Unique Structural and Functional Properties of the ATP-binding Domain of Atypical Protein Kinase C-ι*

Atypical protein kinase C-ι (aPKCι) plays an important role in mitogenic signaling, actin cytoskeleton organization, and cell survival. Apart from the differences in the regulatory domain, the catalytic domain of aPKCι differs considerably from other known kinases, because it contains a modification within the glycine-rich loop motif (GXGXXG) that is found in the nucleotide-binding fold of virtually all nucleotide-binding proteins including PKCs, Ras, adenylate kinase, and the mitochondrial F1-ATPase. We have used site-directed mutagenesis and kinetic analysis to investigate whether these sequence differences affect the nucleotide binding properties and catalytic activity of aPKCι. When lysine 274, a residue essential for ATP binding and activity conserved in most protein kinases, was replaced by arginine (K274R mutant), aPKCι retained its normal kinase activity. This is in sharp contrast to results published for any other PKC or even distantly related kinases like phosphoinositide 3-kinase γ, where the same mutation completely abrogated the kinase activity. Furthermore, the sensitivity of aPKCι for inhibition by GF109203X, a substance acting on the ATP-binding site, was not altered in the K274R mutant. In contrast, replacement of Lys-274 by tryptophan (K274W) completely abolished the kinase activity of PKCι. In accordance with results obtained with other kinase-defective PKC mutants, in cultured cells aPKCι-K274W acted in a dominant negative fashion on signal transduction pathways involving endogenous aPKCι, whereas the effect of the catalytically active K274R mutant was identical to the wild type enzyme. In summary, aPKCι differs from classical and novel PKCs also in the catalytic domain. This information could be of significant value for the development of specific inhibitors of aPKCι as a key factor in central signaling pathways.

Protein kinase C (PKC) 1 (1) is a family of Ser/Thr kinases involved in signal transduction pathways triggered by numerous extracellular and intracellular stimuli. PKC isoenzymes have been shown to play an essential role in a broad range of cellular functions including mitogenic signaling (1)(2)(3)(4)(5), cytoskeleton rearrangement (6,7), glucose metabolism (8 -12), differentiation (13)(14)(15), and regulation of cell survival and apoptosis (16 -22). At least 11 different members of the PKC family have been identified so far. Based on structural similarities and cofactor requirements they have been grouped into three subfamilies: 1) the classical or conventional PKCs (cPKC␣, ␤ 1 , ␤ 2 and ␥), activated by Ca 2ϩ , diacylglycerol and phosphatidylserine; 2) the novel PKCs (nPKC␦, ⑀, , and ), independent of Ca 2ϩ but still responsive to diacylglycerol; and 3) the atypical PKCs (aPKC and /, where PKC is the mouse homologue of human PKC). aPKCs differ significantly from all other PKC family members in their regulatory domain in that they lack both the calcium-binding domain and one of the two zinc finger motifs required for diacylglycerol binding (reviewed in Ref. 23). These differences lead to a different requirement of cofactors for activation; aPKCs are insensitive to Ca 2ϩ and diacylglycerol and exhibit an elevated basal enzymatic activity. Instead of Ca 2ϩ and lipids there exist several protein regulators like -interacting protein and -interacting protein, both activators of aPKCs, and PAR-4 (prostate apoptosis response), an inhibitor of aPKCs (24,25).
Despite the fact that the regulatory domain of PKCs seems to serve essential regulatory functions, constructs with mutations within the regulatory domain rendering the PKC isoenzymes constitutively active still maintain their in vivo selectivity that must in part reside within the catalytic domain. However, although the isoenzyme-specific differences in the regulatory domain and their role in the activation of the three different PKC subfamilies are well understood, little is known about the functional implication of the subtle structural differences in the catalytic domain. The catalytic domain is characteristic for all protein kinases, with a remarkable degree of sequence identity between the more than 180 known mammalian kinases, and highly conserved from yeast to man. The 200 -300 amino acid residues of the catalytic core region are predicted to fold into a common three-dimensional structure, as revealed by crystallization of several protein kinases (26 -29). Some protein kinase motifs are extremely well conserved. One of these motifs is the glycine-rich loop with the consensus sequence GXGXXG located in the first of 12 subdomains of the protein kinase catalytic domain. The three glycine residues of the glycine-rich loop are conserved in over 95% of the known human kinases known so far and even in other nucleotide-binding proteins such as the small G proteins Ras, Rac, and Rho. They fold into a ␤-strandturn-␤-strand structure and form a flexible clamp that covers and anchors the nontransferrable phosphates of ATP (reviewed in Ref. 30). The glycine residues provide the flexibility necessary for anchoring the ATP molecule and excluding the water; for this reason they are essential both for effective catalytic activity and low ATPase activity because of phosphate transfer on water molecules. The third glycine also forms a hydrogen bond with the ATP ␤-phosphate oxygen. Point mutations in any of the conserved glycines generally lead to a loss of enzymatic activity and have been shown to be responsible for enzymatic defects leading to human diseases. Examples include a form of diabetes where the third glycine in the ATP-binding site of the insulin receptor is substituted by valine, leading to insulin resistance (31), and the Ras-V12 oncogene involved in tumor development, where the second glycine is replaced by valine (32,33). It is of special interest that in aPKCs the third glycine is substituted by alanine, suggesting a unique mechanism of ATP binding with respect to other PKCs and unrelated kinases.
