Catalytic Domain Crystal Structure of Protein Kinase C-θ (PKCθ)*♦

A member of the novel protein kinase C (PKC) subfamily, PKCθ, is an essential component of the T cell synapse and is required for optimal T cell activation and interleukin-2 production. Selective involvement of PKCθ in TCR signaling makes this enzyme an attractive therapeutic target in T cell-mediated disease processes. In this report we describe the crystal structure of the catalytic domain of PKCθ at 2.0-Å resolution. Human recombinant PKCθ kinase domain was expressed in bacteria as catalytically active phosphorylated enzyme and co-crystallized with its subnanomolar, ATP site inhibitor staurosporine. The structure follows the classic bilobal kinase fold and shows the enzyme in its active conformation and phosphorylated state. Inhibitory interactions between conserved features of staurosporine and the ATP-binding cleft are accompanied by closing of the glycine-rich loop, which also maintains an inhibitory arrangement by blocking the phosphate recognition subsite. The two major phosphorylation sites, Thr-538 in the activation loop and Ser-695 in the hydrophobic motif, are both occupied in the structure, playing key roles in stabilizing active conformation of the enzyme and indicative of PKCθ autocatalytic phosphorylation and activation during bacterial expression. The PKCθ-staurosporine complex represents the first kinase domain crystal structure of any PKC isotypes to be determined and as such should provide valuable insight into PKC specificity and into rational drug design strategies for PKCθ selective leads.

regulatory domains and second messenger cofactor requirements. PKB/AKT contains an N-terminal pleckstrin homology domain regulated by phosphoinositide second messengers, a central catalytic kinase domain, and a C-terminal regulatory region facilitating key protein-protein interactions with signaling molecules like Src kinase (2). PKC kinases can be regulated by calcium, diacylglycerol, and phorbol esters and are divided into three subfamilies based on their cofactor requirements (3): conventional (PKC␣, PKC␤I, PKC ␤II, PKC␥), novel (PKC␦, PKC⑀, PKC, PKC), and atypical (PKC, PKC, PKC, PKC) isoforms. PKC kinases have a C-terminal catalytic kinase domain and an N-terminal regulatory co-factor-binding domain. The N-terminal motifs comprise the phosphatidylserine-and diacylglycerol-binding C1 motifs and calcium-binding C2 domain, in addition to a pseudosubstrate sequence motif that is regulated by cofactor binding.
The closely related PKC isoforms have been shown to have important roles in T cells (PKC␣, PKC), B cells and mast cells (PKC␤, PKC␦), and macrophages (PKC⑀), contributing to adaptive and innate immunity (3). Both PKC and PKB/AKT are implicated in T cell signaling leading to T cell activation and survival (4 -6). However, the expression and role of PKC are relatively restricted to T cells, with signaling in response to TCR stimulation contributing to T cell activation and cytokine production (7)(8)(9). PKC co-localizes to the immunological synapse in response to T cell activation (10). Thus, PKC inhibition is potentially desirable in T cell leukemias (11) and T cellmediated allergic and autoimmune disorders.
Among AGC superfamily kinases, the kinase domain crystal structures have been determined for both PKB/AKT-and cAMPdependent PKA (12) but not for a PKC isoform. The homologies in the kinase domain ATP-binding site have been a challenge in the development of highly specific inhibitors as disease therapies (1,13). Structural elucidation of kinase active sites and comparison with that of closely related family members greatly increases our understanding of the mechanism of enzyme action and divulges issues regarding selectivity. A Rho kinase (AGC superfamily) inhibitor Fasudil/HA1077/1-(5-isoquinolinesulphonyl)homopiperazine HCl, belonging to the isoquinoline sulfonamide class of compounds, also inhibits both PKA and PKC in a reversible and ATP competitive manner (14). This kinase inhibitor is a therapeutic drug in treating cerebral vasospasm and has recently been co-crystallized with the PKA catalytic subunit to define key interactions of the kinase inhibitor within the ATP binding site (15).
