p23/Tmp21 Associates with Protein Kinase Cδ (PKCδ) and Modulates Its Apoptotic Function*

There is emerging evidence that C1 domains, motifs originally identified in PKC isozymes and responsible for binding of phorbol esters and diacylglycerol, interact with the Golgi/endoplasmic reticulum protein p23 (Tmp21). In this study, we investigated whether PKCδ, a kinase widely implicated in apoptosis and inhibition of cell cycle progression, associates with p23 and determined the potential functional implications of this interaction. Using a yeast two-hybrid approach, we found that the PKCδ C1b domain associates with p23 and identified two key residues (Asp245 and Met266) implicated in this interaction. Interestingly, silencing p23 from LNCaP prostate cancer cells using RNAi markedly enhanced PKCδ-dependent apoptosis and activation of PKCδ downstream effectors ROCK and JNK by phorbol 12-myristate 13-acetate. Moreover, translocation of PKCδ to the plasma membrane by phorbol 12-myristate 13-acetate was enhanced in p23-depleted LNCaP cells. Notably, a PKCδ mutant that failed to interact with p23 triggered a strong apoptotic response when expressed in LNCaP cells. In summary, our data compellingly support the concept that C1 domains have dual roles both in lipid and protein associations and provide strong evidence that p23 acts as an anchoring protein that retains PKCδ at the perinuclear region, thus limiting the availability of this kinase for activation in response to stimuli.

p23/Tmp21 (hereafter referred to as p23) is a type I transmembrane protein that belongs to the p24 endoplasmic reticulum/Golgi cargo family and plays critical roles in protein transport, organization of membrane structure, and the secretory pathway. Members of the p24 family have been widely implicated as COP (coat protein) vesicle cargo receptors, and they are involved in COP vesicle budding and the organization of the Golgi apparatus (1). In addition to the well established role of p23 in vesicle formation and cargo transport in the endoplasmic reticulum/Golgi (2)(3)(4)(5), this protein has been implicated in several other cellular functions, including the recruit-ment of the small GTPase ADP-ribosylation factor 1 to the Golgi apparatus (6,7) and ADP-ribosylation factor 1-dependent assembly of actin in the Golgi (8). p23 also is a component of the presenilin complex at the plasma membrane and modulates ␥-secretase cleavage (9).
In a previous yeast two-hybrid screening, we established that p23 physically associates with ␣and ␤-chimaerins, Rac GTPase-activating proteins (GAPs) 4 regulated by the lipid second messenger diacylglycerol (DAG) and DAG mimetic analogs such as the phorbol esters (10 -13). Mutational and deletional studies revealed that this interaction is mediated by the C1 domain present in chimaerins (10). The 50 -51-amino acid cysteine-rich C1 domains were identified originally in PKC isozymes as the sites for DAG and phorbol ester binding. C1 domains drive the recruitment of proteins to internal membranes and, in this manner, facilitate their access to docking proteins and effectors. More recently, we identified key residues within the C1 domain in ␤2-chimaerin responsible for binding p23 and perinuclear targeting of this Rac-GAP (14). Interestingly, isolated C1 domains from PKC isozymes bind to p23, suggesting a role for this p24 family member in perinuclear docking of C1 domain-containing proteins. Yeast two-hybrid, co-precipitation, and localization studies revealed a high degree of specificity for this interaction. Indeed, p23 binds tandem C1 domains from classical PKCs (cPKC␣) and novel PKCs (nPKC␦ and nPKC⑀) but not the C1 domain from atypical aPKC (14). The fact that chimaerins and PKC␦ have been found in the cellular perinucleus raised the possibility that p23 is implicated in intracellular targeting of these signaling proteins. Such scenario seems to be true for ␤2-chimaerin, as disruption of its interaction with p23 impairs ␤2-chimaerin perinuclear association (14).
