Generation and Characterization of ATP Analog-specific Protein Kinase Cδ*

Background: A traceable form of PKCδ is needed to investigate PKCδ functions. Results: Analog-specific PKCδ could utilize N6-(benzyl)-ATP to phosphorylate PKCδ substrates and was specifically inhibited by PP1 analogs. Conclusion: Binding to the catalytic domain glycine-rich loop, Lys-378, Glu-428, Leu-430, and Phe-633 is the likely mechanism by which N6-(benzyl)-ATP and PP1 analogs interact with analog-specific PKCδ. Significance: This study provides tools to investigate PKCδ-mediated pathways. To better study the role of PKCδ in normal function and disease, we developed an ATP analog-specific (AS) PKCδ that is sensitive to specific kinase inhibitors and can be used to identify PKCδ substrates. AS PKCδ showed nearly 200 times higher affinity (Km) and 150 times higher efficiency (kcat/Km) than wild type (WT) PKCδ toward N6-(benzyl)-ATP. AS PKCδ was uniquely inhibited by 1-(tert-butyl)-3-(1-naphthyl)-1H-pyrazolo[3,4-d]pyrimidin-4-amine (1NA-PP1) and 1-(tert-butyl)-3-(2-methylbenzyl)-1H-pyrazolo[3,4-d]pyrimidin-4-amine (2MB-PP1) but not by other 4-amino-5-(4-methylphenyl)-7-(t-butyl)pyrazolo[3,4-d]pyrimidine (PP1) analogs tested, whereas WT PKCδ was insensitive to all PP1 analogs. To understand the mechanisms for specificity and affinity of these analogs, we created in silico WT and AS PKCδ homology models based on the crystal structure of PKCι. N6-(Benzyl)-ATP and ATP showed similar positioning within the purine binding pocket of AS PKCδ, whereas N6-(benzyl)-ATP was displaced from the pocket of WT PKCδ and was unable to interact with the glycine-rich loop that is required for phosphoryl transfer. The adenine rings of 1NA-PP1 and 2MB-PP1 matched the adenine ring of ATP when docked in AS PKCδ, and this interaction prevented the potential interaction of ATP with Lys-378, Glu-428, Leu-430, and Phe-633 residues. 1NA-PP1 failed to effectively dock within WT PKCδ. Other PP1 analogs failed to interact with either AS PKCδ or WT PKCδ. These results provide a structural basis for the ability of AS PKCδ to efficiently and specifically utilize N6-(benzyl)-ATP as a phosphate donor and for its selective inhibition by 1NA-PP1 and 2MB-PP1. Such homology modeling could prove useful in designing molecules to target PKCδ and other kinases to understand their function in cell signaling and to identify unique substrates.

PKC is a family of 10 serine-threonine kinases that regulate a broad spectrum of cellular functions (1,2). In general, PKC isozymes contain a regulatory domain in the amino-terminal region, followed by a flexible hinge region and a conserved catalytic domain in the carboxyl-terminal tail. The catalytic domain is composed of two lobes. The amino-terminal lobe contains a glycine-rich loop of the consensus sequence GXGXXGX and an invariant Lys that positions ATP for phosphoryl transfer. The carboxyl-terminal lobe contains an activation loop that binds protein substrates for catalysis. The sequence linking these two lobes also contributes to ATP binding and contains a gatekeeper amino acid residue that limits the size of a hydrophobic region within the ATP binding pocket and confers selectivity for binding nucleotides and small molecule inhibitors (1,2). In PKC isozymes, this gatekeeper is a large hydrophobic residue, either Met or Ile.
Based on their amino-terminal structures and sensitivities to Ca 2ϩ and diacylglycerol, PKCs are classified into conventional PKCs (␣, ␤I, ␤II, and ␥), novel PKCs (␦, ⑀, , and ), and atypical PKCs ( and /). Of interest to our laboratory is PKC␦, a member of the novel PKC subfamily, which we found to regulate behavioral responses to ethanol (3) as well as promote reperfusion injury after cerebral ischemia (4). To understand the molecular and cellular actions of PKC␦ in physiological and pathophysiological states, it would be desirable to generate a form of PKC␦ that can be specifically inhibited and can be used to identify PKC␦ substrates for mapping downstream signaling pathways.
