A Novel Diacylglycerol-lactone Shows Marked Selectivity in Vitro among C1 Domains of Protein Kinase C (PKC) Isoforms α and δ as Well as Selectivity for RasGRP Compared with PKCα*

Although multiple natural products are potent ligands for the diacylglycerol binding C1 domain of protein kinase C (PKC), RasGRP, and related targets, the high conservation of C1 domains has impeded the development of selective ligands. We characterized here a diacylglycerol-lactone, 130C037, emerging from a combinatorial chemical synthetic strategy, which showed substantial selectivity. 130C037 gave shallow binding curves for PKC isoforms α, β, γ, δ, and ϵ, with apparent Ki values ranging from 340 nm for PKCα to 29 nm for PKCϵ. When binding to isolated C1 domains of PKCα and -δ, 130C037 showed good affinity (Ki = 1.78 nm) only for δC1b, whereas phorbol 12,13-dibutyrate showed affinities within 10-fold for all. In LNCaP cells, 130C037 likewise selectively induced membrane translocation of δC1b. 130C037 bound intact RasGRP1 and RasGRP3 with Ki values of 3.5 and 3.8 nm, respectively, reflecting 8- and 90-fold selectivity relative to PKCϵ and PKCα. By Western blot of Chinese hamster ovary cells, 130C037 selectively induced loss from the cytosol of RasGRP3 (ED50 = 286 nm), partial reduction of PKCϵ (ED50 > 10 μm), and no effect on PKCα. As determined by confocal microscopy in LNCaP cells, 130C037 caused rapid translocation of RasGRP3, limited slow translocation of PKCϵ, and no translocation of PKCα. Finally, 130C037 induced Erk phosphorylation in HEK-293 cells ectopically expressing RasGRP3 but not in control cells, whereas phorbol ester induced phosphorylation in both. The properties of 130C037 provide strong proof of principle for the feasibility of developing ligands with selectivity among C1 domain-containing therapeutic targets.

following the activation of receptor-coupled phospholipase C or indirectly from phosphatidylcholine via phospholipase D (1). Most but not all effects of DAG reflect its interaction with proteins containing C1 domains, resulting in their activation and/or driving their membrane translocation. Reflecting the importance and diversity of its downstream effectors, DAG is involved in signal transduction of numerous physiological and pathological processes, including proliferation, differentiation, apoptosis, angiogenesis, and drug resistance (2). These functions have focused attention on C1 domain-containing proteins as molecular targets for cancer chemotherapy (3).
The interaction between DAG and its receptors is typically mediated by a DAG-responsive motif called a "C1 domain" (4). The highly conserved C1 domain (ϳ50 amino acids) is a cysteine-rich zinc finger structure (5) that was first identified in protein kinase C (PKC) as the interaction site for DAG and the phorbol esters (6). The PKC family of serine/threonine protein kinases comprises the best studied mediators of DAG signaling. 8 of its 11 family members have DAG-responsive C1 domains: (i) the conventional PKCs (␣, ␤I, ␤II, and ␥) and (ii) the novel PKCs (␦, ⑀, , and ). Both the classic and novel PKCs contain a C-terminal kinase domain and an N-terminal regulatory domain. The regulatory domain contains a pseudosubstrate domain, which occupies the catalytic site of the kinase domain and inhibits the kinase activity. Binding of DAG to the C1 domain completes a hydrophobic surface on the C1 domain, favoring its interaction with the membrane. The twin consequences are membrane translocation, controlling access to substrates, and a conformational change in PKC, removing the pseudosubstrate domain from the catalytic site, thereby activating the enzyme (7).
In addition to the PKC family, five other families of proteins (PKDs, RasGRPs, chimaerins, Munc13s, and DGKs) have been recognized with C1 domains responsive to DAG (7). The protein kinase D (PKD/PKC) family represents kinases superficially similar to PKC; however, the kinase domains are not homologous and show different selectivity. The PKDs lack the pseudosubstrate domain, show different spacing between their C1 domains, and contain a membrane-interacting pleckstrin homology domain (8). PKD is activated upon phosphorylation by PKC, with its C1 domains driving membrane localization. The Ras guanyl nucleotide-releasing protein family members (RasGRP1 to 4) function as guanine nucleotide exchange factors for Ras or Rap, leading to their activation (9, 10), as well as being subjected to PKC phosphorylation (11)(12)(13). The chimae-rins are GTPase-activating proteins for Rac, leading to Rac inhibition. The Munc13 proteins are involved in the priming of vesicle fusion, and finally the DAG kinases function to abrogate DAG signaling, thus providing a negative feedback regulatory loop for the DAG signaling pathway (7).
Not only may DAG receptor families have complementary or antagonistic functions, but the same may be true within families. For example, PKC␦ is growth-inhibitory in NIH3T3 cells, whereas PKC␣ and PKC⑀ are growth-stimulatory (2,3). Thus, complementary therapeutic strategies are to inhibit a specific PKC isoform or to stimulate an antagonistic isoform. For this latter approach, activators selective for different DAG receptors are needed.
A further level of complexity is that the conventional and novel PKCs, unlike the other classes of DAG receptors, have tandem C1 domains. It is becoming increasingly clear that different C1 domains may have different recognition properties and functions (14,15). Therefore, compounds selective for individual C1 domains may have unique effects compared with less selective ligands (14).
