Characterization of AJH-836, a diacylglycerol-lactone with selectivity for novel PKC isozymes

Diacylglycerol (DAG) is a key lipid second messenger downstream of cellular receptors that binds to the C1 domain in many regulatory proteins. Protein kinase C (PKC) isoforms constitute the most prominent family of signaling proteins with DAG-responsive C1 domains, but six other families of proteins, including the chimaerins, Ras-guanyl nucleotide–releasing proteins (RasGRPs), and Munc13 isoforms, also play important roles. Their significant involvement in cancer, immunology, and neurobiology has driven intense interest in the C1 domain as a therapeutic target. As with other classes of targets, however, a key issue is the establishment of selectivity. Here, using [3H]phorbol 12,13-dibutyrate ([3H]PDBu) competition binding assays, we found that a synthetic DAG-lactone, AJH-836, preferentially binds to the novel PKC isoforms PKCδ and PKCϵ relative to classical PKCα and PKCβII. Assessment of intracellular translocation, a hallmark for PKC activation, revealed that AJH-836 treatment stimulated a striking preferential redistribution of PKCϵ to the plasma membrane relative to PKCα. Moreover, unlike with the prototypical phorbol ester phorbol 12-myristate 13-acetate (PMA), prolonged exposure of cells to AJH-836 selectively down-regulated PKCδ and PKCϵ without affecting PKCα expression levels. Biologically, AJH-836 induced major changes in cytoskeletal reorganization in lung cancer cells, as determined by the formation of membrane ruffles, via activation of novel PKCs. We conclude that AJH-836 represents a C1 domain ligand with PKC-activating properties distinct from those of natural DAGs and phorbol esters. Our study supports the feasibility of generating selective C1 domain ligands that promote novel biological response patterns.