The crystal structure of PKA revealed that the invariant amino acids in the highly conserved kinase catalytic core are clustered at the sites of nucleotide binding and catalysis (13). The PKB/AKT active enzyme structure complexed with AMP-PNP and substrate peptide revealed mechanistic implications of key phosphorylations of the kinase domain (16). More re-* 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 1 The abbreviations used are: PKC, PKA, and PKB, protein kinase C, A, and B, respectively; PKI, protein kinase inhibitor; AMP-PNP, adenosine 5Ј-(␤,␥-iminotriphosphate); DTT, dithiothreitol; bis-tris, 2-[bis(2hydroxyethyl)amino]-2-(hydroxymethyl)propane-1,3-diol; PDB, Protein Data Bank; N-and C-lobe, N-and C-terminal lobe, respectively; HM, hydrophobic motif. cently, the comparison of the PKB/AKT structure with PKA structure (17) provided explanations for distinct substrate specificities of the similar active kinase domain conformations. Yang et al. (16) have also speculated potential PKC, SGK, p70, and p90 S6 kinase substrate interactions based on homologies with the GSK-3 substrate binding residues of PKB/AKT. Till now these mechanistic insights remained to be confirmed by a PKC crystal structure, and the studies presented here attempt to define PKC kinase domain characteristics. We report the x-ray structure elucidation of the PKC catalytic domain, with mechanistic insights into similarities and distinctions from the closely related PKB/AKT and PKA catalytic domains. In addition, the structural information of the staurosporine-complexed PKC kinase domain presented here will aid in the rational design and optimization of selective small molecule inhibitors for therapeutic use for inhibiting PKC in specific targeting of T cells.

EXPERIMENTAL PROCEDURES
Expression and Purification-The C-terminal catalytic domain of PKC (residues 362-706) was cloned into a pET-16b expression vector. This vector introduced a hexahistidine tag to the C terminus of the expressed protein, and a methionine-glycine amino acid pair was introduced at the N terminus by cloning. The plasmid was used to transform Escherichia coli strain BL21-DE3 for overexpression. A 10-liter cell culture was expanded at 37°C to an A 600 of about 0.4. The temperature was then lowered to 25°C before addition of isopropyl ␤-D-thiogalactopyranoside to a final concentration of 0.1 mM to induce expression. The cells were grown for an additional 4 h before they were harvested.
Harvested cells were resuspended in 25 mM Tris, pH 8.0, 25 mM NaCl, 5 mM 2-mercaptoethanol, 5 mM imidazole, 50 M ATP, and protease inhibitors and lysed using a microfluidizer. The lysate was applied to 20 ml of nickel-nitrilotriacetic acid resin for 1 h at 4°C. The resin was subsequently poured as a chromatography column and washed extensively with the same buffer including 25 mM imidazole. Protein bound to the resin was eluted with 200 mM imidazole buffer and then immediately loaded onto an anion exchanger HQ. The column was washed with 25 mM Tris, pH 8.0, 25 mM NaCl, 5 mM DTT, 50 M ATP before being resolved by the application of a linear gradient from 25 to 500 mM NaCl. Fractions containing PKC were selected by SDS-PAGE, pooled, and diluted 2-fold with 25 mM Tris, pH 8.0, 5 mM DTT and loaded onto a heparin chromatography column. The flow-through was applied to a hydroxyapatite column and washed extensively with 25 mM Tris, pH 8.0, 50 mM NaCl, 5 mM DTT. A linear gradient of sodium phosphate from 0 to 100 mM eluted the target protein. The protein was then sized as a monomer on a Superdex 200 size exclusion chromatography column. The purified protein was dialyzed overnight against 25 mM Tris, pH 8.0, 50 mM NaCl, 5 mM DTT and concentrated to 7 mg/ml (determined by the Bradford assay) before being used for crystallization experiments.