Among the DAG receptors, particular attention has focused on PKC family members as therapeutic targets, because PKC isozymes play important roles in cell proliferation, tumor growth, and apoptosis (2). For example, elevated levels of PKC␣, PKC␥, PKC⑀, and PKC have been found in central nervous system tumors (2). Overexpression of PKC␣ and PKC␤ has been reported in breast cancer cells (2), and PKC␤ is elevated in chemotherapy-resistant diffuse large B-cell lymphoma (16). On the other hand, there are data suggesting that some PKC isoforms, such as PKC␦, in many contexts promote apoptosis (17).
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. Dramatic variation in PKC behavior is provided through variability in the residues at the binding site in the C1 domains themselves, variation in other portions of the C1 domain that interact with the membranes, variations in other regions of the receptor that affect the membrane interactions (e.g. the pseudosubstrate region, the C2 domain, or the C terminus), differences in the lipid composition of different subcellular membranes, and differences among cells in membrane composition and in interacting proteins (3).
Multiple classes of high affinity ligands for PKCs have been described. These ligands include the diterpenes such as the phorbol esters, macrocyclic lactones such as the bryostatins, polyacetates such as aplysiatoxin, or indole alkaloids such as teleocidin (3). Unfortunately, with the exception of the indole alkaloids, the complicated structures of these ligands have impeded their further development through medicinal chemistry. An alternative strategy, therefore, has been to use DAG derivatives in which the flexibility of the structure has been constrained to reduce the entropic loss due to binding (3). Using DAG-lactones, the Marquez group has been able to achieve potencies approaching those of phorbol esters (3).
The earlier natural products studies have highlighted the important role of the side chains in determining biological function. For example, 12-deoxyphorbol 13-tetradecanoate is a tumor promoter, whereas 12-deoxyphorbol 13-acetate (prostratin) is an inhibitor of tumor promotion (18), and these compounds induce distinct patterns of localization of PKC␦ in CHO cells (19). Similarly, unsaturated side chains promote inflammatory activity by phorbol esters but diminish their tumor promoting activity (20). Recently, the Marquez group has de-veloped a combinatorial approach for exploring extensively the chemical space represented by these side chains (21). Here, we described our characterization of one of the early compounds emerging from this combinatorial chemistry approach. We report that the novel DAG-lactone derivative 130C037 displays marked selectivity among the recombinant C1a and C1b domains of PKC␣ and PKC␦. Likewise, 130C037 displays substantial selectivity for RasGRP relative to PKC␣. These results provide strong encouragement for this drug discovery strategy directed at targets with C1 domains. . Phosphatidyl-L-serine was purchased from Avanti Polar Lipids (Alabaster, AL). Reagents for expression and purification of glutathione S-transferase (GST) fusion proteins were obtained from Pierce. Cell culture medium, reagents, and all of the DNA primers were obtained from Invitrogen. RasGRP3-GFP was generated as previously described (22). The mouse monoclonal anti-GFP antibody was purchased from Roche Applied Science. The rabbit polyclonal anti-PKC␣ and anti-PKC⑀ antibodies were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). The anti-mouse IgG and anti-rabbit IgG were purchased from Bio-Rad. The recombinant human PKC␣, PKC␤, PKC␥, PKC␦, and PKC⑀ were purchased from Invitrogen.

Materials
Construction of GST-and GFP-fused C1 Domains of PKC␣ and PKC␦-The C1a and C1b domains of PKC␣ and PKC␦ were generated by PCR using the Platinum® Pfx DNA polymerase (Invitrogen). The full-length cDNA clones of bovine PKC␣ and murine PKC␦ were used as the templates. The following oligonucleotides were used as the PCR primers to pull out the targeted C1 domains: (i) forward and reverse primers for ␣C1a were 5Ј-AACGTGCACGAGGTGAAGA-3Ј and 5Ј-CT-TGCTCCTCGGGTCATCT-3Ј; (ii) forward and reverse primers for ␣C1b were 5Ј-TGTCCGGGTGCGGATAAG-3Ј and 5Ј-CTTCTCTGTGTGAT-CCATTCCG-3Ј; (iii) forward and reverse primers for ␦C1a were 5Ј-AA-ACAGGCCAAGATCCACTACA-3Ј and 5Ј-GGTGTCCCGGCTATTGG-T-3Ј; (iv) forward and reverse primers for ␦C1b were 5Ј-CAGAAAGA-ACGCTTCAACATCG-3Ј and 5Ј-GGCCTCAGCCAAGAGCTTT-3Ј. 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 fragment. This fragment was then ligated into the appropriate fluorescent protein-containing vectors (i.e. pEGFP-C1, pEGFP-C2, and pEGFP-C3) (Clontech, Palo Alto, CA) and GST-containing vectors (i.e. pGEX-5X-1, pGEX-5X-2, and pGEX-5X-3) (Amersham Biosciences), using the EcoRI restriction sites with the sequence of the insert in the intended reading frame. The DNA sequence of each construct was finally confirmed by sequencing analysis (DNA Minicore, Center for Cancer Research, NCI, National Institutes of Health).
Expression and Purification of the GST-tagged C1 Domains of PKC␣ and PKC␦-The recombinant plasmids of individual C1 domains of PKC␣ and PKC␦ were transformed into BL-21-Gold (DE3) E. coli competent cells (Stratagene, La Jolla, CA). The expression of the GST fusion proteins was induced by the addition of 0.5 mM isopropyl-O-Dthiogalactopyranoside when the OD of the LB medium (Quality Biological, Inc., Gaithersburg, MD) reached 0.5-0.7. The bacteria were harvested after 4 h of induction at 37°C. The expressed GST-tagged C1 protein was purified using a B-PER GST spin purification kit (Pierce). The purity of the protein was verified by SDS-PAGE and staining with Coomassie Blue. The protein concentration was measured using the Bio-Rad protein assay kit. The purified GST-C1 proteins were stored in 30% glycerol at Ϫ70°C.
