Effects on Ligand Interaction and Membrane Translocation of the Positively Charged Arginine Residues Situated along the C1 Domain Binding Cleft in the Atypical Protein Kinase C Isoforms*

The C1 domain zinc finger structure is highly conserved among the protein kinase C (PKC) superfamily members. As the interaction site for the second messenger sn-1,2-diacylglycerol (DAG) and for the phorbol esters, the C1 domain has been an important target for developing selective ligands for different PKC isoforms. However, the C1 domains of the atypical PKC members are DAG/phorbol ester-insensitive. Compared with the DAG/phorbol ester-sensitive C1 domains, the rim of the binding cleft of the atypical PKC C1 domains possesses four additional positively charged arginine residues (at positions 7, 10, 11, and 20). In this study, we showed that mutation to arginines of the four corresponding sites in the C1b domain of PKCδ abolished its high potency for phorbol 12,13-dibutyrate in vitro, with only marginal remaining activity for phorbol 12-myristate 13-acetate in vivo. We also demonstrated both in vitro and in vivo that the loss of potency to ligands was cumulative with the introduction of the arginine residues along the rim of the binding cavity rather than the consequence of loss of a single, specific residue. Computer modeling reveals that these arginine residues reduce access of ligands to the binding cleft and change the electrostatic profile of the C1 domain surface, whereas the basic structure of the binding cleft is still maintained. Finally, mutation of the four arginine residues of the atypical PKC C1 domains to the corresponding residues in the δC1b domain conferred response to phorbol ester. We speculate that the arginine residues of the C1 domain of atypical PKCs may provide an opportunity for the design of ligands selective for the atypical PKCs.

The protein kinase C (PKC) 3 isozymes comprise a family of serine/threonine protein kinases that are involved in the transduction of a large number of signals important for the regulation of proliferation, differentiation, apoptosis, and other physiological functions (1,2). The family is divided into three subfamilies: 1) the conventional PKCs (cPKCs) ␣, ␤ I, ␤ II, and ␥; 2) the novel PKCs (nPKCs) ␦, ⑀, , and ; and 3) the atypical PKCs (aPKCs) and / (human PKC and mouse PKC are orthologs). These three subfamilies share a common requirement for phospholipids for their kinase activity but differ in their dependence on other activators (3). The cPKCs (␣, ␤ I, ␤ II, and ␥) meet the original definition of PKC as a Ca 2ϩ and phospholipid-dependent protein kinase. Indeed, they require phosphatidylserine (PS), Ca 2ϩ , and diacylglycerol (DAG) (or phorbol esters) for their activation. The nPKCs (␦, ⑀, , and ) require only DAG/phorbol ester and PS. The activation of aPKCs ( and /) is much different from that of the cPKCs and nPKCs. They require only PS but not Ca 2ϩ and DAG/phorbol ester (4). As an important second messenger, DAG plays the major role in transducing cellular signals to the PKC molecules (5). A highly conserved cysteine-rich motif (the so-called "C1 domain") in the regulatory region of the PKCs acts as the specific receptor for the DAG signal (6,7). All of the three PKC subfamilies contain this C1 domain. The cPKCs and nPKCs have two tandem C1 domains in their N termini, the C1a and C1b domains. They both showed high binding affinities for DAG or phorbol esters in vitro as isolated fragments (8,9). However, although the aPKCs, PKC as well as PKC/, also possess a C1 domain at their N termini, these atypical C1 domains are not sensitive to either DAG or phorbol esters (10).
The activation mechanisms for the cPKCs and nPKCs have been well studied. Briefly, ligand binding to many receptor tyrosine kinases or G-protein-coupled receptors leads to phospholipase C activation and the consequent production of the DAG second messenger at the plasma membrane. The DAG binds to the C1 domains of cPKC and nPKC, inducing the translocation of PKC to the plasma membrane (3). The C1-DAG interaction and membrane association trigger a conformational change of the PKC molecule, leading to the release of the pseudosubstrate domain from the catalytic site and the final activation of the enzyme. The activation of the aPKCs also involves the release of the pseudosubstrate domain from the catalytic site of the kinase domain (11)(12)(13). However, the molecular mechanism of how their activation process is regulated is not clear. Likewise, little is known about whether the C1 domain plays any role in the activation of the aPKCs.
Among its multiple roles in various physiological processes, the important role of the PKC family in tumorigenesis has attracted much attention (14,15). Many recent studies have found that the atypical PKCs are essentially involved in tumorigenesis. For example, PKC has recently been found overexpressed in human non-small cell lung cancer cells (16,17) and ovarian cancer cells (18,19). It has also been proposed to be an attractive target to develop novel therapeutics against colon cancer (20) and chronic myelogenous leukemia (21). PKC has been reported recently to be associated with chemotaxis of human breast cancer cells (22). The atypical PKC is thus viewed as a new promising therapeutic target for cancer treatment.
An underlying problem with the inhibition of PKC kinase activity as a therapeutic strategy is achievement of sufficient selectivity among serine/threonine-specific protein kinases with homologous catalytic sites. A complementary strategy that we are pursuing has therefore been to design modulators targeted to the C1 domains (23). Using DAG derivatives in which the flexibility of the structure has been constrained to reduce the entropic loss due to binding, we have actually developed a novel DAG lactone derivative, 130C037, which displayed marked selectivity among the recombinant C1a and C1b domains of PKC␣ and -␦ as well as substantial selectivity for RasGRP (another family of the C1 domain-containing proteins) relative to PKC␣ (24). By combining the techniques of computational chemistry and molecular biology, we have also successfully modified the structure of some potent DAG lactones and obtained novel compounds (named dioxolanes), which gained an additional point of contact in their binding with the C1b domain of PKC␦. 4 These previous studies provide encouraging evidence that by manipulating the interactions between the C1 domains and the ligands, we might be able to design DAG derivative compounds that can specifically recognize different C1 domains of different PKC isoforms.
The atypical C1 domains are DAG/phorbol ester-insensitive. Although the sequence identity between the PKC␦ C1b domain and the atypical C1 domain of PKC is only 32%, we can feel confident that these domains adopt the same fold, because the zinc-binding residues (His 1 , Cys 14 , Cys 17 , Cys 31 , Cys 34 , His 39 , Cys 42 , and Cys 50 ) are conserved, as are critical structural residues Phe 3 , Gln 27 , and Val 38 . Many of the ligand binding residues are also conserved: Trp 22 , Gly 23 , and Leu 24 , along with a conservative substitution of isoleucine for leucine at position 21 (25). All of the PKC C1 domains carry a net positive charge, but an interesting characteristic of the atypical C1 domains is that they have an overall positive charge of ϩ9, which is more than twice the average charge of the conventional and novel C1 domains. This is mainly due to four extra arginine residues that line the edges of the binding site: Arg 7 , Arg 10 , Arg 11 , and Arg 20 . No charged residues are found at these positions in the conventional or novel C1 domains.