A chemical-genetics approach has been developed to identify immediate phosphorylation substrates of kinases and to study results of kinase inhibition by selective, cell-permeable, small molecule inhibitors (5,6). This approach targets the structurally conserved ATP-binding pocket within all kinases to generate mutant alleles that can utilize specific ATP analogs in addition to ATP. The mutation creates a "cavity" by replacing a bulky gatekeeper with a smaller residue (alanine or glycine) in the ATP-binding pocket. The engineered "cavity" is located where the N6 amine of ATP usually sits, and thus allows for binding of structurally modified ATP analogs with bulky substitutions attached at the N6 position. Only the analog-specific (AS) 3 kinase, and not the WT kinase, can efficiently use N 6 -substituted ATP analogs as phosphate donors. Therefore, only unique substrates of the AS kinase are labeled by the ATP analogs.
To further facilitate the identification and purification of substrates, an affinity tagging strategy was developed (7). First, an AS kinase mutant is used to thiophosphorylate the substrates with N 6 -(benzyl)-ATP␥S. The thiophosphate group is then alkylated by para-nitrobenzyl mesylate to create thiophosphate ester epitopes that can be recognized by specific antibodies. The tagged substrates can be isolated by immunoprecipitation or immunoaffinity purification. The approach has been successfully used to identify direct substrates of several kinases, including JNK (8), v-Src (9), ERK2 (10), CDK1 (11), Raf-1 (12), CDK7 (13), and PKC⑀ (14,15).
AS kinase mutations are designed to be functionally silent with respect to kinase activity and substrate specificity. The engineered AS kinases are also uniquely sensitive to novel kinase inhibitors, such as analogs of PP1 (16). Based on this approach, we generated an AS PKC␦ that can be regulated by specific ATP and PP1 analogs (3,17). Here we studied specific interactions of these analogs with residues in the ATP binding pocket in silico. Our study provides valuable insights for the rational design of molecules to regulate and study PKC␦ signaling.

Expression and Purification of WT and AS PKC␦-COS-7
and Neuro2A cells were transfected using the SuperFect Transfection Reagent (Qiagen) with amino-terminal FLAG epitopetagged rat WT or AS PKC␦ (M425A) in pcDNA3 (17). The corresponding gatekeeper for human, rat, and mouse PKC␦ is Met-427, Met-425, and Met-425, respectively. After culture for 3 days, cells were washed once with PBS and lysed in buffer containing 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, phosphatase inhibitor mixture 1, and cOmplete TM protease inhibitor mixture (Roche Applied Science). After 20 min of incubation at 4°C, the lysates were centrifuged at 12,000 ϫ g for 10 min. The abundance of WT and AS PKC␦ in the lysate was determined by Western blotting using anti-PKC␦ antibodies (BD Biosciences). To purify WT and AS PKC␦, the supernatants were incubated with anti-FLAG M2 antibody-conjugated agarose (Sigma-Aldrich) at 4°C for 3 h. The agarose beads were washed three times with the lysis buffer. WT and AS PKC␦ were eluted using a storage buffer containing FLAG peptide (20 mM HEPES, pH 7.4, 0.1 mM EGTA, 25% glycerol, 0.03% Triton X-100, 150 ng/l FLAG peptide) and stored at Ϫ80°C until use. The concentrations of WT and AS PKC␦ were determined by ELISA using recombinant PKC␦ prepared in SF9 cells (PanVera) as a standard.
To detect substrates in neutrophils, we isolated neutrophils from bone marrow by Percoll density gradient centrifugation (4,19). Neutrophils were lysed by freeze/thaw treatment in modified radioimmune precipitation buffer containing 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 5 mM EDTA, 5 mM EGTA, phosphatase inhibitor mixtures I and II (Sigma-Aldrich), and cOmplete TM protease inhibitor mixtures (Roche Applied Science) and centrifuged at 20,000 ϫ g for 15 min at 4°C. The supernatant (100 g) was incubated in 60 l of PKC reaction buffer containing 20 mM HEPES, pH 7.4, 0.1 mM EGTA, 0.03% Triton X-100, and 10 mM MgCl 2 at 27°C for 30 min with 1 mM GTP, 100 ng of AS PKC␦ purified from transfected COS-7 cells, 200 M N 6 -(benzyl)-ATP␥S, and 1 M PMA. The thiophosphate groups on substrate proteins in the supernatant were then alkylated with 2.5 mM para-nitrobenzyl mesylate (Epitomics Inc.) for 2 h at room temperature. The reaction was stopped by the addition of SDS-PAGE sample buffer and analyzed by gel electrophoresis and silver staining. Thiophosphate esters on putative substrate proteins were detected by Western blotting using rabbit primary antibodies (1:15,000 dilution, Epitomics Inc.) and HRP-conjugated secondary antibodies (Jackson ImmunoResearch).