[ 3 H]PDBu Binding Assay-[ 3 H]PDBu binding to the recombinant full-length PKC isoforms or the C1 domains of PKC␣ and PKC␦ was measured using the polyethylene glycol precipitation assay developed in our laboratory (23). Briefly, the assay mixture (250 l) contained 50 mM Tris-HCl (pH 7.4), 100 g/ml phosphatidylserine, 4 mg/ml bovine immunoglobulin G, [ 3 H]PDBu, 0.1 mM CaCl 2 (for PKC␣, -␤, and -␥) and various concentrations of competing ligand. Incubation was carried out at 37°C for 5 min (for full-length PKC isoforms) or 18°C for 10 min (for C1 domains). Samples were chilled on ice for 7 min, 200 l of 35% polyethylene glycol in 50 mM Tris-HCl (pH 7.4) was added. The samples were mixed and incubated on ice for an additional 10 min. The tubes were centrifuged in a Beckman Allegra 21R centrifuge at 4°C (12,200 rpm, 15 min). A 100-l aliquot of the supernatant was removed for the determination of the free concentration of [ 3 H]PDBu, and the pellet was carefully dried. The tip of the centrifuge tube containing the pellet was cut off and transferred to a scintillation vial for the determination of the total bound [ 3 H]PDBu. Cytoscint (ICN, Costa Mesa, CA) was added both to the aliquot of the supernatant and to the pellet. Radioactivity was determined by scintillation counting. Specific binding was calculated as the difference between total and nonspecific binding. Standard Scatchard analysis was performed to determine the dissociation constants (K d ) of the individual C1 domains, and the inhibitory dissociation constants (K i ) were calculated using our standard method as described previously (23).
Expression and Imaging of the GFP-tagged Fluorescent Protein in Live Mammalian Cells-LNCaP cells (obtained from ATCC, Manassas, VA) were cultured at 37°C in RPMI 1640 containing 10% fetal bovine serum, penicillin (50 units/ml), and streptomycin (0.05 mg/ml) in a 5% CO 2 atmosphere. The plasmid DNA of GFP-fused individual C1 domains of PKC␣ and PKC␦ or the plasmid DNA of RasGRP3-GFP, PKC␣-GFP, and PKC⑀-GFP was transfected into LNCaP cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. The expression of the fluorescent protein was detected after 24 h of transfection. Confocal fluorescent images were collected with a Bio-Rad MRC 1024 confocal scan head (Bio-Rad) mounted on a Nikon microscope with a ϫ 60 planapochromat lens. Excitation at 488 nm was generated by a krypton-argon gas laser with a 522/32 emission filter for green fluorescence. For kinetics of GFP-C1a/C1b, RasGRP3-GFP, PKC␣-GFP, and PKC⑀-GFP translocation in live cells, cells plated on a 40-mm round coverslip were enclosed in a Bioptechs Focht Chamber System (Bioptechs, Butler, PA). The chamber was inverted and attached to the microscope stage with a custom stage adapter. A temperature controller set at 37°C was connected, and medium was perfused through the chamber with a Lambda microperfusion pump. Sequential images of the same cells were collected at various time points using LaserSharp Software (Bio-Rad).

Preparation of Cell Fractions and Immunoblot Analysis of Membrane Translocation of RasGRP3, PKC␣, and PKC⑀ in CHO-K1 Cells-
CHO-K1 cells (obtained from ATCC, Manassas, VA) were seeded in 60-mm Petri dishes. After drug treatment, the cell monolayer was washed once with culture medium followed by washing once with icecold Dulbecco's phosphate-buffered saline (KD Medical, Inc., Columbia, MD). The cells were harvested with 400 l of 20 mM Tris-HCl (pH 7.4) plus protease inhibitor mixture (Sigma). The cell suspension was then sonicated in an Eppendorf tube three times with a pulse of 6 s each time. 200 l of the sonication mixture was transferred to a Beckman ultracentrifuge tube and centrifuged at 200,000 ϫ g for 1 h with a Beckman ultracentrifuge to separate the cytosolic and membrane fractions; the remaining 200 l was used as the total cell fraction. After ultracentrifugation, the supernatant was designated as the cytosolic fraction. The pellet was resuspended in 200 l of 20 mM Tris-HCl (pH 7.4) plus protease inhibitor mixture (Calbiochem) and 1% Triton X-100. The pellet suspension was put on ice for 1 h. 100 l of lysis buffer plus 1% Triton X-100 was then added to the suspension. The mixture was sonicated three times with a pulse of 6 s each time. 200 l of the sonication mixture was again subjected to centrifugation (200,000 ϫ g for 1 h). This supernatant was designated as the Triton X-100-soluble membrane fraction.
Membrane translocation of RasGRP3, PKC␣, and PKC⑀ in response to 130C037 was determined by Western blot and analyzed by quantitation of the reduction in the amount in the cytosolic fraction. Because of the unavailability of anti-RasGRP3 antibody, the response of RasGRP3 to 130C037 in CHO-K1 cells was detected by the overexpressed RasGRP3-GFP using anti-GFP antibody. Briefly, the samples were subjected to SDS-PAGE and were transferred to polyvinylidene difluoride membranes (Millipore Corp.). Blotting of membranes was performed with antibodies against PKC␣, PKC⑀ (rabbit polyclonal, Santa Cruz), or GFP (for Ras-GRP3-GFP, mouse monoclonal; Roche Applied Science). Following washing, membranes were incubated with horseradish peroxidase-conjugated goat secondary antibody (Bio-Rad), developed with ECL Western blotdetecting reagent (Amersham Biosciences) and exposed to films (Eastman Kodak Co.). Bands were quantified densitometrically using ImageJ (NIH Image). In a typical membrane translocation assay, five concentrations of ligand were used. The ED 50 of the ligand was calculated using nonlinear multiple regression to the Hill equation based on the reduction of proteins in the cytosolic fractions.
Establishment of an HEK-293 Cell Line Stably Overexpressing Ras-GRP3-The coding sequence of human RasGRP3 (KIAA0846) was subcloned into the modified pLenti6/V5-D-TOPO® vector by using the pLenti6/V5-D-TOPO® cloning kit (Invitrogen) according to the manufacturer's instructions. This vector provides a V5 epitope tag at the C terminus of the RasGRP3. Lentivirus-containing culture supernatant was generated by using the ViraPower TM Lentiviral Expression System (Invitrogen) according to the manufacturer's instructions. HEK-293 cells (ATCC, Manassas, VA) were grown on 10-cm tissue culture plates with 10 ml of complete medium containing Dulbecco's modified Eagle's medium (ATCC) and supplemented with 10% fetal bovine serum (ATCC). After 80% confluence had been obtained, the culture medium was removed, and 10 ml of the collected retroviral supernatant containing 6 g/ml Polybrene (Sigma) was added to the plate. The cells were incubated at 37°C for 12 h. The medium-containing virus was then replaced by 10 ml of complete medium. After incubation for a further 24 h, the complete medium was replaced by the selection medium (complete medium containing 10 g/ml blasticidin (Invitrogen)). The fresh selection medium was replaced every 2 days until antibioticresistant colonies were identified. Studies were carried out on the pooled, antibiotic-resistant cells.
Erk Activation in Response to PMA or 130C037-Cells were grown in 60-mm tissue culture plates and serum-starved overnight before treatment with a series of concentrations of 130C037 or PMA for 30 min at 37°C. The cells were washed once with Dulbecco's phosphate-buffered saline and then lysed with 0.2 ml of M-PER mammalian protein extraction reagent (Pierce). The lysates were collected and transferred to 1.5-ml centrifuge tubes. Cell debris was pelleted by centrifugation at 14,000 ϫ g for 10 min. The supernatants were transferred to new tubes, and the protein concentration was measured using Bio-Rad protein assay reagent. The samples, each containing 30 g of total protein, were separated by electrophoresis on 10% SDS-polyacrylamide gels, and the protein bands were transferred onto Immobilon-P membranes (Millipore Corp.). After the membranes were blocked with 1ϫ phosphatebuffered saline containing 0.1% Tween 20 (Bio-Rad, Hercules, CA) and 5% nonfat milk for 1 h, the blots were probed with anti-phospho-44/42 mitogen-activated protein kinase monoclonal antibody (Cell Signaling Technology). Then the blot was stripped and reblotted with anti-44/42 mitogen-activated protein kinase monoclonal antibody (Cell Signaling Technology) and anti-V5 antibody (Invitrogen). The blots were then developed with the ECL system (Amersham Biosciences) and imaged on BioMax MAX films (Kodak).

RESULTS
Chemical Structure of 130C037 and 130C032 as Novel DAGlactones-As described in detail elsewhere, we have developed a combinatorial chemical approach for the synthesis of DAGlactones (21). As part of our further analysis of some of the initial compounds emerging from that program, two compounds showed atypical binding curves with PKC␣ and were therefore examined in detail. The structures of these compounds, 130C037 and 130C032, are illustrated in Fig. 1. Both compounds differ from most DAG-lactone or phorbol ester derivatives that have been described in the literature in that the sn-1 and sn-2 substituents possess polar groups (3).
130C037 and 130C032 Yielded Biphasic Binding Curves with PKC Isoforms in Vitro-The binding of 130C037 and 130C032 to recombinant PKC␣, PKC␤, PKC␥, PKC␦, and PKC⑀ was determined in vitro by competition with [ 3 H]PDBu (Fig. 2). Unlike typical competitive binding curves, these curves were shallow and multiphasic both for 130C037 ( Fig. 2A) and 130C032 (Fig. 2B). In addition, particularly at higher ligand concentrations, they showed appreciable differences between PKC isoforms. This is most evident comparing the curves for the classic PKCs ␣, ␤, and ␥ and for the novel PKCs ␦ and ⑀ at Novel Diacylglycerol-lactone with Marked Selectivity higher concentrations of 130C037. Of these two compounds, 130C037 showed a greater range of apparent potencies. When we analyzed the binding curves using a single site model, the apparent K i values of 130C037 and 130C032 for the five PKC isoforms ranged from ϳ30 to 350 nM (as shown in Table I).
Because of the shallow nature of the binding curves, the calculated K i values should be regarded as approximations.
Comparison of the Binding Affinities of 130C037 for Isolated C1a and C1b Domains of PKC␣ and PKC␦-Since some ligands have been found to show selectivity among individual C1 domains (14,15,24), we measured the binding of 130C037 to the purified, individual C1a and C1b domains of PKC␣ and PKC␦, which we expressed in E. coli as GST fusion proteins. We found dramatic differences in the binding to the different C1 domains ( Fig. 3 and Table II). 130C037 bound with high potency (K i ϭ 1.78 Ϯ 0.51 nM, mean Ϯ S.E., n ϭ 3 experiments) to the PKC␦ C1b domain, whereas its affinity for the PKC␦ C1a domain was 3 orders of magnitude less, with a K i of 2780 Ϯ 900 nM (mean Ϯ S.E., n ϭ 3 experiments). This difference markedly contrasts with that for PDBu, which bound with high affinity to both C1 domains, although its affinity for the ␦C1a domain was slightly weaker (6.2-fold) than for ␦C1b. Thus, whereas 130C037 was modestly less potent than PDBu for ␦C1b (5.4-fold), it was 1360-fold less potent for ␦C1a.
130C037 bound with substantially weaker affinity to either C1 domain of PKC␣ than it did to the ␦C1b domain (Fig. 3B, Table II). Its affinity for ␣C1a was 610 nM; for ␣C1b no binding was detectable. PDBu, in contrast, bound to both ␣C1a and ␣C1b with high potency (K i ϭ 0.40 and 3.40 nM, respectively). Thus, 130C037 was 1530-fold less potent than PDBu at ␣C1a and was Ͼ2900 less potent at ␣C1b.
We conclude that 130C037 demonstrated marked selectivity between C1 domains, having high affinity only for the ␦C1b among the four domains examined. This selectivity contrasts with the very modest selectivity observed for PDBu, arguing against artifacts in our experiments that would simply render a C1 domain inactive for binding. This is the first synthetic DAG analogue to have been reported so far with such high selectivity (Ͼ1000-fold difference in K i values) for the individual C1 domains. An important note of caution is required, however. Comparison of the binding affinity values for the individual C1 domains of PKC␣ and PKC␦ with those for the intact PKC isoforms did not provide good agreement (also see below). This is consistent with the expectation that the binding properties of the intact isoforms will depend on the full complement of structural features that determine the energetics of formation of the ternary/quaternary complex between PKC, ligand (at one or two sites), membrane, and calcium. Although evaluation of the binding characteristics of isolated C1 domains may be informative, ultimate evaluation will require assessment of the behavior of the intact protein, preferably in the context of the target cell.
Translocation of Individual C1a and C1b Domains of PKC␣ and PKC␦ by 130C037 in Single Live LNCaP Cells-The C1 domain selectivity of 130C037 was further examined in live cells. We prepared fusion constructs between the C1a and C1b domains of PKC␣ and PKC␦ and GFPs. The constructs were transfected into LNCaP cells, and the translocation of the overexpressed GFP-C1 was monitored by confocal microscopy as a function of time after the addition of 130C037. GFP-␣C1a, GFP-␣C1b, and GFP-␦C1b were all almost evenly distributed inside the cell (Fig. 4, A and B, first column), whereas GFP-␦C1a predominantly accumulated inside the nucleus (first column in panel 3 of Fig. 4A). After the application of 130C037 (20 M), translocation was only detected in the case of GFP-␦C1b (Fig. 4B). No apparent translocation was seen for GFP-␦C1a or for GFP-␣C1a and GFP-␣C1b under the same experimental conditions (Fig. 4A). In contrast, PMA induced a robust translocation for all four C1 domains (last column of Fig. 4, A and B). We conclude that the in vitro results for the binding of 130C037 to the C1 domains were mirrored in vivo in its ability to selectively induce translocation of the C1 domains to the membrane in intact, live LNCaP cells.
Comparison of the Binding Potency of 130C037 and 130C032 to RasGRP1/3 Versus PKC␣ and PKC⑀-We extended our characterization of 130C037 and 130C032 from PKC family members to the RasGRP family of DAG/phorbol ester-responsive signaling proteins. Unlike the PKC isoforms, RasGRP family members contain only one single C1 domain near their C terminus. In contrast to their weak to moderate binding to the PKC family members, 130C037 and 130C032 bound to both RasGRP1 and RasGRP3 with high affinity (Fig. 5, A and B, respectively). For comparison, the dose-response curves for PKC␣ and PKC⑀ are also shown on the same graphs. The K i values of 130C037 for RasGRP1 and RasGRP3 were 3.51 Ϯ 0.06 nM (mean Ϯ S.E., n ϭ 3 experiments) and 3.80 Ϯ 0.10 nM (mean Ϯ S.E., n ϭ 4 experiments) (Table III), respectively, reflecting ϳ90-fold more potent binding than to PKC␣ and

Binding of [ 3 H]PDBu to the individual C1 domains of PKC␦ (A) and PKC␣ (B) was measured in the presence of different concentrations of 130C037.
Binding assays were carried out as described under "Experimental Procedures." Results are from single, representative experiments. Each experiment was repeated at least two additional times with similar results.  The time in each panel represents the period after the drug administration. Images are from single, representative experiments.
Each experiment was repeated at least two additional times with similar results. 8-fold more potent binding than to PKC⑀. 130C032 behaved similarly but showed slightly less selectivity. Unlike the doseresponse curves for binding of 130C037 and 130C032 to the PKCs, the dose-response curves for binding to RasGRP1/3 were monophasic. We conclude that 130C037 and 130C032 show marked selectivity for RasGRP1 and RasGRP3 relative to PKC family members, especially PKC␣ and less so for PKC⑀, as determined by in vitro binding.
Comparison of 130C037 and Its Positional Isomer 130C045 for Binding to PKC and RasGRP-Computer modeling suggests that DAG-lactones can bind in either of two distinct orientations to C1 domains, depending on whether the binding utilizes the carbonyl of the sn-1 ester or the carbonyl of the sn-2 lactone (3). A critical consequence of the binding orientation is that it determines which side chain of the DAG-lactone projects away from the C1 domain into the lipid bilayer, and it would be expected that the more negatively charged p-nitrophenyl group would be disfavored for insertion into the bilayer. We therefore examined compound 130C045, the positional isomer of 130C037 in which the nitro-and dimethylamino-groups were swapped (see the insets in Fig. 6A). We compared the binding potency of this positional isomer for PKC␣ and RasGRP3 (Fig.  6, A and B) as well as for the C1a and C1b domains of PKC␦ and PKC␣ (Fig. 6, C and D). In most respects, 130C045 behaved similarly to 130C037. The average K i value of 130C045 for PKC␣ was 215 Ϯ 14 nM (mean Ϯ S.E., n ϭ 3 experiments). This value was comparable with that of 130C037 (i.e. 343 Ϯ 35 nM). For RasGRP3, 130C045 bound with a K i of 7.98 Ϯ 0.94 nM (mean Ϯ S.E., n ϭ 3 experiments); it was thus about 2-fold less potent than 130C037. The above results demonstrate that the selectivity between RasGRP3 and PKC␣ of 130C037 does not depend particularly on the positional isomerism of its two side chains.
Although 130C045 resembled 130C037 in its selectivity between RasGRP3 and PKC␣, its binding curves for PKC␣ were monophasic, unlike the biphasic curves observed for 130C037. We therefore examined the binding affinity of 130C045 for the isolated C1a and C1b domains of PKC␣ and PKC␦ to probe the relationship between the patterns of binding of the individual C1 domains and of the intact PKC. Like 130C037, 130C045 showed different binding potencies for the C1a and C1b domains of both PKC␦ (Fig. 6C) and PKC␣ (Fig. 6D). The average K i values indicated that 130C045 had about 800-fold higher binding potency for ␦C1b (K i ϭ 1.54 Ϯ 0.22 nM, mean Ϯ S.E., n ϭ 3 experiments) than for ␦C1a (K i ϭ 1229 Ϯ 25 nM, mean Ϯ S.E., n ϭ 3 experiments). This difference was similar to that of 130C037 for ␦C1a and ␦C1b. The difference between ␣C1a and ␣C1b was less, only about 50-fold between the two K i values (81.6 Ϯ 6.3 nM for ␣C1a and 4240 Ϯ 500 nM for ␣C1b, mean Ϯ S.E., n ϭ 3 experiments respectively). Since 130C037 had no detectable binding affinity for the C1b domain of PKC␣, quantitative comparison of 130C045 with 130C037 for selectivity between the ␣C1a and ␣C1b domains was not possible. However, the weaker binding potency of 130C045 for ␣C1b as compared with ␣C1a agreed qualitatively with that of 130C037. 130C045 thus had C1 domain selectivity for PKC␦ and PKC␣ similar to that of 130C037. This result argues against the apparent biphasic inhibition curves of 130C037 for PKC␣ and PKC␦ resulting simply from selectivity between C1 domains.
Western Blot Analysis Showed Preferential Translocation of RasGRP3 by 130C037 as Compared with PKC␣ and PKC⑀ in CHO Cells-The in vitro analysis on the binding potency of 130C037 for PKC and RasGRP was carried out in the presence of 100 g/ml phosphatidylserine, and it is well recognized that PKC isoforms may show different affinities and different structure activity relations in intact cells. Indeed, structure activity relations even depend on the specific cell types (25,26). We therefore examined three measures of response to 130C037 in intact cells. First, we determined the dose response for translocation of RasGRP3, PKC␣, and PKC⑀ by 130C037 in CHO-K1 cells using Western blot analysis. Because of the unavailability of anti-RasGRP3 antibody, the translocation of RasGRP3 was determined by using the overexpressed RasGRP3-GFP with anti-GFP antibody. The response of the endogenous PKC␣ and PKC⑀ was monitored using anti-PKC␣ and anti-PKC⑀ antibodies. Under our experimental conditions, much of the examined  proteins was already in the membrane fraction of the control samples; we therefore quantitated the reduction of the cytosolic fraction as an index of membrane translocation of the proteins. No change in cytosolic PKC␣ distribution was observed at any concentration of 130C037 over the range investigated (Fig. 7B). Partial loss of RasGRP3 and PKC⑀ from the cytosol was observed (Fig. 7, A and C). However, the quantification of the bands in the cytosolic fractions demonstrated that 130C037 was much more potent for translocating RasGRP3 than it was for translocating PKC⑀ (Fig. 7D). The average ED 50 for Ras-GRP3 from five independent experiments was 286 Ϯ 12 nM (mean Ϯ S.E.); the ED 50 for PKC⑀ was more than 10 M. Only minor loss of RasGRP3 or of PKC␣ from the total fraction was seen, arguing against down-regulation or cell loss under these conditions of time and concentration accounting for the response. Although PKC␣ did not translocate in response to 130C037, it translocated in response to PMA (10 nM, 100 nM, 1 M, and 10 M) under these conditions (three experiments, data not shown), indicating that the lack of translocation of PKC␣ in response to 130C037 was not an artifact. We conclude that, as in vitro, in intact cells RasGRP3 was the most sensitive receptor for 130C037 when comparing with PKC⑀ and PKC␣. It is also clear from this analysis that 130C037 was much less potent in the intact cell system than in the in vitro assays. This reduction in apparent potency of DAG-lactones in cellular systems is typical (e.g. see Ref. 27).
Translocation of RasGRP3-GFP, PKC␣-GFP and PKC⑀-GFP in Response to 130C037 in Live LNCaP Cells-As a complementary approach to confirm the in vivo selectivity of 130C037 for RasGRP3, we investigated the dynamic translocation of GFP-tagged RasGRP3, PKC␣, and PKC⑀ in LNCaP cells, a prostate cancer cell line. The dynamic redistribution of Ras-GRP3-GFP, PKC␣-GFP, and PKC⑀-GFP in live LNCaP cells after the application of 130C037 was monitored using a Bio-Rad MRC 1024 confocal microscope. As a positive control, response to PMA was determined. RasGRP3-GFP translocated to internal membranes after the application of 130C037 (20 M), similar to its response to PMA (Fig. 8A). PKC␣-GFP did not show any membrane translocation in response to 130C037 but showed a clear response to PMA (Fig. 8B). Unlike PKC␣-GFP, PKC⑀-GFP showed a slow, partial plasma membrane translocation in response to 130C037 (20 M). This live cell imaging confirmed again, in a second cell type under different conditions, that in vivo 130C037 was most sensitive for RasGRP3, least active on PKC␣, and of intermediate activity on PKC⑀.
Erk1/Erk2 Phosphorylation Induced by 130C037 but Not by PMA Was Dependent on RasGRP3-As a third measure of selectivity of 130C037 for RasGRP3 in intact cells, we examined the functional response of Erk phosphorylation. Early studies of PMA-induced Erk activation indicated that stimulation of PKC could lead to activation of Raf in a manner that depends on basal levels of Ras activity but does not involve overt Ras activation (28). Although the mechanistic details of this process have not been fully elucidated, it may reflect a ubiquitous process, because PMA activates Erk in many cell types. In contrast, in select cell types, including lymphocytes, PMA activates RasGRP family members. This process probably depends on direct recruitment of RasGRPs to membranes where it can catalyze conversion of Ras-GDP to Ras-GTP that contributes to Ras activation (10,29,30). Besides these two distinct mechanisms of PMA-induced Erk activation, there is growing evidence that PKC and RasGRP signaling mechanisms can intersect. In particular, RasGRP3 requires phosphorylation by PKC for full activation, which may be mediated by co-recruitment of these proteins to common membranes (11)(12)(13).
We compared the ability of 130C037 and PMA to induce Erk1/Erk2 phosphorylation in control HEK-293 cells and in HEK-293 cells stably overexpressing RasGRP3 (Fig. 9). PMA induced Erk phosphorylation in the control cells, reflecting the response mediated by the endogenous PKC (Fig. 9A, left half). This response was inhibited by treatment with the PKC inhibitor GF109203X (data not shown). In the HEK-293 cells heterologously expressing RasGRP3, PMA likewise induced Erk phosphorylation as expected, reflecting its combined effect on endogenous PKC and the heterologously expressed RasGRP3 (Fig. 9A, right half). In contrast to PMA, 130C037 up to a concentration of 40 M was unable to induce Erk phosphorylation in the control HEK-293 cells (Fig. 9B, left half), whereas it was able to do so in the RasGRP3-overexpressing cells (Fig.  9B, right half). Controls for Erk protein confirmed equal load- ing in all lanes, and controls for the V5 epitope confirmed equal levels of the RasGRP3 in the lanes with the RasGRP3-expressing cells. To eliminate an alternative explanation, that PKC was up-regulated in the RasGRP-overexpressing cells, we examined the protein level of different PKC isoforms in both control and RasGRP3-overexpressed cells by Western blot analysis. No difference was observed (data not shown). We conclude that 130C037 is selective for inducing functional responses through a RasGRP3-dependent pathway in this intact cell system.

DISCUSSION
The diacylglycerol signaling pathway has gained great attention as a central regulatory mechanism in cells, contributing to the control of cell proliferation, differentiation, and apoptosis. PKC, the major receptor for DAG and the tumorpromoting phorbol esters, has thus emerged as an attractive therapeutic target for cancer and a range of other conditions. Complementing the design of inhibitors targeted to the catalytic domain of PKC, we and others have pursued the design of ligands targeted to the C1 domains involved in PKC activation (3,14). The rationale is that different PKC isoforms may produce functional antagonism, and activation of an antagonistic PKC may achieve the same result as inhibition of the complementary isoform. With the identification of other families of signaling molecules containing DAG-responsive C1 domains (e.g. RasGRP, PKD, or chimaerin), both the challenges for designing specificity and the opportunities have increased.