The most conserved and probably best characterized residue of protein kinases is the so-called "invariant lysine" in subdomain II of the catalytic domain, corresponding to Lys-274 in human PKC. It is directly involved in the phosphotransfer reaction and interacts with the ␣and ␤-phosphates of ATP, thereby anchoring and orienting the nucleotide. Additionally, for the catalytic subunit of PKA this residue has been shown to form a salt bridge with the carboxyl group of the nearly invariant Glu-91 in subdomain III (corresponding to Glu-293 in aPKC). Replacement of the invariant lysine by any other amino acid, including arginine, generally leads to a catalytically inactive kinase, and site-directed mutagenesis of this residue has almost become a standard approach to generate kinase-dead mutants. This has been shown for PKC, , ␣, and ⑀ (1, 7, 34), Src (35), the epidermal growth factor receptor (36), the insulin receptor (37), Mos (38), Fps (39), and many other kinases. Interestingly, the invariant lysine is also conserved in the lipid kinase phosphoinositide 3-kinase, where a conservative K799R mutation not only completely abolishes ATP binding and kinase activity but also blocks the interaction with the phosphoinositide 3-kinase inhibitor wortmannin (40,41).
Despite the fact that the ATP-binding site in the catalytic domain belongs to the most conserved sequences within the protein kinase family, the most specific inhibitors of PKC bind to this part of the molecule and act by blocking ATP binding to the enzyme. This is the case for the bisindolylmaleimide GF109203X and other homologues of the potent PKC inhibitor staurosporine. It is intriguing that even within the PKC family these inhibitors exhibit clear isoenzyme specificity, affecting atypical isoforms more than 100-fold less than classical or novel PKCs (7). This fact together with the remarkable difference within the glycine-rich loop compared with other kinases suggests that, surprisingly, atypical PKCs differ significantly in the structure of their catalytic domain from other PKC isoforms and protein kinases in general. Because such a difference on one hand could in part explain the in vivo isoenzyme specificity and, on the other hand, should be useful for the development of specific aPKC inhibitors, we decided to exactly characterize the enzymatic properties of atypical PKC. To proof the assumption of a unique ATP-binding domain of ␣PKCs and to see whether this is functionally reflected in the mode of ATP binding, we used site-directed mutagenesis to replace the central ATP-binding residue, the invariant lysine (Lys-274), by the chemically similar arginine or by tryptophan and measured the enzymatic activity of the resulting mutant proteins. The PKC-K274R mutant exhibited normal enzymatic activity, and kinetic properties are very similar to the wild type enzyme. In contrast, the substitution of Trp for the invariant Lys-274 (aPKC-K274W) resulted in a catalytically inactive enzyme. The same behavior could be observed when the PKC mutants were studied in intact cells. This is in sharp contrast to virtually all other protein kinases, where the invariant lysine seems to be essential for correct folding of the kinase domain and interaction with ATP. To our knowledge, similar results have only been published for the very distantly related CDK-activating kinase from budding yeast (42). These results confirm that aPKC possesses a unique ATP-binding domain, and they show that this results in an unusual mechanism of ATP binding and probably catalysis. This knowledge should be useful for the development of specific inhibitors of atypical PKCs as central mediators of essential cellular signaling pathways.