The PKC serine/threonine kinases have been implicated widely in multiple physiological and pathological processes, including differentiation, mitogenesis, tumor progression, cell survival, and apoptosis (15). PKC␦ has been established as a prominent proapoptotic kinase in various cellular models, including prostate cancer, glioma, and thyroid cancer cells (16 -18). Our laboratory established that activation of PKC␦ by phorbol esters leads to apoptosis of androgen-dependent prostate cancer cells (18 -24). PKC␦ activity is also required for cell death induced by DNA-damaging agents such as etoposide or doxorubicin (25,26). The proapoptotic effect of PKC␦ in prostate cancer cells is mediated by the autocrine secretion of death factors and involves the activation of JNK and RhoA/ROCK pathways (21,22).
In the present study, we examined the potential implication of p23 in PKC␦ function. We found that PKC␦ interacts with p23 via its C1 domain and that this association has significant functional implications for PKC␦-mediated apoptosis and signaling in prostate cancer cells.
Yeast Two-hybrid Assay-pLexA expression vectors, which contain a His marker, were co-transformed into the yeast strain EGY48 together with pB42AD-HA-tagged expression vectors, which have a Trp marker, and the p8OP-LacZ reporter vector (Ura marker). Transformants were plated on yeast dropout medium lacking Trp, Ura, and His, thereby selecting for the plasmids encoding proteins capable of a two-hybrid interaction, as evidenced by transactivation of the LacZ reporter gene. A detailed explanation of the methodology used for these experiments has been described previously (10).
Co-immunoprecipitation-Cells were washed once with PBS and then lysed for 10 min at 4°C in 800 l of a lysis buffer containing 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 0.5 mM EDTA, 0.5 mM EGTA, 0.5% Triton X-100, 1 mM ␤-glycerophosphate, and protease inhibitor mixture (Sigma). Cell lysates were centrifuged at 14,000 ϫ g, and the supernatants were precleared with 15 l of protein A-agarose beads (Invitrogen) for 1 h at 4°C. After a short centrifugation, supernatants were incubated with either PKC␦ antibody (2 g) or control IgG (at 4°C for 2 h). Protein A-agarose beads (20 l) were then added, and the mixture was incubated at 4°C for 2 h. Beads were extensively washed in lysis buffer and boiled. Samples were resolved in 12% SDS-polyacrylamide gels and transferred to polyvinylidene difluoride membranes for Western blot analysis.
Apoptosis Assays and Flow Cytometry-The incidence of apoptosis was determined as we described previously (18). Briefly, cells were trypsinized, mounted on glass slides, fixed in 70% ethanol, and then stained for 20 min with 1 mg/ml 4Ј,6diamidino-2-phenylindole. Apoptotic cells were characterized by chromatin condensation and fragmentation when examined by fluorescence microscopy. The incidence of apoptosis in each preparation was analyzed by counting ϳ300 cells.
In other experiments, LNCaP cells were transfected with pEGFP vectors encoding WT-PKC␦ or D245G/M266G-PKC␦. After 72 h, cells were fixed in 70% ethanol and stained with propidium iodide (1 mg/ml). The percentage of cells (GFPtransfected and nontransfected) in sub-G 0 /G 1 was determined in an EPICS XL flow cytometer (Coulter Corp., Hialeah, FL). For each treatment, ϳ7500 events were recorded.
Immunostaining and Confocal Microscopy-Cells were washed twice with PBS, fixed in 4% paraformaldehyde in PBS for 10 min, and then washed briefly with PBS. For immuno-staining, cells were incubated with PBS containing 0.5% SDS/5% ␤-mercaptoethanol (37°C, 30 min) and blocked in 5% BSA in PBS for 30 min. After washing twice with PBS, cells were incubated for 2 h with anti-PKC␦, anti-p23, or anti-V5 antibodies (1:200). Secondary goat FITC-conjugated anti-rabbit or donkey Cy3-conjugated anti-mouse antibodies (1:1000) were used. For studies on GFP-PKC␦ localization, LNCaP cells were transfected with pEGFP-PKC␦ using the Amaxa Nucleofector and treated 48 h later with PMA. In all cases, cover slides were mounted with Fluoromount-G (SouthernBiotech, Birmingham, AL) and visualized with a laser-scanning fluorescence microscope (LSM 410 or 510; Carl Zeiss). For quantification of co-localization, images (RGB) from red and green channels were converted into 8-bit images in ImageJ software (National Institutes of Health, Bethesda, MD), and Pearson's r (Rr) was calculated using this software as described previously (14).