Lipid second messengers, in particular diacylglycerol (DAG) 2 and phosphoinositides, have emerged as key signaling molecules downstream of innumerable cellular receptors. Responses to DAG are mediated by seven families of proteins, protein kinase C (PKC), chimaerin, RasGRP, protein kinase D, MRCK, Munc13, and DAG kinase, which combine a DAG recognition motif, termed a (typical) C1 domain, along with disparate functional domains and other regulatory domains. Among these families of signaling proteins, the PKC isoforms have been studied most intensely (1)(2)(3).
PKC isozymes are key elements in signaling cascades that control multiple cellular functions, including proliferation, survival, motility, and gene expression, and have been widely implicated in cancer progression and other diseases. Of the 10 human PKC isoforms, both the "classical/conventional" (cPKCs) (␣, ␤I, ␤II, and ␥) and "novel" PKCs (nPKCs) (␦, ⑀, , and ) contain twin C1 domains that recognize DAG, whereas only the cPKCs contain a C2 domain that responds to elevated internal calcium. For both C1 and C2 domains, ligand binding leads to insertion/bridging of the C1/C2 domains to the membrane with formation of the ternary complex of domain, ligand, and phospholipid. Typically, there is a requirement for acidic phospholipids such as phosphatidylserine for this membrane interaction (3)(4)(5).
Phorbol esters have become the most extensively characterized nonphysiological natural compounds capable of activating PKCs. These diterpenes bind with high affinity to the C1 domains located in the N-terminal regulatory domain of cPKCs and nPKCs, competing with DAG for binding to C1 domains (6,7). The C1 domains from PKCs and other proteins with phorbol ester/DAG-binding capabilities possess an overall similar globular structure, stabilized by twin zinc fingers, with the phorbol ester/DAG-binding site formed from a hydrophilic cavity bounded by a hydrophobic rim. Phorbol ester/DAG binding completes the hydrophobic surface, while the specific substituents on the phorbol ester/DAG contribute further hydrophobic character. The increased hydrophobicity drives/ stabilizes insertion of the C1 domain-ligand complex into the membrane lipid bilayer. This insertion is further stabilized by interactions between charged residues on the surface of the C1 domain with acidic phospholipid headgroups (8,9). The redistribution to membranes induced by phorbol esters or related C1 domain ligands, i.e. the translocation from the cytosolic compartment to the plasma membrane or internal membranes, is a hallmark of PKC activation and can be readily visualized in cellular models by using fluorescently tagged PKCs (10, 11).
As might be anticipated from their central role in cellular signaling, the signaling network downstream from the PKC isoforms and the other families of signaling proteins with typical C1 domains is extraordinarily complex. Thus, individual PKC isoforms have been shown to confer distinctive patterns of cellular responses, and the differences between isoforms may be dramatic. Depending on cellular context, PKC␣, PKC␦, and PKC⑀, the most commonly expressed PKCs, can trigger mitogenic/tumor-promoting or conversely antimitogenic/ tumor suppressor responses (3,12,13). A prediction is that ligands binding to typical C1 domains with differential selectivity could therefore induce distinct patterns of biological response.
Although the relevant mechanisms may be uncertain, there are clear examples of C1 domain-targeted ligands inducing distinct patterns of response. Whereas phorbol 12-myristate 13-acetate (PMA) is the archetypical mouse skin tumor promoter, prostratin (12-deoxyphorbol 13-acetate) was shown to be anti-tumor promoting. Bryostatin 1, a DAG-mimetic macrocyclic polyacetate, activates PKCs in vitro but paradoxically antagonizes many phorbol ester responses. Ingenol 3-angelate only partially stabilizes PKC-membrane interactions. Such diversity of response underlies therapeutic potential. Ingenol 3-angelate (under the trade name Picato TM ) has been approved for treatment of actinic keratosis. Bryostatin 1 has been the subject on numerous clinical trials for cancer and more recently for dementia. Prostratin has been proposed as an agent for conferring sensitivity to drug treatment for cells latently bearing HIV (12, 14 -17).
A daunting challenge in the development of inhibitors for kinases has proven to be the large size of the kinome and the high homology in the ATP-binding site in the kinase domain (12,18). Although the high structural similarities among the binding clefts of C1 domains might a priori suggest similar problems, several factors provide encouragement. Most importantly, natural products such as prostratin, bryostatin 1, and ingenol 3-angelate have unambiguously proven that C1-targeted ligands can achieve differential biological outcomes (12)(13)(14)(15)(16)(17)(18). Deeper consideration of the mechanism provides a rationale. First, the ligand forms a ternary complex of C1 domain-ligand inserted into the lipid bilayer. The interaction surface of this complex is thus much more extensive than simply that formed between the ligand and the binding cleft of the C1 domain, and there are extensive differences in the surfaces of the typical C1 domains that interact with the lipid headgroups. Second, it is appreciated that different ligands can cause different depths of insertion, changing these interactions. Furthermore, there is considerable diversity between cellular membranes or between microdomains within the membrane, changing the lipid environment with which the interactions will take place (5,12,19,20).
A powerful approach for generating structural diversity among C1 domain-targeted ligands has been through a DAGlactone template, affording compounds approaching the phorbol esters in potency (19 -23). Cyclization of DAGs to rigid lactone structures reduces the entropic penalty associated with the flexibility of the glycerol backbone and provides a scaffold with rich opportunities for decoration at positions homologous to the side chains of DAG. The power of this approach has been shown in the analysis of combinatorial libraries of DAG-lactones where changes in the patterns of acyl and alkyl substitution yielded marked differences in the patterns of biological outcome (24). Likewise, targeted synthesis of DAG-lactones, extending insights from the combinatorial chemistry, have yielded compounds with marked selectivity for the RasGRP family of C1 domain-containing signaling proteins relative to their affinity for PKC (20).
It has been previously reported that natural DAGs bind with similar affinities to all cPKCs and nPKCs (25). Toward the objective of generating isozyme-specific PKC ligands, Ann et al. (26) had described the incorporation of linoleic acid derivatives as well as saturated and unsaturated alkyl chains into the side chains of DAG-lactones and characterized these derivatives for their in vitro binding affinity toward PKC⑀, an oncogenic member of the PKC family, relative to PKC␣ (27)(28)(29)(30)(31)(32). Following up on this analysis, here we describe that the DAG-lactone AJH-836 showed marked selectivity for PKC⑀ and PKC␦ relative to PKC␣ in multiple cell types, unlike the typical phorbol ester PMA, and correspondingly displayed a distinct pattern of biological activity. The pattern of selectivity among PKC isoforms of AJH-836 is novel, further emphasizing the rich opportunities afforded by C1 domain-targeted ligands.
Here, our initial objective was to examine the ability of AJH-836 to bind to recombinant PKC␣, PKC␤II, PKC␦, and PKC⑀ (four of the main phorbol ester-responsive PKCs expressed in cancer cells) under parallel conditions using a competition [20-3 H]phorbol 12,13-dibutyrate ([ 3 H]PDBu) binding assay with phosphatidylserine vesicles as cofactor (5). Under these condi-