EXPERIMENTAL PROCEDURES
Plasmids and Reagents-The plasmids PKC-WT, PKC-A120E, and PKC-K274R were kindly provided by T. Biden (Garvan Institute of Medical Research, Sydney, Australia) (8). All PKC cDNAs were Cterminally tagged with RGS-His 6 using standard polymerase chain reaction procedures and subcloned into the expression vector pEF-neo. The plasmids PKC␣-AE, RasL61, and fos-Luc have been described previously (1). GF109203X was obtained from Calbiochem, and FSBA was from Sigma.
Construction of the PKC Mutants-All point mutations were generated using polymerase chain reaction with the PKC-wild type construct pEF-PKC-WT as a template, except for PKC-K274W/T555E. The mutagenesis primers were 5Ј-GGC AGA TCG TAT TTA TGC AAT GTG GGT TGT GAA AAA AGA GCT TGT TAA TG-3Ј and 5Ј-CAT TAA CAA GCT CTT TTT TCA CAA CCC ACA TTG CAT AAA TAC GAT CTG CC-3Ј for PKC-K274W, 5Ј-CTA ATG AAC CTG TCC AGC TCG AGC CAG ATG ACG ATG ACA TTG TG-3Ј and 5Ј-CAC AAT GTC ATC GTC ATC TGG CTC GAG CTG GAC AGG TTC ATT AG-3Ј for PKC-T555E. PKC-K274W/T555E was created with PKC-K274W as a template, using the T555E mutagenesis primers. The anchor primers for all polymerase chain reactions were 5Ј-AGT CAG GAG ATG CCG ACC CAG AGG-3Ј and 5Ј-TCT AGA TCA TCA ATG ATG ATG ATG GTG ATG GGA TCC CCG ACT AGT GAC ACA TTC TTC TGC AGA CAT-3Ј. The polymerase chain reaction products were inserted into the pEFneo vector using the restriction enzymes EcoRI (5Ј) and SpeI (3Ј).
Cell Culture and Transfection Protocols-COS-7 cells were grown in Dulbecco's minimal essential medium (Biochrom KG, Berlin, Germany) containing 1.028 g/liter N-acetyl-L-alanyl-L-glutamine and 4.5 g/liter D-glucose, supplemented with 10% heat-inactivated fetal calf serum and 50 g/ml gentamycin. HC-11 cells were cultivated in RPMI 1640 medium containing 25 mM HEPES and 446 mg/liter L-alanyl-L-glutamine supplemented with 10% fetal calf serum, 50 g/ml gentamycin, 5 g/ml insulin (Sigma), and 10 ng/ml epidermal growth factor (Sigma). For transfection, cells were seeded in 6-well plates at a density of 200,000 cells/well. After 6 h, the medium was replaced by 1 ml of Optimem medium (Life Technologies, Inc.). 1.5 g of transfection plasmid/well was mixed with 0.5 ml of Optimem and 4.5 l of Lipofectin (Life Technologies, Inc.) according to the manufacturer's protocol, added to the cells and incubated overnight. After removal of the transfection mix, cells were washed with phosphate-buffered saline, 1 ml of fresh medium was added, and the cells were incubated for additional 24 h.
Purification of RGS-His 6 -tagged PKC-Transiently transfected cells were scraped off in phosphate-buffered saline and centrifuged at 1,000 ϫ g for 5 min. The pellet was resuspended in 500 l of lysis buffer containing 150 mM NaCl, 50 mM HEPES, pH 7.5, 1% Nonidet P-40, 50 g/ml leupeptin, 50 g/ml aprotinine, and 1 mM phenylmethylsulfonyl fluoride and incubated on ice for 10 min. Cell lysates were centrifuged at 10,000 ϫ g for 10 min, and the supernatant was transferred to a fresh tube. 100 l of Ni 2ϩ -NTA agarose (Qiagen) equilibrated in lysis buffer was added, and the tubes were rotated for 1 h to allow binding of RGS-His 6 -tagged PKCs to Ni 2ϩ -agarose. The purified PKCs were then washed four times with lysis buffer supplemented with 50 mM imidazole and eluted with 500 mM imidazole in 20 mM Tris-HCl, pH 7.5.
Measurement of in Vitro PKC Activity-100 ng of purified PKC were added to 100 l of assay buffer containing 20 mM Tris-HCl, pH 7.5, 20 mM MgCl 2 , 50 M substrate peptide PKC␣-19 -31/Ser-25 (Alexis), 40 M ATP (Roche Molecular Biochemicals), and 1 Ci of [␥-33 P]ATP (PerkinElmer Life Sciences). The reaction was incubated at 30°C for 10 min and stopped on ice. 50 l of the reaction mix were transferred to phosphocellulose disc sheets (Life Technologies, Inc.). The phosphocellulose sheets were washed three times with 1% phosphoric acid and twice with distilled water and then transferred to a scintillation vial with 4 ml of scintillation fluid. The bound radioactivity was then measured in a liquid scintillation counter.