Time-lapse Microscopy-LNCaP cells were transfected with pEGFP-PKC␦ using the Amaxa Nucleofector and seeded into collagen-coated, glass-bottomed culture dishes (MatTek, Ashland, MA) for 48 h. Cells were cultured for 20 h in phenol redfree RPMI 1640 medium containing 10% FBS and 5 mM HEPES, and monitored under a fluorescence microscope (Nikon Eclipse TE2000U) at a 488-nm excitation wavelength with a 515-nm long pass barrier filter.
Statistical Analysis-Data were analyzed using the Student's t test. p Ͻ 0.05 was considered to be statistically significant.

RESULTS
p23 Is a PKC␦-interacting Protein-In previous studies, we showed that the perinuclear protein p23 associates with the Rac-GAP ␣and ␤-chimaerin as well as with PKC⑀, an interaction that is mediated through their C1 domains (10,14). The C1 domains in PKC␦, a kinase widely implicated in cell growth arrest and apoptosis, play essential roles in intracellular targeting. Due to the relevance of PKC␦ in normal physiology and disease, it is highly important to understand the mechanisms that control its activity. We reasoned that PKC␦ interacts with p23 through the C1 domain region. To test this hypothesis, we used a yeast two-hybrid system. A pLexA construct comprising both C1 domains (C1a and C1b) from PKC␦ was generated (pLexA-C1␦ab, amino acids (aa) 123-297) (Fig. 1A, upper panel). This plasmid was co-transformed with pB42AD-p23 (HA-tagged, aa 108 -208, a fragment that interacts with the C1 domain in chimaerins) into EGY48 yeast containing p8OP-LacZ vector (10). As a positive control, we used pLexA-C1␤2-ch (aa 179 -281), which encodes the C1 domain in ␤2-chimaerin. As a negative control, we used a similar construct encompassing the PKC C1 domain (pLexA-C1, aa 95-197), which fails to interact with p23 (14). The triple vector co-transformants were selected by plating the yeast into SD/ϪUra/ϪHis/ϪTrp dropout plates. As expected, expression of pLexA-fused proteins was observed both in galactose/raffinose (ϩGal/Raf) plates and in glucose plates (ϪGal/Raf), whereas HA-tagged pB42AD-p23 (aa 108 -208) was only expressed in ϩGal/Raf induction plates (Fig. 1A, lower panel). Similar to C1␤2-ch, C1␦ab strongly interacts with p23, as revealed by induction of the LacZ reporter (blue). On the other hand, C1 does not interact with p23 (Fig. 1A, middle panel). As expected, no interaction was observed in ϪGal/Raf (glucose) plates. To determine which C1 domain in PKC␦ associates with p23, we generated pLexA constructs for C1a or C1b (aa 159 -208 and 231-280, respectively). We found that the C1b domain strongly interacts with p23 in a yeast two-hybrid assay. Unfortunately, the C1a construct transactivated the LacZ promoter gene in ϪGal/Raf (glucose) plates, which precluded an analysis using this approach (supplemental Fig. S1).
To further establish a role of the PKC␦ C1b domain in the interaction with p23, we generated GFP-fused constructs encoding C1␦ab or isolated C1 domains from PKC␦ (C1␦a or C1␦b) using the vector pEGFP. These plasmids were co-transfected into HeLa cells together with pcDNA3-V5-p23, a plasmid encoding V5-tagged p23, and localization determined by confocal microscopy. As shown in Fig. 1B, GFP-C1␦ab and GFP-C1␦b co-localized with p23 in the perinuclear region, as judged by the yellow color observed in the overlapped images. On the other hand, GFP-C1␦a was diffusely expressed in the cytoplasm and failed to co-localize with p23. Quantification of co-localization and Pearson's r (Rr) by ImageJ software confirmed these results.
