Direct Interaction of All-trans-retinoic Acid with Protein Kinase C (PKC)

Protein kinase C (PKC) regulates fundamental cellular functions including proliferation, differentiation, tumorigenesis, and apoptosis. All-trans-retinoic acid (atRA) modulates PKC activity, but the mechanism of this regulation is unknown. Amino acid alignments and crystal structure analysis of retinoic acid (RA)-binding proteins revealed a putative atRA-binding motif in PKC, suggesting existence of an atRA binding site on the PKC molecule. This was supported by photolabeling studies showing concentration- and UV-dependent photoincorporation of [3H]atRA into PKCα, which was effectively protected by 4-OH-atRA, 9-cis-RA, and atRA glucuronide, but not by retinol. Photoaffinity labeling demonstrated strong competition between atRA and phosphatidylserine (PS) for binding to PKCα, a slight competition with phorbol-12-myristate-13-acetate, and none with diacylglycerol, fatty acids, or Ca2+. At pharmacological concentrations (10 μm), atRA decreased PKCα activity through the competition with PS but not phorbol-12-myristate-13-acetate, diacylglycerol, or Ca2+. These results let us hypothesize that in vivo, pharmacological concentrations of atRA may hamper binding of PS to PKCα and prevent PKCα activation. Thus, this study provides the first evidence for direct binding of atRA to PKC isozymes and suggests the existence of a general mechanism for regulation of PKC activity during exposure to retinoids, as in retinoid-based cancer therapy.

lipids (phosphatidylserine (PS)) and diacylglycerol (DAG) or phorbol esters for activation (1)(2)(3)(4). The classic and novel PKC isozymes contain an amino-terminal regulatory domain and a carboxyl-terminal catalytic domain (4). The regulatory domain of classic PKC isozymes contains two common regions, C1 and C2. The C1 domain mediates DAG and phorbol ester binding through distinct low and high affinity binding sites (5), while the C2 domain mediates Ca 2ϩ and PS binding and contains the receptor for activated C kinase binding site (6,7). PKC requires acidic phospholipids for its activity, and, in the presence of activators, the enzyme has the highest binding affinity for membranes containing PS. Upon activation, PKC isozymes are translocated to distinct subcellular compartments (membranes) and cell structures to phosphorylate their respective substrates (4,8). This compartmentalization is required for the phosphorylation of specific substrates and the regulation of different physiological functions (4,9). Preventing PKC translocation from the cytoplasm to membranes inhibits PKC function and the subsequent phosphorylation of specific substrates (7,9).
PKC isozymes regulate gene expression and a variety of cellular functions, including growth, differentiation, tumor promotion, aging, and apoptosis (1,2,5); however, the biological significance of the heterogeneity in the PKC family is not clear. The distinct subcellular distribution, the presence of several isozymes in the same cell, and differential activation or inhibition by different stimuli suggest that each isozyme is involved in the regulation of different functions and has a unique role in the cell (2,4,9,10).
Retinoids are vitamin A derivatives that regulate cellular growth, differentiation, development, and apoptosis (11,12). Retinoids control gene expression and synthesis of a variety of proteins through nuclear retinoic acid receptors and are directly involved in regulation of numerous signaling molecules, including PKC (13)(14)(15)(16)(17)(18). Specifically, retinoids have been implicated in the mechanism of reversion from the malignant to the normal cell phenotype, inhibition of cancer invasiveness, and are currently used to treat various types of cancer, including promyelocytic leukemia, head and neck, skin, breast, and ovarian cancers (19 -22). It has been postulated that retinoids exert some of their effects on cellular differentiation and reversion of malignant phenotype through interactions with PKC isozymes (13,16,23,24). It has been shown that retinoic acid alters the cellular localization of PKC␣, consistent with PKC␣ inactivation, in endometrial adenocarcinoma cells concurrent with the induction of differentiation of these cells (24). Collectively, the data suggest that alterations in PKC activity induced by retinoids are associated with reversion of the malignant phenotype (25)(26)(27). However, the mechanism by which retinoic acid regulates PKC is not known.
In this work, we have begun the characterization of the biochemical events underlying this regulation. We have previously identified an amino acid motif common to the putative RA binding sites of several RA-utilizing proteins (28). The presence of this motif in several PKC isoforms led us to postulate and examine the existence of cellular mechanisms that mediate receptor-independent signaling by retinoids through a direct interaction with various PKC isozymes. We have used photoaffinity labeling to demonstrate that purified, recombinant PKCs bind atRA with high affinity. Using PKC activity assays, we have shown that preincubation of PKC␣ with micromolar concentrations of atRA decreases PKC activity.