Detection of PKC Levels by Western Blot Analysis-20 ng of purified PKC was separated on a 10% polyacrylamid gel and transferred to a polyvinylidene fluoride membrane (Immobilon-P TM , Millipore) by electroblotting. The membrane was equilibrated with Tris-buffered saline (10 mM Tris-HCl, pH 8.0, 150 mM NaCl) and blocked with Tris-buffered saline containing 1% Tween-20 and 1% skim milk powder overnight. The first antibody was added in a concentration of 0.1 g/ml for the anti-RGS-His 6 antibody (Qiagen) and 0.25 g/ml for the anti-PKC antibody (BD Transduction Laboratories), and the membranes were incubated for 1 h. After washing with Tris-buffered saline containing 1% Tween-20, the blot was incubated for 1 h with the secondary antibody (anti-mouse horseradish peroxidase, 0.5 l/ml; Amersham Pharmacia Biotech). Detection of the protein bands was performed using Supersignal Reagent (Pierce).
Detection of PKC Levels by Slot-blot Analysis-The indicated amounts of protein were directly pipetted onto a polyvinylidene fluoride membrane (Immobilon-P TM , Millipore) using a slot-blot apparatus (Amersham Pharmacia Biotech). Detection of PKC was performed as described above for the Western blot analysis.
Luciferase Assays-HC-11 mouse mammary epithelial cells were seeded into 6-well plates. 24 h later they were transfected for 1 h with 1.5 g of DNA and 6 l/well of Transfast (Promega) in 1 ml of Optimem medium. 24 h after transfection, the cells were serum-starved with RPMI 1640 medium containing 0.5% fetal calf serum and 5 g/ml insulin (Sigma) for additional 24 h. Following this incubation, the cells were collected and lysed according to the manufacturer's prescription. Luciferase activity was measured using the Dual-Luciferase TM Reporter Assay (Promega) in a MicroBeta Trilux 1450 counter (Perkin-Elmer). The fos-SRE-luciferase activity values were standardized by the activity of a cotransfected pRL-SV40 Renilla luciferase (Promega).

Effects of the Substitution of the Invariant Lysines in cPKC␣ and aPKC on the in Vitro Enzymatic
Activity-To characterize specific properties of the catalytic domain of atypical PKC, we have used site-directed mutagenesis to replace the invariant lysine in the ATP-binding pocket by the chemically similar amino acid residue arginine or by the unrelated tryptophan residue (K274R and K274W mutants, respectively), as described under "Experimental Procedures." All constructs were inserted into the eukaryotic expression vector pEF-neo containing a COOH-terminal RGS-His 6 tag to allow protein purification from whole cell lysates. The constructs were transiently transfected into COS-7 cells using Lipofectin. 48 h later, cells were lysed, and RGS-His 6 tagged aPKC was purified from the lysates by addition of Ni ϩ2 agarose. After elution with imidazole the protein concentration was measured, and the samples for the in vitro kinase assay were adjusted to equal amounts of protein. The activity of the purified enzymes was measured in vitro by phosphorylation of a standard PKC substrate peptide derived from the cPKC␣ pseudosubstrate sequence (PKC␣-19 -31/Ser-25). The in vitro activity of the aPKC mutants was compared with active and kinase-dead mutants of cPKC␣ purified in the same way. Fig. 1A shows that in contrast to the Lys 3 Arg mutant of cPKC␣, aPKC-K274R exhibits in vitro enzymatic activity similar to the wild type enzyme. Additionally, like the wild type aPKC enzyme, aPKC-K274R shows a very high basal activity in the absence of cofactors that is comparable with the activity of constitutively active cPKC␣-AE or the corresponding Ala 3 Glu mutant of aPKC (aPKC-A120E). The results also show that the high basal enzymatic activity of aPKC cannot be further increased by this mutation of the pseudosubstrate domain.
To control the expression of the different mutants, the purified proteins were separated on a polyacrylamid gel, blotted onto nitrocellulose membrane, and detected with a RGS-His 6specific antibody. Fig. 1B shows that all proteins were expressed in the COS-7 cells, although the expression level of kinase-dead aPKC-K274W is significantly lower than that of the other constructs.