Next, we investigated whether p23 associates with full-length PKC␦. COS-1 cells were transfected with pEBG control vector (which encodes GST alone) or pEBG-p23 (10) and subsequently infected with an adenovirus for full-length PKC␦ (multiplicity of infection, 3 pfu/cell). After 24 h cells were subject to GST pulldown using glutathione-Sepharose 4B beads. As shown in Fig. 1C, PKC␦ was detected in complex with GST-p23 but not with GST, suggesting that PKC␦ associates with p23 in cells. As a second approach, HeLa cells were transfected with GFP-PKC␦ full-length together with pcDNA3-V5-p23. Fig. 1D shows that GFP-PKC␦ co-localized with p23 as judged by the yellow color observed in the overlapped image as well as by quantification of co-localization and Pearson's r (Rr ϭ 0.406) using ImageJ software.
Asp 245 and Met 266 in PKC␦ C1b Are Critical for Interaction with p23-Schultz et al. (27) showed that Asp 257 and Met 278 in the PKC⑀ C1b domain (amino acids 15 and 36 in the C1 motif, respectively) are implicated in the perinuclear localization of PKC⑀ or its isolated C1b domain in neuroblastoma cells. Our previous study (14) showed that an acidic amino acid in position 15 and a hydrophobic amino acid in position 36 in the ␤2-chimaerin C1 domain were signatures for its association with p23. We speculated that similar positions in the PKC␦ C1b domain (Asp 245 and Met 266 , Fig. 2A) might be crucial for the interaction with p23. To test this hypothesis, we generated the double mutant C1␦ab (D245G/M266G) and inserted it into the pLexA vector for yeast two-hybrid analysis. pLexA-C1␦ab (WT) or pLexA-D245G/M266G-C1␦ab were co-transformed with pB42AD-p23 (HA-tagged, aa 108 -208) into EGY48 yeast containing p8OP-LacZ vector, and the triple vectors co-transformants were selected by plating the yeast into SD/ϪUra/ ϪHis/ϪTrp dropout plates. C1⑀ab and C1 were used as positive and negative controls, respectively (14). As shown in Fig.  2B, only C1␦ab (WT) but not D245G/M266G-C1␦ab interacted with p23, as revealed by the induction of the LacZ reporter (blue). , and ␤2-chimaerin (C1␤2-ch) fused to pLexA. Middle panel, EGY48 yeast (containing the p8OP-LacZ vector) was co-transformed with pLexA-C1␦ab, pLexA-C1, or pLexA-C1␤2-ch, together with pB42AD-HA-tagged p23 (aa 108 -208). ␤-Galactosidase activity was determined after induction with galactosidase/raffinose (ϩGal/Raf) or no induction (ϪGal/Raf) 72 h after transformation. Lower panel, expression of pLexA-fused C1 domains, as determined by Western blot using an anti-pLexA antibody and pB42AD-HA-p23 (aa 108 -208) using an anti-HA antibody. B, the C1b domain of PKC␦ co-localizes with p23 at the perinuclear region. HeLa cells were co-transfected with pEGFP-fused C1␦ab, C1␦a, or C1␦b, and pcDNA3.1-V5-p23. After 48 h, cells were fixed and visualized by confocal microscopy. Similar results were observed in three independent experiments. Co-localization and Pearson's r (Rr) were generated by ImageJ software. Scale bar, 10 m. C, full-length PKC␦ complexes with p23. COS-1 cells were transfected with either pEBG (empty vector) or pEBG-p23. Twenty-four h later, cells were infected with a PKC␦ adenovirus (multiplicity of infection, 3 pfu/cell). After 24 h, GST or GST-p23 were precipitated with glutathione-Sepharose 4B beads, and PKC␦ was detected by Western blot using an anti-PKC␦ antibody. Left panel, representative experiment. Right panel, densitometric analysis of three independent experiments, expressed as fold-change relative to GST. **, p Ͻ 0.01 versus GST. D, full-length PKC␦ co-localizes with p23. HeLa cells were co-transfected with pEGFP-PKC␦ and pcDNA3.1-V5-p23. After 48 h, cells were fixed and visualized by confocal microscopy. Similar results were observed in three independent experiments. Co-localization and Pearson's r (Rr) were generated by ImageJ software. Scale bar, 10 m.