Differential translocation of PKC␣ and PKC⑀ by AJH-836 in cellular models
In vitro binding assays using reconstituted phospholipid vesicles will necessarily fail to capture the full range of differences in C1 domain ligand recognition occurring in a cellular context. To tackle this issue, we next examined subcellular translocation of the individual PKC isoforms expressed in cells as GFP-fused proteins. We hypothesized that the differential binding properties observed in vitro for AJH-836 and AJH-1512 could be reflected in characteristic patterns of PKC isozyme translocation in cellular models.
Treatment of HeLa cervical adenocarcinoma cells with AJH-836 fully translocated PKC⑀ to the plasma membrane in a concentration-dependent manner ( Fig. 2A). The EC 50 for this effect was 0.23 M. Conversely, much higher AJH-836 concentrations were required to induce PKC␣ plasma membrane translocation (EC 50 ϭ 9.8 M). The ratio EC 50 PKC␣/ EC 50 PKC⑀ of 43 indicates a notable selectivity for activation of PKC⑀ by AJH-836 in this cellular context. Unlike AJH-836, AJH-1512 displayed similar potencies for translocation of these PKCs (EC 50 PKC␣ ϭ 26.1 M; EC 50 PKC⑀ ϭ 17.1 M; ratio EC 50 PKC␣/EC 50 PKC⑀ ϭ 1.5) (Fig. 2B), which was consistent with the lack of selectivity observed in in vitro binding assays. As a control, PMA, which potently activates cPKCs and nPKCs, fully translocated both PKC␣ and PKC⑀ to the plasma membrane.
Although PKC␣ is the major classical PKC isoform in HeLa cervical adenocarcinoma cells, PKC␦ is the second major, wellstudied novel PKC isoform along with PKC⑀ in these cells (33). Because our in vitro binding assays revealed similar preferential   affinities for PKC⑀ and PKC␦ for AJH-836, we wished to address whether this DAG-lactone displays similar selectivity for PKC␦ translocation in the cellular context. A limitation, however, is the predominant pattern of PKC␦ translocation to internal membranes (perinuclear/Golgi region and nuclear membrane) (11, 22, 34 -36), which makes it less objective to quantify upon direct visualization. Moreover, prominent perinuclear PKC␦ localization could be observed under basal conditions in HeLa cells (Fig. S1). As expected, AJH-836 treatment led to a quite heterogeneous pattern of PKC␦ translocation with

PKC isozyme selectivity of a DAG-lactone
cells showing staining in plasma membrane, nuclear membrane, and/or the perinucleus. If only plasma membrane translocation were assessed, results revealed a reduced ability of AJH-836 to redistribute PKC␦ relative to PKC⑀ (EC 50 ϭ 11 M; ratio EC 50 PKC␦/EC 50 PKC⑀ ϭ 48) (data not shown).
As the pattern of PKC isozyme translocation may vary depending on the cellular context, we also evaluated PKC␣ and PKC⑀ relocalization by AJH-836 in A549 and H358 non-smallcell lung cancer (NSCLC) cells. We used a concentration of AJH-836 of 1 M, which showed the maximum difference in translocation between PKC␣ and PKC⑀ in HeLa cells. As clearly shown in Fig. 2, C and D, AJH-836 fully translocated PKC⑀ in both A549 and H358 NSCLC cells, whereas translocation of PKC␣ was negligible. As expected, PMA caused full translocation of both PKC␣ and PKC⑀ to the plasma membrane in NSCLC cells.
To further corroborate the differential translocation pattern of PKC isozymes by AJH-836, we carried out a quantitative densitometric analysis of PKC cellular localization using ImageJ. As shown in Fig. 3, PMA promoted a marked shift in cytosolic to membrane intensity for both PKC␣ and PKC⑀ in HeLa cells with a 4.4-and 4.0-fold increase in the membrane/cytosolic PKC ratio, respectively. In contrast, the pattern of redistribution of PKC␣ and PKC␦ induced by AJH-836 was drastically different. Indeed, at 1 M, AJH-836 treatment caused a 3.5-fold increase in the membrane/cytosolic ratio for PKC⑀, whereas it did not appreciably affect this ratio for PKC␣, thus confirming the selectivity of this DAGlactone for PKC⑀.