Taken together, this is the first evidence for a direct, high affinity binding of atRA to PKC␣ that results in a decrease in PKC␣ activity. Given the significance of these in vitro results, we hypothesize that atRA might also be involved in regulating PKC␣ activity in vivo by competing with PS for binding to PKC molecules. The exact localization of the atRA binding sites is under investigation.

Amino Acid Alignment
Polypeptides for CRABP I (29), CRABP II (29), and RAR (30) were selected on the basis of the crystal structures of these proteins with atRA bound. A polypeptide for UGT2B7 was selected according to modeling of a RA binding site in this protein (28). Polypeptides for PKC proteins were chosen through a homology comparison to CRABP and RAR␥ proteins and represent COOH-terminal fragments of their C2 domains. Sequences were aligned using programs resident in the GCG Protein and DNA Analysis Package (31).

Photoaffinity Labeling with [ 3 H]atRA
Direct photoaffinity labeling with [ 3 H]atRA was done using modifications of the method described by Bernstein (32). For screening of the different PKC isoforms, 7 pmol each of PKC␣, ␤I, ␤II, ␥, ␦, ⑀, and in 20 mM HEPES, pH 7, were incubated for 2 min at room temperature with or without 30 M atRA. The PKC catalytic subunit was labeled according to the same protocol, except that only 2 pmol of protein was used. Following electrophoresis, gels were stained with Coomassie Blue, destained, washed thoroughly with water, treated with Autofluor autoradiography enhancer, (National Diagnostics, Manville, NJ), dried, and subjected to autoradiography at Ϫ80°C for 4 -7 days. Results were quantitated by densitometric analysis of the autoradiograms using an AlphaInnotech IS-1000 Digital Imaging System (AlphaInnotech, San Leandro, CA).
For more detailed studies with PKC␣, photoaffinity labeling was carried out in essentially the same manner using 0.2 g (2.6 pmol) of protein/sample in a final volume of 10 l with an ethanol concentration of 5%. Any other modifications are given in the descriptions of specific experiments.

Enzymatic Assay
Preparation of Mixed Micelles-PS (2.8 mg) and PMA (final concentration: 100 M) dissolved in chloroform were transferred to a glass tube. Following evaporation of chloroform under a stream of nitrogen, 1 ml of 3% Triton X-100 (prepared fresh) was added and the mixture was sonicated.

RESULTS AND DISCUSSION
Amino Acid Alignment-Amino acid alignments and comparison of the crystal structures of several atRA-utilizing proteins, such as CRABP I and II (cytosolic retinoic acid-binding protein I and II), nuclear receptor RAR␥ and UGT2B7 (the enzyme catalyzing glucuronidation of atRA) with the C2 domain of PKC was carried out as the initial step in the investigation. A motif, previously designated as comprising part of the atRA-binding site, was present in three PKC isozymes examined (␣, ␤I, and ␥) (Fig. 1).
The amino acid distribution among the analyzed sequences revealed three categories of homologous amino acids ( Fig. 1): conserved in CRABP and UGT2B7 (yellow blocks), conserved in PKCs and RAR␥ (green blocks), and those maintained across all the analyzed sequences (red blocks). For the purpose of this alignment, amino acids in hydrophobic set (Leu, Val, Ile, Ala), hydrophilic set (Thr, Ser, Asn), acidic set (Asp, Glu) and basic set (Lys, Arg, His) were considered homologous. The alignment shows a significant homology of amino acids in the C2 domains to amino acids constituting RA binding pockets for CRABPs, RAR␥, and that predicted for UGT2B7 (double underlined sequences). Arg-249 of PKC␣ and PKC␥ corresponds to Arg-111 involved in the interaction with the RA carboxyl in CRABPs and Lys-229 of RAR␥, involved in electrostatic interaction with the ligand (star). A lysine cluster that constitutes a phosphatidylserine binding site in C2 domains is positioned 35 amino acids upstream from the center of the predicted RA binding site of PKCs.
On the basis of the significant sequence homologies among the analyzed sequences, we propose that the amino acids shown on the consensus line might represent all or part of the RA binding pocket of PKC␣. The conserved motif in CRABP I contains five amino acids, Arg-111, Leu-120, Phe-122, Arg-131, and Tyr-133, which have been identified from analysis of the crystal structure as being a part of the atRA binding pocket (29,34).
Additionally, significant homologies between a PKC␣ sequence pattern containing the previously defined PS binding site and the predicted atRA binding site (amino acids 172-260) of the C2 domain and the PKC C1 domain sequence were not detected.