Do these arginine residues interfere with ligand binding? Can we take advantage of these positively charged arginine residues to design compounds that form specific contacts with them? In this study, using the C1b domain of PKC␦ as a template, we mutated the residues Asn 7 , Ser 10 , Pro 11 , and Leu 20 in the binding cleft into arginines corresponding to those in the atypical C1 domains. We wanted to study the roles of these arginine residues in phorbol ester binding both in vivo and in vitro, to broaden our general understanding of ligand-C1 domain interactions as well as to develop insights into the unique properties of the C1 domains of the atypical PKCs. We expect that with these atypical C1-like mutants, which still maintain some binding activity for the phorbol esters, we will be able to screen DAG analogue compounds that have been specifically designed for interaction with the arginine residues of the atypical PKC C1 domains.
Site-directed Mutagenesis of the C1b Domain of PKC␦-Sitedirected mutagenesis of the Asn 7 , Ser 10 , Pro 11 , and Leu 20 residues was performed in both the GST-␦C1b and GFP-␦C1b fusion proteins using the QuikChange II site-directed mutagenesis kit (Stratagene, La Jolla, CA), according to the manufacturer's instructions. The ␦C1b domain containing-pGEX plasmid and pEGFP plasmid, which had been constructed previously in this laboratory (24), were used as templates for the mutagenesis reactions. For generation of mutants at multiple sites, the mutagenesis was performed stepwise. The mutations were confirmed by DNA sequence analysis (DNA Minicore, Center for Cancer Research, NCI, National Institutes of Health).
Construction of the GFP-fused C1 Domains of PKC and PKC-The C1 domains of the atypical PKC and -were generated by PCR using the Platinum Pfx DNA polymerase (Invitrogen). The full-length cDNA clones of human PKC and -(obtained from OriGene (Rockville, MD)) were used as the templates. The blunt-ended PCR products were ligated into a pCR-Blunt vector using the Zero Blunt PCR cloning kit (Invitrogen). The pCR-Blunt vector was digested with EcoRI (New England BioLabs, Inc., Beverly, MA) to produce adhesive ends of the C1 fragments. These fragments were then ligated into the appropriate pEGFP vectors (Clontech) using the EcoRI restriction sites with the sequence of the insert in the intended reading frame. The DNA sequence of the constructs was confirmed by sequence analysis.
Site-directed Mutagenesis of the C1 Domains of PKC and PKC-Site-directed mutagenesis was employed to generate the back mutants of the atypical C1 domains of PKC and -. Briefly, the arginine residues at positions 7, 10, 11, and 20 were mutated back to the corresponding residues of Asn 7 , Ser 10 , Pro 11 , and Leu 20 , as in the binding cleft of the ␦C1b domain. The mutagenesis was performed in the GFP-tagged PKC and -C1 domains using the QuikChange II site-directed mutagenesis kit. For generation of the quadruple back mutants at multiple sites, the mutagenesis was performed stepwise. The mutations were confirmed by DNA sequence analysis.
Expression and Purification of GST Fusion Proteins from Escherichia coli-The recombinant plasmids containing the arginine mutants of the GST-␦C1b domains were expressed and purified from BL-21 E. coli as described elsewhere (24).
[ 3 H]PDBu Binding Assays-Scatchard analysis was performed in this study to determine the dissociation constant (K d values) of the arginine mutants in binding to [ 3 H]PDBu as described elsewhere (26). Competitive binding assays were also performed in this study to determine K d values of some mutants with weak potencies for PDBu binding as described elsewhere (27).
Expression and Imaging of the Fluorescent Protein-labeled C1 Domains in Live LNCaP and CHO Cells-LNCaP and CHO-K1 cells (obtained from ATCC, Manassas, VA) were cultured at 37°C in RPMI 1640 medium or F-12 medium containing 10% fetal bovine serum, penicillin (50 units/ml), and streptomycin (0.05 mg/ml) in a 5% CO 2 humidified atmosphere. The plasmids of GFP-fused C1 proteins were transfected into the cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. The expression of the fluorescent protein was detected 24 h after transfection. Confocal fluorescent images were collected with a Zeiss MRC 1024 confocal scan head (Zeiss) mounted on a Nikon microscope with a ϫ60 planapochromat lens as described before (24).
Molecular Modeling-Structures for the single mutants N7R, S10R, P11R, and L20R were generated from the crystal structure of the PKC␦ C1b domain (28) by mutating the selected residue to arginine while preserving the backbone and 1 angles. A conformational search was performed on the four angles of each new arginine residue using 5000 steps of systematic torsional sampling in Macromodel 9.1 (29). The search used the OPLS 2005 force field with implicit water solvent. Each arginine structure in the resulting set was energy-minimized to convergence at a gradient of 0.05, while fixing the rest of the protein structure in place. Duplicate structures, defined as those with less than a 0.5 Å difference in the arginine side chain atoms, were deleted. Structures for the double, triple, and quadruple mutants were built using the lowest energy conformer found for each single arginine side chain. A structure for a phorbol-binding quadruple mutant was built in the same way, with an alternate conformer of Arg 20 .
A homology model for the PKC C1 domain was built on the backbone coordinates of the crystal structure of the PKC␦ C1b domain (28). Side chains were constructed using the program SCWRL (30), which uses a backbone-dependent rotamer library to place residues in their most likely conformation given the backboneangles at that position. Residues homologous to ␦C1b were left unchanged from their crystallographic positions, and the binding site arginine residues at positions 7, 10, 11, and 20 were given the same lowest energy conformation found in the conformational search above. The model was energy-minimized with harmonic positional restraints on the backbone atoms to eliminate steric clashes in the side chains without inducing deformations in the backbone. The resulting structure was inspected carefully by hand to ensure that residue packing was reasonable and that the orientations of conserved but nonidentical charged residues were similar to the crystal structure of ␦C1b.
Electrostatic potentials were calculated for the wild-type ␦C1b, C1, and all of the arginine mutants in Grasp (31), using an interior dielectric of 2, an exterior dielectric of 80, a solvent radius of 1.4 Å, an ionic strength of 0.145 M, and an ionic radius of 2.0 Å. PARSE3 atomic charges and radii were used (32), modified slightly to include values for zinc (ϩ2 charge, 0.74 Å radius) and to modify the values for the eight zinc-coordinating residues, which were each given a partial charge of Ϫ0.5 to set the net charge for each zinc-binding motif to zero.

Site-directed Mutagenesis of PKC␦ C1b Domain to Generate Atypical PKC C1 Domain-like Arginine Mutants
Compared with the DAG/phorbol ester-responsive PKC C1 domains, one of the most striking characteristics of the atypical PKCs (PKC as well as PKC) is that there are several positively charged arginine residues located in the loops that make up the binding site. As shown in Fig. 1, the Arg residues that line the binding cleft in the atypical C1 domains of the human are Arg 7 , Arg 10 , Arg 11 , and Arg 20 , whose corresponding residues in the PKC␦ C1b domain are Asn 7 , Ser 10 , Pro 11 , and Leu 20 (see the three-dimensional structure in Fig. 1B). Arg 7 in the atypical C1 domain is polar in other C1 domains (Asn, Thr, and Tyr) and in fact the adjacent residue 6 is positively charged in many C1 domains. In the folded structure, Arg 7 is also close to Arg 26 , which is a conserved positively charged residue across all PKC C1 domains. Thus, this region of the structure is generally already positively charged. Arg 10 is at the apex of the first loop and points out away from the binding site. Some non-PKC C1 domains (protein kinase D C1b and RasGRP) also have positively charged Arg or Lys at position 10. We expected that this residue was probably the least likely to interfere with natural ligand binding. According to previous studies (25), Pro 11 and Leu 20 are partially relevant for binding. Mutating these two residues into glycine caused 100-and 15-fold reductions, respectively, in the in vitro binding affinity for PDBu. However, it was not known whether the introduction of Arg residues at these two positions, such as is found in the atypical C1 domains (Arg 11 and Arg 20 ), would cause any change in the ligand binding affinity.
To explore the role of these Arg residues experimentally, we used the potent phorbol ester-sensitive C1b domain of PKC␦ as the template, and we mutated the four residues Asn 7 , Ser 10 , Pro 11 , and Leu 20 in the binding cleft into Arg, generating all possible combinations of 1-4 mutated residues. The rationale for our approach was that the ␦C1b domain is biochemically well behaved and extensively characterized by us both structur-ally and with regard to its ligand interactions. The site-directed mutagenesis was carried out stepwise. The single-site mutations yielded N7R, S10R, P11R, and L20R; the double-site mutations provided N7R/S10R, N7R/P11R, N7R/L20R, S10R/P11R, S10R/L20R, and P11R/L20R; the triple-site mutations yielded N7R/S10R/P11R, N7R/S10R/L20R, N7R/P11R/L20R, and S10R/P11R/L20R; and finally the quadruple-site mutation yielded N7R/S10R/P11R/L20R. These mutants were fused with GST or GFP for use in either in vitro binding assays or in vivo translocation assays. With the GST-tagged mutants, we determined how the binding affinity for [ 3 H]PDBu in vitro was affected by the Arg residues. Then, with the GFP-tagged mutants, we evaluated the effect of the introduction of the Arg residues on the subcellular distribution of the C1 domain in living cultured mammalian cells in the absence of PMA and its response as a function of time upon the addition of PMA.

Using [ 3 H]PDBu Binding Assays to Characterize the Arginine Mutants in Vitro
The Single-site Mutations Showed No to Moderate  Reduction in the Binding Affinity for PDBu; the Dependence of Binding on PS Was Differentially and Significantly Increased-We first examined whether the binding affinity for [ 3 H]PDBu was affected by the introduction of an Arg residue along the rim of the binding cleft. The GST-tagged ␦C1b mutants were expressed and purified from E. coli. A Scatchard assay was employed to measure the binding affinities (represented by K d values) of these mutants for [ 3 H]PDBu. Under these in vitro assay conditions, binding is evaluated in the presence of 100 g/ml phosphatidylserine. Our results demonstrated that all the single mutants N7R, S10R, P11R, and L20R maintained considerably potent binding affinities for the phorbol ester PDBu ( Table 1).
As we expected, the mutation at Ser 10 (S10R) had no effect on ligand binding. It showed a K d value (0.32 Ϯ 0.10 nM) similar to that of the wild-type ␦C1b (0.33 Ϯ 0.05 nM). Strikingly, the P11R mutant retained potent binding affinity for [ 3 H]PDBu with a K d value of 1.19 Ϯ 0.19 nM. This K d was thus only 4-fold weaker than that for the wildtype (K d ϭ 0.33 Ϯ 0.05 nM). Pro 11 is highly conserved across all the typical DAG/phorbol ester-responsive C1 domains but not the atypical C1 domains. Previous studies had shown that mutation of the Pro 11 residue into glycine (P11G) reduced the binding affinity for PDBu more than 100-fold (25). However, we demonstrated here that replacing the proline with a positively charged Arg residue at this site did not significantly interfere with the phorbol ester interaction. This suggests that it may be the increased flexibility of glycine at this site that interferes with binding rather than a change in the overall structure of the loop. In fact, the backbone angles of this proline in the PKC␦ C1b crystal structure (28) are ϭ Ϫ58.6, ϭ 118.0, ϭ Ϫ176.7, which is a ␤-turn conformation that is compatible with any amino acid. Like Ser 10 , Asn 7 is located in a region of the binding site that can be positively charged. However, unlike Ser 10 , the replacement of Asn 7 with an Arg residue (N7R) slightly decreased the binding affinity (4-fold) for PDBu. The most significant change in PDBu binding occurred with the mutation at site 20. The introduction of an Arg residue to this site caused a 24-fold decrease in the binding affinity for PDBu. The K d was increased from 0.33 Ϯ 0.05 nM (for the wild type) to 8.0 Ϯ 1.8 nM (for the L20R mutant). In general, we conclude that none of the individual Arg residues along the binding cleft of the atypical C1 domain by itself can account for the loss of the potency of the atypical C1 domains for phorbol ester binding.
The binding of phorbol ester to the C1 domain represents a ternary interaction between ligand, C1 domain, and phospholipid. We therefore also examined the effect of the introduction of Arg residues along the rim of the binding cleft on the ability of phospholipid to support phorbol ester binding. We found that the dependence of binding on the proportion of PS in the phospholipids was significantly increased when either Asn 7 , Pro 11 , or Leu 20 was replaced with an Arg, which is positively charged. The total lipid concentration was fixed at 100 g/ml, and the ratio of PS/PC was varied. Results were normalized to the maximal level of binding at the optimal PS/PC ratio for each mutant. The wild-type ␦C1b displayed little dependence on PS for binding to PDBu ( Fig. 2A). In the absence of PS (with 100% PC in the reaction mixture), it could already bind PDBu to an extent of about 50% of the maximal binding (i.e. in the presence of 100% PS). S10R behaved similarly to the wild type ( Fig. 2C). In contrast, marked PS dependence was observed for the N7R (Fig. 2B), P11R (Fig. 2D), and L20R ( Fig. 2E) mutants. In the absence of PS, binding was barely detected for N7R and L20R and was only about 20% of the maximal binding for P11R. The proportions of PS supporting 50% of the maximal binding (EC 50 values) were 15% for N7R, 6% for P11R, and 24% for L20R. The change of the PS dependence on binding suggests that increasing the number of positive charges around the binding site may increase the requirement for negatively charged PS to facilitate the interaction with the membranebound ligand.
Combinations of the Mutations to Arg at the Four Residues Caused Different Degrees of Loss of the Binding Affinity for PDBu-Since single mutations with any of the four residues had only limited effects (K d values of 1-24-fold that of the wild type) on the binding affinity for PDBu in vitro, we next evaluated the consequences of introducing multiple Arg residues into the rim of the binding cleft. We prepared the double, triple, and quadruple mutants of the four residues. Like the single-site mutants, the multiple-site mutants were tagged with GST, purified from E. coli, and analyzed for [ 3 H]PDBu binding in vitro. The K d values are summarized in Table 1. It can be seen that mutation to Arg at any two sites

TABLE 1 Binding affinities for PDBu of ␦C1b and the arginine mutants
An in vitro binding assay was employed to measure the K d values. Each mean value was obtained from three independent experiments. The free energy of binding, ⌬G, the individual energy penalty for single arginine mutations, ⌬G i , and the joint contribution energy for multiple mutations, ⌬G ϩ , were calculated as described under "Results." NA, not applicable. . When Leu 20 was mutated together with Pro 11 , the K d was decreased to 374 Ϯ 20 nM, which was a 1100-fold loss of affinity compared with the wild type. At the other extreme, the mutation of Ser 10 together with the other sites did not change the binding affinity very much compared with the single mutants. This result is consistent with our finding for the single S10R mutant that mutation at this site had little effect on binding.

Receptor
The triple and quadruple mutants displayed a much more severe disruption of their ligand interaction. Again, triple mutants with the S10R mutation showed generally similar binding activity to those of the corresponding double mutants without S10R. Thus, the K d values of N7R/S10R/P11R, N7R/ S10R/L20R, and S10R/P11R/L20R were 0.6, 2.8, and 1.0 times those of their respective double-site mutants N7R/P11R, N7R/ L20R, and P11R/L20R. However, the triple mutant N7R/P11R/ L20R and the quadruple mutant N7R/S10R/P11R/L20R showed no detectable binding affinity for PDBu under our in vitro assay conditions.
To analyze the thermodynamic effects of multiple mutations on ligand binding of the C1 domains, it is helpful to convert the K d values into the free energies of binding using the equation ⌬G ϭ RT ln K d . These ⌬G values are shown in the second column of Table 1. By subtracting the free energy of binding for the wild-type ␦C1b domain from that of the single mutants, we can also calculate an individual energy penalty, ⌬G i , for each arginine mutation, ⌬G R i ϭ ⌬G R Ϫ ⌬G wt , shown in the third column of Table 1.
It is tempting to assume that the decrease in binding affinity of the double and triple mutants ought to be equal to the sum of the individual energy penalties of each mutation, but this is not theoretically correct (33). The binding of PDBu to a double or triple mutant can be separated into three parts: ⌬G wt , ⌺(⌬G i ) for each mutation, and a joint contribution energy ⌬G ϩ that arises because the mutations are located together on the same molecule (34). For example, an equation for the free energy of binding of the N7R/P11R mutant can be written as follows: The ⌬G ϩ term accounts for the (favorable) difference in entropy of binding one molecule with two mutations versus two molecules with one mutation each, as well as any (unfavorable) extra conformational strain or desolvation energy that occurs from binding the second mutation in the presence of the first that is not seen with the second mutation alone (33). Values of ⌬G ϩ for each multiple mutant are shown in the fourth column of Table 1.
The joint contribution energy for multiple arginine mutations ranges from essentially zero, suggesting that entropy effects and any strain/desolvation penalties are canceling each other out, to more than ϩ2 kcal/mol. The mutation at position 10 is particularly interesting, because, as discussed above, the penalty for adding S10R as a second mutation is generally less than that of the others; N7R/S10R, S10R/P11R, and S10R/L20R have the smallest ⌬G ϩ values. However, adding S10R as a third mutation can apparently act either to reduce or increase the joint contribution of the other two mutations. ⌬G ϩ for N7R/ P11R is 1.23 kcal/mol, which is larger than the value of 0.92 kcal/mol for N7R/S10R/P11R, whereas the ⌬G ϩ value of 1.67 kcal/mol for N7R/L20R is smaller than the 2.29 kcal/mol value for the N7R/S10R/L20R mutant. This suggests that the relative locations of the mutations on the binding site loops have an important effect on interactions with PDBu and phosphatidylserine, and their contributions to the free energy of binding should not be seen as independent.

Modeling and Analysis of the Arginine Mutants in Silico
Molecular Modeling of the Single-site Mutations Showed That They Can Fully or Partially Block the Binding Site-To model the effects of the arginine mutations on the structure of the C1 domain, we began with the crystal structure of the C1b domain of PKC␦ (28). We first built four single mutant structures: N7R, S10R, P11R, and L20R. Arginine residues are quite flexible, with four side chain angles. To explore the conformational space available to each mutated arginine residue and to find the lowest energy rotamer for each residue, we performed a conformational search for each mutant, varying the angles of the arginine while keeping the rest of the protein structure frozen.
The structures of the lowest energy rotamer for each mutant, along with several other representative low energy conformations, are shown in Fig. 3. The L20R mutant (Fig. 3E) adopts a conformation in which it is buried in the binding site and forms a hydrogen bond to Leu 21 . This would completely block the interaction of phorbol with the binding site (Fig. 3A). The lowest energy conformations of the N7R and P11R mutants do not block the binding site, but each of these mutants has at least one rotamer that does. An arginine at position 7 (Fig. 3B) can partially block the back end of the binding site by forming a hydrogen bond to the backbone of Tyr 8 . The P11R (Fig. 3D) mutant can adopt a conformation in which the arginine is tilted over the top of the binding site, blocking ligand access, or it can "snorkel" down into the binding site to form a hydrogen bond with Leu 21 . In contrast, none of the low energy conformations of the S10R mutant (Fig. 3C) will block the binding site or interfere with phorbol binding.
These structures provide an explanation for both the binding affinities of the single-site mutants and their phosphatidylserine dependence. The occupancy of a rotamer, or the amount of time that a residue spends in a given conformation, will be proportional to the energy of that conformation. Thus, for L20R, where the lowest energy conformation of Arg 20 is buried in the binding site, the binding site will be blocked a large part of the time, and phorbol must compete with the arginine residue for access. This is why L20R has the weakest binding affinity for PDBu among all of the mutants. The arginine residues in the N7R and P11R mutants will not block the binding site most of the time, but the binding site will be at least partially blocked some of the time, reducing the binding affinity for the phorbol ester. The binding site of the S10R mutant is unaffected by the presence of the arginine residue, and S10R is also not as strongly affected by membrane composition as are the other mutants. This may be because in structures where the binding site is blocked or partially blocked by an arginine, interactions with charged lipid molecules are required to bias the average conformation of the arginine residue away from the binding site.
Using the lowest energy conformation for each residue, we built structures for each of the double, triple, and quadruple mutant C1 domains. The mutated arginine residues do not clash with one another or come close enough for their positively charged nitrogen atoms to experience a strong repulsion. Replacing the lowest energy rotamer of Arg 20 , which blocks the binding site, with a different rotamer, we can insert phorbol ester back into the binding site (Fig. 3F). The phorbol ester can easily be accommodated even in the quadruple mutant, and it does not form any strong favorable or unfavorable contacts with the arginines. Once the phorbol ester has successfully competed with the arginine residues for entrance into the binding site, the arginine residues do not need to adopt unfavorable strained conformations. Therefore, the binding affinity data and the joint contribution energy penalties for the multiple mutants cannot be explained by conformational strain or by repulsions between arginines distorting the binding site.
Since the drop in binding affinity with multiple arginine substitutions is not due to problems within the C1 domain or between the C1 domain and PDBu, this suggests that the joint contribution energy penalty of multiple mutations arises from the interaction between the C1 domain and the lipid membrane. We can imagine a model for C1 domain binding to membrane-bound PDBu in which, as the C1 domain approaches the membrane, any arginine residues around the binding site must find favorable interactions with the phospholipid headgroups in order to break their interactions with the binding site and open it for phorbol binding. At the same time, however, the C1 domain must adopt the correct orientation and penetration depth relative to the membrane interface to locate and productively bind PDBu. Furthermore, the arginine mutations will have altered the charge distribution around the C1 domain as a whole, which will affect its electrostatic interactions with the negatively charged membrane. This juggling act must become increasingly difficult as additional arginine residues are added around the binding site.
Visualization of the Electrostatic Potential around Mutated C1 Domain Structures-To visualize how mutating residues around the binding cleft to arginine changes the overall electrostatic profile of the ␦C1b domain, we calculated the electric field around each mutant structure by solving the Poisson-Boltzmann equation as implemented in the program Grasp (31). The Poisson-Boltzmann equation describes the electrical field surrounding a collection of fixed point charges in a continuum ionic solvent. A protein molecule is modeled as a set of atomic partial charges, each with Cartesian coordinates and a radius. The radii define a surface for the molecule, and the dielectric constant of the solvent outside the surface is set to 80, for highly polarizable water, whereas the dielectric inside the surface is set to 2, because dipoles inside the protein are generally fixed in place by hydrogen bonding. The electrostatic potential around the protein can be visualized by drawing a three-dimensional contour through the field at a constant value of electrostatic potential energy. This is shown in Fig. 4, looking down on the binding site from above, with contouring at ϩ1 kT/e drawn in blue mesh and contouring at Ϫ1 kT/e in red.
As discussed earlier, the C1 domains all carry a net positive charge. In the case of ␦C1b (Fig. 4a), this charge is concentrated around the right and back side of the domain (relative to the orientation shown in Fig. 1B). As would be expected, the charge distribution in the single mutants (Fig. 4, b-e) is fairly similar to that in the wild-type ␦C1b domain (Fig. 4a). The N7R mutant adds positive potential to the back of the C1 domain (Fig. 4b), the S10R mutant (Fig. 4c) adds to the top and left-hand side, and the P11R (Fig. 4d) and L20R (Fig. 4e) mutants add to the front and top of the domain. These differences in the distribution of the electrostatic potential for the single mutants mean that each set of double and triple mutants also has a differently shaped and oriented electrostatic field (Fig. 4, f-o). The extent of the positive potential at ϩ1 kT/e becomes larger as positively charged arginine residues are added one by one around the binding site. With the quadruple mutant (Fig. 4p) and C1 (Fig.  4q), the positive potential at ϩ1 kT/e extends completely over the top of the binding site.
It is believed that the C1 domains bind to membranes in two stages: first a nonspecific electrostatic attraction between the positively charged protein and the negatively charged membrane, followed by the insertion of the binding loops into the interfacial region of the membrane to bind DAG or phorbol esters (35). The shape of the electrostatic potential around the binding site will affect the orientation of the C1 domain relative to the plane of the membrane as it binds nonspecifically and thus perhaps the extent of the conformational change required for interfacial insertion and phorbol ester binding. The size of the electrostatic potential, or the amount of positive charge around the binding site, will affect the kinetics of membrane association as well as the relative energies of the nonspecifically bound surface state versus the phorbol ester-bound inserted state. Since the shape and size of the electrostatic potential around each mutated C1 domain is different, we expected to see strong differences between the mutants in subcellular distribution and in membrane translocation in response to phorbol ester. The extent to which the electrostatic potential is qualitatively more like ␦C1b or more like C1 or somewhere in between echoes the cellular behavior of the mutants, as will be discussed below.

Characterization of the Arginine Mutants in Vivo
The Distribution Pattern in LNCaP Cells of the GFP-tagged Arginine Mutants as Compared with That of the Wild-type GFP-␦C1b-The in vitro binding assays described above demonstrated that the arginine mutants displayed decreases in binding affinity for PDBu when one, two, three, or four Arg residues were introduced to replace the Asn 7 , Ser 10 , Pro 11 , and Leu 20 residues on the rim of the binding cleft of ␦C1b. How, then, would the introduction of the Arg residues affect the behavior of the ␦C1b domain in the intracellular environment? In order to explore the in vivo characteristics of the mutated C1 domains, we cloned the genes of the Arg mutants into the pEGFP vector (Clontech) that contains the gene of GFP. The . Electrostatic isopotential contours around wild-type ␦C1b, the arginine mutants, and a homology model of the PKC C1 domain. The orientation is rotated forward 90°about the x-axis relative to Fig. 1B, looking down on the binding site from above. a, wild-type ␦C1b; b-e, single mutants (N7R, S10R, P11R, and L20R); f-k, double mutants (N7R/S10R, N7R/P11R, N7R/L20R, S10R/P11R, S10R/L20R, and P11R/L20R); l-o, triple mutants (N7R/S10R/P11R, N7R/S10R/L20R, N7R/P11R/L20R, and S10R/P11R/L20R); p, quadruple mutant (N7R/S10R/P11R/L20R); q, homology model of C1. Contours are drawn at ϩ1 kT/e (blue) and Ϫ1 kT/e (red). The protein backbone is shown in white.
fusion proteins were expressed in LNCaP cells. The distributions of the GFP-tagged Arg mutants were then observed using a confocal microscope, and a typical image of the subcellular distribution pattern of each mutant is displayed in Fig. 5.
In LNCaP cells, the isolated C1b domain of PKC␦ (wild type) was almost uniformly distributed in the cytosol with a small portion associated with the internal membranes (Fig. 5a). Interestingly, when Asn 7 or Ser 10 was replaced with an Arg, plasma membrane association started to appear (Fig. 5, b and c). The plasma membrane association became stronger when both of the Asn 7 and Ser 10 sites were replaced with Arg residues, as demonstrated in the double mutant of N7R/S10R (Fig. 5f). These two sites are located toward the back left side of the binding cleft (relative to the orientation shown in Fig. 1B). The increase in positive charge in that region of the structure may facilitate interaction with negatively charged lipids on the inner layer of the plasma membrane. Pro 11 and Leu 20 are located toward the front right side of the binding cavity. An Arg residue there did not seem to cause membrane association but did lead to some distribution to the nucleus in addition to the cytoplasm (Fig. 5, d and e). A nuclear distribution was also seen to some extent with the N7R mutant.
Another interesting change of the distribution pattern was that with more Arg residues in the motif at positions P11R, L20R, and N7R, nuclear accumulation was enhanced. The nuclear accumulation of the protein started to appear in some double mutants (Fig. 5, g-k). It became quite prominent in the triple mutants (Fig. 5, l-o) and the quadruple mutant (Fig. 5p). In the triple mutant N7R/S10R/L20R and the quadruple mutant N7R/S10R/ P11R/L20R, both nuclear accumulation and plasma membrane association could be seen (Fig. 5, m and  p), reflecting the effects of P11R, L20R, and N7R enhancing nuclear association and the effects of N7R and S10R enhancing plasma membrane localization. It has been reported that positively charged residues are abundant in nuclear localization signals in general, since some of these positive residues bind to importins, a family of proteins for nuclear translocation (36).
The Plasma Membrane Translocation Dynamics of the GFP-tagged Arginine Mutants as Compared with That of the Wild-type GFP-␦C1b in Response to PMA in Living LNCaP Cells-To examine the activity of these mutants for phorbol esters in vivo, we first observed the response of the single mutants (N7R, S20R, P11R, and L20R) to PMA using real time confocal microscopy. The GFP-tagged mutants were transfected into LNCaP cells. The cells were then incubated at 37°C to express the fluorescent proteins for 24 h. Then the cells were exposed to 1 M PMA, and the subcellular redistribution of these mutants was recorded as a function of time after PMA application. The images were taken every 15 or 30 s. The wild-type ␦C1b showed a very quick response to the 1 M PMA. It translocated to the plasma membrane within 5 min after the drug application (Fig. 6A). However, the plasma membrane translocation was transient, and the fluorescent signal began to return to the cytosol within 10 -20 min (Fig. 6A). This redistribution of the ␦C1b may reflect kinetics of uptake of the PMA into the cell, where it first is taken up by the plasma membrane and subsequently equilibrates with the internal membranes (37).
Similar redistribution dynamics were seen in the four single mutants N7R, S10R, P11R, and L20R. As shown in Fig. 6, B-E, all the single mutants translocated quickly to the plasma membrane within 5 min after the application of the 1 M PMA. With the exception of N7R, membrane translocation appeared to be b-e, single mutants (N7R, S10R, P11R, and L20R); f-k, double mutants (N7R/S10R, N7R/P11R, N7R/L20R, S10R/P11R, S10R/L20R, and P11R/L20R); l-o, triple mutants (N7R/ S10R/P11R, N7R/S10R/L20R, N7R/P11R/L20R, and S10R/P11R/L20R); p, quadruple mutant (N7R/S10R/P11R/ L20R). Each image represents a typical example from at least three independent experiments. slightly faster for the single mutants than for the wild type, as we expected due to the increase in positive charge. Similar to the behavior seen with the wild-type ␦C1b, after the initial plasma membrane translocation, the mutants likewise partially relocated back to internal locations (Fig. 6, compare B-E with A). For the mutants for which the C1 domain is initially present in part in the nucleus (N7R, P11R, and L20R), the C1 domain translocated out of the interior of the nucleus to the nuclear or plasma membranes, but the response of the protein in the nucleus was slower than that in the cytosol. Consistent with our previous binding assays, our living cell imaging experiments demonstrated that the introduction of a single positively charged Arg residue at any of the four sites Asn 7 , Ser 10 , Pro 11 , and Leu 20 did not significantly affect the binding activity for the phorbol ester in vivo, but some differences were seen in the dynamics of translocation.
Differences in response to phorbol ester were magnified in vitro when multiple Arg residues were introduced into the binding cleft (Table 1). Similarly, marked differences were seen in the response to PMA of the multiple mutants in the living cells. We first looked at the double mutants. Like the single mutants, the redistribution of the GFP-tagged double mutants (N7R/S10R, N7R/P11R, N7R/L20R, S10R/P11R, S10R/L20R, and P11R/L20R) was monitored after the application of 1 M PMA. As shown in Fig. 7A, the N7R/S10R, N7R/P11R, S10R/ P11R, and S10R/L20R mutants translocated to the plasma membrane very quickly (within 10 min) after the application of 1 M PMA. However, N7R/L20R and P11R/L20R showed a much slower and weaker plasma membrane translocation under these conditions (Fig. 7B). These double mutants also showed a significant accumulation of protein in the nucleus. From Table 1, it is clear that the N7R/L20R and P11R/L20R were much less potent (about 50 -400-fold lower) for binding to the phorbol ester PDBu in vitro compared with the other four double mutants. Thus, our results from the in vivo translocation assay were consistent with those of the in vitro data.
Similar results were seen with the triple mutants. As shown in Fig. 8A, N7R/S10R/P11R displayed a very quick and clear The images were captured every 30 s. A, time series images of N7R/S10R, N7R/P11R, S10R/P11R, and S10R/L20R after the addition of PMA. These four double mutants translocated to the plasma membrane quickly in response to 1 M PMA; B, time series images of N7R/L20R and P11R/L20R after the addition of PMA. These two double mutants showed a much slower and weaker plasma membrane translocation under the same experimental conditions. Each panel represents a typical example of three independent experiments. plasma membrane translocation in response to the 1 M PMA application. This mutant showed the best binding affinity for PDBu in vitro among the triple mutants, better even than the N7R/P11R double mutant. N7R/S10R/L20R and S10R/P11R/ L20R also showed some plasma membrane translocation, with somewhat more response for the N7R/S10R/L20R mutant (Fig. 8A). However, compared with N7R/S10R/P11R, their responses were less potent. These results fit with the lower binding affinities of the two mutants for PDBu as well as the trend toward greater membrane association for N7R and S10R. We were unable to detect any response of the triple mutant N7R/P11R/L20R and the quadruple mutant N7R/S10R/P11R/ L20R to the application of 1 M PMA (data not shown). However, when we increased the concentration of PMA to 10 M, a very weak response was displayed in these two mutants, as shown in Fig. 8B. We could see the decrease of the fluorescent intensity in the nucleus of the cells with N7R/P11R/L20R (Fig.  8B, upper panel) and the appearance of a weak signal at the plasma membrane in the cells with N7R/S10R/P11R/L20R (Fig.  8B, lower panel). In the previous in vitro binding assays, we were unable to detect any binding activity of either N7R/P11R/L20R or N7R/S10R/P11R/L20R ( Table 1). The in vivo results emphasize that the recognition of phorbol ester by the two mutants is attenuated rather than lost entirely. All of the triple mutants and the quadruple mutant were highly concentrated in the nucleus, especially those containing both P11R and L20R, suggesting again that a positive charge in this region of the structure may contribute to a nuclear distribution. The N7R/S10R/ P11R and N7R/S10R/L20R mutants have somewhat higher basal cytosolic concentrations, and this initial distribution Cells expressing the triple and quadruple mutants were treated with 1 or 10 M PMA. The redistribution of the fluorescent proteins was monitored with a Zeiss MRC 1024 confocal microscope as a function of time after the addition of PMA. The images were captured every 30 s. A, time series images of N7R/S10R/P11R, S10R/P11R/L20R, and N7R/S10R/L20R in response to 1 M PMA. These three triple mutants showed a response to 1 M PMA. However, the triple mutant N7R/P11R/L20R and the quadruple mutant N7R/S10R/P11R/L20R did not respond to 1 M PMA (data not shown). B, time series images of the triple mutant N7R/P11R/L20R and the quadruple mutant N7R/S10R/P11R/L20R N7R/L20R in response to 10 M PMA. These two mutants showed a weak response to the high dose of PMA. Each panel represents a typical example of three independent experiments. appeared to permit an initial membrane response, followed by a secondary membrane response, which was sustained for at least 20 min as more protein made its way out of the nucleus. The translocation of the N7R/P11R/L20R and S10R/P11R/L20R mutants appeared much slower, since there was little to no cytosolic protein to give an initial response.

The Isolated C1 Domains of the Atypical PKC and -Showed Similar Characteristics in Vivo to Those of the Quadruple Arginine Mutant
In the above studies, we have demonstrated that when four Arg residues were introduced into the binding cleft of the C1b domain of PKC␦ corresponding to the sites in the atypical C1 domains, the potency of this mutated ␦C1b domain (i.e. the quadruple mutant) for ligand interaction was completely abolished in vitro or dramatically reduced in vivo. Does the quadruple mutant reflect the characteristics of the real atypical C1 domains? To compare the multiple mutants with the real atypical C1 domains, we further investigated the characteristics of the isolated C1 domains of the atypical PKC and -expressed in cultured mammalian cells. The sequences encoding the 50 amino acids of the C1 domains of PKC and -were obtained from those of the full-length proteins and fused with the gene for the fluorescent protein enhanced GFP. The constructs were then expressed in LNCaP cells. The subcellular distribution of the fluorescent proteins was examined using a confocal microscope. The characteristics of the distribution patterns of both C1 and C1 turned out to be very similar to those of the quadruple mutant of the ␦C1b domain. We observed three patterns of subcellular distributions of the isolated atypical C1 domains, as illustrated in Fig. 9: 1) the C1 domains were exclusively localized in the nucleus (as shown in Fig. 9, A (a) and B (a)) and associated with the nucleoli; 2) plasma membrane association of the C1 domains was evident, although much of the protein was still localized in the nucleus (Fig. 9, A (b) and B (b)); 3) the nuclear accumulation of the fluorescent proteins was not as strong but the plasma membrane association was clear (Fig. 9, A (c) and B (c)). In terms of the distribution patterns, C1 and C1 showed some differences. Over 99% of the cells expressing C1 had a distribution of exclusively nuclear localization, like pattern 1 in Fig. 9A (a). Only very few cells showed the patterns of plasma membrane association as in Fig. 9, A (b and c). However, for C1, many more cells (ϳ40%) displayed plasma membrane association, as in Fig. 9B, b and c. Interestingly, these three patterns of subcellular distribution were seen in our quadruple mutant of ␦C1b (N7R/S10R/P11R/ L20R). As we have shown in Fig. 5p, the quadruple mutant displayed a significant nuclear accumulation and plasma membrane association of the protein, more like that of the C1 domain than that of C1. It had about 40% of the cells displaying plasma membrane association plus nuclear accumulation of the proteins (Fig. 9C, b and c), whereas about 60% of the cells displayed exclusively nuclear localization ( Fig. 9C (a)). Nucleolar association was also seen in the quadruple mutant. Similar subcellular distribution patterns were also seen in another cell line, CHO cells (data not shown).
It is generally believed that the atypical PKCs are not DAG/ phorbol ester-sensitive. Since the quadruple arginine mutant showed some weak response to the phorbol ester PMA at high dose (10 M), we also treated the GFP-C1-and C1-expressing LNCaP and CHO cells with 10 M PMA. Interestingly, we observed a very weak response of the GFP-C1 in CHO cells (but not LNCaP cells), reflected by a small and slow change of the fluorescence intensity ratio between the nucleus and the cytosol (Fig. 10A). However, unlike the quadruple Arg mutant (Fig. 8B), the C1 domain was unable to translocate to the plasma membrane in response to the high dose of PMA (Fig.  10A). Unlike GFP-C1, GFP-C1 did not show a discernable response to the high dose of PMA (10 M) in either CHO cells or LNCaP cells (Fig. 11A). We therefore conclude that, like the quadruple mutants, the atypical C1 domains may have maintained the basic structures for ligand binding. However, because of the presence of the positively charged Arg residues along the binding cleft and other factors, their structure may no longer effectively support binding with the traditional ligands for typical PKCs, at least over the range of ligand concentrations examined.