Kinetic Analysis of AS PKC␦ with ATP Analogs and PP1 Inhibitors-The kinase activity of AS PKC␦ was measured by fluorescence polarization using the Protein Kinase C Assay Kit with fluorescein-labeled substrate peptides (Invitrogen) (20). PKC␦ (0.109 ng) or AS PKC␦ (0.175 ng) was added to the PKC kinase buffer (20 mM HEPES, pH 7.4, 10 mM MgCl 2 , 0.02% Nonidet P-40, 0.03% Triton X-100, 0.1 g of phosphatidylserine, and 0.05 mM sodium vanadate) with 250 nM substrate peptide (RFARKGSLRQKNV) modified from the PKC␣ pseudosubstrate domain (1, 2), 12.5 nM PMA, 1ϫ fluorescein-labeled phosphopeptide, and 1ϫ anti-phosphoserine antibody, according to the manufacturer's instructions, in a final volume of 90 l. The kinase reaction was initiated by the addition of ATP (1-100 M) or N 6 -(benzyl)-ATP (0.5-200 M) in 10 l. The plate was incubated at 22°C, and the plane-polarized fluorescence was measured every 5 min using an Analyst HT plate reader (Molecular Devices) with excitation at 485 nm and emission at 530 nm. Polarization, Michaelis-Menten constant (K m ), maximum velocity (V max ), and catalytic constant (k cat ) were calculated as described (20).
The effect of inhibitors was also measured using fluorescence polarization. Different concentrations of each inhibitor (0.0012-2.0 ϫ 10 5 nM) were added to the reaction mixture in a final volume of 40 l. The mixture was incubated at 22°C for 5 min in the dark. The kinase reaction was initiated by the addition of ATP (final concentration 2.5 M) and then allowed to proceed for 90 min in the dark at 22°C. Next, 50 l of solution containing 2ϫ fluorescein-labeled phosphopeptide and 2ϫ anti-phosphoserine antibody were added according to the manufacturer's instructions. After incubation for 30 min at 22°C in the dark, plane-polarized fluorescence was measured as described above. The percentage of inhibition was calculated with the equation, (signal without inhibitor Ϫ signal with inhibitor)/(signal without inhibitor Ϫ signal in the presence of 20 M staurosporine) ϫ 100. Data from three triplicate experiments were used to calculate IC 50 values (nM) by nonlinear regression analysis using Prism version 5.0c (GraphPad Software).
Generation of AS PKC␦ Knock-in Mice-Knock-in mice were generated by Taconic Biosciences using homologous recombination with a targeting vector derived from the C57BL/6J mouse genomic BAC clone RP23-133F24. The targeting vector was composed of a 3.1-kb short arm carrying the M425A mutation, a Neo cassette flanked by two loxP sites, a 5.4-kb-long arm, and a diphtheria toxin A gene. NotI-linearized vector was electroporated into C57BL/6 ES cells and selected with 200 g/ml G418. Surviving ES clones were screened by Southern blotting, and a PCR fragment encompassing the M425A mutation was generated and sequenced to confirm the mutation. The floxed-Neo r cassette used for selection was deleted by electroporation of a Cre recombinase plasmid. Chimeric mice were generated following blastocyst injection of targeted ES cells. Heterozygous mutant mice were obtained by breeding chimeras with C57BL/6NTac mice. Heterozygous offspring were intercrossed to generate homozygous knock-in mutant mice. Mouse genotyping was performed by PCR using the primers G8-PCR-F (5Ј-GCTTTGGCTGAGTGTACTGGCAGAC) and G8 -35-R (5Ј-GCCCACCAGTCCCATCGCC-3Ј). PKC␦ and actin were detected in mouse tissues by Western blot analysis using a mouse monoclonal antibody against PKC␦ (1:1000 dilution; BD Biosciences) and actin (1:2000 dilution; Sigma-Aldrich). All procedures were conducted in accordance with Institutional Animal Care and Use Committee policies.
