Phorbol ester-induced apoptosis of C4-2 cells requires both a unique and a redundant protein kinase C signaling pathway.

Phorbol 12-myristate 13-acetate (PMA) potently induces apoptosis of LNCaP human prostate cancer cells. Here, we show that C4-2 cells, androgen-hypersensitive derivatives of LNCaP cells, also are sensitive to PMA-induced apoptosis. Previous reports have implicated activation of protein kinase C (PKC) isozymes alpha and delta in PMA-induced LNCaP apoptosis using overexpression, pharmacological inhibitors, and dominant-negative constructs, but have left unresolved if other isozymes are involved, if there are separate requirements for individual PKC isozymes, or if there is redundancy. We have resolved these questions in C4-2 cells using stable expression of short hairpin RNAs to knock down expression of specific PKC isozymes individually and in pairs. Partial knockdown of PKCdelta inhibited PMA-induced C4-2 cell death almost completely, whereas near-complete knockdown of PKCalpha had no effect. Knockdown of PKCepsilon alone had no effect, but simultaneous knockdown of both PKCalpha and PKCepsilon in C4-2 cells that continued to express normal levels of PKCdelta inhibited PMA-induced apoptosis. Thus, our data indicate that there is an absolute requirement for PKCdelta in PMA-induced C4-2 apoptosis but that the functions of PKCalpha and PKCepsilon in apoptosis induction are redundant, such that either one (but not both) is required. Investigation of PMA-induced events required for LNCaP and C4-2 apoptosis revealed that p38 activation is dependent on PKCdelta, whereas induction of retinoblastoma protein hypophosphorylation requires both PKC signaling pathways and is downstream of p38 activation in the PKCdelta pathway.

First, LNCaP cells and fibroblasts from an osteosarcoma were co-injected into a male nude mouse, followed by castration and excision of a tumor 8 and 12 weeks later, respectively, and the C4 cell line was derived from in vitro culture of the tumor. Co-injection of C4 cells and osteosarcoma fibroblasts into a castrated mouse, removal of a tumor 12 weeks later, and culture of the tumor yielded the C4-2 line. C4-2 cells grow faster than LNCaP cells in culture and, in contrast to LNCaP cells, form moderately metastatic tumors in castrated mice when injected without any bone or other matrix (6). Recent evidence has shown that C4-2 cells exhibit increased stabilization and nuclear localization of the androgen receptor relative to LNCaP cells and are hypersensitive to growth stimulation by androgens (7). Thus, C4-2 cells approximate the result of progression of most human prostate tumors after androgen withdrawal. Here, we report that C4-2 cells have retained sensitivity to PMA-induced apoptosis similar to LNCaP cells, and we have clarified the PKC isozyme signaling pathways controlling the process.
PMA activates at least 26 proteins of seven classes in human cells by mimicking the endogenous activator diacylglycerol, including four classical (␣, ␤I, ␤II, and ␥) and four novel (␦, ⑀, , and ) PKC isozymes, three protein kinase D (PKD) isozymes (PKC (PKD), PKD2, and PKC), five calcium-and diacylglycerolactivated guanine nucleotide exchange factors, four chimaerins, three Munc13 isozymes, two myotonic dystrophy kinase-related Cdc42-binding kinases (MRCKs), and diacylglycerol kinase-␥ (8,9). Of these, LNCaP and C4-2 cells express PKC␣, PKC␦, PKC⑀, and PKC (3,10); all three PKD isozymes; Munc13-2; and MRCK␤. 2 PMA-induced apoptosis of LNCaP cells is accompanied by prolonged translocation of PKC␣, PKC␦, and PKC⑀ to nonnuclear membranes (10). Previous reports from our (3) and other (2,4,11,12) laboratories have implicated PKC␣ and PKC␦ in PMA-induced LNCaP apoptosis, but the questionable isozyme specificities of the techniques used in those studies have left ambiguities about specific PKC isozyme requirements. Sequencespecific targeting of mRNAs by vector-directed expression of short hairpin RNAs (shRNAs) is a highly promising approach for stable inhibition of expression of specific proteins. We have used this approach to generate clones of C4-2 cells with PKC␣, PKC␦, and PKC⑀ knocked down individually and in pairs and have used these clones to clarify the PKC isozyme requirements of PMAinduced C4-2 cell death. Our data show a requirement for PKC␦ and a redundant pathway that can be effected by either PKC␣ or PKC⑀, a more complex combination of isozymes than recognized previously.