p23/Tmp21 Association with PKC␦
Next, we examined whether mutations in positions 15 and 36 in the PKC␦ C1b domain could alter the interaction with p23 in mammalian cells. COS-1 cells were co-transfected with pEBG or pEBG-p23 together with GFP-PKC␦ (WT) or GFP-PKC␦ (D245G/M266G) and subjected 24 h later to a GST pulldown assay using glutathione-Sepharose 4B beads. As shown in Fig.  2C, the mutant D245G/M266G-GFP-PKC␦ showed a reduced association with p23 compared with WT GFP-PKC␦. Altogether, these results established that Asp 245 and Met 266 in the PKC␦ C1b domain are critical for the interaction with p23.
p23 Depletion Potentiates PKC␦-mediated Apoptosis in LNCaP Cells-To determine any potential functional implication of the PKC␦-p23 association, we turned to the LNCaP prostate cancer model. Previous studies from our laboratory established that phorbol esters such as PMA triggered an apoptotic response in LNCaP cells through PKC␦. Silencing PKC␦ from LNCaP cells using RNAi essentially abolishes PMA-induced apoptosis in these cells (21)(22)(23).
First, we wished to establish whether the interaction between PKC␦ and p23 occurs in LNCaP cells. An anti-PKC␦ antibody was used to immunoprecipitate "endogenous" PKC␦ from LNCaP cells, and we examined whether endogenous p23 can be detected in the immunoprecipitate by Western blot. As shown in Fig. 3A and the corresponding densitometric analysis in Fig.  3B, p23 can be detected readily in immunoprecipitates using an anti-PKC␦ antibody but not with control IgG. Furthermore, confocal microscopy revealed that endogenous p23 co-localizes with both endogenous PKC␦ (Fig. 3C) or with pEGFP-PKC␦ (supplemental Fig. S2) in the perinuclear region of LNCaP cells, as judged by the yellow color in the overlapped images as well as by quantification of co-localization and Pearson's r (Rr ϭ 0.45) using ImageJ software (Fig. 3C). Thus, PKC␦ interacts with p23 in LNCaP prostate cancer cells.
To determine a potential role of p23 in PKC␦-mediated apoptosis, we silenced p23 from LNCaP cells using RNAi. Transient delivery of a p23 dsRNA duplex into LNCaP cells reduced p23 expression by ϳ90%. p23 RNAi depletion caused a remarkable potentiation of PMA-induced apoptosis in LNCaP cells. p23 depletion by itself did not cause any significant apoptosis (Fig. 3D). To further confirm this effect and also to minimize the chances of "off-target" effects of RNAi, we generated two LNCaP cell lines in which p23 was stably depleted using shRNA lentiviruses. Depletions of ϳ58 and ϳ80% were achieved using p23 shRNAs 1 and 2, respectively. In agreement with results observed using transient RNAi, a markedly higher PMA apoptotic response was observed in the p23 stable knockdown cell lines relative to control cells. In the absence of PMA treatment, no apoptosis could be observed (Fig. 3E). Thus, p23 negatively modulates PMA-induced apoptosis in LNCaP prostate cancer cells.
Our laboratory recently identified the RhoA/ROCK and JNK pathways as mediators of PMA-induced apoptosis in prostate cancer cells. PMA activates ROCK and JNK in a PKC␦-dependent manner, and inhibition of either pathway impaired the apoptotic response of the phorbol ester (22). Notably, we found a significant potentiation of JNK activation by PMA in p23- depleted LNCaP cells compared with control cells, as determined using an anti-phospho-JNK antibody. Likewise, phosphorylation of the ROCK substrate myosin phosphatase target subunit 1 was markedly elevated in LNCaP subject to p23 depletion relative to control cells (Fig. 3F). Densitometric analyses of these Western blots are shown in Fig. 3G. Taken

p23/Tmp21 Association with PKC␦
together, our data suggest that p23 depletion enhances PKC␦ responses induced by PMA stimulation.