Immunofluorescence Staining of Neutrophils-Neutrophils isolated by Percoll density gradient centrifugation (4,19) were plated on glass coverslips coated with 20% fetal calf serum (FCS) for 10 min at 37°C. The coverslips with attached neutrophils were treated with or without 200 nM PMA for 2 min at 37°C and fixed in 2% paraformaldehyde in PBS for 10 min at room temperature (21). After permeabilization in 0.1% Triton X-100 in PBS for 5 min, neutrophils were blocked in 10% normal donkey serum, 0.2% BSA in PBS for 1 h and stained with mouse anti-PKC␦ (1:200 dilution; BD Biosciences) antibody diluted in PBS containing 2% normal donkey serum and 0.2% BSA overnight at 4°C. After three washes with PBS, neutrophils were incubated with the appropriate donkey fluorochromeconjugated secondary antibodies (1:200 dilution; Jackson ImmunoResearch) and coverslipped in mounting medium containing DAPI (Vector Laboratories) to localize the nuclei. Images were captured using a Zeiss LSM 510 laser-scanning confocal microscope.
Neutrophil Superoxide Anion (O 2 . ) Production-Neutrophils isolated by Percoll density gradient centrifugation (4,19) were treated with 5 M 1NA-PP1 for 15 min. O 2 . production was then measured by a cytochrome c reduction assay using 100 nM PMA as a stimulus (19). Statistical Analysis-Quantitative data were expressed as mean Ϯ S.E. and analyzed using Prism version 5.0 (GraphPad). Analysis of variance with post hoc tests was used to determine statistical significance between means. A value of p less than 0.05 was considered to be statistically significant.
Homology Modeling-Human PKC␦ homology models (amino acids 341-669) were generated using MOE version 2011.10 software and PKC with ATP bound in the active kinase domain (Protein Data Bank (PDB) code 3A8W) (22) as the reference structure. During the homology modeling procedure in MOE, 10 models were developed, and the final model was used for the studies. The homology model developed by MOE did not have the ATP or bound water molecules for correct ATP positioning in the binding pocket. We therefore transferred the ATP and binding pocket water molecules from PKC into the binding pocket of the PKC␦ homology model. The complex of WT PKC␦ and ATP/water molecules was then energy-minimized, and the pH of the system was set at pH 7.4, to correlate with experimental conditions. After the human PKC␦ model was developed, we mutated the Met-427 to Ala using the side chain mutation function in MOE with energy minimized. AS PKA was generated by mutating Met-120 to Ala with energy minimized in the crystal structure of PKA (PDB code 1ATP) (23,24). The interactions of ATP and inhibitors with the residues of the nucleotide-binding pocket were analyzed using the two-dimensional ligand-receptor interaction diagrams of Discovery Studio version 3.5.
Docking Studies-Docking of N 6 -(benzyl)-ADP (derived from PDB entry 1KSW) and 1NM-PP1 (PubChem code 5154691) was performed using the MOE software, which utilizes an induced fit model of the ligand and receptor for docking. Docking of 1NA-PP1 (PubChem code 4877) and 2MB-PP1 (modified from the co-crystal structure of Src-AS1/3MB-PP1 (PDB code 4LGG)) was performed using AutoDock Vina (25) integrated with Chimera version 1.7 (26). The pH of the induced fit docking procedure used in the MOE was set at 7.4. The top 10 poses returned from the docking studies were evaluated by visual inspection.

RESULTS
Generation of AS PKC␦-PKC isozymes belong to the AGC (PKA, PKG, and PKC) kinase group sharing a highly conserved catalytic domain. Based on the sequence homolog of amino acid residues in the catalytic domain of PKC isozymes and PKA (Fig. 1A), we identified Met-427 as the gatekeeper residue in human PKC␦ and substituted it with alanine in AS PKC␦. Because mutations can destabilize proteins and reduce their abundance in cells, we expressed WT and AS PKC␦ in two different cell lines, COS-7 (Fig. 1B) and Neuro2A (Fig. 1C). We found that both proteins expressed abundantly in these cell lines and that AS PKC␦ showed levels of expression similar to WT PKC␦.