PMA-induced apoptosis of LNCaP and C4-2 cells is preceded by activation of the three major families of mitogen-activated protein kinases (MAPKs), p38 (12), extracellular signal-regulated kinase (ERK)-1 and ERK2 (3,12), and c-Jun N-terminal kinase (JNK)-1 and JNK2 (12,13), and by induction of p21 WAF1/CIP1 and hypophosphorylation of the retinoblastoma protein (Rb) (14). Dephosphorylation of Akt (12), truncation of E-cadherin (15), and induction of ceramide synthase activity (16) by PMA have been described in LNCaP cells. All of these events, except activation of ERKs and possibly JNKs, appear to be necessary for PMA-induced apoptosis to occur (12)(13)(14)(15). Examination of some of these events during PMA treatment of our PKC knockdown clones revealed that p38 activation requires PKC␦, whereas Rb hypophosphorylation requires both PKC␦ and either PKC␣ or PKC⑀. Furthermore, p38 activation is required for PMA-induced Rb hypophosphorylation, indicating that PKC␦ acts through a separate signaling pathway than PKC␣ and PKC⑀ to activate Rb. Our results show that PMAinduced apoptosis of C4-2 cells requires a more complex combination of PKC isozymes than recognized previously and that different PKC signaling pathways control different events required for apoptosis.

EXPERIMENTAL PROCEDURES
shRNA Expression Constructs-The shRNA expression vector pSHAG-1 (17), a derivative of pENTR/D-TOPO (Invitrogen) containing a human U6 small nuclear RNA promoter upstream of BseRI and BamHI cloning sites, was a gift from Dr. G. J. Hannon (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY). For selection of stably transfected clones, we inserted a neomycin or blasticidin resistance cassette under the control of the herpes simplex virus thymidine kinase promoter and polyadenylation signals into the EcoRV site of pSHAG-1 to yield pSHAG-1neo and pSHAG-1bla. Double-stranded oligodeoxynucleotides encoding shRNAs targeting human PKC␣, PKC␦, and PKC⑀ were designed using the online program shRNA Retriever, developed by R. Sachidanandam and J. Faith (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY), and following suggestions of the Tuschl laboratory (29). All had the following structure: mRNA target ϭ N 1-27 C 28 ; oligonucleotide A ϭ (reverse complement of N 1-27 )-GAAGCTTGN 1-27 C 28 , with every third A or C of the sense N 1-27 converted to G or T, respectively, up to a maximum of four base changes; and oligonucleotide B ϭ exact reverse complement of the entire oligonucleotide A, bounded by GATC (BamHI overhang) at its 5Ј-end and CG (BseRI overhang) at its 3Ј-end. Each target sequence was compared with all other sequences in the Celera Human Genome Database, and any with Ͼ10 continuous matching nucleotides or fewer than four mismatches relative to any mRNA sequence in the data base were rejected. Negative control shRNAs with 4 mismatched bases relative to the target were designed for PKC␣ and PKC␦ and rechecked for homologies to other sequences in the Celera Database.
5Ј-Phosphorylated polyacrylamide gel-purified oligodeoxynucleotides were purchased from Integrated DNA Technologies (Coralville, IA). Equimolar amounts of oligonucleotides A and B were annealed and ligated to BseRI-BamHI-double-digested calf intestinal alkaline phosphatase-treated and agarose gel-purified pSHAG-1neo or pSHAG-1bla. Ligation mixtures were transformed into TOP10 competent cells (Invitrogen), and plasmid DNA was isolated with a Wizard Plus SV miniprep kit (Promega, Madison, WI). Inserts were checked by restriction digestion and sequenced at the Utah State University Biotechnology Center. To increase knockdown efficacy in stably transfected clones, two shRNA expression cassettes targeting different regions of the same mRNA were inserted into one pSHAG vector. This was done by excising the U6 shRNA promoter expression cassette as a NotI-AscI fragment from one shRNA expression construct, filling in the overhangs with Klenow fragment, and inserting it into the HpaI site of another shRNA expression construct (Fig. 1). Only constructs with both expression cassettes in the same orientation were used. In each PKC mRNA, the two target sequences chosen for the shRNA expression constructs used for selection of stable clones were as follows: PKC␣ target 1, 5Ј-TGAT-TCAGGATGATGACGTGGAGTGCAC; PKC␣ target 2, 5Ј-TGATTCAG-GATGATGACGTGGAGTGCAC; PKC␦ target 1, 5Ј-CATCAAGAACCA-TGAGTTTATCGCCACC; PKC␦ target 2, 5Ј-ATGCATCGACAAGATC-ATCGGCAGATGC; PKC⑀ target 1, 5Ј-CCTACATTGCCCTCAATGTG-GACGACTC; and PKC⑀ target 2, 5Ј-CACTTCGAGGACTGGATTGAT-CTGGAGC. For overexpression of wild-type PKC⑀, a cDNA containing the complete coding sequence of human PKC⑀ was obtained from American Type Culture Collection (Manassas, VA) and subcloned into the expression vector pIREShyg. A point mutation at nucleotide 102 of the coding region of that cDNA was reverted to the wild type G, using a QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA).
Cell Culture and Transfections-LNCaP and C4-2 cells were obtained from American Type Culture Collection and UroCor Labs (Oklahoma City, OK), respectively. Cells were maintained as monolayer cultures in RPMI 1640 medium supplemented with 10% fetal bovine serum and 2 mM L-glutamine (complete RPMI 1640 medium) in a humidified atmosphere of 5% CO 2 at 37°C. shRNA expression constructs were transfected into C4-2 cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Cells were incubated with plasmid DNA and Lipofectamine 2000 in complete RPMI 1640 medium for 18 h; the medium was replaced with fresh complete RPMI 1640 medium; and 12-18 h later, cells were replated in 100-mm dishes. 24 -36 h after replating, selection was initiated by adding 500 g/ml Geneticin or 10 g/ml blasticidin. After ϳ2 weeks of selection, individual colonies were isolated and expanded; PKC isozyme levels were determined by immunoblot analyses; and aliquots of selected clones were promptly frozen in liquid nitrogen. Human wild-type PKC⑀ in pIREShyg or the pIREShyg vector alone was transfected into C4-2 clone ␣ϩ⑀1, with PKC␣ and PKC⑀ knocked down. Selection of these transfectants in increasing concentrations of hygromycin (100 -600 units/ml) yielded pools of cells used in the experiments of Fig. 6. In all cases, cells subjected to multiple transfections and drug selections were kept in culture for the minimal time to obtain stable clones for analysis and were otherwise kept frozen in liquid nitrogen.
Immunoblot Analyses-All cell lysis buffers contained a protease inhibitor mixture (Roche Applied Science) and the phosphatase inhibitors sodium fluoride (50 mM), activated sodium orthovanadate (0.2 mM), and p-nitrophenyl phosphate (10 mM). For immunoblotting of PKC isozymes, poly(ADP-ribose) polymerase, and phosphorylated and total c-Jun and Rb, total cell lysates were prepared in radioimmune precipitation assay buffer (18). For analyses of MAPKs, cells were lysed by Dounce homogenization in hypotonic buffer (20 mM Tris-Cl (pH 7.5), 1 mM EGTA, and inhibitors) and centrifuged at 20,000 ϫ g for 10 min, and the supernatants were analyzed. Cytoplasmic and membrane fractions were prepared by Dounce homogenizing cells in buffer without detergent (10), pelleting nuclei and unbroken cells by centrifugation at 800 ϫ g, and then centrifuging post-nuclear supernatants at 100,000 ϫ g for 1 h. The supernatants (cytoplasm) were recovered, and the pellets were Dounce-homogenized in buffer containing 1% Triton X-100 and centrifuged at 20,000 ϫ g for 15 min to separate non-nuclear membranes (supernatants) from Triton-insoluble particles (pellets).
Cell Viability Assay-Viable cells were quantified by incubating the cells with the tetrazolium compound 3-(4,5-dimethylthiazol-2-yl)-5-(3-FIG. 1. Modified pSHAG-1 (17) containing two shRNA expression cassettes. ϩ1 indicates the start of transcription from the U6 promoters. U6 prom, human U6 small nuclear RNA promoter; Neo and Bla, neomycin and blasticidin resistance cassettes, respectively. carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium inner salt (MTS) (Promega) and measuring the absorbance of the resulting formazan compound. Cells were plated in flat-bottom 96-well tissue culture plates (8 wells/condition), and PMA (10 nM) or ethanol vehicle (final concentration of 0.006%) was added 24 h later. At the indicated times after drug addition, 0.2 volume of MTS was added; the cells were incubated at 37°C for 1 h; and the absorbance at 490 nm was measured in a Molecular Devices UV max kinetic microplate reader. Standard curves were constructed by counting viable LNCaP and C4-2 cells by trypan blue exclusion, assaying different numbers of viable cells with MTS, and plotting A 490 versus cell number.
Cell Death Enzyme-linked Immunosorbent Assay-Quantification of cytoplasmic histone-associated DNA fragments characteristic of cells undergoing apoptosis was performed with a cell death detection ELISA PLUS kit (Roche Applied Science). Cells (2 ϫ 10 4 /well) were plated in 96-well plates with complete RPMI 1640 medium, and PMA (10 nM) or Me 2 SO vehicle (0.006%) was added 24 h later. After 24 h of drug treatment, plates were centrifuged at 200 ϫ g for 10 min at room temperature to pellet detached cells. The medium was removed; cells were lysed; and the enzyme-linked immunosorbent assay was performed according to the manufacturer's instructions. Enrichment of histone-associated mono-and oligonucleosomes released into the cytoplasm was calculated as the absorbance at 405 nm of PMA-treated cells divided by the absorbance of vehicle-treated cells.
Other Reagents-PMA was purchased from LC Laboratories (Woburn, MA). The p38 inhibitor SB203580 and the JNK inhibitor SP600125 were purchased from Calbiochem; the MAPK/ERK kinase (MEK) inhibitor U0126 was from Promega.

Induction of C4-2 Cell
Death by PMA-Continuous treatment of C4-2 cells with several concentrations of PMA yielded substantial reductions in viability at doses of 1-100 nM PMA, with maximal death induction by 10 nM (Fig. 2A), similar to the optimal dose in LNCaP cells (1,19). A comparison of LNCaP and C4-2 cells revealed a similar time course of PMA induction of death in both cell lines (Fig. 2B). When numbers of PMAtreated cells were expressed as a percentage of vehicle-treated cells 72 h after addition of PMA, C4-2 cells (3.7%) were about as sensitive to 10 nM PMA as LNCaP cells (3.5%). We determined the effects on cell death of limiting the time of incubation of C4-2 cells with 10 nM PMA by removing the culture medium containing PMA at different times after its addition and replacing the medium with fresh medium without PMA. Fig. 2C shows that maximal induction of cell death by PMA required at least some signaling (either by PMA or by one or more factors released into the medium in response to PMA) that was prolonged for several hours.
Stable shRNA-directed Knockdown of PKC Isozymes-We determined that C4-2 cells express the same subset of PMAactivated proteins as LNCaP cells. As shown by RNase protection and immunoblot analyses, C4-2 cells express PMA-activated PKC␣, PKC␦, PKC⑀, and PKC, but not PKC␤I, PKC␤II, PKC␥, or PKC. LNCaP and C4-2 cells also express mRNAs for the three known PKD isozymes: PKD/PKC, PKC, and PKD2. Regarding other PMA-activated proteins, both cell lines express Munc13-2 and MRCK␤, but not Munc13-1, Munc13-3, or MRCK␣. We also detected very low amounts of calcium-and diacylglycerol-activated guanine nucleotide exchange factor-II by RNase protection, but were unable to detect this protein with available antibodies (data not shown).
Because previous data implicated PKC␣ and PKC␦ in PMAinduced LNCaP apoptosis, we initially targeted these isozymes for shRNA-directed knockdown in C4-2 cells. We also targeted PKC⑀ because we found that overexpression of PKC⑀ in LNCaP cells changed the effect of bryostatin-1 from inhibition of PMAinduced apoptosis to induction of apoptosis, 3  bryonic kidney 293 cells until several days after transfection and usually required two sequential transfections of the cells, 5 days apart. The two most potent shRNAs targeting PKC␣ were expressed from the same pSHAG-1neo vector and yielded greater knockdown of PKC␣ than either one alone as assessed by competitive reverse transcription-PCR (data not shown). The construct for expressing these two shRNAs was transfected into C4-2 cells. Because sequential transfections yielded significant toxicity and because transfection efficiency is usually low in C4-2 cells, stably transfected clones were selected. The same strategy was used to generate C4-2 clones stably transfected with pSHAG-1bla encoding two shRNAs targeting PKC␦ or PKC⑀. The levels of PKC␣, PKC␦, and PKC⑀ in the knockdown clones are shown in Fig. 3A. PKC␣ was knocked down to barely detectable levels (clones ␣1 and ␣2), whereas knockdowns of PKC␦ and PKC⑀ were less complete, but Ͼ60% (clones ␦1, ␦2, ⑀1, and ⑀2). Because we had observed previously that PMA induces higher expression of PKC␣, but not of PKC␦ or PKC⑀, in LNCaP cells (10), we also measured PMA-induced membrane translocation of PKC␣ in ␣1 cells. As shown in Fig.  3B, minimal PKC␣ was translocated to non-nuclear membranes by PMA in ␣1 cells relative to parental C4-2 cells. PKC␣ knockdown clone ␣1 was subsequently transfected with the PKC␦ or PKC⑀ shRNA expression constructs and selected with blasticidin to yield clones with PKC␣ and PKC␦ knocked down (clone ␣1ϩ␦) or with PKC␣ and PKC⑀ knocked down (clones ␣ϩ⑀1 and ␣ϩ⑀2) (Fig. 3A). Also, clone ␦2 was subsequently transfected with the PKC␣ shRNA expression construct and selected with Geneticin to yield clone ␦2ϩ␣, with PKC␣ and PKC␦ knocked down (Fig. 3A).
