Identification of an Amino Acid Residue in the Protein Kinase C C1b Domain Crucial for Its Localization to the Golgi Network*

Protein kinase C (PKC) isoforms have been reported to be targeted to the Golgi complex via their C1 domains. We have shown recently that the regulatory domain of PKC (cid:1) induces apoptosis in neuroblastoma cells and that this effect is correlated to Golgi localization via the C1b domain. This study was designed to identify specific residues in the C1 domains that mediate Golgi localization. We demonstrate that the isolated C1b domains from PKC (cid:2) , - (cid:3) , - (cid:4) , - (cid:5) , and - (cid:1) are targeted to the Golgi complex, whereas the corresponding C1a domains localize throughout the cell. Sequence alignment showed that amino acid residues corresponding to Glu-246 and Met-267 in PKC (cid:1) are conserved among C1b but absent from C1a domains. Mutation of Met-267, but not of Glu-246, to glycine abolished the Golgi localization of the isolated C1b domain and the regulatory domain of PKC (cid:1) . The mutated PKC (cid:1) regulatory domain constructs lacking Golgi localization were unable to induce apoptosis, suggesting a direct correlation between Golgi localization and apoptotic activity of PKC (cid:1) regulatory domain. Mutation of analogous residues in the C1b domain of PKC (cid:4) abrogated its Golgi localization, demonstrating that this effect is not restricted to one PKC isoform. The abolished Golgi localization did not affect neurite induction by PKC (cid:4)

The members of the protein kinase C (PKC) 1 family are central components in pathways regulating a wide variety of cellular processes. This family of serine/threonine kinases contains at least 10 different isoforms, which can be subgrouped into classical PKCs (␣, ␤I, ␤II, and ␥), novel PKCs (␦, ⑀, , and ), and atypical PKCs ( and ) (1,2). The functions of the different isoforms are believed to be largely controlled by the localization of the proteins. Activation of both classical and novel isoforms, for instance, is frequently correlated to a translocation to the plasma membrane (3,4). The Golgi apparatus is another site at which different PKC isoforms can be located. In some cases, the Golgi localization has been suggested to be crucial for the effect of the PKC isoform (5)(6)(7). Several reports indicate that the targeting to the Golgi network is mediated by the C1 domains in the PKC molecule (5,8,9). C1 domains are found in many different proteins and they are generally divided in two subgroups: typical C1 domains that bind phorbol esters and atypical C1 domains that do not bind phorbol esters (10). The classical PKCs and novel PKCs have in their regulatory domains (RD) a tandem repeat of typical C1 domains C1a and C1b. A C1 domain consists of a conserved amino acid sequence of about 50 residues, with six cysteine and two histidine residues that co-ordinate two Zn 2ϩ ions per C1 domain (11,12). At the tip of typical C1 domains, there is a hydrophilic ligand-binding cleft that is surrounded by hydrophobic residues. Binding of diacylglycerol or phorbol esters caps the hydrophilic cleft and generates a continuously hydrophobic surface, which enables the C1 domain to penetrate membranes (13). Several studies have investigated the relative importance of residues within C1 domains for structural maintenance and ligand interaction (14 -17). The Zn 2ϩ -coordinating residues are essential for the integrity of the tertiary structure, and the loops of amino acids 6 -12 and 20 -27 in PKC␦ C1b domain have been characterized as essential for phorbol ester binding (15,16).
For several of these proteins, targeting to the Golgi network has been suggested to be important for the function of the protein. For instance, RasGRP1 translocates to the Golgi apparatus and activates Ras at this site (23). Protein kinase D regulates the fission of transport carriers destined for the cell surface (24). Furthermore, PKC⑀ modulates secretion from the Golgi complex (5,6), whereas PKC␤ has been implicated to be involved in the regulation of coat assembly on Golgi membranes (7). We have recently shown that the RD of PKC induces apoptosis in neuroblastoma cells and this effect is correlated to a Golgi localization via the C1b domain (9).
The tendency of many C1 domains to localize to the Golgi led us to hypothesize that these domains have residues in common that mediate a specific interaction with structures in the Golgi network. Our aim was to identify such residues. This study shows that mutation of Met-267 in the PKC C1b domain or Asp-257 and Met-278 in the PKC⑀ C1b domain abolishes their Golgi localization. The mutations also inhibit the Golgi localization and apoptosis-inducing capacity of PKCRD and suppress the responsiveness of full-length PKC⑀ to ceramide.

EXPERIMENTAL PROCEDURES
Plasmids-Plasmids encoding PKCRD and PKC⑀FL (25), the C1ab regions of PKC␣ and -⑀ (26), and the C1a, C1b, and C1ab domains of PKC (9) fused to EGFP have been described. cDNA encoding the C1a and C1b domains of PKC␣, -␦, -⑀, and -and the C1ab regions of PKC␦ and -were generated by PCR using plasmids encoding the full-length PKC isoforms as template. The PCR reactions were performed with Pfu polymerase (Promega) to minimize introduction of mutations. Restriction enzyme sites were introduced in the primers (Table I) enabling the ligation of the PCR product in the EGFP-N1 vector (BD Biosciences Clontech). cDNA encoding the PKC mutants E246G, M267G, and the double mutant E246G/M267G and the PKC⑀ mutants D257G, M278G, and the double mutant D257G/M278G were generated by QuikChange site-directed mutagenesis kit (Stratagene) according to the supplier's protocol, using plasmids encoding PKC-EGFP and PKC⑀-EGFP as templates. Table I lists the primers used to generate the PKC and PKC⑀ mutant fragments. To create myc-tag fusions of the PKCC1b domain, PKCC1b and PKCC1b-M267G cDNA were cut out from the pEGFP-N1 vector by digestion with SalI and BglII and ligated into pcDNA4/myc-His vector (Invitrogen). All PKC cDNAs were sequenced to ensure that only the desired mutations were introduced in the mutagenesis reactions.
Cell Culture and Transfections-Human neuroblastoma SK-N-BE (2)C cells were cultured in minimum essential medium (Invitrogen) supplemented with 10% fetal bovine serum, 100 IU/ml penicillin, and 100 g/ml streptomycin. For transfection experiments, cells were trypsinized and seeded at a density of 150,000 cells per 35-mm cell culture dish on glass cover slips in medium containing serum and antibiotics. Transfections were initiated 24 h after seeding. Before transfection, cells were washed with serum-free medium. Cells were transfected using 3 l of LipofectAMINE 2000 (Invitrogen) and 1.6 g of DNA essentially according to the supplier's protocol. When indicated, cells were treated with 16 nM 12-O-tetradecanoylphorbol-13-acetate (TPA; Sigma) or 50 M C 2 -ceramide (Sigma). Ethanol or Me 2 SO was added to the control at the appropriate final concentration.
Analysis of Apoptosis-Sixteen hours after transfection, neuroblastoma cells were fixed in 4% paraformaldehyde in PBS for 4 min and permeabilized with 0.1% Triton X-100 and 0.1% sodium citrate in PBS for 2 min on ice. Cells were washed in PBS and then incubated with reaction mixture from in situ cell death detection kit TMR (Roche) at 37°C for 1 h. Thereafter, cells were washed in PBS and mounted on microscopy slides using 9.6% polyvinyl alcohol, 24% glycerol, and 2.5% 1,4-diazabicyclo (2.2.2) octane in 67 mM Tris-HCl, pH 8.0. 200 transfected cells, visualized by the fluorescence of EGFP, were counted per experiment, and the percentage of TUNEL-positive cells was determined. Cells were examined by fluorescence microscopy using standard fluorescein isothiocyanate and TRITC filters.
Morphology Studies-Sixteen hours after end of transfections, cells were fixed in 4% paraformaldehyde in PBS for 4 min, washed in PBS, and mounted on microscopy slides using polyvinyl alcohol-diazabicyclo octane. 200 transfected cells per experiment, visualized by the fluorescence of EGFP, were counted. The transfected cells were considered to have neurites if the length of the process exceeded that of two cell bodies.
Confocal Microscopy-SK-N-BE(2)C cells were fixed with 4% paraformaldehyde in PBS for 4 min, permeabilized, and blocked with 5% normal goat serum and 0.3% Triton X-100 in TBS for 30 min. Syntaxin 6 was detected with primary monoclonal mouse antibodies (BD Transduction Laboratories) diluted 1:25 in TBS followed by the secondary antibody Alexa Fluor 546-conjugated goat-anti-mouse (Molecular Probes) diluted 1:400 in TBS. Myc-tagged PKC constructs were detected with primary monoclonal anti-c-myc antibodies (Calbiochem) diluted 1:50 in TBS followed by the secondary antibody Alexa Fluor 488-conjugated goat-anti-mouse (Molecular Probes) diluted 1:400 in TBS. Both incubations were 1 h in length. Extensive washing was done between all steps and the cover slips were mounted on object slides. The cells were examined using a 60ϫ objective (numerical aperture, 1.4) and a Bio-Rad Radiance 2000 confocal system with excitation wavelengths at 488 nm (EGFP and Alexa Fluor 488) and 543 nm (Alexa Fluor 546) and emission filters 515HQ30 (EGFP and Alexa Fluor 488) and 600LP (Alexa Fluor 546).
Live Confocal Experiments-Live SK-N-BE(2)C cells were examined by confocal microscopy on the day after transfection. The cover slips were washed twice with buffer H (20 mM HEPES, 137 mM NaCl, 3.7 mM KCl, 1.2 mM MgSO 4 , 2.2 mM KH 2 PO 4 , 1.6 mM CaCl 2 , and 10 mM glucose, pH 7.4) and mounted on the heated stage of a Nikon microscope. To study the translocation of PKCC1b-EGFP constructs, 1 M phorbol 12,13-dibutyrate (Sigma) was added to the buffer and the localization of PKC-EGFP was followed for 2 min. The cells were examined as described for fixed cells.

PKC C1 Domains Localize to the Golgi Complex-It has
previously been demonstrated that overexpressed C1 domains of PKC⑀ (5,8) and PKC (9) are enriched in the Golgi complex. To identify a putative specific motif that targets C1 domains to the Golgi network, we first compared the localization of C1 domains from several PKC isoforms. SK-N-BE(2)C neuroblastoma cells were therefore transfected with plasmids encoding the two C1 domains of PKC␣, -␦, -⑀, -, and -fused to EGFP. The cells were fixed and the trans-Golgi network (TGN) was visualized by immunofluorescence toward the TGN marker (27) syntaxin 6 ( Fig. 1). Examination of the cells with confocal microscopy confirmed the previous observations that both PKC⑀C1 (5) and PKCC1 (9,21) almost exclusively localize to the TGN. Furthermore, the C1 domains of PKC␣, -␦, andlocalized to the TGN as well. All PKC C1 domains responded to TPA with an increased localization at the plasma membrane. Taken together, these results suggest that the localization of PKC C1 domains to the Golgi complex is not isoform-specific and a Golgi-targeting motif is likely to be conserved among these isoforms.
The C1b Domains of PKCs Are Important for Golgi Localization-Because no differences in localization pattern could be detected between the tandem C1 domains of the different isoforms, our next approach was to compare the localization of the individual C1a and C1b domains. SK-N-BE(2)C cells were transfected with vectors encoding the isolated C1a and C1b domains of PKC␣, -␦, -⑀, -, and -fused to EGFP. Transfected cells were fixed, and the TGN was visualized by staining with antibodies toward syntaxin 6 ( Fig. 2). Analysis with confocal microscopy revealed that the PKC C1a fusion proteins generally localized throughout the cell, and there was no detectable localization to the Golgi complex. In contrast, PKC␦C1b, PKC⑀C1b, and PKCC1b all showed a distinct co-localization with the TGN marker, and these fusion proteins were not found in the nuclear matrix. PKC␣C1b and PKCC1b both localized throughout the cell, but there was an enrichment of the fusion proteins in the TGN. Cells expressing the PKC C1a and C1b domains were also treated with TPA, and the localization of the fusion proteins was examined with confocal microscopy. The C1a domains of PKC␣, -⑀, and -were translocated from the nuclear matrix, whereas neither PKC␦C1a nor PKCC1a responded to TPA. PKC␣C1b and PKCC1b were translocated from the nuclear matrix after TPA treatment. PKC␦C1b, PKC⑀C1b, and PKCC1b lost their distinct Golgi localization and distributed to other cytoplasmic structures as well after treatment with TPA. There was also a tendency to increased localization to the nuclear and plasma membrane of some C1b domains. These results indicate that a Golgi localization motif in PKC mainly resides in the C1b domain. Such a motif is conceivably common for all the investigated C1b domains but not present in the C1a domains.
Sequence Alignment of PKC C1 Domains-The differential localization pattern of the C1a and C1b domains led us to search for divergences between the domains. We aligned the protein sequences of the C1a and C1b domains of PKC␣, -␦, -⑀, -, and -(NCBI protein database accession numbers CAA36718, AAA03175, CAA46388, NP_006246, and AAA60101, respectively) and looked for residues that are conserved only among the C1b domains (Fig. 3A). We also took into consideration the published crystal structure of the PKC␦ C1b domain (PDB code 1PTQ) ( Fig. 3B) (13). One residue, Met-266 in PKC␦, which is conserved only among the C1b domains, caught our attention as an interesting candidate. Met-266 is located outside the hydrophobic part of the domain that is presumably inserted in the membrane and has its side chain exposed on the protein surface. In this context, it is interesting that conservation of methionine on the protein surface has been suggested to imply a putative protein binding site (28). We also noted that the side chain of Asp-245, which is also conserved only among the C1b domains, is near Met-266. Therefore, we decided to explore whether these amino acid residues are important for the localization of C1b domains to the Golgi complex.
Mutation of Glu-246 and Met-267 in PKCC1b-Our previous results have indicated a correlation between localization to the Golgi and induction of apoptosis by the isolated RD of PKC (9). We therefore focused on the C1b domain of PKC and altered the residues in the domain (Glu-246 and Met-267) corresponding to PKC␦ Asp-245 and Met-266 with site-directed mutagenesis. Three mutated variants of C1b-EGFP were created in which Glu-246, Met-267, or both residues were replaced by glycine (denoted C1b-E246G, C1b-M267G, and C1b-DM, respectively). These vectors and a vector encoding the wild-type C1b domain fused to EGFP (C1b-wt) were transfected into SK-N-BE(2)C cells and expression of the fusion proteins was confirmed by Western blotting (Fig. 4B).
To confirm that the mutations do not lead to a disruption of the overall conformation of the C1b domains, we investigated whether TPA could alter the subcellular localization of the EGFP-fused C1b mutants (Fig. 4A). As previously seen, C1b-wt localized throughout the cell but was clearly enriched in perinuclear structures. TPA treatment influenced the localization as seen by the clearance of nuclear fusion proteins. The same pattern was observed for the C1b-E246G mutant. In contrast, C1b-M267G and C1b-DM were evenly distributed throughout the cells with no perinuclear enrichment. Treatment with TPA translocated the fusion proteins from the nuclear matrix to the cytoplasm and gave rise to a minor perinuclear enrichment.
The phorbol ester responsiveness of the C1b domains was further confirmed in a live confocal experiment. SK-N-BE(2)C cells expressing C1b-wt, C1b-M267G, or C1b-DM were treated with 1 M phorbol 12,13-dibutyrate and the localization of the fusion proteins was followed. In untreated cells, C1b-wt, C1b-M267G, and C1b-DM were all distributed throughout the cells, with a perinuclear enrichment in the case of C1b-wt (Fig. 4C). After treatment with phorbol 12,13-dibutyrate for 2 min, the fusion proteins translocated to the nuclear membrane. The fact that all fusion proteins were sensitive to phorbol esters indicates that they are properly folded.
To exclude the possibility that the localization of the C1b domains to the Golgi network is influenced by EGFP, cells were transfected with constructs encoding wild-type C1b and the corresponding M267G mutant, both fused to a myc tag. The fusion proteins were visualized by immunofluorescence toward myc, and their localization was examined with confocal microscopy. This experiment demonstrated that wild-type C1b was enriched in the perinuclear region, whereas C1b M267G resided mainly in the nucleus (Fig. 4D).
Met-267 Is Necessary for Localization of PKCC1b to the Golgi Complex-The previous experiments indicated an altered localization of the C1b-M267G and C1b-DM mutants compared with the wild-type protein. To characterize the differences in localization between C1b-wt and C1b-E246G on one hand and C1b-M267G and C1b-DM on the other hand, transfected and fixed SK-N-BE(2)C cells were stained with antibodies toward syntaxin 6. Co-localization of the fusion proteins with the TGN marker was then analyzed with confocal microscopy. This experiment showed a clear enrichment of C1b-wt and C1b-E246G in syntaxin 6-positive structures, whereas only small amounts of C1b-M267G and C1b-DM were present in the TGN (Fig. 5). Thus, Met-267 seems to be crucial for proper Golgi localization of the C1b domain.
To establish whether mutation of Met-267 also abolishes the Golgi localization of the entire RD, we made three new constructs encoding RD fused to EGFP, in which Glu-246, Met-267, or both residues were replaced by glycine (denoted RD-E246G, RD-M267G, and RD-DM, respectively). SK-N-BE (2)C cells were transfected with these vectors or a vector encoding the wild-type RD (RD-wt), and fixed cells were subjected to staining with antibodies toward syntaxin 6. This experiment demonstrated that both RD-wt and RD-E246G almost exclusively localized to syntaxin 6-positive structures (Fig. 5). However, RD-M267G and RD-DM were primarily targeted to the nucleus and localized to the TGN only to a minute extent. Taken together, these results further demonstrate that the localization of PKCRD to the Golgi complex mainly is mediated by the C1b domain and points out Met-267 as an important residue for this interaction.
Mutation of Met-267 Abolishes the Apoptotic Effect of PKCRD-We next sought to determine whether the abrogated Golgi localization caused by mutation of Met-267 is correlated to a diminished apoptotic activity of RD. SK-N-BE(2)C cells were transfected with vectors encoding RD-wt, RD-E246G, RD-M267G, RD-DM, or empty EGFP vector (EGFP). After transfection, cells were fixed and scored for apoptosis using a TUNEL assay. As seen before, expression of RD-wt caused a significantly higher rate of apoptosis compared with cells expressing EGFP alone (Fig. 6A). A similar effect was obtained by the RD-E246G variant. However, expression of neither RD-M267G nor RD-DM led to an altered rate of apoptosis compared with control cells. The apoptotic effects of RD-wt and RD-M267G was further analyzed in the presence of TPA. This experiment revealed that although TPA causes an increased rate of apoptosis in cells expressing RD-M267G, the percentage of apoptotic cells is significantly lower than for TPA-treated cells expressing RD-wt (Fig. 6B). To exclude that the apoptotic effects of RD-wt and RD-E246G, compared with the RD-M237G mutant, were caused by higher expression levels, SK-N-BE(2)C cells were also transfected with half or a quarter of the amount of these vectors. Analysis of the protein expression by Western blotting demonstrates that transfection with a smaller amount of vector reduces the amount of expressed protein (Fig. 6A). Although a lower expression level of the fusion protein correlates with a slightly lower rate of apoptosis, RD-wt still induced a significantly higher rate of apoptosis than RD-M267G or RD-DM. RD-E246G also maintained apoptotic activity at lower expression levels.
To confirm that the mutations do not impede the capacity of RD to respond to phorbol esters and thus that the general characteristics of the C1 domains are maintained, we examined whether the RD mutants translocate after TPA treatment. RD-wt primarily localizes to the perinuclear region in untreated cells, and TPA treatment does not alter this localization (Fig. 6C). A similar pattern is observed for RD-E246G. However, RD-M267G and RD-DM, which both primarily localize to the nucleus before TPA treatment, are translocated to the plasma membrane and to punctuate cytoplasmic structures after addition of TPA. In some cells, these structures are slightly enriched in a perinuclear region. Because the mutated RD fusion proteins were able to respond to TPA treatment, the protein conformation is probably maintained. The fact that the two PKCRD constructs lacking Golgi localization (i.e. RD-M267G and RD-DM) were unable to induce apoptosis implies that Golgi localization is important for the apoptotic activity of PKCRD.
Analogous Mutations of PKC⑀C1 Alter the Localization but Not the Neurite-inducing Capacity of the Protein-The finding that mutation of Met-267, and Glu-246/Met-267 in combination, abrogates the Golgi localization of both the RD and the isolated C1b domain of PKC, raised the question of whether this effect is isoform-specific. To test this, mutated variants of PKC⑀C1ab-EGFP and PKC⑀C1b-EGFP were created in which the residues analogous to PKC-Glu-246 and PKC-Met-267, PKC⑀-Asp-257, and PKC⑀-Met-278, were replaced by glycine (denoted ⑀C1ab-DM and ⑀C1b-DM). We chose to study the double mutant because initial experiments indicated that PKC⑀C1ab with only Met-278 mutated still displayed some minor enrichment in the Golgi network (data not shown). SK-N-BE(2)C cells were transfected with these vectors or the corresponding wild-type vectors (⑀C1ab-wt and ⑀C1b-wt) and protein expression was confirmed by Western blot analysis (Fig.  7B). We next investigated the subcellular localization of the PKC⑀C1 mutants and compared it with the wild-type variants. Transfected SK-N-BE(2)C cells were examined with confocal microscopy (Fig. 7A). As previously shown, both ⑀C1b-wt and ⑀C1ab-wt have distinct Golgi localization. In contrast, ⑀C1b-DM is located throughout the cell, whereas ⑀C1ab-DM localizes uniformly in the cytoplasm and is absent from the nucleus. Treatment with 16 nM TPA increases the enrichment of ⑀C1b-wt in the perinuclear region, whereas ⑀C1b-DM translocates from the nuclear matrix. Both the wt and DM of ⑀C1ab translocate to the plasma membrane in response to TPA. It has previously been shown that a structure encompassing the C1 domains of PKC⑀ induces neurites independently of the kinase activity of the enzyme (25). To investigate whether the mutations that abolish Golgi localization of PKC⑀C1 fusion proteins also affect the neurite-inducing capacity, SK-N-BE(2)C cells overexpressing ⑀C1ab-wt, ⑀C1ab-DM, or empty EGFP vector were treated with 16 nM TPA or vehicle for 16 h. The morphological effects were visualized with fluorescence microscopy and quantified by counting the number of transfected cells with neurites longer than two cell bodies. The results demonstrate that both ⑀C1ab-wt and ⑀C1ab-DM induced neurites in more than 30% of the transfected cells, compared with 2% of cells expressing EGFP alone (Fig. 7C). TPA treatment further potentiated the neurite-inducing effect of both PKC⑀C1ab constructs.
Taken together, the results demonstrate that the mutated residues are important for localization of C1b domains to the Golgi complex and that this is not restricted to one PKC isoform. Furthermore, the fact that ⑀C1ab-DM induced neurites as efficiently as ⑀C1ab-wt, despite its lower expression levels, indicates that Golgi localization is not important for the neurite-inducing capacity of PKC⑀C1 domains.
Localization of Full-length PKC-We have previously shown that TPA treatment induces apoptosis in neuroblastoma cells expressing full-length PKC (FL) (9). It was of interest to determine whether this effect is correlated to a Golgi localization and to study the effect of the C1b mutations on the apoptosis-inducing capacity of holo-PKC. We therefore investigated the localization of FL in TPA-treated cells, the conditions under which it induces apoptosis (Fig. 8). However, examination of the cells with confocal microscopy revealed that FL does not localize to the Golgi complex. This experiment suggests that Golgi localization is not important for the apoptotic effect of FL. Thus, FL and RD are probably exerting their apoptotic effects via different mechanisms in neuroblastoma cells.
Mutation of the C1b Domain Makes Full-length PKC⑀ Less Responsive to Ceramide-Because we could not study the effect of the C1b mutations on full-length PKC, we turned to PKC⑀. We have previously shown that treatment with C 2 -ceramide induces a relocation of PKC⑀ to perinuclear structures that is accompanied by a suppression of the neurite-inducing capacity of PKC⑀ (29). The results in Fig. 7 suggest that the Asp-257/ Met-278 residues of ⑀C1b are important for Golgi localization; we therefore wanted to clarify whether these mutations also affect the ability of ⑀FL to translocate to the Golgi network in response to ceramide. SK-N-BE(2)C cells were transfected with constructs encoding wild-type ⑀FL (⑀FL-wt) or ⑀FL with mutated Asp-257/Met-278 residues (⑀FL-DM), both fused to EGFP. After transfection, cells were treated with 16 nM TPA and/or 50 M C 2 -ceramide for 16 h and the localization of the fusion proteins was examined with confocal microscopy. In untreated cells, ⑀FL-wt localizes throughout the cytoplasm with a minor enrichment in the perinuclear region (Fig. 9A). Treatment with TPA relocates the fusion protein to the plasma membrane, and, as reported previously (8,29), treatment with C 2 -ceramide induces a clear enrichment of ⑀FL in the perinuclear region. Combined treatment of C 2 -ceramide and TPA leads to a primarily cytoplasmic localization of ⑀FL-wt, which is analogous with studies on other cell types (8). The C1b mutant, ⑀FL-DM, also distributes evenly in the cytoplasm and translocates to the plasma membrane by TPA treatment. However, in contrast with the wild-type protein, ⑀FL-DM still localizes in the cytoplasm after treatment with C 2 -ceramide. Moreover, C 2 -ceramide does not suppress the plasma membrane localization induced by TPA.