Kinase Activity of AS PKC␦-To characterize AS PKC␦, we purified WT and AS PKC␦ from COS-7 cells and found that both proteins utilized [␥-32 P]ATP to phosphorylate a model substrate (histone 3) in mixed micelle PKC assays (Fig. 2, A and  B). WT and AS PKC␦ showed similar binding affinity (K m ) and catalytic rate (k cat ) for ATP in fluorescence polarization assays using a peptide substrate derived from the psuedosubstrate sequence of PKC␣ ( Table 1). The similarities in K m and k cat  bottom, Coomassie-stained histone 3. C, mouse neutrophil lysates (100 g) were labeled with AS PKC␦ and N 6 -(benzyl)-ATP␥S. After paranitrobenzyl mesylate (PNBM) alkylation, the labeled substrates were detected by Western blot analysis using an antibody that detects thiophosphate esters. Immunoreactivity was increased in samples treated with the PKC activator PMA. D, the same amounts of neutrophil lysates as in D were loaded and visualized by silver staining. In contrast, AS PKC␦ was able to utilize an ATP analog with a bulky substitution (benzyl ring) at the N6 position (N 6 -(benzyl)-ATP) to phosphorylate histone 3 (Fig. 2, A and B). A similar result was observed in fluorescence polarization assays, which showed ϳ200-fold lower K m and ϳ150-fold higher efficiency (k cat /K m ) for AS PKC␦ than WT PKC␦ when N 6 -(benzyl)-ATP was the phosphate donor (Table 1).
We previously found that AS PKC␦ can utilize N 6 -(benzyl)-ATP␥S to thiophosphorylate histone 3, which can be detected after alkylation with para-nitrobenzyl mesylate by immunoblot analysis using an antibody against thiophosphate esters (17). We decided to use this approach to determine whether we could use AS PKC␦ and N 6 -(benzyl)-ATP␥S to identify PKC␦ substrates in cells. We were particularly interested in PKC␦ substrates in neutrophils because of our work demonstrating an important role for neutrophils in cerebral ischemia and reperfusion injury (4). Because N 6 -(benzyl)-ATP␥S does not cross cell membranes, we incubated neutrophil lysates in kinase buffer containing recombinant AS PKC␦ (Fig. 2, C and D). Despite the pres-ence of endogenous WT PKC␦ in these lysates, we found several proteins that were specifically labeled, and the labeling intensity of some of these proteins was increased when lysates were incubated with the PKC activator PMA, suggesting that they are PKC␦ substrates.
Sensitivity of AS PKC␦ to Kinase Inhibitors-We next compared the ability of different kinase inhibitors to reduce phosphorylation of histone 3 or the PKC␣ pseudosubstrate peptide. WT PKC␦, like other wild type PKC isozymes we tested, was potently inhibited by the general kinase inhibitor staurosporine and by the more PKC-selective inhibitor bisindolylmaleimide I but not by PP1 analogs (Fig. 3 and Table  2). AS PKC␦ was uniquely inhibited by 1NA-PP1, weakly inhibited by 1NM-PP1, and not at all inhibited by 2NM-PP1. Staurosporine and bisindolylmaleimide I inhibited AS PKC␦ much less potently than WT PKC␦. Both WT and AS PKC␦ were unable to phosphorylate histone 3 in the absence of phosphatidylserine and PMA (Fig. 3B). PMA mimics diacylglycerol, an endogenous activator of PKC␦ (1, 2). Phosphatidylserine, an important phospholipid membrane component, is required for membrane binding and kinase activity of PKC. This result indicates that, like WT PKC␦, AS PKC␦ requires known PKC lipid regulators for activation.