Kinase Assays and Data Analysis-ATP, ADP, phosphoenolpyruvate, NADH, pyruvate kinase/lactate dehydrogenase enzymes, staurosporine, acetonitrile, and the buffer HEPES were purchased from Sigma. Peptide substrate was purchased from AnaSpec (San Jose, CA), Syn Pep (Dublin, CA), or Open Biosystems (Huntsville, AL). The enzymatic activity determined using the coupled pyruvate kinase lactate dehydrogenase assay followed spectrophotometrically at 340 nm. The standard reaction, except where indicated, was carried out in 25 mM HEPES, pH 7.5, 10 mM MgCl 2 , and 2 mM DTT, 0.008% Triton, 100 mM NaCl, 20 units of pyruvate kinase, 30 units of lactate dehydrogenase, 0.25 mM NADH, 2 mM phosphoenolpyruvate. The PKC concentration was in the range of 0.156 -0.312 g/ml. The kinetic analysis was carried out in a 384-well plate at 25°C on a Molecular Devices spectrophotometer in a final volume of 0.080 ml. The steady state kinetic parameters were determined in the buffer described above containing varied sucrose or Ficoll 400. Data were fitted to Equation 1 for normal Michaelis-Menten kinetics, where [S] is the substrate, V max is the maximum enzyme velocity, and K m is the Michaelis constant. For inhibition kinetics the data were fitted to a competitive inhibition model in Equation 2, where K is is the slope inhibition constant. The data were analyzed using Sigma Plot 2000 Enzyme Kinetics Module from SPSS Science (Richmond, CA). Crystallization and Structure Determination-Crystals with staurosporine were obtained at 18°C from hanging drops containing 1 l of protein:staurosporine solution (at ϳ1:1 molar ratio) and 1 l of precipitating solution (2 M ammonium sulfate, 40 mM DTT, and 0.1 M bis-tris, pH 5.0). The crystals belong to the monoclinic space group C2, with one protein-staurosporine complex in crystallographic asymmetric unit. Prior to data collection, crystals were stabilized in solution containing mother liquor plus 25% glycerol and flash-frozen in a 100 K nitrogen stream. The x-ray diffraction data were collected to 2-Å resolution at the Advanced Light Source (Berkeley, CA) using a Quantum-4 CCD detector (Area Detector Systems) and then reduced and scaled with HKL2000 (18). The structure was solved by molecular replacement with AMORE (19) using the structure of PKA (PDB code: 1STC (20)), in which the A-helix, glycine-rich loop, activation loop, and C-terminal tail were omitted from the search model. The rotation and translation function solutions were found using data from 8 to 3.5 Å. The BUSTER where ͗I h ͘ is the average intensity over symmetry equivalents. Numbers in parentheses reflect statistics for the last resolution shell (2.07-2.0 Å). b R work ϭ ¥͉͉F obs Ϫ ͉F calc ͉/¥͉F obs ͉, where R free is equivalent to R work but calculated for a randomly chosen 4% of reflections omitted from the refinement process.  (22) were applied in generating maximum entropy omit maps to overcome model bias and to produce a more detailed map for the bound inhibitor. Several rounds of rebuilding (QUANTA, Molecular Simulations, Inc.) and refinement were performed. To further reduce model bias and to generate better maps, an "average map" was calculated using CNS (23) by overlapping seven protein kinase coordinates including those of PKA. The resulting electron density maps were of better quality, especially for loop regions in the N-terminal lobe (N-lobe). The model was further rebuilt and refined, and the quality of the model was judged by the decrease in R-factors. Refinement converged after many rebuilding cycles to an R-factor of 0.201 and R free of 0.216. Crystallographic data collection and refinement statistics are summarized in Table I. The final model contains protein residues 377-649 and 688 -696, two phosphate groups attached at Thr-538 and Ser-695, one staurosporine molecule, and 115 water molecules. Residues 362-376 from the N terminus, C-terminal region 650 -687, and residues 697-706 at the very C terminus were not detected in the electron density maps due to disordering. Structural figures were generated using PyMOL (24) and QUANTA (Molecular Simulations, Inc.).

Modeling of Peptide Substrates Bound to PKC-
The PKC-staurosporine complex structure was aligned to a structure of PKB-␤ (AKT-2) in complex with a GSK-3 peptide (16), PDB code: 106L) via automatic alignment (using weights of 1.0 for both sequence homology and structural homology) in the Protein Design module of Quanta (Accelrys (2004), San Deigo, CA). The alignment was further adjusted manually to improve the overlap of the "hinge region" backbone (residues 459 -461 in PKC) between the two structures. The initial positioning in the PKC structure of the ATP analog AMP PNP in the ATP binding site, the peptide in its binding site, and the glycine-rich loop was determined by the alignment. The GSK-3 peptide was mutated to the PKC activation segment ( 532 GDAKTNTFCG 541 ) peptide in one case and the hydrophobic motif peptide ( 689 NMFRNFSFMN 698 ) in the other case. The conformation and position of residues 386 -395 (the glycine-rich loop plus two residues on either side) was taken from the PKB-␤ structure. First, the attachment points for the loop, residues 386 -388 and 696 -698, were energy minimized keeping the remainder of the structure fixed. Next, the peptide, the glycine-rich loop, AMP-PNP, and the surrounding residues (with an atom within 8 Å) were minimized subject to decreasing harmonic constraints. Finally, the peptide was

Analysis of Protein Construct and Overall Structure-
The bacterially expressed PKC kinase domain used for structure determination (residues 362-706) showed higher molecular weight than expected. Treatment by -phosphatase and subsequent molecular mass determination by electrospray ionization-mass spectrometry indicated that the protein is phosphorylated at either six or seven amino acid residues (roughly a 50:50 mixture). The purified PKC kinase domain was shown to have a higher specific activity than the full-length, commercially available PKC. Further characterizations of its catalytic activity demonstrated that it has an apparent K m of ϳ49 M for ATP and K m of ϳ6.5 M for a peptide derived from the PKC␣ pseudosubstrate, as indicated in Table II. Details of mass spectrometric analysis of phosphorylation sites and catalytic domain enzymatic characterization will be published elsewhere. 2 Consistent with the above data and as discussed further below, the enzyme crystallized in its phosphorylated state and in an active conformation.
As outlined in the Introduction, the kinase domains of PKA, PKB/AKT, and PKC are highly homologous. Within the PKC subfamily, isozymes display more than 60% sequence identity in the kinase domain and share three conserved phosphorylation motifs (Fig. 1). The overall fold of the catalytic domain of PKC is very close to other protein kinase structures solved, with most similarities to those of PKB/AKT and PKA (Fig. 2). The conserved core of the structure is made of a small Nterminal lobe (residues 377-461) and a large C-terminal lobe (C-lobe) (residues 466 -696), connected by a hinge linker (res-idues 462-465). The N-lobe is based on a five-stranded ␤-sheet (␤1-␤5) and two ␣-helices (␣B and ␣C), and the C-lobe is mostly helical consisting of eight ␣-helices (␣D-␣K). The ATP-binding site, occupied by staurosporine, with the adjacent peptide-substrate binding site open to solvent, constitute the active site cleft at the interface of the two lobes. The glycine-rich phosphate-binding loop (GXGXXG), which shows a broad range of conformations (12), including multiple conformations observed within a single crystal (15), connects the ␤1 and ␤2 strands (residues 386 -394) and adopts a fixed and closed conformation.
Catalytic key residues (Lys-409, Asp-504, and Asp-522), invariant in all protein kinases, preserve intramolecular interactions observed in active kinase structures, in accordance with the structural criteria used to define catalytically active kinase conformations (12). As in most Ser/Thr kinase structures reflecting active enzymes (28), helix ␣C (residues 421-437) is properly aligned for substrate binding and catalysis, and the activation loop (residues 522-544) bearing the essential phosphothreonine Thr-538(P) (29) is well ordered and in an extended conformation.
The C-terminal hydrophobic motif (HM) FXXFS* (residues 691-695), another conserved feature across AGC family, is adjacent to the hydrophobic groove of the N-lobe, in a location similar to the FXXF-binding pocket in PKA and PKB. Like in PKB, but in contrast to PKA that terminates with the FXXF sequence, HM in PKC contains phosphoserine at position 695 (Ser-695(P)), a phosphorylation that substantially effects the PKC kinase activity (29).
PKC has an additional conserved phosphorylation site referred to as the turn motif (residues 662-685 and Fig. 1). In the structure, a long polypeptide linker between the kinase domain and the C-terminal HM (residues 650 -687) is disordered; therefore the region corresponding to the turn motif is not defined in the electron density maps. There is also no observable electron density for residues C-terminal to HM (residues 697-706). In both PKA and PKB, the corresponding C-terminal linker is structurally ordered extending across the ATP-bind- ing cleft (Fig. 2B). Compared with PKB, where HM is contained in the 17-residue-long stretch, HM seen in the PKC structure is considerably shorter (residues 688 -696) and displays high b-factor values. Together these observations indicate that the C-terminal tail of PKC encompassing the turn motif and HM is intrinsically flexible either in this particular crystal form or in the absence of other functional domains or substrates.
Binding of Staurosporine-The natural broad-spectrum kinase inhibitor staurosporine, with micromolar potency against only few kinases and low nanomolar potency against most kinases, has been shown to have a higher degree of selectivity toward PKC kinases (as reviewed in Ref. 30). Our kinetics data on PKC activity in the presence of staurosporine indicate it to be a strong, ATP competitive inhibitor with a K i value of 0.33 nM (Table II). Associations maintaining this tight interaction are clearly and well defined by electron density (Fig. 3A) and will be described in comparison with the binding of staurosporine to PKA (PDB code: 1STC (20)).
As expected, these high affinity ligands bind to the related enzymes in a generally similar binding mode. Staurosporine resides in the ATP-binding pocket, forming four potential hydrogen bonds with the protein backbone (Fig. 3A) and extensive van der Waals contacts with the surrounding residues from both lobes and the hinge linker (Table III) Table III (contacts calculated within 4 Å). For example, relatively more staurosporine to main chain van der Waals contacts are observed in PKC than in PKA (43.5 versus 36.5%), which is largely attributed to the atoms in the phosphate binding loop. In PKA, on the other hand, additional side chain contacts to the inhibitor are provided from the residue of the C-terminal tail (Phe-327), a region that is of high disorder in our structure.
Despite this and other differences, the total number of polar and van der Waals interactions between staurosporine and PKC (4 polar and 85 van der Waals) is identical to the number calculated for staurosporine-PKA (Table III). In addition, the buried surface areas of staurosporine are comparable for the two enzymes (569 Å 2 in PKC and 563 Å 2 in PKA, calculated using a solvent probe radius of 1.4 Å). It seems unlikely, therefore, that these rather subtle structural variations are the only factors that account for at least a 20-fold difference in inhibition and are sufficient to drive selectivity (staurosporine K i against PKA ϳ8 nM). Below, we analyze other structural features such as ligand-dependent conformational states of kinase domains that could be a contributing factor to inhibitor affinity/specificity.
The aligned structures of the PKC-and PKA-staurosporine complexes display remarkably good agreement in overall conformation (Fig. 2B). Accordingly, comparative analysis with three so far reported main conformational states of protein kinases, "open," "intermediate," and "closed" (reviewed in Ref. 31) indicates that, except for the glycine loop, relative disposition of the N-and C-lobes in PKC represents intermediate lobe structures (32), (20). In most intermediate kinase structures, including complexes with staurosporine, the glycine-rich loop also adopts a position that is intermediate between open and closed states (31). In contrast, and as represented in this crystal structure, the tip of the glycine-rich loop (residues 389 -392) moves deep inside the phosphate-binding subsite, approaching staurosporine and assuming positions that would clash with the nucleotide phosphate moiety. As a result, two atoms of Ser-390 engage in hydrogen bonding with the metal binding residue Asp-522 (Asp-184 in PKA), and Phe-391 swings into the active site approaching the catalytic Lys-409 (Lys-72 in PKA) and helix ␣C (Fig. 3B). These interactions shield the phosphate-binding site, promoting and stabilizing its closure so that the resulting conformation of the glycine-rich loop is even more closed than is apparent in PKA and other kinase complex structures (see legend to Fig. 3 and Ref. 31). Due to this, and in contrast to PKA, access to the residues that are crucial for catalysis is largely restricted. Consequently, on binding to PKC staurosporine appears to be able to exploit not only adaptational changes in lobe orientation but also changes in the glycine loop structure that ensure that the phosphate recognition site is effectively blocked.
The binary complex represented in this crystal structure highlights flexibilities of protein kinases to accommodate ligand-induced binding effects and offers an explanation of the higher inhibitory activity of staurosporine against PKC. Whether the binding mode of staurosporine to other PKC isotypes, at least to their active phosphoryalted forms, employs the same combination of general and specific features as observed here for PKC remains to be seen. Further structural analysis of inhibitors complexed with PKC kinases in both active and inactive states is required to address the latter possibility.
Activation Loop and Helix ␣C-The activation loop and the central ␣C-helix share crucial roles in catalysis. The activation loop provides part of the binding surface for peptide substrates and together with helix ␣C also serves as a docking site for activating or inactivating co-factors (28). In most kinases these two highly variable structural regions fold into an active conformation as a consequence of phosphorylation on the activation loop (12). For PKCs, this is a single phosphorylation site at threonine located in the activation loop sequence between the invariant DFG and TPD motifs (Fig. 1). It corresponds to Thr-538 in PKC and represents the only site, phosphorylation of which is essential for enzyme function (29). It is also the most critical site for in vitro kinase activity, as substitution of Thr-538 for alanine (T538A) results in a significant ϳ100-fold decline in activity (29). By comparison with structures of other kinases, including those that are inactive due to lack of phosphorylation, the structure of PKC clearly reveals that the essential phosphate has a major impact on the conformation of the active site.