Complete Absence of Enzymatic Activity in aPKC-K274W-The difference in the expression level between wild type aPKC and aPKC-K274W was quantified by a slot-blot analysis using different amounts of purified protein. Fig. 2A shows that the expression level of wild type aPKC is about 10-fold higher than that of the kinase-inactive aPKC-K274W mutant. Because equal amounts of purified protein were used for both wild type aPKC and aPKC-K274W in the slot-blot analysis, these results also show that the Ni 2ϩ -agarose used for purification exhibits significant unspecific binding to His-rich proteins, although these proteins do not have significant kinase activity (Fig. 1A, vector control). We wanted to confirm that the absence of kinase activity in the aPKC-K274W samples was not caused by a lower PKC content in the final reaction mix because of the lower expression level. For this purpose, different amounts of purified aPKC-WT and aPKC-K274W were adjusted to equal protein concentrations with Ni ϩ2 -agarose-treated samples from vector-transfected control cells containing unspecifically bound proteins and measured for in vitro kinase activity. Fig.  2B shows that aPKC-K274W is unable to phosphorylate the peptide substrate even at high concentrations, whereas wild type aPKC leads to a concentration-dependent increase in enzymatic activity.
aPKC-K274W Is Enzymatically Inactive in Intact Cells-In contrast to other PKC isoforms, aPKCs exhibit high basal enzymatic activity in vitro without the addition of cofactors. Although other PKCs have been shown to require diacylglycerol and acidic phospholipids like phosphatidylserine for optimal activity and the maintenance of their active conformation, for atypical PKCs a basal activation by protein activators like -interacting protein (43) or -interacting protein (25) seems to be sufficient. The enzymatic activation of aPKCs is dependent on the autophosphorylation of a threonine residue in the C terminus that is conserved in all PKCs (44). Introduction of a negative charge in this position (T555E) leads to an enzyme with elevated basal enzymatic activity (Fig. 3). To verify whether aPKC-K274W retains any residual activity possibly activable in vivo, the same point mutation was introduced into the aPKC-K274W protein. Fig. 3 shows that this double mutant is still completely inactive when measured in vitro. Because the T555E is capable to effectively increase the enzymatic activity of the wild type enzyme, the lack of activation of aPKC-K274W indicates that this mutant is missing any residual activity.
Because a kinase-dead mutant of aPKC is a valuable tool for studying signal transduction pathways in vivo, we next confirmed its enzymatic inactivity using a biological readout system. HC-11 mouse mammary epithelial cells were transiently transfected with a fos-SRE-luciferase reporter plasmid and constitutively active Ras-L61 together with wild type aPKC or the aPKC-K274W mutant. As demonstrated in Fig. 4, the K274W mutant is completely unable to confer an increase in the fos induction by Ras-L61 that can be observed with the wild type construct. The data also show that aPKC-K274R behaves like the wild type enzyme not only in vitro but also in vivo using the fos promoter as a biological readout system. ATP Dependence of PKC-WT and -K274R-Because the replacement of the invariant lysine by arginine generally leads to a complete abolishment of ATP binding and thus enzymatic activity of many protein kinases, e.g. PKC␣, it was of interest whether the kinetic properties of the PKC-K274R mutant differ from the wild type enzyme. For this purpose, the kinase activity of purified PKC-WT and -K274R was measured at different ATP concentrations. As shown in Fig. 5A, both the wild type and the K274R mutant of aPKC show very similar Michaelis-Menten kinetics for ATP. The Lineweaver-Burk plot shown in Fig. 5B illustrates that the affinity of aPKC-K274R for ATP is comparable with that of the wild type enzyme, with K m (ATP) values of 11.4 Ϯ 0.6 and 13.5 Ϯ 4.25 M, respectively. By comparison, the ATP affinity of the classical PKC isoenzyme PKC␣ is lower even in the constitutively active form, with a K m value of 38.8 M (data not shown).
In contrast to the K m value, the V max of the phosphorylation reaction is significantly decreased with aPKC-K274R (7.4 Ϯ 1.6 versus 14.0 Ϯ 2.3 pmol⅐min Ϫ1 ⅐mg protein Ϫ1 with wild type aPKC). This indicates that in aPKC the invariant lysine plays a less central role for ATP binding than in other kinases and can functionally in part be replaced with arginine. Again, the V max value for PKC-KR was still in the range of constitutively active PKC␣-AE with a V max of 8.4 pmol⅐min Ϫ1 ⅐mg protein Ϫ1 . For the aPKC-K274W mutant no kinetic studies could be executed because there was no enzymatic activity detectable; for the same reason no kinetic data for the corresponding cPKC␣-KR mutant could be determined.