Generation and Characterization of AS PKC␦ Knock-in
Mice-To investigate the functions of PKC␦ in vivo, we used homologous recombination in embryonic stem cells to generate knock-in mice endogenously expressing AS PKC␦ (Fig. 4, A  and B). AS PKC␦ mice were normal in appearance and behaved normally in a cage environment. Western blot analysis of neutrophil lysates from these knock-in mice demonstrated a 78-kDa PKC␦ immunoreactive band of similar intensity in samples from WT mice (Fig. 4C). PKC␦ immunoreactivity was present in the neutrophil cytoplasm of both genotypes and translocated to cell membranes after PMA activation (Fig. 5). We previously found that the superoxide anion (O 2 . ) production from neutrophils is significantly reduced in PKC␦ null mice (4). To confirm this phenotype and assess the effect of 1NA-PP1 in intact cells, we treated neutrophils from WT and AS PKC␦ mice with 1NA-PP1 and found that it specifically reduced O 2 . production from AS PKC␦ but not from WT neutrophils (Fig. 6). N 6 -(Benzyl)-ATP Fits in the Nucleotide-binding Pocket of AS PKC␦ but Not WT PKC␦-To understand the molecular basis for differences in binding affinity and specificity of ATP and PP1 analogs in WT and AS PKC␦, we generated PKC␦ models in silico. Because the x-ray crystal structure of PKC␦ is not available, PKC␦ homology models were created based on the crystal structure of human PKC (PDB code 3A8W) (22). The root mean square deviation for the overlaid structure of WT and AS PKC␦ is 0.279 Å (Fig. 7A), suggesting that the M427A mutation did not significantly change the three-dimensional  structure of PKC␦ (28). The position of ATP was nearly identical in the nucleotide-binding pockets of WT and AS PKC␦, with the adenine ring pointing toward the gatekeeper residues (Fig.  7B). In both WT and AS PKC␦, the adenine ring of ATP formed hydrogen bonds with Glu-428 and Leu-430 near the gatekeeper. The ATP phosphate groups were positioned by the charge interactions with the invariant Lys-378 and by hydrogen bonds with Ser-359, Phe-360, and Gly-361 in the glycine-rich loop (Fig. 7, C and D). These interactions are similar to those reported in previous studies of the crystal structure of ATPbound PKA (PDB code 1ATP) (23,24,29,30), PKC (PDB code 3A8W) (22), and PKC␤ (PDB code 3PFQ) (31), suggesting the accuracy of our PKC␦ homology models. ATP interacted with the same residues in WT and AS PKC␦, but the electrostatic interaction with the gatekeeper residue (Met-427) in WT PKC␦ was replaced by a weaker van der Waals interaction with Ala-427 in AS PKC␦. This finding suggests a molecular basis for the 2-fold lower specificity (k cat /K m ) for ATP of AS PKC␦ compared with WT PKC␦ (Table 1). Because N 6 -(benzyl)-ADP has been crystalized with an analog-specific allele of c-Src (T338G) (32), we extracted N 6 -(benzyl)-ADP from the co-crystalized structure (PDB code 1KSW) and docked it in the PKC␦ homology models. N 6 -(benzyl)-ADP and ATP docked at similar positions within the nucleotidebinding pocket of AS PKC␦ (Fig. 8A). The benzyl ring of N 6 -(benzyl)-ADP was positioned deeper in the nucleotidebinding pocket and placed in a space enlarged by mutating Met-427 to Ala-427. This change allowed the phosphate groups of N 6 -(benzyl)-ADP to interact with the glycine-rich loop for potential phosphoryl transfer (Fig. 8C). In contrast, N 6 -(benzyl)-ADP was shifted outward in the nucleotide-binding pocket of WT PKC␦ (Fig. 8B), preventing its phosphate groups from interacting with the glycine-rich loop (Fig. 8D). Moreover, the hydrogen bonds formed between ATP and Glu-428 and Leu-430 in AS PKC␦ (Fig. 7D) were lost with N 6 -(benzyl)-ADP (Fig.  8C). This might explain why ATP showed 2 times higher affinity (K m ) and specificity (k cat /K m ) than N 6 -(benzyl)-ATP for AS PKC␦ (Table 1).
Mechanistic Studies of Inhibition of AS PKC␦ by PP1 Analogs-To investigate the basis for specific 1NA-PP1 inhibition of AS PKC␦, we compared docking of 1NA-PP1 and ATP in AS PKC␦ (Fig. 9A) and WT PKC␦ (Fig. 9B). The adenine rings of 1NA-PP1 and ATP were aligned in AS PKC␦ but not in WT PKC␦. Specifically, the adenine rings of 1NA-PP1 and ATP   formed hydrogen bonds with the same residues (Glu-428 and Leu-430) in AS PKC␦ (Figs. 7D and 9C). Moreover, the naphthyl ring of 1NA-PP1 fits in the space created by the M427A mutation, forminginteractions with the invariant Lys-378 in AS PKC␦ (Fig. 9C). Neither of these interactions was found with 1NA-PP1 docked in WT PKC␦ (Fig. 9D) or with 1NM-PP1 or 2NM-PP1 docked in AS PKC␦ (Fig. 10).