Photoaffinity Labeling-To test our hypothesis of a direct interaction between atRA and PKC, we used direct photoaffinity labeling of PKC␣ with [11, H]atRA, which covalently modified the PKC␣ protein within the atRA-binding site and provided the direct evidence for atRA binding to PKC. Because atRA is one of the several common retinoids used experimentally and clinically, it was the retinoid of choice to investigate the direct interaction between retinoids and PKC. The photoaffinity labeling occurred in a concentration-dependent manner, was dependent on UV irradiation, and was effectively protected by unlabeled atRA, atRA glucuronide (atRAG), and various other retinoids (Fig. 2). It has been demonstrated here that PKC␣ exhibits different affinities for various retinoids and that the presence of a carboxyl function on the ligand was an obligatory requirement for binding to PKC. All RA derivatives tested, with the exception of 13-cis-RA, strongly inhibited binding of atRA to PKC␣. These results also indicate that atRA derivatives such as 9-cis-RA and atRAG might be functional ligands for PKC␣. Thus, our data demonstrate that atRA binds to PKC␣ through a RAbinding site and that this process is highly specific.
To characterize atRA interactions with PKC, we examined the ability of atRA to compete with different PKC activators such as PS, PMA, DAG, and Ca 2ϩ for binding to PKC␣. Fig. 3 showed that there was a strong competition between PS and atRA for binding to PKC␣ but only a slight competition between PMA and atRA and no competition between DAG and atRA. These results indicated that atRA does not interact with PKC␣ through the PMA or DAG binding sites. They also confirmed that atRA binds specifically to PKC␣ and competes with PS for binding. This is consistent with our hypothesis based on the amino acid analysis shown in Fig. 1 and with a model proposing that PS interacts with PKC within a different domain than DAG and PMA. Calcium had no effect on atRA binding to PKC␣ in the presence of PS, PMA, or DAG (data not shown), which demonstrates that, although PS and Ca 2ϩ binding sites  are localized to the same domain (C2) of PKC␣, atRA does not interact with PKC␣ through the Ca 2ϩ binding site (Fig. 3).
In order to identify whether other PKC isozymes contain atRA-binding sites, PKC␣, ␤I, ␤II, ␥, ␦, ⑀, and were photolabeled under the same conditions with equimolar concentrations of PKC protein and the ligand. The data presented in Fig. 4 show that all PKC isozymes investigated were photolabeled; however, clear differences in the degree of incorporation of label into individual isozymes were observed. Thus, PKC␤II had the highest affinity for atRA, while PKC had the lowest affinity.
The catalytic subunit of PKC was used for photoaffinity labeling to determine other possible sites of atRA binding to the PKC molecule. There was some atRA binding to the catalytic subunit of PKC (ϳ3% of atRA binding to PKC␣). However, this binding was not UV-light dependent nor was it protected by unlabeled atRA (data not shown). These results indicated that there is no specific interaction between atRA and the PKC catalytic subunit and suggested that atRA binding occurs in the regulatory subunit of PKC. However, the efficiency of this binding differs depending on the PKC isoform being used for photolabeling.
Direct comparison of the photolabeling of PKC␣ and PKC by atRA and protection by PS and arachidonic acid were also carried out and the data are presented in Fig. 5. PKC was photolabeled by atRA with 6-fold lower efficiency than PKC␣. Like photolabeling of PKC␣, the photolabeling of PKC by atRA was light-sensitive and protected by increasing concentrations of unlabeled atRA, which demonstrated that atRA binding to PKC is specific. However, in contrast to PKC␣, the photolabeling of PKC was not protected by preincubation with PS (Fig. 5).
Furthermore, there was no competition between arachidonic acid and atRA for binding to PKC␣, but the photolabeling of PKC by atRA was dramatically diminished by all concentrations of arachidonic acid used in the study (Fig. 5). No effect of PMA on PKC photolabeling by atRA was found (data not shown).
Taken together, these data suggest that: 1) the majority of atRA binds to the regulatory subunit of PKC and 2) the major atRA binding site in PKC␣ may be localized in close proximity to the PS binding site. These results also show that there is no apparent affinity of atRA for the fatty acid binding site of PKC␣, presumably localized to the C1 domain. In the case of PKC, which lacks the C2 domain and is strongly regulated by fatty acids (35), atRA may bind to the C1 domain either in, or in close proximity to, the fatty acid binding site.