Recent reports have shown that even 21-nucleotide doublestranded RNAs can activate the interferon system and the double-stranded RNA-activated protein kinase PKR in certain cells, including up-regulation of numerous interferon-stimu- indicates untransfected C4-2 cells. Stable C4-2 transfectants were as follows: ␣m, a clone transfected with an shRNA construct containing four mismatched nucleotides relative to PKC␣ target 1 (see "Experimental Procedures"); ␣1 and ␣2, two independent clones, each stably transfected with pSHAG-1neo containing shRNA expression cassettes for PKC␣ targets 1 and 2; ␦m, a clone transfected with an shRNA construct containing four mismatched nucleotides relative to PKC␦ target 1; ␦1 and ␦2, two independent clones, each stably transfected with pSHAG-1bla containing shRNA expression cassettes for PKC␦ targets 1 and 2; ␣1ϩ␦, clone ␣1 further transfected with pSHAG-1bla containing shRNA expression cassettes for PKC␦ targets 1 and 2; ␦2ϩ␣, clone ␦2 further transfected with pSHAG-1neo containing shRNA expression cassettes for PKC␣ targets 1 and 2; ␣ϩ⑀1 and ␣ϩ⑀2, two independent subclones of clone ␣1, each further transfected with pSHAG-1bla containing shRNA expression cassettes for PKC⑀ targets 1 and 2; and ⑀1 and ⑀2, two independent clones, each stably transfected with pSHAG-1bla containing shRNA expression cassettes for PKC⑀ targets 1 and 2. B, PMA-induced membrane translocation of PKC␣ in PKC␣ knockdown clone ␣1. lated genes and activation of Jak-Stat signaling (20,21). Thus, it was important to determine whether expression in C4-2 cells of our shRNAs targeting PKCs induced an interferon response. The parental LNCaP cell line, from which C4-2 cells were derived, has been reported to be insensitive to interferon-mediated Stat activation and growth inhibition (22,23). Consistent with these reports, we found that stable expression of PKC shRNAs in C4-2 cells did not yield increased expression of p56, the most abundant protein induced by interferon in human cells, as assessed by immunoblot analyses. Also, microarray analyses conducted in the laboratory of Dr. B. R. G. Williams (Department of Cancer Biology, Lerner Research Institute, Cleveland Clinic Foundation) revealed no substantial changes in expression of 850 putative interferon-stimulated genes in our PKC shRNA-expressing clones relative to untransfected C4-2 cells (data not shown).
Effects of PKC Isozyme Knockdowns on PMA-induced C4-2 Cell Death-All of the knockdown clones grew in complete RPMI 1640 medium at similar rates compared with untransfected C4-2 cells (data not shown). PMA-induced C4-2 cell death was unaffected by near-complete knockdown of PKC␣ alone (Fig. 4A), but was inhibited markedly by partial knockdown of PKC␦ (Fig. 4B). Subsequent partial knockdown of PKC␦ in a clone with PKC␣ depleted (clone ␣1ϩ␦) yielded substantial inhibition of PMA-induced death (Fig. 4A), whereas subsequent knockdown of PKC␣ in a clone with PKC␦ knocked down (clone ␦2ϩ␣) did not yield additional resistance to PMA (Fig. 4B). Knockdown of PKC⑀ alone had no effect on PMAinduced C4-2 cell death (Fig. 4C), but subsequent knockdown of PKC⑀ in a clone with PKC␣ knocked down (clone ␣ϩ⑀1 or ␣ϩ⑀2) yielded moderate inhibition of PMA-induced death (Fig. 4C). Although the numbers of cells in clones ␣ϩ⑀1 and ␣ϩ⑀2 did not increase substantially over the 4 days of the experiment, observation of the plates revealed a mixture of healthy and dying cells on each day, suggesting that the resistant cells continued to proliferate. Together, these results indicate an absolute requirement for PKC␦, but not for PKC␣ or PKC⑀, in PMAinduced C4-2 cell death. However, partial inhibition of PMAinduced death by simultaneous partial inhibition of PKC␣ and PKC⑀ indicates a redundant pathway, such that at least one of these isozymes is required in addition to PKC␦.
We were unable to achieve near-complete knockdown of PKC␣ and PKC⑀ together. Substantial knockdown of PKC⑀ in clone ␣1, which previously exhibited near-complete depletion of PKC␣, was accompanied by partial recovery of PKC␣ expression (clone ␣ϩ⑀1) (Fig. 3). On the other hand, in clone ␣ϩ⑀2 (Fig. 3), which retained near-complete depletion of PKC␣, knockdown of PKC⑀ was only partial. Similar results were observed in three other ␣ϩ⑀ knockdown clones.
The effects of PKC isozyme depletions on induction by PMA of two markers of apoptosis were measured. PMA induced extensive accumulation of cytoplasmic histone-associated DNA fragments (Fig. 5A) and cleavage of poly(ADP-ribose) polymerase (Fig. 5B) in C4-2 cells. Both effects were inhibited by knockdown of PKC␦ or by simultaneous knockdown of PKC␣ and PKC⑀, but not by knockdown of PKC␣ or PKC⑀ alone, confirming that both the PKC␦ and PKC␣/⑀ pathways mediate PMA-induced apoptosis as well as loss of cell viability.
Overexpression of PKC⑀ in C4-2 cells with PKC␣ and PKC⑀ Knocked Down-Because PKC⑀ had not previously been impli- cated in PMA-induced apoptosis and because its overexpression has been reported to render LNCaP cells resistant to PMA (24), in contrast to our own observations, 3 it was important to determine whether exogenous expression of PKC⑀ could overcome the effects of the shRNAs targeting PKC␣ and PKC⑀. Clone ␣ϩ⑀1, with PKC␣ and PKC⑀ knocked down, was transfected further with human wild-type PKC⑀ in pIREShyg or with the pIREShyg vector alone, and pools of hygromycinresistant subclones were selected and labeled ␣ϩ⑀1⑀wt and ␣ϩ⑀1pIRES, respectively. We chose to express human PKC⑀ mRNA rather than another mammalian PKC⑀ mRNA that has mismatches relative to the PKC⑀ shRNAs to achieve replacement of PKC⑀ levels without excessive overexpression. As shown in Fig. 6, exogenous expression of PKC⑀ reverted the partial PMA resistance of clone ␣ϩ⑀1 back to PMA sensitivity similar to that of parental C4-2 cells. Northern blot analyses revealed that expression of the two PKC⑀ shRNAs was markedly lower in the cells overexpressing PKC⑀ than in clones ␣ϩ⑀1 and ␣ϩ⑀1pIRES, whereas expression of the two PKC␣ shRNAs remained high in all of these clones (data not shown). Possibly, the overexpressed PKC⑀ overtaxed and led to degradation of the shRNAs targeting it. Thus, although the reversion of ␣ϩ⑀1⑀wt cells to PMA sensitivity ruled out the possibility that clone ␣ϩ⑀1 developed resistance to PMA spontaneously, the possibility that an effect of the PKC⑀ shRNAs unrelated to knockdown of PKC⑀ contributed to the PMA resistance of clone ␣ϩ⑀1 could not be ruled out completely. However, the latter possibility is unlikely since the ␣ϩ⑀1⑀wt cells continued to express the two PKC␣ shRNAs, and expression of the PKC⑀ shRNAs alone in C4-2 cells had no effect on PMA sensitivity.
Effects of MAPK Inhibitors on PMA-induced C4-2 Cell Death-As in LNCaP cells (3,12,13), PMA activated MAPKs of the p38 (presumably the ␣ isozyme), JNK (JNK1 and JNK2), and ERK (ERK1 and ERK2) families in C4-2 cells as assessed by induction of dual Thr/Tyr phosphorylation (Fig. 7). We measured the effects of pharmacological inhibitors of p38, JNK, and ERK activation on PMA-induced C4-2 cell death. Under our conditions of pretreatment with inhibitor for 1 h, followed by incubation of C4-2 cells with inhibitor ϩ 10 nM PMA for 4 h, we found that p38 was required for PMA-induced cell death (Fig. 8A), in agreement with the findings of Tanaka et al. (12) in LNCaP cells. Inhibition of PMA activation of JNK1 and JNK2 by SP600125, measured by inhibition of PMA-induced phosphorylation of c-Jun, had a small but significant inhibitory effect on PMA-induced C4-2 cell death (Fig. 8B). (PMA-induced FIG. 7. PMA-induced activation of MAPKs. Cells were plated with complete RPMI 1640 medium in 6-well plates; PMA was added 24 h later; and lysates were prepared as described under "Experimental Procedures" at the indicated times after PMA addition or 1 h after addition of 0.006% Me 2 SO vehicle (0). Immunoblot analyses were performed as described under "Experimental Procedures." Phospho p38, Phospho JNK, and Phospho ERK indicate detection with antibodies against dual phosphorylated (Thr/Tyr) p38, JNKs, and ERKs, respectively.

FIG. 8. Effects of MAPK inhibitors on PMA-induced death.
Cells were plated with complete RPMI 1640 medium in 96-well plates (for MTS assays) or 6-well plates (for immunoblot analyses). MAPK inhibitors were added 24 h later, 30 min before addition of PMA (10 nM) or Me 2 SO vehicle (0.006%), and then both drugs were removed 4 h after addition of PMA. The numbers of viable cells, including adherent and detached cells, were counted 72 h after addition of PMA by MTS assay. For immunoblot analyses of phosphorylated and total c-Jun and ERKs, lysates were prepared as described under "Experimental Procedures" at the indicated times after PMA addition. A, p38 inhibitor SB203580. B, JNK inhibitor SP600125. p values were determined using an unpaired t test with Welch's correction. C, MEK inhibitor U0126. expression of c-Jun was not inhibited by SP600125.) A larger effect may have been masked by the inhibitory effect of SP600125 alone on C4-2 cell growth. Inhibition of PMA activation of ERK1 and ERK2 by the MEK inhibitor U0126 had no effect on PMA-induced C4-2 cell death (Fig. 8C).
PKC Isozyme Dependence of PMA-induced p38 Activation and Rb Hypophosphorylation-PMA-induced dual phosphorylation of p38 was inhibited markedly by knockdown of PKC␦, but was unaffected by simultaneous knockdown of PKC␣ and PKC⑀ or by knockdown of PKC␣ or PKC⑀ alone (Fig. 9A). Fig.  9B shows that PMA-induced hypophosphorylation of Rb was inhibited by knockdown of PKC␦ and partially by simultaneous knockdown of PKC␣ and PKC⑀, but was unaffected by knockdown of either PKC␣ or PKC⑀ alone. Thus, PMA-induced Rb activation requires both the PKC␦ and PKC␣/⑀ signaling pathways. Inhibition by SB203580 of PMA-induced p38 activation also prevented Rb hypophosphorylation (Fig. 9C), indicating that the latter requires p38 activation. Thus, PKC␦ signals through p38, whereas PKC␣ and PKC⑀ must signal through a separate pathway to activate Rb. These conclusions are summarized in Fig. 10.
Further evidence for separate PKC␦ and PKC␣/⑀ signaling pathways has been obtained from preliminary microarray analyses that revealed some PMA-induced changes in gene expression affected by knockdown of PKC␦ and others affected by simultaneous knockdown of PKC␣ and PKC⑀. One such gene is CYR61 (cysteine-rich 61), which encodes a protein shown to inhibit Taxol-induced apoptosis of breast can-cer cells (25). Induction of CYR61 expression by 10 nM PMA for 12 h was unaffected by knockdown of PKC␦, PKC␣, or PKC⑀ alone, but was 3.5-7-fold higher when both PKC␣ and PKC⑀ were knocked down. Thus, either PKC␣ or PKC⑀ can markedly attenuate PMA-induced expression of CYR61. PMA-induced E-cadherin expression and truncation, shown previously to be required for PMA-induced LNCaP apoptosis (26), was not affected by knockdown of PKC␦, PKC␣, PKC⑀, or PKC␣ ϩ PKC⑀ (data not shown), suggesting that these responses to PMA involve either additional redundancy among PKC isozymes or one or more non-PKC PMA receptors. DISCUSSION PMA-induced apoptosis of LNCaP androgen-sensitive human prostate cancer cells has been studied for several years, and different reports have implicated PKC␣ and/or PKC␦ as a PMA receptor in death induction (2)(3)(4)12). Although the consensus from the previous studies put forth in a recent review (8) is that activation of either PKC␣ or PKC␦ is sufficient to trigger LNCaP apoptosis, we reasoned that ambiguities in the previous data left such a conclusion open to question. In particular, we questioned the PKC isozyme specificities of pharmacological inhibitors at high doses and kinase-negative PKC constructs that inhibit PMA-induced apoptosis. For example, we found that overexpression of one PKC isozyme markedly altered PMA-induced membrane translocation of other isozymes and that inhibition of PMA-induced LNCaP cell death by exogenous expression of the PKC␣ regulatory domain was accompanied by inhibition of membrane translocation of PKC␣, PKC␦, and PKC⑀. 2 In our report on bryostatin-1-induced death of LNCaP cells overexpressing PKC␣, we noted that overexpression of high amounts of PKC␣ in membranes of these cells has no effect on cell growth in the absence of drug and that bryostatin-1 induces prolonged membrane translocation of PKC␦ in both parental and PKC␣-overexpressing LNCaP cells (3). This left the possibility that neither PKC␣ nor PKC␦ alone could induce LNCaP cell death, although other possibilities were discussed. Here, we have shown that C4-2 cells, androgen-hypersensitive metastatic derivatives of LNCaP cells, are as sensitive as LNCaP cells to PMA-induced apoptosis and that apoptosis requires a more complex combination of PKC isozymes than indicated previously for LNCaP cells. Partial knockdown of PKC␦ and near-complete knockdown of PKC␣ showed clearly that PMA-induced C4-2 cell death requires PKC␦, but not PKC␣. However, a contributory role for PKC␣ and the necessity of a second PKC signaling pathway were shown by simultaneous knockdown of PKC␣ and PKC⑀.
Although PMA-induced C4-2 apoptosis requires multiple PKC signaling pathways, we cannot rule out the possibility that novel, more selective PKC activators might induce death through a single pathway. For example, Kazanietz and coworkers (4,12) reported that the diacylglycerol lactone HK654 translocates PKC␣ and PKC␦ to non-nuclear and nuclear membranes of LNCaP cells, respectively, while inducing apoptosis and concluded that apoptosis induction is through selective activation of PKC␣. A slight possibility exists that activation of PKC␦ by PMA induces both pro-and anti-apoptotic signals and that the anti-apoptotic signals must be offset by PKC␣ or PKC⑀, whereas activation of PKC␣ alone induces only apoptotic signals in LNCaP cells. However, given that HK654 activates PKC␦ as well as PKC␣ in vitro (4) and that we have detected substantial amounts of PKC␦ in non-nuclear membranes of LNCaP cells in the absence of drug (3), it seems more likely that HK654 acts through PKC␦ as well as PKC␣ in inducing LNCaP apoptosis.
The observation that simultaneous knockdown of PKC␣ and PKC⑀ yields only partial inhibition of PMA-induced C4-2 cell death may reflect our inability to achieve near-complete depletion of both isozymes in the same cells or may indicate that there is additional redundancy among PMA receptors in this signaling pathway. The reasons for the inability to knock down PKC␣ and PKC⑀ may be technical, such as competition among shRNAs for RNA interference components, or may reflect a requirement for one of these isozymes to maintain cell growth. The latter possibility would not be incompatible with a requirement for PKC␣ or PKC⑀ in PMA-induced cell death since PMA yields abnormally prolonged activation of these isozymes, whereas growth factors in serum yield only transient activation via tightly regulated generation of diacylglycerol. We have also found that, whereas PKC⑀ is eventually down-regulated by PMA in LNCaP cells, especially in the minority of cells refractory to PMA-induced apoptosis, PMA-resistant LNCaP cells developed in our laboratory recover normal PKC⑀ expression (10).