PKA, another member of the AGC kinase family, has been engineered following the same approach (33). The AS PKA mutant (M120A) is sensitive to both 1NA-PP1 and 1NM-PP1 inhibitors. To investigate the structural basis for PP1 inhibition in AS PKA as well as to compare the mechanism with AS PKC␦, we docked 1NA-PP1, 1NM-PP1, and ATP in AS PKA. The adenine rings of 1NA-PP1 and 1NM-PP1 aligned with the adenine ring of ATP in AS PKA, with the naphthyl and naphthylmethyl rings situated in the engineered cavity (M120A) (Fig. 11, A and B). Interestingly, the adenine rings of 1NA-PP1 and 1NM-PP1 also formed hydrogen bonds with Glu-121 and Val-123 (corresponding to Glu-428 and Leu-430 in PKC␦) (Fig. 11, C and D).
To assess whether the in silico AS PKC␦ model can be used to screen potential inhibitors, we docked novel PP1 analogs (6) and found that 2MB-PP1 (Fig. 12A) is better aligned with the adenine ring of ATP than 1NA-PP1 (Fig. 12B). 2MB-PP1 interacts with the same residues of AS PKC␦ as 1NA-PP1, but the interaction is further strengthened by an additionalinteraction with Phe-633 (Fig. 12C). Thisstacking interaction has been implicated in stabilizing nucleic acid-protein complexes (34,35). To test whether this enhancedinteraction improves the inhibitory effect of 2MB-PP1, we measured its IC 50 using fluorescence polarization assays and found that the IC 50 for 2MB-PP1 (40.6 nM) was almost 4 times lower than the IC 50 for 1NA-PP1 (154.0 nM) (Fig. 12D). The results demonstrate that in silico modeling can predict the relative potency of inhibitors and support the use of this model to design potential modulators of PKC␦ activity. Taken together, we conclude that PP1 analogs inhibit AS PKC␦ and AS PKA by competing with ATP for interaction with Lys, Glu, Leu/Val, and Phe residues within the purine binding pockets of these mutant kinases.

DISCUSSION
In this report, we demonstrate that AS PKC␦ can use N 6 -(benzyl)-ATP as a phosphate donor to phosphorylate immediate substrates of PKC␦. One of the limitations of this approach is that ATP analogs, such as N 6 -(benzyl)-ATP, are not cell-permeable (5,6). One approach is to identify substrates in cell lysates, which may result in identifying proteins that do not have access to the kinase in intact cells. To compensate for this potential pitfall, ATP analogs have been successfully delivered to detergent-permeabilized cells to detect substrates within intact intracellular compartments (17,36,37).
We also found that AS PKC␦ can be selectively inhibited by 1NA-PP1 and 2MB-PP1 in vitro and in intact neutrophils. Generation of O 2 . by neutrophils is an important mechanism in innate immunity and contributes to the pathogenesis of many diseases (38 . production (41). Although this peptide inhibitor was designed to disrupt protein-protein interactions with the regulatory domain of PKC␦, its design was not based on knowledge of PKC␦ binding partners specific to neutrophils. Therefore, such a negative result with this peptide does not eliminate a role for PKC␦ in neutrophil activation. Instead, our approach, which targets the catalytic activity of PKC␦, does not depend on prior knowledge about binding partners that may be cell type-specific. Indeed, 1NA-PP1 specifically reduced the O 2 . production from the neutrophils of AS PKC␦ knock-in mice (Fig. 6), sug- gesting that the kinase activity of PKC␦ is indispensable in generating O 2 . in neutrophils.
PP1 analogs are rapid, stable, reversible, cell-permeable inhibitors (5,6), as shown for a single application of 100 nM 1NM-PP1 or 1NA-PP1, which gave sustained inhibition within 20 min and lasted for at least 3 days in cortical neurons from AS TrkB knock-in mice (44). Additionally, removal of 1NM-PP1 for 2 h restored the kinase activity of AS TrkB. Similarly, in mouse embryo fibroblasts from AS JNK2 knock-in mice, JNK2 kinase activity was completely inhibited by 10 M 1NM-PP1 in 15 min and recovered 15 min following 1NM-PP1 removal (45).