Recent site-directed mutagenesis studies from another laboratory carried out on several residues in the C2 domain of PKC␣, including the arginine residues, Arg-249 and Arg-252, indicated that all mutants exhibited full membrane binding affinity and enzymatic activity at saturating concentrations of PS, DAG, and Ca 2ϩ (43). These studies also have shown that amino acids Arg-249 and Arg-252 in the C2 domain of PKC␣ are involved in electrostatic interactions with membrane anionic phospholipids. The two arginines are positioned in the center of our amino acid alignment of proteins that bind RA (Fig. 1), with Arg-249 being highly conserved. Particularly, Arg-249 aligns with Arg-111 of CRABP I and II, which is involved in electrostatic interaction with carboxyl function of RA (29). It is likely that RA may interact with Arg-249 and/or Arg-252 of PKC in a similar fashion and, thus, could be responsible for the interplay between PS and RA demonstrated in our studies. This explanation does not exclude other possible interpretations for RA-mediated modifications of PKC activity, including a RA-induced allosteric effect.
The general conclusion from these experiments is that all the PKC isoforms studied directly interact with atRA but the binding of atRA to different isoforms occurs with different affinity. Although the major site of atRA binding in PKC␣ is postulated to be on the regulatory domain, in PKC, the binding occurs in the C1 domain of the regulatory subunit. The exact location and the nature of this binding and its relation to PS and arachidonic acid binding are currently under investigation in our laboratory. Whether the interaction between PKC and RA results in up-or down-regulation of PKC activity clearly depends on the biological system being studied (36,37).
Enzymatic Activity-To examine the effects of atRA binding on PKC catalytic activity, the activity of purified PKC␣, the most common PKC isoform, was determined in the presence or absence of atRA and various PKC activators. Pre-incubation of PKC␣ with PS and PMA, followed by the addition of increasing concentrations of atRA, resulted in a 23% and 61% decrease in PKC␣ activity in the presence of 50 and 100 M atRA, respectively ( Fig. 6; data not shown for 100 M). At concentrations lower than 50 M, atRA had no effect if added following preincubation with the PS/PMA complex. Simultaneous incubation of PKC␣ with PS, PMA, and atRA resulted in a slight (20%) decrease in PKC␣ activity in the presence of 1-10 M atRA and a 30% decrease in PKC␣ activity in the presence of 20 -50 M atRA (Fig. 6). In contrast, pre-incubation of PKC␣ with atRA prior to addition of the PS/PMA complex resulted in a dramatic decrease in PKC␣ activity (40% and 62% in the presence of 1 and 20 M atRA, respectively) and total inhibition of the kinase activity at 50 M atRA (Fig. 6). These results show that: 1) atRA decreases PKC␣ activity, 2) atRA competes with PS and/or PMA for binding to PKC␣, and 3) atRA binds to PKC␣ with a high affinity and cannot be displaced by an addition of PS and/or PMA.
To characterize specific interactions of atRA with individual activators of PKC␣, the enzyme was pre-incubated in the presence of 20 M atRA or the vehicle and the activity measured following addition of increasing concentrations of PS, PMA, DAG, and Ca 2ϩ . In the absence of atRA, PKC␣ activity was stimulated by PS in a concentration-dependent manner and reached a plateau at concentration of 350 g/ml PS (23-fold stimulation in comparison with PKC␣ activity in the absence of PS) (Fig. 7). In the presence of atRA, the stimulation of PKC␣ activity by PS was decreased to 30% of the stimulation in the absence of atRA. These results show that atRA decreases stimulation of PKC␣ activity by PS and suggest that atRA and PS compete for binding to PKC␣.
In the absence of atRA, PKC␣ activity increased gradually in the presence of increasing concentrations of PMA and reached a plateau at a concentration of 1 M PMA (10-fold stimulation in comparison with PKC␣ activity in the absence of PMA) (Fig.  8A). atRA decreased the activity of PKC␣ by 44% in the absence of PMA but did not prevent subsequent stimulation of the PKC␣ activity by increasing concentrations of PMA (Fig. 8A). Likewise, atRA decreased PKC␣ activity in the absence of DAG but did not prevent PKC␣ stimulation by increasing concentrations of DAG (Fig. 8B). These results show that atRA does not inhibit the activation of PKC␣ by PMA or DAG and suggest that atRA does not compete with PMA or DAG for binding. Finally, atRA decreased the overall PKC␣ activity by 53% in the absence of Ca 2ϩ and the presence of PS and PMA but did not prevent PKC␣ stimulation by increasing concentrations of Ca 2ϩ (Fig. 9). These results demonstrate that atRA prevents the stimulation of PKC␣ activity by PS and suggest that atRA alters the interaction of PS with PKC␣.