We are still in the early stages of exploring the full potential of constrained DAG derivatives as probes of the function of C1 domain containing receptors and as lead molecules in drug discovery. Extensive analysis has identified a potent DAGlactone template and an appropriate pattern of substitution for generating potent ligands (3). These structures have now been incorporated into a combinatorial chemistry approach to evaluate in detail the influence of extensive, novel modifications of the side chains. In parallel, we are developing appropriate screening approaches to capture the potential diversity of functional responses to such compounds. In this paper, we have presented our characterization of two compounds that have emerged from our initial efforts. We described here a novel DAG-lactone (130C037) that was selective in vitro for binding to RasGRP1/3 as compared with PKC␣. This is the one of the first studies so far reporting a compound selective for RasGRP and the first example of a compound with such a degree of selectivity for a DAG receptor other than PKC. Previously, we have shown that the iridal NSC631939 bound to RasGRP with 5-fold selectivity relative to PKC␣ (32). We have shown in this study that the novel DAG-lactone 130C037 had a much stronger binding affinity for RasGRP1/3 (K i ϭ 3.8 Ϯ 0.1 nM and K i ϭ 3.51 Ϯ 0.06 nM) than for PKC␣ (K i ϭ 343 Ϯ 35 nM). The Western blot analysis of RasGRP3 and PKC␣ translocation in CHO-K1 cells also demonstrated that under similar experimental conditions 130C037 could only translocate RasGRP3 and not PKC␣ to the membranes (Fig. 7, A and B). Although the novel PKC isoform ⑀ could respond to 130C037, the potency of 130C037 to induce membrane translocation of RasGRP3 (ED 50 ϭ 286 nM) was appreciably stronger than for PKC⑀ (ED 50 Ͼ 10 M) (Fig. 7, C and D). The live cell imaging confirmed the selectivity of 130C037 in inducing RasGRP3-GFP translocation but not that of PKC␣-GFP (Fig. 8, A and B). Under these conditions, PKC⑀-GFP showed weak plasma membrane translocation (Fig. 8C), which was consistent with the Western blot result. Finally, we have demonstrated that 130C037 could only induce Erk phosphorylation in HEK-293 cells exogenously expressing RasGRP3 (Fig. 9), whereas PMA could also induce its phosphorylation subsequent to the activation of the endogenous PKC in the cells.
The identification of 130C037 provides strong proof of principle for the potential of our combinatorial chemistry strategy and for the feasibility of designing compounds selective for members of the families of signaling molecules with DAGresponsive C1 domains. On the other hand, 130C037 has its limitations. Its activity in intact cellular systems is at the level of 10 Ϫ7 M, in contrast to an in vitro potency of 10 Ϫ9 M. It is somewhat less selective relative to PKC⑀ than it is relative to PKC␣ or PKC␦. Finally, it appears to have a problem with solubility at micromolar concentrations, which we think contributes to the shallowness of the dose-response curves at the higher concentrations. The cells were washed with Dulbecco's phosphate-buffered saline and then lysed. The cell lysates were separated by electrophoresis on 10% SDSpolyacrylamide gels and subjected to Western blotting with anti-phospho-Erk antibody. Equal loading of all lanes was confirmed by blotting with anti-Erk antibody and, for the RasGRP3-overexpressing cells, with antibody directed against the V5 epitope tag incorporated into the RasGRP3 construct.
Besides the selectivity of 130C037 for RasGRP versus PKC␣ and to a lesser degree PKC⑀, our study emphasizes that the behavior of C1 domains is strongly modulated by the context in which they are found in the intact proteins. As we have shown in the in vitro binding assays, the binding affinities of 130C037 for full-length PKC␦ (or PKC␣) and its isolated C1a and C1b domains clearly did not match. For example, the K i value of 130C037 for PKC␦ was 91 Ϯ 16 nM, but for ␦C1b it was 1.78 Ϯ 0.51 nM. For ␦C1a, the K i was 2780 Ϯ 900 nM. For PKC␣, the K i of 130C037 (343 Ϯ 35 nM) was 4-fold weaker than that for PKC␦ (91 Ϯ 16 nM), but its affinity for the better binding C1 domain, ␣C1a (610 Ϯ 170 nM), was 350-fold weaker than that for ␦C1b (1.78 Ϯ 0.51 nM). Although the analysis of structure activity relations for the binding of ligands to isolated C1 domains, as so elegantly done by Irie et al. (14,24,33), should continue to be informative, particularly for identifying positive interactions, great caution should be exercised in interpreting negative results.
Presumably, a basis for the disparity between the binding to isolated C1 domains and to the intact receptors reflects the complexity of the interaction being measured, which is the formation of (at the very least) a ternary complex in which the ligand, the phospholipids, and the receptor all interact. As a simplified example, the addition of extra positively charged residues at the end of the C1 domain, in a position where they will not affect the direct interaction of ligand with the binding cleft but where they can affect charged interactions with the negatively charged phospholipids, had a substantial effect on the measured binding affinity (33). In the intact PKC, moreover, multiple domains contribute to the membrane interaction, including the pseudosubstrate region, the C1 domains, and Ca 2ϩ -C2 domain complex. Furthermore, as suggested by Oancea and Meyer (34), the C1 domains appear to be buried and require a conformational change of the enzyme to become accessible. This same conformational change may help drive the extraction of the pseudosubstrate domain from the catalytic site of the enzyme. All of these coupled changes in interactions and conformational effects will necessarily be incorporated into the global energy change reflected in the apparent binding affinity of the ligand (35).
In the case of the PKC and PKD families of receptors, a further source of divergence in the behavior between the individual C1 domains and the intact proteins is that they contain two C1 domains. The relative contributions of these domains to the intact receptor remain uncertain, reflecting different analytic methodologies as well as different receptors being analyzed, for which the actual contributions may be different. For example, initial studies (6,36) showed that both C1 domains of PKC␥ bound PDBu with high affinity, suggesting that C1a and C1b of PKC␥ are functionally equivalent. In contrast, other investigations (15,31) suggested that the C1 domains in PKC␣ and PKC␦ played somewhat different roles, depending on the specific ligand. Our results from 130C037 binding support nonequivalent roles of the C1a and C1b domains of PKC␣ and PKC␦, at least in terms of binding affinity for the DAG ligand. The development of ligands with marked selectivity among C1 domains should assist in the evaluation of their roles in the functioning of the intact receptors.
130C037 provides a glimpse into the opportunities that the combinatorial chemistry approach should afford for manipulating and dissecting the function of the members of the DAG receptor superfamily. Given its access to chemical modification, the DAG-lactone template provides a powerful platform for structural exploration.