positioned as a pseudosubstrate in the catalytic site. Short peptides based on the Thr-286 region, such as AC-2, may therefore bind to the activated kinase at both the active site and this second site whereas other substrates, such as syntide-2, may only bind at the catalytic site, as suggested by kinetic analyses (46). Arg-283 at P(3) to Thr-286 in AC-2 would interact with Glu-96 and Glu-139 in the catalytic site and with Asp-238 at the non-catalytic site. A longer peptide that includes residues 281-309 may mimic the auto-inhibitory domain and position itself across both binding sites; its pseudosubstrate segment, which lacks a phosphorylatable residue, binding to the active site while its Thr-286 segment positioned at the non-catalytic binding site. This would explain why such a peptide binds with high affinity yet is poorly phosphorylated at Thr-286 (47). Binding of Ca²⁺/calmodulin to this peptide masks the pseudosubstrate segment and allows Thr-286 to bind to the active site where it can be phosphorylated (47). Second, the suggestion by the model that Thr-286 faces inward explains why it is not available for intersubunit phosphorylation until calmodulin binds the auto-inhibitory domain that is to be phosphorylated (4, 6). The dual requirement for calmodulin, to both activate subunits and present Thr-286 for autophosphorylation by a neighboring subunit, is the basis for a functional cooperativity that underlies a switch in kinase sensitivity to the frequency of Ca²⁺ oscillations that is elicited by autophosphorylation (12). Third, autophosphorylation generates an autonomous enzyme because phospho-Thr-286 does not seem to be accommodated in the hydrophobic groove and the enzyme cannot return to the auto-inhibited state. This is supported by the finding that introduction of positively charged residues near the hydrophobic groove attenuates autonomy whereas introduction of negatively charged residues increases the autonomous state (Fig. 6). In addition to the repulsion of phospho-Thr-286 by the hydrophobic environment, there is a cluster of negative charges, especially Glu-236, which provide additional repulsion with phospho-Thr-286 and with Glu-285 and thereby enhance the autonomous activity (Fig. 6).

The region amino-terminal to the core pseudosubstrate sequence, rather than the pseudosubstrate itself, appears most critical for auto-inhibition. Furthermore, although numerous contacts in this region contribute to the positioning of the pseudosubstrate sequence at the peptide-binding site and for constraining the ATP binding site, mutational modification of any of these contacts generates constitutive activity, as if the orientation and inhibitory function of this domain is on a hare trigger. Nature has placed a Thr residue in the middle of this domain so that its autophosphorylation would produce an extended conformational change that disables inhibition of peptide and ATP binding by this domain after calmodulin is no longer bound. The markedly enhanced affinity of calmodulin for autophosphorylated kinase (7, 48) and the enhanced and prolonged binding of this form of the kinase for synaptic sites (13-15) may also be a consequence of this conformational change.

Our hybrid model structure of CaM kinase II (Fig. 5) exhibits similarities to the crystal structure of CaM kinase I (17). The inhibitory domain of CaM kinase II resembles that of CaM kinase I (and of PKA) in its secondary structure and orientation. It fits into a channel on the surface of the protein that is created by the edge of the D helix along with an extensive region stretching from the end of F, through G, to H. The position of the inhibitory segment also appears similar in auto-inhibited MLCK. Analogous electrostatic interactions across the surface of the lower lobe and at the cleft between the two lobes are likely to be critical in MLCK auto-inhibition (36). However, some acidic residues corresponding to those shown to participate in the auto-inhibition of MLCK do not seem to be critical in the CaM kinase II auto-inhibition (Table III). Since more distal basic residues act as specificity determinants for MLCK (37), it would be reasonable to postulate that additional charged interactions might be involved in the MLCK auto-inhibition. Two acidic residues, Glu-96 and Glu-139, are likely to provide a recognition site for the P(3) basic residue in substrates of CaM kinase II, but only one of these, Glu-96, interacts with a basic residue (Arg-297) in the auto-inhibitory domain. The two homologous acidic residues in MLCK provide auto-inhibitory interactions (17, 36, 37, 49), suggesting that its relative arrangement of the two lobes might be different from CaM kinase II. Another substrate-mimicking interaction in CaM kinase II occurs at the P(5) site; Phe-293 and Asn-294 interact with Phe-98 and Ile-205, similarly observed at Phe-298 in CaM kinase I. Major differences in the inhibitory interactions of CaM kinases I and II arise in the region amino-terminal to Thr-286 of CaM kinase. Two electrostatic interactions, His-282 with Lys-148 (in the 7 and 8 loop) and Arg-283 with Asp-238 (in the G and H loop), do not permit homology modeling of helix D based on the coordinates of CaM kinase I but allow modeling based on those of PKA. The relative position of the D helix in the two CaM kinases seems to be achieved by these subtle differences in the auto-inhibitory interactions.

In summary, based on mutational analysis and homology modeling using constraints, we have developed a relatively detailed model of the three-dimensional structure of CaM kinase II that can account for the rules governing its autophosphorylation and for the molecular basis for generation of autonomous activity by autophosphorylation. As in any such study, the model presented here should be viewed strictly as a testable hypothesis, based as much as possible on currently available data. However, the insights obtained suggest a number of avenues for further study and form a framework for rational strategies for additional site-directed mutagenesis.

	ACKNOWLEDGEMENTS

We are grateful to Drs. Angus Nairn and Andy Hudmon for helpful comments on the manuscript.

	FOOTNOTES

^* This work was supported by National Institutes of Health Grant GM30179.The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

^§ To whom correspondence should be addressed. Tel.: 650-723-7668; Fax: 650-725-3958; E-mail: schulman@cmgm.stanford.edu.

Present address: Pharmaceutical Discovery Division, SRI International, Menlo Park, CA 94025.

	ABBREVIATIONS

The abbreviations used are: CaM kinase, Ca²⁺/calmodulin-dependent protein kinase; AC-2, autocamtide-2; AC-3, autocamtide-3; BSA, bovine serum albumin; HA, hemagglutinin; MLCK, myosin light chain kinase; PIPES, 1,4-piperazinediethanesulfonic acid; PKA, protein kinase A; PKI, protein kinase inhibitor.

	REFERENCES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES