Independence of Protein Kinase C-δ Activity from Activation Loop Phosphorylation

Activation loop phosphorylation plays critical regulatory roles for many kinases. Unlike other protein kinase Cs (PKC), PKC-δ does not require phosphorylation of its activation loop (Thr-507) for in vitro activity. We investigated the structural basis for this unusual capacity and its relevance to PKC-δ function in intact cells. Mutational analysis demonstrated that activity without Thr-507 phosphorylation depends on 20 residues N-terminal to the kinase domain and a pair of phenylalanines (Phe-500/Phe-527) unique to PKC-δ in/near the activation loop. Molecular modeling demonstrated that these elements stabilize the activation loop by forming a hydrophobic chain of interactions from the C-lobe to activation loop to N-terminal (helical) extension. In cells PKC-δ mediates both apoptosis and transcription regulation. We found that the T507A mutant of the PKC-δ kinase domain resembled the corresponding wild type in mediating apoptosis in transfected HEK293T cells. But the T507A mutant was completely defective in AP-1 and NF-κB reporter assays. A novel assay in which the kinase domain of PKC-δ and its substrate (a fusion protein of PKC substrate peptide with green fluorescent protein) were co-targeted to lipid rafts revealed a major substrate-selective defect of the T507A mutant in phosphorylating the substrate in cells. In vitro analysis showed strong product inhibition on the T507A mutant with particular substrates whose characteristics suggest it contributes to the substrate selective defect of the PKC-δ T507A mutant in cells. Thus, activation loop phosphorylation of PKC-δ may regulate its function in cells in a novel way.

Protein kinase C (PKC) 2 is a family of 9 genes that can be further divided into classical, novel, and atypical PKCs, depending on their structural characteristics and their requirement for activation (1,2). Each of them is autoinhibited by an intramolecular interaction of the kinase domain with an N-terminal regulatory domain, whose organization differs between subfamilies. Classic PKCs have C1 and C2 domains that bind diacylglycerol and Ca 2ϩ , respectively, for their activation. Novel PKCs have a C1 domain that binds diacylglycerol, but their C2-like domain does not bind Ca 2ϩ . Atypical PKCs do not have the ability to bind either Ca 2ϩ or diacylglycerol, but are activated by other lipids or small G-proteins. Binding of the regulatory region with appropriate cofactors causes a conformational change that releases the autoinhibition and results in activation.
Besides the co-factor-induced conformational change, PKC activity is also regulated by phosphorylation on its kinase domain, most importantly on its activation loop (3,4). The activation loop is a stretch of 20 -30 amino acids located in the catalytic cleft of the kinase domain of all eukaryotic protein kinases that form part of the substrate peptide binding surface. The activation loop is relatively flexible, and undergoes varied forms of conformation regulation between the active and inactive states (5)(6)(7). One of the most common modes of kinase regulation is by phosphorylation of residues in the activation loop (8,9). When phosphorylated, the negatively charged phosphate forms critical interactions with charged residues on the kinase domain, stabilizing an active conformation. In some cases, there are no phosphorylation sites on the activation loop, but instead a negatively charged residue serves as a functional replacement for the phosphate group. There are also kinases that do not need negative charge on the activation loop to be active. Generally, PKCs depend on activation loop phosphorylation for their kinase activity. It is believed that most PKCs are well phosphorylated at its kinase domain shortly after synthesis (2,4). So most cellular PKC is "competent," but its activity is autoinhibited until it is released by a conformational change caused by co-factor binding.
But PKC-␦ is a special case; the dependence of its activity on activation loop phosphorylation is still a point of controversy. Some studies have found that the activation loop-dephosphorylated mutant of PKC-␦ was still fully active in the in vitro kinase assay (10 -12). Others found that PKC-␦ only had minimal activity when not phosphorylated, and its activity increased about 10-fold when phosphorylated on its activation loop (13). Although various possible explanations have been proposed, the cause of the discrepancy is still unclear. In contrast to PKC-␦, the catalytic activity of the closest isoform, PKC-, is strictly dependent on activation loop phosphorylation (12). Because the two kinases have 87% sequence similarity in the kinase domain, they constitute an informative pair for investigating the structural basis of the unique functional properties of PKC-␦.
As noted above, activation loop phosphorylation is generally constitutive in novel and classical PKCs in cells. In contrast, although PKC-␦ is constitutively phosphorylated on its activation loop in some cell types, in a variety of contexts it undergoes regulated activation loop phosphorylation or dephosphorylation in cells (13)(14)(15)(16), which therefore resembles the regulatable phosphorylation that is so critical for many other kinases, including AGC kinases such as PKC-, PRK1, and AKT. It is appealing that this phosphorylation has an important regulatory role in cells, but that possibility has been clouded by the controversy regarding the functional effect of phosphorylation at this site.
We therefore investigated in greater detail the structural basis of activity regulation by phosphorylation of Thr-507 and the functional consequences thereof. Using a combination of mutational analysis, structural modeling, and molecular dynamics we have identified the molecular basis for stabilization of the activation loop of PKC-␦ in the absence of phosphorylation. Moreover, our assessment of its functional capacity demonstrates both an altered range of substrates in cells and increased susceptibility to product inhibition in vitro. Of particular interest, we find that phosphorylation of Thr-507 is not required for apoptosis, one of the most notable functions of PKC-␦ in cells.

MATERIALS AND METHODS
DNA Constructs-PKC-␦ cDNA constructs have been described previously (12). PKC-␦ kinase domain and substrate (MARCKS and CREB) constructs were created using the Gateway cloning system (Invitrogen). The kinase domains (with or without the 20-residue N-terminal extension) are from human protein: PKC-␦ (aa 329 -676 or 349 -676). Targeted substrate constructs were made by cloning the annealed oligonucleotides coding for the following peptide sequence into the pDest516 Gateway destination vector with Lck targeting motif and HA tag at the N-terminal and green fluorescent protein tag at the C terminus: human MARCKS aa 154 -168 (KKKRFSFKKSFKLSG) and CREB aa 126 -140 (EILSRRPSYRKILND) (supplemental Fig. 6A).
Transfection, Apoptosis Assay, Immunoprecipitation, and Western Blot-HEK293T cell transfection was done with calcium phosphate using standard procedures. For apoptosis assay, 20 h after transfection the cells were fixed with 4% paraformaldehyde, permeabilized, stained with HA mAb and 4,6-diamidino-2-phenylindole, and observed with a Zeiss LSM510 confocal microscope. The transfected cells were lysed with lysis buffer: 1% Triton X-100, 150 mM NaCl, 50 mM Tris-HCl, pH 7.4, 2 mM Na 3 VO 4 , 5 mM Na 4 P 2 O 7 , and protease inhibitor mixture tablet. The lysate was cleared by centrifugation at 14,000 ϫ g for 10 min. For immunoprecipitation, lysate from one 10-cm dish was combined with 2 g of HA mAb and then 20 l of recombinant protein G-agarose (Invitrogen). For Western blot, the immunoprecipitate or cell lysate in the SDS sample buffer were separated by SDS-PAGE and transferred to nitrocellulose membrane. Western blot was done using the Amersham Biosciences ECL kit following the manufacturer's protocol. NF-B and AP-1 reporter assays in transfected Jurkat cell were done as described previously (12).
In Vitro Kinase Assay-In vitro kinase assays with PKC-␦ were generally done using eluted purified kinase (Figs. 4 -6 and 8) or in a few experiments by an immune complex assay ( Figs. 1 and 2). The immune complex kinase assay was done as described previously (12). Briefly, PKC constructs were transfected into HEK293T cells, and immunoprecipitated using HA mAb. The immune complex kinase assay was done using 20 M PKC-␣ pseudo-substrate peptide as substrate, which was captured onto P81 paper and emissions were counted. The kinase buffer contains 100 mM HEPES, pH 7.4, 0.05% Triton X-100, 1 mM CaCl 2 , 20 mM MgCl 2 , 200 g/ml phosphatidylserine, 100 ng/ml phorbol 12-myristate 13-acetate, 100 M ATP, 1 Ci of [␥-32 P]ATP. A fraction of the immunoprecipitated PKC was blotted with HA mAb, quantitated, and assays adjusted to equal concentrations.
For determining kinetic parameters of PKC-␦ and its mutants, the transfected kinase was immunoprecipitated using HA mAb and eluted with HA peptide, and the purified kinase was used in the kinase assay with the same procedure as in the immune complex kinase assay. Comparisons of the relative kinase activity of different constructs were made from in vitro kinase assays performed at equal kinase concentrations as determined by Western blots. For kinase assays using biotinylated peptides (Figs. 4 -6), phosphorylation was assessed by capture of peptides onto streptavidin-coated plates as described previously (17). K m for ATP was assessed using PKC-␣ pseudo-substrate peptide as substrate and ATP concentrations ranging from 0.25 to 400 M.
In-cell Phosphorylation Assay-Lck motif-targeted PKC-␦ kinase domain constructs and substrate (MARCKS and CREB) constructs were co-transfected into HEK293T cells. 24 h after transfection, the cells were harvested and lysed. The cell lysate was blotted with HAtagged mAb, to determine expression of both kinase domain and substrate constructs, and pMARCKS/pCREB antibody, to determine phosphorylation of the substrates. For all assays, all results shown are representatives of at least two independent experiments with similar results.
Molecular Modeling-The spatial structure model of PKC-␦ was constructed based on the known structure of protein kinases: PKC-, Protein Data Bank codes 1XJD; AKT2, 1O6L; and PKA, 1L3R, using the COMPOSER subroutine of SYBYL (Tripos, St. Louis, MO). Models of T507A and T507A/F500L mutants were made by creating point mutations. The molecular coordinates of the atoms in those mutants were first optimized by potential energy minimization using SYLBL subroutines and AMBER Force Fields. Terms of the energy function used for optimization include bond stretching, angle bending, torsional energy, van der Waals energy terms, electrostatic energy terms, and hydrogen bonding terms. For each of the mutants 10 short independent molecular dynamic simulations were carried out in the presence of solvent (box 77 ϫ 62 ϫ 62 Å with ϳ8,000 water molecules) at 300 K. These minimized models were then further optimized in SYBYL by "heating," which allows atoms to further reposition into even more energy-optimal conformational states. The total time of each molecular dynamic simulation was 20 ps; the Verlet algorithm, one of numerical algorithms for integrating the equations of motion in potential energy function (i.e. for calculating the trajectories of particles), was used with 1-fs steps.

Two Phenylalanines In/Near the Activation Loop of PKC-␦ Are Involved
in Maintaining Kinase Activity-We sought a structural explanation for the dramatic difference shown previously between PKC-and -␦ in their dependence on activation loop phosphorylation. We explored the possibility that sequence differences between PKC-␦ and -, which are critical to maintaining an active conformation without phosphorylation are: 1) residues in the activation loop itself; and 2) residues that directly contact the activation loop. Although no structure is available for the PKC-␦ kinase domain, the similarity between PKC-␦ and PKC-made it straightforward to find such residues in a "first-pass" homology model. The notable differences are in the activation loop itself and ϳ15 residues thereafter that are located in the C-lobe. Two Phe residues (Phe-500 and Phe-527) are notable among the few non-conservative changes (Fig. 1). Three attributes made them particularly interesting: 1) Phe is a very hydrophobic residue often involved in hydrophobic stacks that provide stable interactions. 2) They are located sufficiently close in the spatial model to make a hydrophobic stack between them plausible. 3) PKC-␦ is the only PKC with a Phe at both positions, which could explain its uniqueness.
As an initial test of whether these Phe were critical to activation loop independence of PKC-␦, functional studies were performed with mutant constructs. The two Phe residues were mutated to the corresponding residue in PKC-(F500L/F527H); mutations were made both in the wild type (WT) PKC-␦ and PKC-␦ mutant whose activation loop is unable to be phosphorylated (T507A). In vitro kinase assays demonstrated that mutation of either Phe (or both) dramatically decreased kinase activity of PKC-␦ T507A (Fig. 1B). These Phe were unusually important because mutation of other residues corresponding to non-conservative changes (I499M, A505T, and T526N) did not have any significant effect on kinase activity of PKC-␦ T507A (Fig. 1B). When the same analysis was conducted with mutants on the WT background, the Phe mutation did not have any defect. Thus the two Phe are critical for kinase activity only when the activation loop is not phosphorylated.
The N-terminal Extension of the Kinase Domain Helps Maintain Kinase Activity by Interactions with the Activation Loop-Two considerations prompted us to consider the possibility that residues N-terminal to the kinase domain contribute to activation loop stabilization. (a) A mutant of PKCin which the residues corresponding to Phe-500 and Phe-527 in PKC-␦ (Leu-531 and His-558) are mutated to Phe is still not active without activation loop phosphorylation, suggesting requirements beyond the two Phe for activity. (b) Modest sequence similarity between PKC-␦ and the A-helix of PKA ( Fig. 2A) suggested that PKC-␦ might have an ␣-helix. This would be relevant because the ␣-helix in PKA ("A-helix") plays a role in stabilizing the kinase domain (18,19). To test the possibility that the sequence N-terminal to the kinase domain plays a role in stabilizing PKC-␦, we made four PKC-␦ kinase domainonly constructs varying with respect to: 1) presence versus absence of 20 residues of the N-terminal sequence (aa 329 -348) and 2) WT versus T507A. The in vitro kinase activity of T507A is markedly impaired in the absence of the N-terminal extension but close to normal with that extension (Fig. 2B). In contrast, the presence of a phosphorylatable activation loop (WT) allows PKC-␦ to function normally without its N-terminal extension. Thus, the N-terminal extension, like the pair of Phe, is necessary for kinase activity only in the absence of activation loop phosphorylation.
Structural modeling was undertaken to determine a structural explanation for the experimental results with the above constructs. Our FIGURE 1. Two phenylalanines (Phe-500 and Phe-527) are critical for PKC-␦ kinase activity when without activation loop phosphorylation. A, sequence alignment between PKC-␦, closely related protein kinases, and CKII showing most of the activation loop and 15 residues C-terminal to it. Symbols above the alignment indicate identity (*) or similarity (:) among all kinases shown except CKII. Bold highlighted residues are nonconservative differences between ␦ and ; the Phe-507 in PKC-␥ is the only aromatic residue in PKCs that is aligned with PKC-␦ Phe-500 or Phe-527; and relevant aromatic residues in CKII. B, immune complex kinase assay comparing phosphorylation by relevant mutant constructs derived from WT PKC-␦ or from T507A. Results are expressed as percent relative to WT control Ϯ S.E. Standard errors for the left panel are too small to be visible. model of the kinase domain per se was based on the solved structure of its closest paralog, PKC-(Protein Data Bank code 1XJD). Two elements that are important for a functional kinase but are missing from that structure were derived from other closely related kinases: part of the C-terminal extension (aa 619 -633) from AKT2 (PDB code 1O6L) and the rest (aa 636 -645) from PKA (PDB code 1L3R), and the A helix (aa 320 -340) from PKA (PDB 1L3R). An important feature of the model is molecular interactions of the A-helix with the "bottom" of the kinase domain (Fig. 3). Its interaction with the C-helix is virtually identical to that observed in PKA, based on the conservation of critical Arg on the C-helix and a Trp on the A-helix. An interesting difference, however, is observed in the remainder of its interactions. In PKA the A-helix has strong hydrophobic interactions with a major hydrophobic groove on the bottom of the C-lobe (18,19). In contrast, in the PKC-␦ model the A-helix does not interact directly with the C-lobe but rather is connected to the C-lobe indirectly via hydrophobic residues in a short "insertion" in the activation loop (i.e. in PKCs but not PKA, Fig. 1A). Specifically, 1) Tyr-334 in the A-helix interacts with Phe-500 in the activation loop; and 2) Ile-499 in the activation loop interacts with Phe-527 in the C-lobe. Thus, the three structural elements found experimentally to be important for PKC-␦ catalysis in the absence of activation loop phosphorylation (Phe-500, Phe-527, and A-helix) are all involved in this chain of hydrophobic interactions unique to PKC-␦. The surface features of these structural elements that create the hydrophobic interactions are depicted in supplemental Fig. 1.
Molecular dynamics was used to investigate in more detail the position and function of critical phenylalanine residues Phe-500 and Phe-527. In the case of WT PKC-␦ an unusual predominance of a single conformation was observed during multiple simulations, suggesting that this conformation is strongly favored. Preference of that conformation also was observed in models of T507A, but when Phe-500 is substituted by Leu the structure becomes more flexible mainly because of the loss of the tight interaction between Phe-500 and Tyr-334, indicating the special role of those aromatic residues in maintaining this stabilization of the activation loop. This favored conformation is controlled by distinctive interactions of Phe-500 with Tyr-334 and Phe-527 with Ile-499. Thus, the two Phe residues unique to PKC-␦ appear to be essential to form that stable conformation. It is notable that the structural elements necessary for this stabilization (the pair of Phe residues and motif in N-terminal region) have been conserved in most vertebrates (mouse, rat, dog, xenopus, and even pufferfish).
Analysis of ATP binding (supplemental Fig. 2) shows that the double Phe mutant F500L/F527H has a ϳ2-fold increase in the K m for ATP, confirming the role of those Phe in optimizing kinase domain conformation. This alteration was less than the increase caused by the T507A mutation, which was ϳ4-fold. The T507A mutant also had altered lipid dependence for activation (supplemental Fig. 3) as previously described (14).
Product Inhibition of PKC-␦ T507A with Some Substrates-Because the activation loop forms part of the peptide substrate binding surface, the presence or absence of phosphorylation on the activation loop of PKC-␦ could potentially change its substrate specificity (14,20). This would be important to know, both for interpretation of results on kinase  activity, and for its potential relevance to the functional effects of regulated phosphorylation of PKC-␦ in cells. To explore this possibility, we compared PKC-␦ WT and T507A substrate specificity by positional scanning with degenerate peptide, which has proved highly informative in investigating kinase peptide specificity (17). T507A was virtually identical to WT as assessed by this method for substrate positions between PϪ7 and Pϩ6 (supplemental Fig. 4).
As an alternative way to assess peptide specificity, we analyzed phosphorylation by PKC-␦ WT, T507A, and the FF mutant (F500L/F527H as a control) of a panel of 96 proteomic peptides. Those peptides were chosen from genomic sequences not only for similarity to the consensus sequence of PKC substrates but also for diversity between peptides in exact sequence. Consequently, the set incorporates a diverse set of candidate PKC phosphorylation sites. Examination of the patterns shows that at a low peptide concentration (1 M) the extent of phosphorylation of all peptides by T507A was very similar to WT (correlation coefficient ϭ 0.94). But at a higher peptide concentration (10 M) rather marked scatter became apparent (correlation coefficient ϭ 0.56) (Fig.  4). The distribution of peptides suggested that many fell within a fairly linear distribution, but that there were a substantial subset of peptides for which phosphorylation by T507A was substantially less efficient. The specificity of the FF mutant protein showed no such deviation either at low or high peptide concentrations (correlation coefficients Ͼ ϭ 0.96 at both concentrations).
To confirm and extend this unexpected observation, we preformed a more careful peptide titration with four potentially informative well phosphorylated peptides for which we had sufficient quantity for analysis (Fig. 5A). Two peptides have classic titration curves (D40 and PARIS, Fig. 4B, peptides 3 and 4) in which phosphorylation increased with increasing peptide concentration and plateaued around 10 M; these peptides are the ones for which the screening assay indicated equivalent phosphorylation by WT and mutant. A very different pattern was observed with the other two peptides (MLK3 and diacylglycerol kinase-), which the screening assay had suggested behave anomalously with T507A (Fig. 4B, peptides 1 and 2). Their phosphorylation at low peptide concentrations by WT and T507A was very similar, but their phosphorylation by WT and T507A differed significantly at a high peptide concentration. Instead of a plateau, phosphorylation by T507A decreased significantly at high peptide concentrations. The results suggest that inhibition of phosphorylation by PKC-␦ T507A occurs with some peptide substrates at high concentrations either as a result of substrate inhibition or product inhibition. To begin to distinguish these two possibilities we analyzed phosphorylation as a function of incubation time (Fig. 5B) Table I). of PARIS peptide increased as a function of time and was similar between T507A and WT, showing a regular time-dependent increase. In contrast, phosphorylation of MLK3 peptide by T507A was similar to that of WT at early time points but dropped off as the reaction continued. The progressive decrease in efficiency of T507A phosphorylation of MLK3 (but not of PARIS) was consistent with inhibition of T507A by the accumulating phospho-MLK3 product, the binding of which to the kinase domain would be facilitated by its phosphate group occupying the positive charge pocket left empty by the pseudo-dephosphorylation of the activation loop phosphorylation site (Thr-507).
To test this, we synthesized phosphorylated and non-phosphorylated "inhibitor" peptides corresponding to: 1) the peptide showing greatest high-dose falloff (MLK3, Fig. 5A); and 2) one of the peptides showing no high-dose falloff (PARIS, Fig. 5A). The ability of each peptide to inhibit was tested over a range of peptide concentrations on WT and T507A (Fig. 6). The inhibitor peptide designed from the well behaved substrate (PARIS) was unremarkable; it was an inefficient inhibitor for both proteins and the corresponding phosphopeptide was a somewhat poorer inhibitor consistent with its less favorable charge. The inhibitor peptide designed from the anomalous substrate (MLK3) behaved in a different manner. It was a much better inhibitor for both WT and T507A than unphosphorylated PARIS. Most notable is the finding that the MLK3 phosphopeptide showed a very strong phosphate-dependent inhibition primarily of T507A (K i ϳ 2.5 M). These results indicated that unphosphorylated PKC-␦ (T507A) was very susceptible to product inhibition by some of its substrates.
PKC-␦ T507A Is Defective in Reporter Activation in Cells-The foregoing studies demonstrate that although the pseudo-dephosphorylated PKC-␦ (T507A) has "normal" in vitro kinase activity on many substrates, it also has significant functional alterations of in vitro phosphorylation of some peptides and susceptibility to inactivation by other mutations. Because non-phosphorylated PKC-␦ is present in cells, we addressed the critical question as to whether non-phosphorylated PKC-␦ (T507A) functions normally in cells using NF-B and AP-1 reporter assays. Those studies demonstrate that constitutively active PKC-␦ is a potent activator of both NF-B and AP-1 reporter constructs in Jurkat T cells (Fig. 7A). In contrast, the T507A mutant of the constitutively active PKC-␦ failed to activate either NF-B or AP-1 despite similar levels of expression. Because such reporter assays require kinase activity (sup- FIGURE 6. Some peptides mediate strong phosphate-dependent inhibition of T507A. In vitro kinase assay of PKC-␦ and its T057A mutant on 3 M biotinylated PARIS peptide in the presence of the graded concentrations of phospho-or non-phospho-MLK3 and PARIS peptides ("inhibitors"). At the end of the assay, biotinylated PARIS substrate peptide was captured onto a streptavidin plate and 32 P incorporation was assessed. Kinase activity without inhibitor peptide was taken as 100%, and activity with various concentrations of inhibitor peptide was expressed as a percentage. FIGURE 7. The T507A mutant cannot activate NF-B and AP-1 in cells. NF-B and AP-1 reporter assays were performed in Jurkat Tag cells using either full-length constitutively active PKC-␦ constructs (A) or kinase domain-only constructs targeted to the membrane via the Lck-lipid binding motif (B). In each context the reporter activity was compared between the WT construct and a construct with a T507A mutation. Graded levels of expression were achieved by transfecting graded amounts of the cDNA and expression was assessed by Western blot. Such cDNA titration provided conditions where constructs were equivalently expressed and therefore activities could be fairly compared.
plemental Fig. 5), the simplest interpretation is that unphosphorylated PKC-␦ is unable to mediate phosphorylation of the specific sites in cells involved in NF-B and AP-1 activation.
To exclude the possibility that activation loop phosphorylation may regulate localization of PKC-␦ in the cells, which might be critical for its function in cells, we made Lck motif-targeted PKC-␦ kinase domain constructs (PKC-␦ aa 329 -676) with Thr-507 mutated or not. The Lck motif contains the N-terminal 15 amino acids of Lck, which has been shown to be the lipid modification site that targets Lck to membrane lipid rafts. The Lck motiftargeted PKC-␦ kinase domain construct was capable of activating both NF-B and AP-1 reporters (Fig. 7B). But constructs in which the Thr-507 was mutated totally lost that capability. Untargeted PKC-␦ kinase domain had little activity in reporter assay (data not shown). Because both constructs were targeted by the same motif, the results indicated that kinase activity rather than localization was the reason the activation loop mutant failed to activate the reporters.
PKC-␦ T507A Has a Substrate Selective Defect in Phosphorylating Substrate in Cells-Clarification of the molecular defect in T507A would be provided by an assay of its capacity to directly phosphorylate its substrate in cells; ideally this assay should be in the same cells used for the reporter assay. However, to our knowledge no physiologically relevant PKC-␦ substrate has been definitively identified in T-cells. We therefore developed an in-cell assay that involves co-transfection of a kinase construct and a fusion protein that includes site(s) suitable for phosphorylation by PKC, such as those in MARCKS (21). Colocalization is a major physiological mechanism for facilitating signal transduction, including phosphorylation (22,23). Therefore, this assay exploits colocalization by incorporating into both the kinase and substrate constructs an Lck targeting motif that localizes them to the same membrane location (see supplemental Fig. 6). To determine whether the in-cells defect of T507A observed in reporter assays reflects a defect in catalysis, we analyzed its ability to phosphorylate in this in-cells co-targeting assay. The results demonstrate a striking defect in its in-cells phosphorylation of the co-targeted MARCKS sites (Fig. 8A). We extended the analysis to a different co-targeted substrate that contained the Ser-133 of CREB-␣, which is described to be a substrate for both PKA and PKC (24). In contrast to the results with the MARCKS sites, the CREB site was phosphorylated similarly by WT and T507A (Fig. 8B); phosphorylation of CREB peptide is detectable both by phosphoantibody and by a decrease in mobility on SDS-PAGE. Thus, the results indicate that PKC-␦ T507A is catalytically active but has a substrate-selective defect in phosphorylating substrates in cells.
We hypothesized that the selective defect of T507A in cells might relate to the product inhibition of the selective substrate in vitro. So we tested PKC-␦ WT and T507A activity on MARKCS and CREB peptide in in vitro kinase assays (Fig. 8, C and D). WT and T507A phosphorylate CREB peptide equally well. At low peptide concentrations, WT and T507A both phosphorylated MARCKS peptide equally well. But at high peptide concentrations, T507A activity is significantly lower than WT. So T507A-specific product inhibition correlates with the defect of phosphorylation in cells.
T507A in Apoptosis-Because PKC-␦ has been demonstrated to mediate the biologically important process of apoptosis (25), we investigated functional activity of T507A in apoptosis. In cellular models of apoptosis, caspase cleavage of PKC-␦ to generate an active catalytic fragment is often required for apoptosis (25,26). We used a well characterized model of apoptosis induced by kinase domain transfection (26); our constructs correspond to the region produced by caspase cleavage. We found that PKC-␦ kinase domain T507A caused apoptosis judging by nuclear DNA condensation and fragmentation (as well as rounding up of cells) (Fig. 9A). T507A-induced apoptosis was comparable with that observed with the WT PKC-␦ kinase domain expressed at comparable levels. As shown previously (26), apoptosis depends on the catalytic activity and was not observed with the kinase-dead K378W mutant. Thus, for inducing apoptosis, activation loop-dephosphorylated PKC-␦ was as effective as the WT, in sharp contrast with NF-B and AP-1 reporter activation.
We considered whether apoptosis induction by the WT PKC-␦ kinase domain might be mediated by a pool thereof that is not phosphorylated on Thr-507. Using in-cell treatment with phosphatase inhibitor, we observed that the amount of Thr-507 phosphorylation could be increased 6-fold within 20 min without a change in PKC-␦ kinase domain expression (Fig. 9B). Thus, at least 85% of WT PKC-␦ kinase domain is not phosphorylated on Thr-507. This unphosphorylated pool  both (B). Expression of the substrate constructs (MARCKS or CREB) and the kinase domain constructs was assessed by Western blot with antibodies against peptide tags. Western blot results were quantitated and represented in a bar graph in which phosphorylation by WT is shown as 100% and phosphorylation by T507A was expressed as percent of the WT. Note that the CREB substrate undergoes a mobility shift after phosphorylation. The arrowhead indicates the position of a slightly lower mobility nonspecific band that is also seen in untransfected cells (not shown). C and D, PKC-␦ WT or T507A protein was purified from transfected HEK293T cells, adjusted to equal protein concentrations by Western blot comparisons, and the immune complex kinase assay was done with serially diluted MARCKS peptide (C) or CREB peptide (D) as substrates.
is about comparable with that present in the T507A transfection and therefore sufficient to explain the apoptosis observed. This raises the possibility that the Thr-507-phosphorylated PKC-␦ kinase domain may not mediate apoptosis.

DISCUSSION
The foregoing studies were initiated to understand the structural basis and significance of the finding that the catalytic activity of PKC-␦ is independent of activation loop phosphorylation, unlike all other PKCs. Three particularly notable aspects of the findings warrant discussion. First, PKC-␦ has a unique structural mechanism for stabilizing its activation loop, which enables it to be catalytically active in the absence of activation loop phosphorylation. Second, the mutant PKC-␦ lacking activation loop phosphorylation shows striking susceptibility to inhibition by certain of its phosphopeptide products. Third, PKC-␦ T507A, which lacks activation loop phosphorylation, differs profoundly in cells from WT PKC-␦; it is defective in activity in two reporter assays, and displays a substrate-selective defect in phosphorylation when co-targeted with substrates in cells but it is as active as WT PKC-␦ in induction of apoptosis.
The structural mechanism by which PKC-␦ stabilizes its activation loop in the absence of phosphorylation is unique among AGC kinases. It involves hydrophobic interactions of the activation loop with both a short N-terminal ␣-helix and with an aromatic residue in the C-lobe. The N-terminal A-helix of the PKC-␦ kinase domain has fundamental similarities to the A-helix of PKA both in terms of its general position and of the detailed mechanism of the A-helix-stabilizing interaction with the C-helix. All novel and classical PKCs have the Arg at the C-terminal of the C-helix that PKA uses for binding to the A-helix. However, only PKC-␦ has the corresponding Trp in a suitable position to contribute to binding of an A-helix in the manner used by PKA. Despite the similarity, unlike PKA, the A-helix of PKC-␦ binds directly to the activation loop through hydrophobic interactions (see "Results" and supplementary figures). Phe-500 plays a critical role as the principal contact between the activation loop and A-helix. Molecular modeling of F500L mutations shows that it is critical to maintenance of activation loop backbone conformation; i.e. conformation changed in T507A/F500L but not in T507A (supplementary Fig. 7). The strategy of an N-terminal extension stabilizing activation loop conformation has been observed in only one other kinase, casein kinase II (CKII) (27), which does not require activation loop phosphorylation for activity and prefers highly acidic protein substrates. The isolated catalytic subunit of CKII is constitutively active. But when interacting residues in the N-terminal extension or activation loop are deleted or mutated, CKII loses activity or its activity become dependent on interaction with the regulatory subunit (28,29). The structural basis of stabilization of PKC-␦ and CKII involves the same three discrete regions and are fundamentally similar. Aromatic residues are present in all three regions: N terminus, activation loop, and beginning of the F-helix. Multiple aromatic stacks provide stable contact of the N terminus to the activation loop and of the activation loop to the F-helix; the end result is an active conformation of the activation loop. Thus, PKC-␦ utilizes an unusual mode of activation loop stabilization that has only been observed in one other context in a distantly related kinase.
Activation loop phosphorylation has been demonstrated to have several possible functional consequences for kinases (5,7,8,30). One paradigm for the function of phosphorylation is "gated activation loops" in which phosphorylation prevents the activation loop from assuming a conformation that blocks substrate access (9). Another more widely relevant paradigm is that phosphorylation stabilizes the activation loop in an optimal conformation that facilitates phosphoryl transfer (9). There are a few citations that suggest an additional paradigm in which activation loop phosphorylation alters substrate specificity. In one instance the activation loop phosphate of CDK2 was shown to be important for substrate binding/alignment via its interaction with the Pϩ3 lysine (31). In contrast, two other reports of altered specificity cannot be explained on such a clearly defined structural basis. One shows that activation loop phosphorylation of calcium/calmodulin-dependent kinase I is not simply an on/off switch but broadens the range of phosphorylated substrates (32); analysis of kinetic parameters demonstrated a 40-fold improvement in binding. Steinberg and colleagues (14) demonstrated that PKC-␦ from activated cells was able to phosphoryl- ate additional substrates not well phosphorylated by PKC-␦ from resting cells, but the molecular basis was not defined. The present studies demonstrate a unique mechanism by which activation loop phosphorylation alters the ability of PKC-␦ to phosphorylate specific substrates, namely by altering its susceptibility to product inhibition. Non-phosphorylated PKC-␦ (T507A) is sensitive to inhibition by some of its phosphorylated products, exemplified by the MLK3 phosphopeptide that was inhibited with a K i of ϳ2.5 M. The inhibition was: 1) sequence-specific because it was not observed with other phosphopeptides; 2) dependent on peptide phosphorylation; and 3) dramatically more evident in the absence of activation loop phosphorylation.
The most likely structural explanation for product inhibition is that the phosphorylated product re-binds to the kinase domain with the phosphate group occupying the site normally occupied by activation loop phosphate. First, because the inhibition is phosphate dependent it is likely that there is a phosphate-binding site. Second, there is an unoccupied phosphate-binding site in T507A because the activation loop phosphate coordinates with positively charged residues Arg-472 and Lys-496. Third, that binding site, when unoccupied, can be used for binding exogenous substrate, as strikingly illustrated by glycogen synthase kinase-3, in which that pocket is used for binding of phosphateprimed substrate (33). Fourth, analysis of the sequences shows marked enrichment of basic residues N-terminal to the phosphorylation site in peptides that mediate high-dose inhibition ( Table 1). All the inhibitory peptides have 4 basic residues at positions between PϪ2 and PϪ5, whereas the non-inhibitory ones have only 2 basic residues at the most critical positions for PKC-␦ phosphorylation (PϪ2 and PϪ3). These "extra" basic residues are likely to participate in the binding. Two possible modes of binding seem plausible. The peptide may remain in the catalytic cleft but displaced by 3-4 residues toward its C terminus, analogous to the phosphate-primed glycogen synthase kinase-3 substrate. In that case the original PϪ5 position would now be PϪ1, which is a position that PKC-␦ prefers basic residues (17) and the original PϪ4 would be positioned to bind the ␤-phosphate of ADP. Alternatively, it may bind in a different manner in which, for example, the four basic residues binding to the nearby highly acidic G-helix and the phosphate group bind to Lys-524, which is located not directly in the catalytic cleft, but in close contact with the phosphate-coordinating group of residues according to our model. Thus, profound susceptibly to product inhibition by some of its substrates is a hallmark of non-phosphorylated PKC-␦.
The present studies demonstrate that PKC-␦ lacking activation loop phosphorylation (i.e. T507A) is functionally active in cells but that its functional capacities are only partly overlapping with capacities of phosphorylated WT PKC-␦. On the one hand T507A resembles WT PKC-␦ in induction of apoptosis by the isolated kinase domain. On the other hand, T507A lacks some functions of WT PKC-␦, namely the capacity to induce AP-1 and NF-B reporter activation. The quest for molecular explanation of these functional affects is complicated by the fact that the direct PKC substrates, which mediate these functional affects, are incompletely defined. We therefore established an assay to monitor in-cell phosphorylation by the constructs by co-targeting them to the membrane with peptide substrates. This novel assay provided a direct demonstration of shared function (phosphorylation of CREB Ser-133) and of T507A loss-of-function (phosphorylation of MARCKS basic effector domain). The loss-of-function change of MARCKS phosphorylation in T507A is plausibly explained by our foregoing studies demonstrating substrate-specific product inhibition for T507A. Notably, the MARCKS substrate that PKC-␦ T507A fails to phosphorylate in cells resembles the substrate class that gives product inhibition in vitro: 1) basic residues at PϪ2 through PϪ5; and 2) high dose inhibition observed with in vitro phosphorylation only by PKC-␦ T507A. Note that in-cells recruitment of substrates can result in concentrations comparable with the high substrate concentrations we tested in vitro (22,23). These data suggest a unifying hypothesis that product inhibition is a mechanism that constrains catalytic function of non-phosphorylated PKC-␦ in cells. Regardless of the mechanism, the foregoing studies demonstrate that the activation loop non-phosphorylated PKC-␦ is catalytically active in cells, but that its pattern of substrate phosphorylation and function is markedly changed. Differences between WT and T507A reflect novel regulatory roles of activation loop phosphorylation, quite different from the traditional role in turning on kinase activity.
We suggest the possibility that the unique design of PKC-␦ among PKCs (i.e. activity without activation loop phosphorylation) evolved due to the combination of: 1) the PKC-␦ catalytic fragment evolving an important function (e.g. in apoptosis); and 2) the vulnerability of catalytic fragments to activation loop dephosphorylation. If induction of apoptosis by the caspase-cleaved kinase domain depended on Thr-507 phosphorylation, it would be quite vulnerable to inactivation by dephosphorylation because the regulatory domain has been shown previously to help protect against PKC activation loop dephosphorylation (34). This vulnerability is borne out in our studies in which the catalytic fragment is poorly phosphorylated in cells (Fig. 9B), unlike full-length PKC-␦ in the same cells whose Thr-507 phosphorylation in cells is not augmented by phosphatase inhibitor (supplementary Fig. 8). Evolution may have fixed this vulnerability by engineering phosphorylation independence into the kinase domain of PKC-␦.