Activation of PKC␦ by AJH-836 in cellular models
To investigate the potential ability of AJH-836 to activate PKC␦, we measured PKC␦ phosphorylation at Ser 299 . Phosphorylation at this site has been reported to be a marker of catalytically active enzyme (37,38). Analysis of Ser 299 -PKC␦ phosphorylation by Western blotting using a specific phosphoantibody revealed concentration-dependent activation by AJH-836 in HeLa, A549, and H358 cells. Our results revealed maximum phosphorylation by AJH-836 at 10 M in all cases. The EC 50 values from the phospho-Ser 299 -PKC␦ densitometric analysis were 0.8 (HeLa cells), 1.3 (A549 cells) and 0.9 M (H358 cells) (Fig. 4).

Differential down-regulation of PKC isozymes by prolonged treatment with AJH-836
Prolonged stimulation of cells with phorbol ester analogues is known to down-regulate the expression of PKCs, a process that occurs upon sustained association with the plasma membrane and is mediated via protein degradation at intracellular compartments (39 -41). We compared the ability of AJH-836 to down-regulate the PKC isoforms as a function of time and concentration using the NSCLC cellular models. Down-regulation of PKC␦ and PKC⑀ by AJH-836 in the A549 and H358 cells was negligible after 3 h and began to be detected at 6 h, partic-

PKC isozyme selectivity of a DAG-lactone
ularly at 1 M AJH-836. Near complete down-regulation for both PKC␦ and PKC⑀ was observed after 16 or 24 h of treatment with 1 M AJH-836 for both the A549 and H358 cells. In contrast, the pattern of PKC␣ down-regulation was notably different. PKC␣ levels in the A549 and H358 cells remained essentially unchanged upon treatment with 1 M AJH-836, even at the longer time points. Only a partial PKC␣ down-regulation could be observed at the highest concentration used in this analysis (10 M) (Fig. 5, A and B). The distinctive pattern of PKC isozyme down-regulation caused by AJH-836 became more evident when compared with that induced by PMA. PMA downregulated all three PKC isoforms in the NSCLC cells with a slight preference toward PKC␣. This differential sensitivity for PKC isozyme down-regulation between PMA and AJH-836 is depicted in Fig. 5, C and D.

AJH-836 induces major changes in cytoskeletal morphology
We have previously reported that specific activation of lung cancer cells with PMA induces the formation of lamellipodia and membrane ruffles, actin-rich cytoskeletal structures implicated in the reorganization of actin cytoskeleton and cell motility. This effect is mediated by the activation of the small G-protein Rac1. PKC⑀ has been shown to be a crucial mediator of ruffle formation in lung cancer cells (31,42,43). To investigate the functional responses triggered by AJH-836, we examined morphological changes induced by this agent in A549 cells. Fig.  6A shows that the DAG-lactone induces the formation of ruffles in A549 cells in a concentration-dependent manner. Next, we examined the effect of PKC inhibitors with distinctive isozyme selectivity on the ability of AJH-836 to induce ruffles.

PKC isozyme selectivity of a DAG-lactone
The widely used pan-PKC inhibitors GF109203X and Gö6983, which inhibit both nPKCs and cPKCs, caused complete inhibition in ruffle formation induced by AJH-836. In contrast, Gö6976, which preferentially inhibits cPKCs over nPKCs (44), only caused a small reduction in ruffle formation (Fig. 6B).
To further assess the involvement of individual PKCs expressed in A549 cells in AJH-836 -induced ruffle formation, we used validated RNA interference (RNAi) duplexes to knock down PKC␣, PKC␦, or PKC⑀ (Fig. 6C). Consistent with the involvement of PKC⑀ in cytoskeleton rearrangement (31), silencing its expression caused a marked reduction in the formation of ruffles induced by the DAG-lactone. Conversely, no inhibition in AJH-836 -induced ruffle formation could be observed upon silencing PKC␣. PKC␦ RNAi depletion in A549 cells caused a marginal reduction in ruffle formation by AJH-836, although not statistically significant (p ϭ 0.06), and did not have any additive effect when knocked down together with PKC⑀.
Lastly, to further confirm the involvement of PKC⑀ as a target for AJH-836 in A549 cells, we took advantage of ⑀V1-2, a TATfused peptide that selectively inhibits PKC⑀ (18,30,31,45). This inhibitor caused a significant inhibition in ruffle formation induced by the DAG-lactone, whereas a TAT control peptide had no effect (Fig. 6D). Although we did not examine the activity of AJH-836 on targets other than the PKC isoforms, these experiments clearly show that the effects of AJH-836 on morphology in the cell systems examined were mediated by the novel PKCs, in particular PKC⑀.

Discussion
In the present study, we characterized a DAG analogue, AJH-836, which has higher affinity in vitro for novel PKC␦ and PKC⑀ isozymes relative to PKC␣. Consistent with this result, analysis of intracellular relocalization revealed that AJH-836 has greater potency for translocation (i.e. activation) of PKC⑀ to the plasma membrane relative to PKC␣. This is quite remarkable because, as demonstrated in early studies, DAG generated physiologically upon activation of receptors displays similar in vitro affinities for cPKCs and nPKCs, a pattern also observed for prototypical phorbol esters. Unlike AJH-836, other DAG-lactones have been reported to have similar activities for cPKCs and nPKCs in vitro and better translocate PKC␣ to the plasma membrane (22,46). Other natural compounds with C1 domain-binding capabilities such as 12-deoxyphorbol esters, mezerein, and thymeleatoxin have higher affinities for cPKCs relative to nPKCs (25).
It is well-established that phorbol esters and related analogues, including DAG-lactones, bind specifically to the C1 domains in cPKCs and nPKCs. These domains are duplicated in tandem (C1a and C1b) in these two PKC subfamilies. Binding occurs in the groove formed where the two strands of a ␤-sheet pull away from each other, and the bound ligand results in a completed hydrophobic surface on the C1 domains that is energetically favorable to interact with the hydrocarbon core of the phospholipid bilayer. Basic residues distal to the phorbol ester/ DAG-binding site contribute by interacting with headgroups from acidic phospholipids (8). In a cellular context, phorbol ester/DAG binding to the C1 domain triggers the translocation of the enzyme to membranes. C1 domains also have additional roles, including binding to autoinhibitory sites, that reflect the potential differential roles of "exposed" versus "nonexposed" C1 domains (9,47,48). The differential binding of AJH-836 to PKCs may be the consequence of different scenarios, one of the obvious being the preferential association between the ligand and nPKC C1 domains dictated by specific residues interacting with the DAG-lactone backbone and/or alkyl chains in the appropriate binding modes. It is probable that specific "ternary complex" interactions that involve membrane phospholipids may contribute to AJH-836 isozyme specificity as well, a setting that is consistent with the dependence of its binding affinity on the membrane composition (26). Although dissecting the individual structural features contributing to the pattern of PKC isoform selectivity will depend on extensive structural analysis, the current data make unambiguously clear that it is possible to generate differential PKC isoform selectivity such as described here along with differential biology.
AJH-836 is structurally simple compared with the highly complex natural PKC ligands such as the phorbol esters, bryostatins, ingenols, and indolactams. It therefore provides a powerful lead compound for probing the features conferring nPKC selectivity. Structure-activity analysis based on modifications of DAG-lactones has proven to be a successful approach for the generation of potent analogues capable of distinguishing between different proteins with C1 domains (20,25,49,50), and it therefore represents a promising strategy for achieving PKC isozyme selectivity. It should be emphasized that absolute selectivity among the families and family members of the signaling proteins that contain typical C1 domains is not necessary to generate useful compounds. Rather, as evidenced by natural products such as bryostatin 1, ingenol 3-angelate, and prostratin, compounds that can generate different patterns from the rich network of responses downstream of these signaling proteins may have utility.
One attractive scenario where selective agents such as AJH-836 may have unique effects is in cancer cell models where individual PKCs seem to exert specific functions. For example, PKC⑀ is largely overexpressed in epithelial cancer cell lines and drives migratory and metastatic phenotypes (3, 27, 29 -32). In this context, we found that AJH-836 promotes the formation of ruffles, a step that is essential for cancer cell motility. Because PKC⑀ has tumor-promoting functions and other PKCs, including PKC␣, have been shown to suppress tumor growth (3,27), differential regulation of PKC isoforms such as PKC⑀ and PKC␣