As shown in Fig. 4A, all three oxygens of the essential phosphate Thr-538(P) are involved in hydrogen-bonding interaction, consistent with the results of the in vitro activity studies of the full-length PKC in which Thr-538 was substituted for glutamate (29). The T538E mutant has shown a 3-fold decrease in activity compared with the wild-type enzyme, indicating that the glutamic acid at this position would provide only a partial mimic of the phosphoamino acid. Thr-538(P) is placed beneath the ␣C-helix to compensate for the conserved cationic cluster formed by Arg-503 and Lys-527 (Arg-165 and Lys-189 in PKA). As in other related kinases, these ionic interactions, on one hand, provide a direct link to the catalytic loop helping stabilize the correct orientation of the catalytic base Asp-504 (Asp-166 in PKA), and on the other, allow for the correct positioning of the ␣C-helix, which is now properly aligned to place the invariant Glu-428 (Glu-91 in PKA) for hydrogen bonding with the DFG motif and with the catalytic Lys-409 (Lys-72 in PKA). As has been shown previously, a K409R mutation eliminates the catalytic activity of PKC (11), confirming that this residue is essential for catalysis. In addition, and similar to PKA, Thr-538(P) hydrogen bonds with the side chain of the preceding Thr-536, an interaction that further tethers the phosphate ion to the activation loop and may help stabilize it.  Much of this networking is similar to the equivalent interactions in phoshorylated PKA and PKB. However, outside of this region, there are several features inherent to the PKC structure and not found in either the PKA or PKB kinases. In both PKA and PKB, helix ␣C presents a histidine side chain to contact the phosphate of threonine on the activation loop (His-87 in PKA and His-196 in PKB). In PKC, in a structurally equivalent position to His-87, at the tip of the helix, lies a strictly conserved cysteine residue, Cys-424 ( Figs. 1 and 4). Because of this, Thr-538(P) is unable to form the equivalent, electrostatic or hydrogen-bonding, contact. Instead, there is an alternative double salt bridge interaction that links the ␣C helix directly to the activation loop but does not engage the phosphate ion: from Arg-430, two turns along the helix, to Glu-528 that sits just after Lys-527 and directly under Arg-430. Along with this, formation of an ion pair between Arg-430 and Glu-423, a helix capping interaction, appears to facilitate the correct orientation of the arginine chain and hence the anchoring of Glu-528 to the outer surface of helix ␣C (see also Fig. 5). Constellation of the three invariant residues of complementary charges, Glu-528, Arg-430, and Glu-423, placed at this interdomain interface is unique to PKC kinases (Fig. 1) and as such is not shared by PKA or PKB. In structural terms, these newly formed, ϳ13 Å away from Thr-538(P), hydrogen-bonding interactions seem to accomplish the same role as a histidine-phosphothreonine contact in PKA and PKB by contributing to the correct positioning of both the ␣C helix and the activation loop. They may also have functional assignments specific for the PKC family, for example, by keeping interaction of the activation loop with the long ␣C helix fixed at this side while rendering easier access to the N-terminal side of the helix and Cys-424 therein, which may be required for substrate/effector binding.
Further comparison with AGC kinase structures shows that the ␣C helix of PKC contains a single residue insertion that creates an additional helical turn at its C terminus and shortens the ␣C-␤4 loop (see inset in Fig. 5). The first residue in the last turn, Trp-436, is largely exposed to solvent and forms part of the surface that faces the C-terminal HM (Fig. 5). While sequence analysis suggests that all PKC members will share a similar whole helical turn insert, a tryptophan insertion is not a conserved feature of this family (Fig. 1). Therefore, PKC may be unique in how this tryptophan residue specifically shapes and stabilizes the surface on which its own HM segments dock. In general, the insertion of an additional turn into the ␣C helix will impose conformational constraints on the flexibility of the ␣C-␤4 loop and the mobility of the ␣C helix thereby helping to tether the critical helix in a productive position. As activation through orientation of the central ␣C helix is common to other protein kinases, it is possible that these additional conformational constraints, not observed in either of the AGC kinase structures, have enzyme-specific roles in the activation or regulation of PKC family.