Although N 6 -(benzyl)-ATP and PP1 analogs are highly selective for AS kinases, off-target effects have been described (5,6). We investigated one possible off-target family of proteins, membrane-bound mammalian adenylyl cyclases (mAC), which possess large hydrophobic pockets in their catalytic domains that accommodate bulky substituents in nucleotides, such as MANT (27,46). Our docking studies suggested that N 6 -ATP and PP1 analogs are unlikely to be specific ligands for mAC (Fig. 13). A recent paper examined off-target effects of PP1 analogs with several hundred WT kinases (6). Although some kinases are weakly inhibited by PP1 analogs in vitro, they are not likely to be bona fide targets in cells Our in silico docking studies support the chemical-genetics approach for AS kinases (5,6). Placement of the benzyl ring of N 6 -(benzyl)-ATP in the cavity created by the M427A mutation allowed stable interaction with the glycine-rich loop for efficient phosphoryl transfer. The naphthyl and 2-methylbenzyl rings of 1NA-PP1 and 2MB-PP1 were similarly accepted in the cavity, permitting 1NA-PP1 and 2MB-PP1 to compete with ATP for stable interactions with Glu-428 and Leu-430 as well as the invariant Lys-378 and Phe-633 of AS PKC␦. In contrast, the insensitivity of WT PKC␦ for PP1 and N 6 -ATP analogs could be explained by steric clash of the naphthyl ring of 1NA-PP1 and the benzyl ring of N 6 -(benzyl)-ATP with the large Met-427 gatekeeper.
The glycine-rich loop (G 356 KGSFGK), invariant Lys-378, gatekeeper Met-427, two residues near the gatekeeper (Glu-428 and Leu-430), and Phe-633 are associated with 1NA-PP1, 2MB-PP1, and N 6 -(benzyl)-ATP binding in AS PKC␦. These residues are in the amino-terminal lobe of the PKC␦ catalytic domain. The catalytic domain for substrate phosphorylation is highly conserved in the AGC family (1,29), and the phosphoryl transfer event has been well characterized in PKA (29,30). The invariant Lys interacts with ␣and ␤-phosphates, and the glycine-rich loop interacts with ␤and ␥-phosphates to orient ATP for catalysis. Thus, disrupting these interactions would be predicted to impede ATP binding and substrate phosphorylation.
The gatekeeper lies in the middle of the catalytic domain and is a key residue for determining the selectivity of inhib-itors (1). This residue is conserved as a large hydrophobic amino acid. No kinases with a small Gly or Ala gatekeeper residue have been found in the human kinome (47). The side chain of the gatekeeper controls the relative accessibility of a hydrophobic pocket located adjacent to the N6 amine of ATP. Moreover, the gatekeeper residue may also regulate the autocatalytic activity of ERK2 (48) and the flexibility of PKA (49).
Staurosporine is a nonspecific kinase inhibitor, inhibiting more than 90% of kinases (50). Interestingly, we found that AS PKC␦ is less sensitive to staurosporine inhibition (Fig. 3). Mutation of the gatekeeper residue usually causes minimal change in kinase activity (47) but might confer resistance to inhibitors. The size of the gatekeeper residue has been correlated with the binding affinity for staurosporine (51). The group of kinases with larger gatekeeper residues (e.g. Met or Phe) tends to have stronger binding affinity for staurosporine, whereas smaller gatekeepers (e.g. Thr or Leu) are associated with weaker binding affinities. Mutations at gatekeeper residues have been reported in BCR-ABL1 (breakpoint cluster region-Abelson murine leukemia viral oncogene homolog 1) (T315I), PDGFR␣ (plateletderived growth factor receptor ␣) (T674I), EGF receptor (T790M), and KIT (T670I) (52). These mutations confer resistance to several kinase inhibitors (53).
PKC isozymes have unique and sometimes opposing roles in both normal cellular physiology and various disease states (1,42). PKC transduces its signals by both protein-protein interactions and substrate phosphorylation. Developing specific PKC isozyme inhibitors has been challenging due to their closely related structures (43). Selective PKC isozyme translocation inhibitors have been developed to disrupt protein-protein interactions with receptors for activated C kinase (RACKs) (42,43), but the respective RACK proteins have not been disclosed for every PKC isozyme (1). The chemical-genetics approach we used here provides not only a method to specifically interrogate the role of individual kinases in cell signaling  but also a method to reveal kinase-specific sets of phosphorylated substrates. The AS PKC␦ models herein will not only be useful in identifying PKC␦-specific substrates and designing modulators in silico to manipulate PKC␦ kinase activity but will also provide a foundation for future application of this chemical-genetics approach to other protein kinases.