Our discovery of a contributory role for PKC⑀ in PMA-induced C4-2 cell death is the first indication of a pro-apoptotic role for this oncogenic isozyme in prostate cancer cells and is in marked contrast with results of another group who reported that overexpression of high amounts of PKC⑀ in LNCaP cells renders the cells androgen-insensitive and resistant to apoptosis induction by PMA (24). We have stably overexpressed several different amounts of human wild-type PKC⑀ in LNCaP cells, and all of our clones remained as sensitive to PMAinduced apoptosis as the parental LNCaP cells, with the highest overexpressors killed by bryostatin-1, 3 similar to our observations of PKC␣-overexpressing LNCaP cells (3). We can only speculate how the other group obtained such markedly different results. One possibility is that their PKC⑀-overexpressing LNCaP cells, which were also androgen-insensitive, had been grown in medium containing insufficient androgen, resulting in co-selection of androgen-and PMA-insensitive cells. Such cross-resistance to PMA has been reported in serum starvation-resistant LNCaP cells developed by Chen et al. (27).
Previous reports on the roles of MAPKs in PMA-induced LNCaP apoptosis have indicated a requirement for one or more p38 isozymes (12), conflicting results on the importance of JNKs (12,13,28), and no effect or a slight inhibitory effect of ERK activation (12,13). Our observation of partial inhibition of PMA-induced C4-2 cell death by the JNK inhibitor SP600125 (Fig. 8A) is in between the results of Tanaka et al. (12), who found that up to 50 M SP600125 has a minimal effect on LNCaP apoptosis induced by 100 nM PMA, and Engedal et al. (13), who reported that expression of the JNK-binding domain of the JNK-interacting protein markedly inhibits PMA-induced LNCaP apoptosis. Also, Chen et al. (28) reported that overexpression of JNK1 induces apoptosis of LNCaP cells. Additional work will be required to confirm a role for JNKs in PMAinduced prostate cancer cell death. The absence of an effect on PMA-induced C4-2 cell death by doses of the MEK inhibitor U0126 that blocked activation of ERK1 and ERK2 (Fig. 8C) differs slightly from the results of Tanaka et al. (12), who reported some potentiation of PMA-induced LNCaP cell death by another MEK inhibitor, PD98059. The difference may reflect higher specificity of U0126 for ERKs or the shorter PMA treatment (1 h) used by Tanaka et al., which does not yield maximal cell death.
Our data showing the dependence of PMA-induced p38 activation on PKC␦ are in agreement with the results of Tanaka et al. (12), who reported that transient transfection of LNCaP cells with a small interfering RNA targeting PKC␦ blocks PMAinduced p38 activation, although they did not show the data. These authors also concluded that either PKC␣ or PKC␦ can act through p38 to induce apoptosis based on data that the p38 inhibitor SB203580 inhibits PMA-induced apoptosis of LNCaP cells overexpressing either PKC␣ or PKC␦. However, PMA still induces endogenous PKC␦ in the PKC␣ overexpression model. Our data showing inhibition of PMA-induced p38 dual phosphorylation by knockdown of PKC␦, but not by knockdown of PKC␣, PKC⑀, or PKC␣ ϩ PKC⑀, indicate that activation of p38 by PMA is directed by PKC␦, not by PKC␣ or PKC⑀. We have further shown that PMA-induced Rb hypophosphorylation requires PKC␦-directed p38 activation as well as a PKC␣/⑀ signaling pathway that does not involve p38 (Fig. 10). These data confirm that PKC␣ and PKC⑀ act through a separate signaling pathway than PKC␦ to activate Rb. The attenuation of PMAinduced CYR61 expression by PKC␣ or PKC⑀, but not by PKC␦, provides further evidence for distinct signaling pathways, although no evidence for an effect of CYR61 on Rb has been reported. To the best of our knowledge, this is the first demonstration of separate PKC signaling pathways, including a unique isozyme and a redundant pair of isozymes, effecting a PMA-induced cellular alteration. Further experiments will be necessary to uncover signals unique to the PKC␣/⑀ pathway that are required for PMA-induced Rb hypophosphorylation.
The possibility that additional PMA receptors are required for PMA-induced C4-2 cell death remains to be determined. Also, since C4-2 cells retain a functional androgen receptor and are androgen-hypersensitive, our results showing PMA-induced C4-2 cell death in the presence of medium containing 10% fetal bovine serum do not rule out a requirement for the androgen receptor. It is possible that PMA-induced death of C4-2 cells, like that of LNCaP cells, requires the androgen receptor to limit activation of NF-B by PMA. Still, it is significant that these cells have retained sensitivity to PMA-induced death after progressing through two rounds of selection in castrated mice to a faster growing subline with much higher metastatic ability compared with LNCaP cells.