Hydrophobic Motif-The HM phosphorylation of PKC was shown to be required for optimal enzyme activity, with a 5-fold reduction of kinase activity in the full-length PKC S695A mutant immunoprecipitated from transfected HEK293 cells (29). Crystallographic data on PKB indicate that the structural role of the hydrophobic phosphorylation site is to tighten intramolecular association between the HM and the N-lobe to align the ␣C-helix favorably for catalysis (16,33). The nature of interactions between the C-terminal HM, including phosphoserine 695, and the hydrophobic groove of the N-lobe is preserved in PKC, indicating that phosphorylation at this site plays an important and similar role in the structure of PKC.
The molecular surface of the N-lobe shows a channel, which hosts three characteristic aromatic side chains of the HM motif, Phe-691, Phe-694, and Phe-696 (Fig. 5). The deeply buried rings form extensive hydrophobic contacts with the surrounding residues from the ␣B and ␣C helices and ␤4 and ␤5 strands. The equivalent phenylalanine residues in PKA and PKB have been shown essential for protein stability and catalytic activity (34,41). There are also several hydrogen-bonding contacts extending on both sides of the channel and anchoring the backbone of the HM (residues 691-696) to the ␣C helix (Lys-429) and to the ␤4 strand (residues 447-449). Finally, the top of the channel is capped by the phosphate ion of Ser-695(P). The latter forms two hydrogen bonds with the invariant Gln-449 ( Fig. 1) from the ␤5 strand, an observation consistent with analysis of PKC activity in vitro, which indicates that the glutamic acid at the phosphor acceptor position (S695E) has slightly reduced and not abolished activity (29). These results also correlate with the least conservation of this phosphorylation site, which is reflected in atypical PKC kinases that share glutamate as a phosphate mimic at this position (35).
Concluding Remarks-With regard to PKC phosphorylation, the results described here strongly suggest that phosphorylation of the PKC kinase domain in E. coli is autocatalytic and are further supported by our mutation analysis in which the catalytically defective PKC kinase domain mutant K409W expressed in E. coli was found unphosphorylated. 2 This appears to be in accord with the previous finding that the kinasedead full-length PKC mutant K409W is not phosphorylated in growing HEK293 cells (29). This study also demonstrated that the PKC mutant T538A had completely lost phosphorylation of its hydrophobic motif and, to a certain extent, of its turn motif, indicating that both motifs represent sites of autophosphorylation and that their phosphorylation is regulated by the activation loop (29).
Interestingly, in contrast to the majority of other PKC isoforms ( Fig. 1), the sequences surrounding the two critical phosphorylation sites of PKC, Thr-538 (KTNT*F) and Ser-695 (RNFS*F), both contain a positively charged residue at position P-3 (Lys-535 and Arg-693, respectively), which is compatible with the preferred substrate sequence R/KXXT*/S*F recognized by PKC family (36). A model of the peptide corresponding to the activation segment of PKC (residues 532-541) and an ATP analog bound to PKC (Fig. 6) explains the preference for positively charged residue at the P-3 position in substrates. In the model, the P-3 Lys-535 of the peptide interacts with Asp-465 and the ribose group of ATP. Also, Thr-538 of the peptide is positioned to react with a phosphate group of ATP as expected. A model of the hydrophobic motif peptide and an ATP analog bound to PKC (not shown) show similar interactions.
The autophosphorylation model for the PKC kinase domain would be analogous to the model described for the catalytic subunit of PKA, in which the phosphates can all be introduced autocatalytically (37). On the other hand, recent results with other PKC isotypes (reviewed in Ref. 38) suggest that the in vivo phosphorylation reaction depends on the universal PKC upstream kinase PDK1 and that PDK1-mediated activation loop phosphorylation followed by autophosphorylation at the HM is a general regulatory mechanism for PKCs (35). Although physical association of PKC with PDK1 has been demonstrated (29), there is still a lot to be determined to indicate whether or not PDK1 is responsible for the constitutive phosphorylation of the activation loop of PKC or whether it is primarily regulated by autophosphorylation.