A Model for PKC/atRA Interaction-A model for the proposed regulation of PKC by phosphorylation, membrane binding, and activation is well established and has been published previously (8). In this model, newly synthesized PKC associates with the detergent-insoluble fraction of cells. Three phosphorylation events convert PKC into a mature form that is re-leased into cytosol but is inactive due to the presence of pseudosubstrate in the substrate-binding cavity. PKC translocation to the membrane occurs once DAG binds to the C1 domain and is followed by phospholipid (PS) binding to the C2 domain. Association of DAG with its binding site in the C1 domain and PS with its binding site on the C2 domain leads to pseudosubstrate release and PKC activation.
Based on our results, we propose a modification of this existing model, which accounts for the direct interaction between atRA and PKC␣ in the cell (Fig. 10). The central hypothesis of our model is that atRA binds to PKC at pharmacological concentrations (1.0 M or higher), a situation that can result from consumption of large amounts of vitamin A or its metabolic precursors or atRA therapy (38). Under normal conditions, atRA and 13-cis-RA are present in the plasma at nanomolar concentrations (39,40) and atRA is considered the most prevalent form of vitamin A in most tissues (41,42). We tested our hypothesis by exploring the effect of different atRA concentrations on both atRA binding and PKC activity. The results show that, in an in vitro system, atRA binds to PKC with high affinity at concentrations above 10 M. Because atRA does not exist in the cell in the free form, we hypothesize that it can react with PKC following binding to CRABP in those cells where PKC and CRABP are both expressed. In preliminary experiments (data not shown) on CRABP and PKC interactions, a significant decrease of CRABP photolabeling by atRA was seen in the presence of PKC, which suggests that there is an interplay between PKC, CRABP and atRA. Experiments are in progress to elucidate the exact mechanism of this interaction and its implications. Previously, it has been shown that apo-CRABP can be phosphorylated by Ca 2ϩand PS-dependent PKC and that holo-CRABP inhibits PKC activity (36,44), which suggests that an interaction between RA-binding proteins and PKC may play a role in the regulation of PKC activity and retinoid action.
Our model is restricted to the regulation of PKC␣ by atRA and is still hypothetical. We propose that atRA interacts with PKC when the atRA concentration in the cell exceeds 1.0 M (pharmacological). Following atRA entry into the cell and binding to CRABP, the complex can react with PKC. As shown in Fig. 10, the interaction of atRA and PKC may occur: 1) when PKC is associated with the detergent-insoluble fraction of cells, 2) following PKC phosphorylation and release into the cytosol, or 3) when PKC translocates to the membrane domain but prior to PKC binding to phospholipids (PS). With atRA bound to the PKC␣ molecule, the final activation by PS cannot occur. A similar model for PKC is under investigation.
Our results show that, in vitro, at concentrations lower than 50 M, atRA can not displace PS from its binding site (Fig. 6). Therefore, atRA binding to PKC␣ probably does not occur following PKC translocation to cellular membranes because the binding site is not accessible in this situation. However, once atRA is bound to PKC prior to PKC translocation, the association of PKC with membrane phospholipids and final activation cannot occur. Our examination of the crystal structure of the C2 domain of PKC␣ showed that a polypeptide of the predicted atRA binding site and a lysine cluster, which constitutes the PS binding site (45), appear to form a shared binding pocket (Fig. 11). A 23-amino acid spacer runs between and behind the two sites, bringing them to this common interface. Whether there are other atRA binding sites on the PKC molecule and the specificity of atRA binding to these sites are subjects currently being investigated.
Conclusions-Our results from PKC photoaffinity labeling and measurements of catalytic activity of PKC␣ are consistent and demonstrate: 1) the existence of a specific atRA-binding site on PKC and 2) a direct interaction between atRA and PKC␣ that decreases PKC activity. The results presented here also suggest that atRA may be involved in the mechanism of regulation of PKC activity in the cell. Previous reports are inconsistent and atRA has been shown to stimulate or inhibit PKC activity depending on the experimental model used (36,44,46). Cope and co-workers (47) have demonstrated an interaction, similar to that shown here, between RA and PKC in vivo in the mouse brain. The precise mechanism of the PS-atRA interaction, such as allosteric effects and/or steric hindrance to PS binding, resulting in a decrease in enzymatic activity has not been unambiguously determined.
The fact that atRA treatment modulates PKC activity could account for some of the diverse effects that atRA treatment has on cancer cells. Using retinoids at therapeutic doses to alter PKC activity could control major cellular processes in malignant cells, including inhibition of cell proliferation and initia-tion of differentiation or apoptosis. Thus, agents that decrease the activity of certain PKC isozymes might be beneficial as pharmacological tools to alter cellular functions regulated by these PKC isozymes, specifically under pathological conditions such as cancer.