The Activity of the Atypical PKC and PKC C1 Domains to Phorbol Ester Was Clearly Recovered When the Four Arginine Residues in the Binding Cleft Were Mutated Back to the Residues Corresponding to Those in the ␦C1b Domain
To further elucidate the role of the four positively charged Arg residues in affecting ligand interaction and membrane translocation of the atypical C1 domains, we mutated the four Arg residues at positions 7, 10, 11, and 20 back to the corresponding residues (Asn, Ser, Pro, and Leu, respectively) in the binding cleft of the ␦C1b domain. The back mutants (R7N/ R10S/R11P/R20L) of the C1 and C1 domains were fused with GFP. Their ability to respond in vivo to the phorbol ester PMA by translocation was then evaluated. The back mutants showed a potent response to PMA (Figs. 10, B-D, and 11, B and C) as compared with the wild-type GFP-C1 (Fig. 10A) and GFP-C1 (Fig. 11A). The back mutant of GFP-C1 translocated immediately (within 5 min) to the plasma membrane in response to 3 and 10 M PMA (Fig. 10, C and D). The lowest concentration of PMA that induced plasma membrane translocation of GFP-C1 within 30 min was 1 M (Fig. 10B). Compared with GFP-␦C1b, for which 10 nM PMA was the lowest dose that caused translocation in the CHO cells (data not shown), the back mutant of the GFP-C1 was thus ϳ100-fold less potent. The back mutant of the GFP-C1 was a little less sensitive compared with the back mutant of GFP-C1. It displayed clear plasma membrane translocation to 10 M PMA within 10 min (Fig.  11C), and the lowest concentration of PMA that caused translocation in 30 min was 3 M (Fig. 11B). Unfortunately, because the back mutants of both atypical C1 domains were apparently unstable in vitro, we were unable to determine their binding affinities (K d values) using [ 3 H]PDBu binding assays in the present study.
Other differences were also observed for the back mutants in comparison with the wild-type C1 and C1 domains. Although the back mutants of both C1 and C1 domains still accumulated in the nucleus, a larger portion of the protein was present in the cytosol compared with the wild-type C1 and C1 domains (first column in Figs. 10 and 11). Second, no initial plasma membrane association was evident in either back mutant, as compared with the wild-type C1 and C1 domains (Fig. 9, A and B). Last, the shape of the nucleus occasionally became irregular or fragmented in the back mutants of the atypical C1 domains (Figs. 10D and Fig. 11, B and C).

DISCUSSION
The PKC superfamily members, the typical PKCs as well as atypical PKCs, have gained great attention as essential players in the control of cell proliferation, differentiation, and apoptosis (3). As the major receptor for DAG and the tumor-promoting phorbol esters, the typical PKCs (the conventional PKCs and novel PKCs) have already become attractive therapeutic targets for cancer and a range of other conditions (14,15). Different strategies have been used to develop PKC modulators, such as using inhibitors to modulate the PKC kinase activity, using antisense oligonucleotides to modulate the PKC protein level, and using compounds that act through the C1 domains (38 -40).
Among these strategies, natural products targeted to the C1 domains are prominently represented in clinical trials. Bryostatin 1, which has a unique biphasic dose response for downregulation and protection of PKC␦ (41,42), is now in clinical trials as a cancer chemotherapeutic agent (14). Ingenol 3-angelate (PEP005), which induces biphasic induction of interleukin-6 through the activation of PKC (43), is in clinical trials for actinic keratosis and nonmelanotic skin cancer. Prostratin, which functions as an inhibitor of phorbol ester tumor promotion despite being an activator of PKC (44), is being evaluated for combination therapy in HIV/AIDS in order to eliminate the latent state of HIV. PMA itself is in clinical trials based on its ability to induce differentiation of leukemia cells (45,46). With the exception of PMA, these PKC activators have been selected based on their unique patterns of biological activity, although the basis for their biological selectivity remains uncertain.
As a strategy for developing ligands interacting with the C1 domains of PKC (and related members of the superfamily), the Marquez group (23) has exploited the structure of the endogenous ligand DAG. Through conformational constraint using an optimized DAG lactone together with appropriate patterns of substitution, this approach has yielded DAG-lactones with in vitro potencies approaching those of the phorbol esters (23). To further advance the development of selective ligands targeted to the C1 domains, we are trying to dissect the structural factors influencing ligand recognition and C1 domain function. As one component of this general effort, we have explored here one aspect of the structural differences that set apart the C1 domains of the atypical PKCs from those of the phorbol ester and DAG-responsive classical and novel PKCs.
A critical conclusion from our studies is that the C1 domains of the atypical PKCs retain the appropriate structure for ligand binding, but with the binding cleft predominantly blocked by occupancy by the arginine residues along its rim. These findings suggest that ligands that can both interact with the binding cleft and specifically interact with these arginine residues may be selective high affinity ligands for the atypical PKCs. Based on these findings, we are now attempting to develop compounds and indeed have preliminary compounds (based on the DAG-lactone template) that showed modest selectivity for the arginine mutants of the ␦C1b domain versus the wild type. 4 These preliminary compounds are now being used as templates to improve both potency and selectivity for the atypical C1 domains.
Our emerging understanding of the function of the C1 domains of the atypical PKCs argues that these domains are highly weakened but still potentially responsive receptors for potent ligands, such as the phorbol esters. We indeed observed a variable, weak translocation response for the GFP-C1 domain in the CHO cells in the presence of 10 M PMA. Others have likewise reported translocation of atypical PKCs at high phorbol ester concentrations (47). Although interpretation of such findings has been unclear, because it could reflect either perturbation of the membrane by the phorbol ester at high concentration or an indirect effect of the phorbol ester treatment consequent to PKC activation, the current results suggest that such responses may indeed be real.
Our results argue that the classification of C1 domains as DAG/phorbol ester-responsive or unresponsive is simplistic. Rather, the "unresponsive" C1 domains should be further distinguished based on whether the underlying binding cleft is retained, as in the atypical PKCs, or is grossly distorted, as is the case with C1 domains such as that of Raf (6,48).
As with the classical and novel PKCs, the isolated C1 domains of the atypical PKCs behave somewhat differently from the full-length PKC proteins. The atypical C1 domains, C1 as well as C1, tend to accumulate in the nucleus and associate with the nucleoli. In addition, in some cells, they also tend to associate with the plasma membrane. Studies have shown that the full-length atypical PKC proteins are diffusely distributed in the cytosol (49). Both plasma membrane translocation (50) and nuclear translocation (51) of the atypical PKCs have been reported in the literature. According to our imaging results, the isolated atypical C1 domains have already been localized to the plasma membrane and the nucleus without any inducer. Therefore, it is possible that the C1 domain itself possesses the ability to translocate the full-length atypical PKC to either the nucleus or the plasma membrane. Perander et al. (49) have reported that the hexapeptide KRFNRR located in the N terminus of the atypical C1 domain (amino acids 6 -11) acts as the core of the nuclear localization signal for translocating the protein into the nucleus. They also found that intramolecular interactions between the N-terminal pseudosubstrate domain and the C-terminal catalytic domain inhibit nuclear translocation of PKC (49). Therefore, our results are in agreement with their conclusions. As an isolated fragment, the C1 domain is exposed to the intracellular environment without the cover from the other parts of the intact protein. The Arg residues, especially Arg 10 and Arg 11 at the tip of the C1 domain, can easily drive the protein to the nucleus. It is also possible that the positive charges at the surface would drive the protein to the plasma membrane through interaction with negatively charged phospholipids like PS.