Inhibition of PKC-WT and -K274R by GF 109203X-One of the most specific PKC inhibitors described in the literature is the bisindolylmaleimide GF 109203X. Like most other potent PKC inhibitors, GF 109203X competes with ATP for binding to the enzyme and therefore is believed to bind directly to the ATP-binding site of PKCs (45). If the three-dimensional structure of the ATP-binding site of aPKC is retained even when the invariant lysine is replaced by arginine, then also the affinity for GF 109203X should be identical with the wild type enzyme. Otherwise, the potency of this substance to inhibit the enzymatic activity should be changed. Such an effect has been demonstrated for the phosphoinositide 3-kinase, where replacement of the invariant lysine by arginine (K802R) completely abolished binding of the PI3K inhibitor wortmannin (40). To confirm the assumption that aPKC has a slightly different structure of the ATP-binding site, the inhibitory po-tencies of GF 109203X and the ATP analogue FSBA against both wild type and K274R-PKC were determined. Fig. 6 shows that GF 109203X and FSBA have identical affinities to both wild type and K274R-mutated aPKC. The fact that these two substances, which differ in their structure from ATP, bind with unchanged affinities to both proteins confirms that arginine is able to substitute the invariant lysine in aPKC with regard to the structure of the corresponding binding site, which is supposed to be located in the ATP-binding domain. DISCUSSION The atypical PKC isoforms aPKC and aPKC play essential roles in signaling pathways involved in mitogenesis, differentiation, malignant transformation, and resistance to apoptosis (1,7,13,46). Malfunction of these kinases has been shown to be implicated in the development of several human diseases, e.g. Alzheimer disease, chronic lymphatic leukemia, and diabetes (8,16,47,48). The main structural differences between aPKCs and the classical and novel isoforms reside predominantly in the regulatory domain; the atypical isoforms lack the calciumbinding region and one of two zinc finger motifs responsible for the interaction with diacylglycerol, a lipid cofactor of all other PKC isoenzymes. Also the pseudosubstrate domain, a conserved sequence motif shown to bind to and block the substrate-binding domain before activation, differs significantly in aPKCs. Much less pronounced are the differences in the primary structure of the catalytic domain. Nevertheless, these subtle differences are supposed to be at least partly responsible for the in vivo selectivity of different PKC isoenzymes, because mutations within the regulatory domain rendering the PKC isozymes constitutively active do not affect their biological specificity (1,6,7). Like for other PKCs, the catalytic domains of aPKC and aPKC share the main structural motifs of all protein kinases, e.g. the invariant lysine responsible for correct folding of the ATP-binding pocket and directly interacting with ATP, the invariant glutamate in subdomain III forming a salt bridge with the invariant lysine and stabilizing its interaction with ATP, and a phosphorylation site in the activation loop (Lys-274, Asp-373, and Thr-403 in aPKC, respectively) (30,44,49). Intriguingly, one structural hallmark of virtually all protein kinases and other nucleotide-binding proteins, the GXGXXG motif, is only partially conserved in atypical PKCs; they contain an alanine residue instead of the third conserved glycine (GrGsyA) (Fig. 7). The glycine-rich sequence is an integral part of the ATP-binding site, it serves as a lid to anchor the ATP and to exclude water from the catalytically active site. Consistently, mutation of one of these conserved glycines normally leads to an enzyme with impaired catalytic activity, as has been shown for the insulin receptor, the nonreceptor tyrosine kinase Fyn, and the Ras oncogene, the latter with respect to GTPase activity (31)(32)(33). This indicates that the GXGXXG motif is essential for correct binding of the nucleotide. An altered glycine-rich motif in aPKCs therefore may be a hint to a slightly different mode of ATP binding compared with other kinases. Intriguingly, aPKCs are quite insensitive to specific PKC inhibitors acting at the ATP-binding site and competing with ATP for binding, like the bisindolylmaleimide GF109203X. This led us to the assumption that these unique properties of aPKCs are caused by a unique ATP-binding domain that may also be reflected in the function of its central residue, the so-called invariant lysine. This residue is conserved throughout the whole kinase family, and even a conservative mutation like a Lys-to-Arg exchange almost inevitably abolishes the enzymatic activity of the protein (35)(36)(37)(38)(39). For the characterization of the aPKC ATP-binding domain we have used a combination of site-directed mutagenesis and in vitro enzymatic analysis. We have replaced the conserved Lys-274 by the chemically similar amino acid Arg or by Trp, which has different chemical and structural properties. The proteins have been expressed in COS cells as RGS-His 6 -tagged forms and purified with Ni 2ϩ -agarose, and their activity was measured in vitro with a pseudosubstrate-derived peptide as substrate. Surprisingly and in contrast to virtually all other reports of protein kinases, an aPKC with an Arg exchange of the invariant lysine retained almost full enzymatic activity (Fig.  1A). There exist only very few reports in the literature about kinases that are insensitive for mutations of the most conserved amino acid in the ATP-binding pocket, e.g. one concerning the very distantly related CDK-activating kinase 1p from Saccharomyces cerevisiae.