PKC isozyme selectivity of a DAG-lactone PKC isozyme selectivity of a DAG-lactone
by C1 domain ligands could promote novel, inhibitory effects on the transformed phenotype. Although we have yet to examine the binding properties of AJH-836 for PKCs expressed in specialized cell types, such as PKC (expressed mainly in T lymphocytes) and PKC (expressed primarily in skin), their homology to PKC␦ and PKC⑀, respectively, argues for potential usefulness of this DAG analogue in specific contexts (51,52).
In summary, the functional characterization of AJH-836 provides a glimpse into the opportunities for PKC isozyme selectivity by manipulation of conformationally constrained DAGlactones. Given the simplicity and accessibility for chemical modification, this compound provides a platform for the rational design of nPKC-selective agents. Individual members of the PKC family, including nPKCs, and other phorbol ester receptor families have been extensively implicated in human disease, not only cancer but also cardiovascular, neurological, and immunological disorders. Therefore, compounds capable of distinguishing among the different proteins with C1 domains could be promising leads for therapeutic purposes.

Synthesis of AJH-836 and AJH-1512
The synthesis of AJH-836 and AJH-1512 is reported in detail elsewhere (23). These agents are described as compounds 104 and 98, respectively, in that study.

In vitro binding assays
PKC␣, -␤II, -␦, and -⑀ were from Life Technologies. Compounds were assayed in vitro by competition for the binding of [ 3 H]PDBu (13.5 Ci/mmol; prepared as a custom synthesis by PerkinElmer Life Sciences) in the presence of 100 g/ml phosphatidylserine and 1 mM EGTA as described previously (5). Incubation was for 5 min at 37°C. In each competition assay, seven concentrations of ligand (DAG-lactone) were used with the increasing concentrations of ligand spaced at half-log intervals. In each experiment, triplicate measurements at each concentration of ligand were performed. K i values were calculated from ID 50 values determined from the competition curves using GraphPad software built-in analysis tools (GraphPad Software, Inc., San Diego, CA).

Translocation of GFP-tagged PKC isozymes in cellular models
For localization studies of GFP-fused PKCs, experiments were carried out essentially as described previously (10). After transfection of cells (5 ϫ 10 4 ) with 1 g of pEGFP-N1-PKC␣, pEGFP-N1-PKC␦, or pEGFP-N1-PKC⑀ using Lipofectamine 3000, the cells were plated on coverslides in 24-well plates. Twenty-four hours later, cells were serum-starved for 24 h and then stimulated with PMA, AJH-836, AJH-1512, or vehicle for 30 min. After cells were washed with PBS and fixed with precooled (Ϫ20°C, 20 min) methanol, samples were stained with DAPI (1 g/ml) for 10 min at 4°C, mounted on a glass slide, and visualized with a Nikon TE2000-U fluorescence microscope. For the quantification of translocated cells, 20 -100 cells/plate were visually assessed in a blinded manner. Translocation was also quantitated using ImageJ. A line was traced across the cytoplasm of individual cells, and the signal intensities and profiles were obtained using the Plot profile tool of the program.

PKC down-regulation assays
For assessment of PKC down-regulation, 1.5 ϫ 10 5 cells were serum-starved for 24 h and then treated with different concentrations of PKC activators. Cells were lysed at different time points (3, 6, 16, and 24 h). Expression of endogenous PKC isozymes was determined by Western blotting.

Cell ruffle formation
Assessment of morphological cytoskeletal changes was done as described before (31). Briefly, A549 lung cancer cells growing on glass coverslides at low confluence were serum-starved for PKC isozyme selectivity of a DAG-lactone cated. Following fixation with 4% formaldehyde, F-actin was stained with phalloidin-rhodamine, and nuclei were counterstained with DAPI. Slides were visualized by fluorescence microscopy, and five random fields were scored for the number of ruffles. Ruffle area was measured by thresholding for signal intensity using ImageJ.

Statistical analysis
Analysis of variance was performed using GraphPad Prism software built-in analysis tools. The confidence interval was set to 95%. A p value Ͻ0.05 was considered statistically significant.