As PKC␦ is the mediator of PMA-induced apoptosis in LNCaP cells (18,20,21), we speculated that the potentiating effect of p23 depletion on PMA-induced apoptosis should be mediated by PKC␦. To test this hypothesis, we first assessed the effect of the "pan" PKC inhibitor GF 109203X on PMA-induced apoptosis in p23 knockdown LNCaP prostate cancer cells. As shown in Fig. 4A, GF 109203X essentially abolished apoptosis induced by PMA in LNCaP cells subjected to either p23 or control RNAi. Moreover, silencing PKC␦ in LNCaP cells subjected to either p23 or control RNAi impaired the death response of PMA. The inhibitory effect was proportional to the level of PKC␦ depletion achieved because the PMA apoptotic effect was negligible in LNCaP cells transiently transfected with the PKC␦ dsRNA1, which reduced PKC␦ expression by ϳ82%, but it was partial in LNCaP cells transfected with PKC␦ dsRNA2, which only reduced PKC␦ expression by ϳ69% (Fig.  4B). Altogether, these results suggest that the potentiating effect of p23 depletion on PMA-induced apoptosis is mediated through PKC␦.
Asp 245 /Met 266 -PKC␦ Mutant Induces Apoptosis in LNCaP Cells-One attractive possibility is that p23 serves as an anchoring protein for PKC␦ in LNCaP cells. p23 may limit PKC␦ effects by sequestering the kinase to the perinuclear region of the cell. A similar scenario was described in our previous studies for the Rac-GAP ␤2-chimaerin (10,14). To begin addressing this possibility, we used the double mutant D245G/M266G-PKC␦, which is unable to bind to p23 (see Fig. 2B). LNCaP cells were transfected with plasmids encoding either GFP-fused WT-PKC␦ or D245G/M266G-PKC␦. Seventy-two h after transfection, the "green" and the "non-green" cell populations were subjected to flow cytometry analysis. Interestingly, we observed a significant population of LNCaP cells expressing the double mutant GFP-D245G/M266G-PKC␦ in sub-G 0 /G 1 (28%), characteristic of apoptosis, which was very low in cells expressing WT GFP-PKC␦ (ϳ3% apoptosis). The percentage of cells in sub-G 0 /G 1 in the "non-transfected" population was negligible (ϳ1%) (Fig. 5). Thus, loss of binding to p23 enhances PKC␦-mediated apoptosis.
Silencing p23 Leads to Enhanced Translocation of PKC␦ to Plasma Membrane-In a recent study, we determined that the plasma membrane pool of PKC␦ was responsible for conferring the apoptotic effect of PMA (23). We decided to examine whether p23 influences PKC␦ translocation to the plasma membrane. We speculated that PKC␦ peripheral translocation is enhanced by depletion of p23. p23-depleted and control LNCaP cells were transfected with a plasmid encoding GFP-  MAY 6, 2011 • VOLUME 286 • NUMBER 18 PKC␦ (WT) and treated with PMA. PKC␦ localization was monitored by time-lapse video microscopy. Using high concentrations of PMA (100 -300 nM), translocation of GFP-PKC␦ could be observed readily in LNCaP cells, and therefore, under these experimental conditions, it was very hard to pursue a comparative study between control and p23-depleted cells, although translocation of PKC␦ was detected at earlier time points in LNCaP cells subjected to p23 RNAi (data not shown). Therefore, to detect a potential enhancement of PKC␦ translocation by p23 depletion, we used a lower PMA concentration. At 30 nM PMA, PKC␦ translocation in control LNCaP cells was barely visible. On the other hand, plasma membrane translocation of GFP-PKC␦ was detected readily in p23-depleted LNCaP cell lines 5-10 min after PMA treatment under the same experimental condition (Fig. 6). These results support the concept that the association of PKC␦ to p23 is an important regulatory mechanism for controlling PKC␦ activation.

p23/Tmp21 Association with PKC␦
p23 Regulates Apoptosis Induced by Doxorubicin-PKC␦ not only mediates the apoptotic effect of phorbol esters but also death triggered by a number of stimuli, including ceramide and DNA-damaging agents (25,26,28). To examine whether the effect of p23 depletion on LNCaP cell apoptosis is not only restricted to PMA, we analyzed the response of the chemotherapeutic agent doxorubicin. As shown in Fig. 7, depletion of PKC␦ reduced by ϳ50% doxorubicin-induced LNCaP cell apoptosis, arguing for the involvement of PKC␦ in this effect. Notably, the number of apoptotic cells observed in response to doxorubicin treatment was essentially doubled in p23-depleted LNCaP cells. This result reinforces the concept that p23 is a negative modulator of PKC␦-mediated apoptosis.

DISCUSSION
Here, we report that the perinuclear protein p23, also known as Tmp21, interacts with PKC␦, a member of the serine-threo-

p23/Tmp21 Association with PKC␦
nine kinase PKC family. This interaction is mediated by the second C1 domain (C1b) of PKC␦. C1 domains originally have been characterized as lipid-binding motifs that drive the relocalization of PKC isozymes, chimaerins, and other signaling molecules to membranes via binding to the second messenger DAG and phospholipids (11, 12, 29 -31). Association with membranes represents a key step in the allosteric activation of PKCs by lipids. At a structural level, the C1 domains are remarkably similar, although only a subset of C1 domains has the ability to bind DAG and DAG mimetics such as the phorbol esters (32)(33)(34). Regardless of their resemblance, a distinctive property of C1 domains is their ability to localize to different intracellular compartments when expressed in cells. Isolated C1a domains of PKC isozymes are generally located diffusely when expressed in mammalian cells, whereas isolated C1b domains preferentially localize at the perinuclear region (14,35,36). The single C1 domain present in chimaerins and Ras-GRPs also show a characteristic localization in the perinucleus (10,13,14,(37)(38)(39). Although the distinctive localization of individual C1 domains may relate to unique interactions with membrane lipids, there is growing evidence for a potential involvement of accessory proteins in C1 domain targeting. For example, the C1b domain of PKC␣ interacts with fascin to drive cell motility (40), the C1b domain in PKC⑀ binds to peripherin and induces its aggregation in cells (41), and the C1b domain of PKC␥ associates with 14-3-3⑀ (42). C1a domains from PKCs also interact with proteins, including pericentrin (43) and an E3 ubiquitin ligase (44). Thus, protein interactions through C1 domains may contribute to the differential compartmentalization of PKC isozymes.
Previous reports from our laboratory identified p23 as an interacting protein for C1 domains. In an early study (10), we found that the C1 domain in ␣and ␤chimaerins associates with p23 in the perinuclear region. Deletion of the C1 domain from ␤2-chimaerin abolished this interaction, and p23 RNAi depletion prevented ␤2-chimaerin relocalization in response to stimuli (10,14). The present study clearly shows that the C1 domain region in PKC␦ associates with p23. The C1 domain region of PKC⑀ also interacts with p23, and this perinuclear association is lost upon activation of PKC⑀ (14). Larsen and co-workers (27) originally postulated that an acidic amino acid in position 15 and a hydrophobic amino acid in position 36 in the C1 domain consensus were key determinants for perinu-    (14) or in positions Asp 245 and Met 266 in PKC␦ (this study) greatly reduces association with p23. Mutation of equivalent residues in ␤2-chimaerin C1 domain (Glu 227 and Leu 248 ) also affects its perinuclear targeting via p23. From structural data, it can be inferred that residues in positions 15 and 36 in C1b domains are not inserted into the membrane bilayer, and therefore, they should be accessible for interacting with other partners (27). C1a domains in PKC␦ and PKC⑀ do not fulfill the requirements for binding to p23 (27). Altogether, our studies strongly support a role for p23 in targeting proteins with C1 domains to the perinucleus and argue for an exquisite selectivity for these interactions.
It is quite striking that silencing p23 expression has a significant impact on responses mediated by PKC␦. Evidence from several laboratories established that PKC␦ inhibits cell proliferation or kills cells upon activation through an apoptotic mechanism. Previous studies from our laboratory established that PKC␦ mediates apoptosis induced by phorbol esters or chemotherapeutic drugs in prostate cancer cells such as the LNCaP model (18,24). The ability of PMA to confer an apoptotic response in LNCaP cells depends on the relocalization of PKC␦ to the plasma membrane. This step seems to be essential to promote the release of factors required for the killing effect of the phorbol ester (18,21,23,24). Bryostatin 1, a C1 domain ligand that is unable to translocate PKC␦ to the plasma membrane but still translocates PKC␣ and PKC⑀ to the cell periphery, fails to provoke an apoptotic response in LNCaP cells. In addition, preventing the access of PKC␦ to the plasma membrane impairs the apoptotic effect of PMA. Moreover, expression of a myristoylated, targeted PKC␦, which specifically localizes at the plasma membrane, is sufficient to trigger apoptosis in LNCaP cells (23). We hypothesize that perinuclear anchoring of PKC␦ through p23 serves as a mechanism that restricts the availability of this kinase, therefore limiting the pool available for accessing the plasma membrane. In support of this concept, an early study established that the magnitude of the PMA apoptotic response in LNCaP cells is proportional to the amount of PKC␦ expressed in the cells, as PKC␦ overexpression markedly potentiates apoptosis, and reduction in PKC␦ cellular levels diminishes the apoptotic effect of the phorbol ester (18,21,23). One would expect that silencing p23, which does not change total levels of cellular PKC␦, would lead to elevated PKC␦ available for plasma membrane translocation. A similar scenario may be possible for the apoptotic effect of chemotherapeutic drugs, with the difference that in this case, the effect depends on the nuclear translocation of PKC␦ (25). Consistent with this notion, we found that translocation of PKC␦ to the periphery of LNCaP cells is more pronounced when p23 has been silenced. A similar paradigm has been described previously for ␤2-chimaerin. In this later case, depletion of p23 leads to enhanced peripheral translocation of this Rac-GAP, elevated Rac-GAP activity, and reduced cellular Rac-GTP levels (14). What becomes obvious from all of these studies is that different intracellular pools exist for PKC␦ and possibly for other DAG/phorbol ester receptors and that the "activatable" pool of the kinase may be limited by its association with anchoring proteins.
Our studies also highlight the diversity of functions regulated by p23. This member of the p24 family has been implicated in the retention and retrieval of cargo proteins in the secretory pathway. Additional roles have been assigned to p23, including recruiting the small GTPase ADP-ribosylation factor 1 to the Golgi, assembling actin in the Golgi, and regulating ␥-secretase activity (7,9). Our studies argue that p23 plays additional roles as a modulator of vital signaling pathways through its association with C1 domain-containing proteins. It remains to be determined whether perinuclear PKC␦ plays any significant functional role. Conceivably, p23 may facilitate the access of PKC␦ to perinuclear substrates and effectors. Indeed, PKC␦ (but not PKC␣) regulates endosome-to-Golgi transport (46). PKC␦ translocates to the Golgi in a C1b-dependent manner (47)(48)(49), and Golgi proteins have been established as PKC substrates (46,48). PKC␦ also has been implicated in endoplasmic reticulum stress (50,51). PKC␦ becomes associated to a Golgi compartment in the G 2 /M phase of the cell cycle, suggesting a role for perinuclear PKC␦ in the phosphorylation of mitotic kinases and growth arrest (52)(53)(54). Interestingly, perinuclear targeting of PKC␦ prevents apoptosis (23,45), which is consistent with the proapoptotic role of this kinase only when it is disassociated from p23.
In summary, our results identified p23 as an anchoring protein for PKC␦. This association occurs via the C1b domain in PKC␦ and influences the availability of the kinase to trigger cellular responses, strongly supporting the concept that this domain plays a key role in driving the perinuclear localization of signaling proteins. Our studies also highlight the concept that C1 domains do not act only as lipid-binding motifs, a role for which they have been known for many years, but also play important roles in protein-protein interactions.