Another conspicuous property of aPKC is its high basal activity in vitro. Although other PKC isoforms require cofactors like phosphatidylserine and diacylglycerol for full activity, even if they have been "preactivated" by an Ala 3 Glu mutation in the pseudosubstrate region, wild type aPKC and even the KR mutant show full enzymatic activity without addition of any cofactor. It has been shown previously that for full activation PKCs have to be phosphorylated on at least three different positions: the first phosphorylation is introduced on a conserved Thr residue in the activation loop by other kinases like the phosphoinositide-dependent kinase (44, 49 -51). Following activation, two other phosphorylations are introduced by autophosphorylation at two Ser/Thr residues near the C terminus. These activating phosphorylations can be mimicked by introduction of a negative charge through mutational exchange of the phosphorylatable Ser/Thr residues for glutamic acid (51). Interestingly, wild type atypical PKCs already contain a glutamate residue instead of the second autophosphorylatable residue (Ser). Therefore, the negatively charged glutamic acid residue could confer a state of constitutive activation and explain the high basal activity of these kinases.
One remarkable phenomenon was the correlation between protein level and activity of aPKC. Although aPKC-WT, -A120E, and -K274R were expressed at comparably high levels (Fig. 1B), the protein level of the inactive K274W mutant was much lower. When quantified by slot-blot analysis, the concentration of aPKC-K274W in extracts from transiently transfected COS cells was about 1 ⁄10 of the wild type protein. Because the wild type gene and both mutants had been cloned into the same expression vector (pEF-neo) and the integrity of the promoter has been controlled by sequencing (data not shown), all proteins are expected to be synthesized in the cell at the same level. Thus, different aPKC concentrations can only be explained by variations in protein stability. It has been shown previously that PKCs are degraded by proteasomes by the ubiquitinylation pathway (52). To see whether inactive aPKC is more susceptible for proteasomal degradation than active forms and therefore present at a lower protein level in the cell, MG-132, an inhibitor of the proteasomal degradation pathway, was used. This substance had no effect on the level of either wild type or kinase-dead aPKC (data not shown). This indicates that at least degradation by the proteasome pathway is not directly linked to enzyme activity. Yet, other degradation pathways not affected by MG-132 could be responsible for this effect.
One other possibility for how the lack of enzymatic activity could cause a reduced protein stability could be the inability of kinase-dead aPKC to autophosphorylate. It has been shown previously for cPKC␣ that activation by phorbol esters can induce its own phosphorylation and that this phosphorylated form is resistant to the degradation process normally induced by phorbol ester treatment (53). To test this hypothesis for aPKC, we replaced the autophosphorylatable threonine by glutamate to mimic autophosphorylation. The potency of this mutation to mimic an autophosphorylated state of the enzyme and thus to stimulate enzymatic activity is shown in Fig. 3. Nevertheless, the protein level of the aPKC-K274W/T555E double mutant is not increased with respect to the K274W mutant (data not shown). Because the second position known to be accessible for autophosphorylation in classical and novel PKCs in aPKCs is already replaced by glutamic acid, autophosphorylation seems not to influence the stability of atypical PKC isoenzymes.
For the use of kinase-dead mutants as dominant negative inhibitors of endogenous kinases in the study of signal transduction pathways, it is essential that the mutants lack any residual enzymatic activity. The K274W/T555E double mutant of aPKC was also used to confirm this property for aPKC-K274W mutant. Because the T555E mutation is able to FIG. 7. Alignment of the core ATP-binding region of aPKC and other kinases and nucleotide-binding proteins. The amino acid sequence surrounding the GXGXXG motif and the invariant lysine in the human forms of the indicated proteins are shown. Sequences were obtained from the GenBank TM data base. strongly activate wild type aPKC (Fig. 3), the same mutation would be expected to increase a potential residual enzymatic activity of aPKC-K274W. As shown in Fig. 3, the K274W/ T555E double mutant completely retained its kinase-dead state when measured in vitro. This confirms the finding that the K274W mutant really lacks any residual protein kinase activity. The same result could also be achieved when increasing amounts of purified protein where used for the in vitro measurement of kinase activity (Fig. 2B). Although wild type aPKC exhibited a clear dependence in the enzymatic activity on the protein concentration in the sample, the enzymatic activity aPKC-K274W samples never rose over the background level of samples from vector control transfected cells.
The fact that the replacement of the invariant lysine in the catalytic domain of aPKC by arginine does not abolish its enzymatic activity is in sharp contrast to findings with several other protein kinases like the epidermal growth factor receptor, the insulin receptor, Src, Mos, Fps, and even the lipid kinase phosphoinositide 3-kinase (35)(36)(37)(38)(39)(40). It is of special interest that for phosphoinositide 3-kinase this Lys-to-Arg mutant not only is unable to bind ATP and thus phosphorylate its substrate phosphoinositide but also lost the capability to bind the inhibitor wortmannin that covalently binds to the invariant lysine (Lys-802), thus inhibiting the activity of the wild type enzyme (40). This mutant also lost its affinity to FSBA, another ATP analogue that covalently binds to and blocks the invariant lysine within the ATP-binding pocket. Because the mutation of the invariant lysine in aPKC only marginally affected the enzymatic activity of the kinase, it was of special interest to see whether the affinity to ATP or its analogues was affected. By biochemical characterization of both the wild type aPKC and the K274R mutant, we could show that the affinity of the mutated protein for ATP is not altered (Fig. 5); the K m value of aPKC-K274R (11.4 M) is comparable with the wild type enzyme (13.5 M); by comparison, the classical isoform PKC␣ has a significantly higher K m value of 38.8 M. Only the velocity of the enzymatic reaction seems to be reduced (V max of 7.4 pmol⅐min Ϫ1 ⅐mg protein Ϫ1 versus 14 pmol⅐min Ϫ1 ⅐mg protein Ϫ1 for the wild type enzyme). However, this is still higher than that of the constitutively active PKC␣-AE (8.4 pmol⅐min Ϫ1 ⅐mg protein Ϫ1 ; data not shown). These results indicate that the Lys 3 Arg mutation induces a slight change in the conformation of the catalytic domain without affecting the affinity. This can be explained by the fact that the invariant lysine plays a dual role in the catalytic domain. In addition to its involvement in the phosphotransfer reaction by direct interaction with the ␣and ␤-phosphates of ATP, it forms a salt bridge with the carboxyl group of Glu-293, thereby stabilizing the three-dimensional structure of the catalytic domain. Therefore, the Lys 3 Arg mutant may bind the ATP with unaltered affinity but may be restrained in its ability to correctly orientate the nucleotide. This could reduce the accessibility of the bound ATP to the amino acid residues mediating the phosphate transfer. To test this hypothesis, we measured the ATP affinity of both the wild type enzyme and the K274R mutant by competition with the staurosporine derivative GF 109203X or the ATP analogue FSBA. Indeed these compounds were equally potent to compete with ATP with both forms of the kinase, showing that the ATP affinity is not altered when the invariant lysine is replaced by arginine. This unique feature confirms that the ATP-binding site of atypical PKCs and probably also the mode of ATP binding differs significantly from other PKCs and kinases in general. Such an unusual enzymatic property has so far been demonstrated only for very few kinases like the CDK-activating kinase from S. cerevisiae. Thus, an inhibitor that specifically binds to this unusual ATP-binding pocket should be highly specific for atypical PKCs without affecting other PKC isoforms or unrelated protein kinases. Our data provide evidence that the ATP-binding pocket of aPKC shows marked differences to other PKC isoforms. This could be the basis for further investigations on the exact three-dimensional structure and ATP binding properties of the catalytic domain of aPKC and for the development of specific inhibitors of atypical PKCs blocking the important signaling pathways mediated by this enzyme.