We next analyzed whether the PKC⑀ mutants also become more resistant to the neurite-inhibiting effect of ceramide. The experiment confirmed that overexpression of ⑀FL-wt induces neurites, an effect that is further potentiated by TPA (Fig. 9B). Treatment with C 2 -ceramide significantly inhibits the induction of neurites, and a combined treatment with TPA cannot restore the neurite-inducing capacity of ⑀FL. Overexpression of ⑀FL-DM induces neurites as potently as ⑀FL-wt, and treatment with TPA enhances this effect. However, compared with the ⑀FL-wt-expressing cells, a significantly higher percentage of ⑀FL-DM-expressing cells have neurites after C 2 -ceramide treatment. Moreover, TPA treatment partially restores the neurite-inducing capacity of ⑀FL-DM after C 2 -ceramide treatment. Taken together, these results indicate that Asp-257/Met-278 is important for the ability of ⑀FL to relocate to the perinuclear region in response to C 2 -ceramide. Furthermore, the Asp-257/Met-278 residues are also important for the functional response of PKC⑀ to ceramide.

DISCUSSION
Typical C1 domains, once identified as the phorbol ester binding site in PKC isoforms, have during the last years also been identified in several proteins outside the PKC family (19). This domain conceivably contributes to both the regulation and the localization of the protein. C1 domains are primarily thought to bind lipids, but recent studies have also identified proteins that directly interact with different C1 domains (30,31). In addition to phorbol esters/diacylglycerol, C1 domains have been shown to directly bind and/or mediate intracellular targeting induced by mediators, including retinoic acid (32), arachidonic acid, and ceramide (8). Binding of these agents results in a translocation and/or an activation of the protein highlighting the importance of the C1 domain as a regulatory element. It has also become clear that C1 domains are differentially sensitive to lipids. For instance, the C1b but not the C1a domain of PKC⑀ is a target for ceramide and arachidonic acid (8), whereas the C1a but not the C1b domain of PKC␣ is sensitive to diacylglycerol (33,34). Our study further underscores the differences between C1 domains, because the C1b domains of PKC␣, -␦, -⑀, -, and -localize to the Golgi complex, whereas the C1a domains of the same isoforms are not enriched in this structure. This is in line with previous studies indicating an important role for the C1b domain of PKC⑀ (5,8) and PKC (9,21) for Golgi localization. Furthermore, all C1a domains and the C1b domains of PKC␣ and PKC also localize to the nucleus, whereas the tandem domains (C1ab) of all isoforms are extranuclear. This suggests that nuclear targeting motifs in many C1 domains may contribute to the nuclear localization that is sometimes observed for different PKC isoforms (35,36). In many cases, however, the putative nuclear localization signal is overridden by other targeting determinants in the holoenzyme.
In this study, we investigated two amino acid residues that are conserved among Golgi-localizing PKC C1 domains, Glu-246 and Met-267 in PKC. The results suggest that Glu-246 is dispensable for Golgi localization, but the data highlight PKCMet-267 as an important residue determining whether a C1 domain localizes to the Golgi network or not. The conclusion is supported by the sequence alignment of PKC C1 domains, which showed that all known Golgi-targeted C1 domains contain a methionine at the corresponding site, whereas of the C1 domains that do not localize to the Golgi, none has a methionine or an analogous amino acid at this place. Furthermore, the C1 domain of RasGRP1 (NCBI accession number NP_005730), which also localizes to the Golgi, has a methionine at the corresponding site, and the C1 domain of ␤2-chimaerin (Swiss-Prot accession number P52757) and the C1a domain of protein kinase D (NCBI accession number CAA53384), which both determine Golgi localization of the fulllength proteins, have a leucine at the analogous place (alignments not shown). Thus, C1 domains that localize to the Golgi have in common a methionine or an amino acid with similar characteristics at the site corresponding to PKCMet-267. This residue is therefore conceivably involved in interaction with a Golgi-specific structure.
It is unlikely that the methionine residue is important for a general interaction with cellular membranes. We base this assumption on the fact that, as revealed by the crystal structure of the PKC␦C1b domain (PDB code 1PTQ) (13), the methionine residue is located outside the hydrophobic surface that is presumably inserted into membranes. Furthermore, as revealed by our experiments, both PKC⑀C1ab and PKCRD variants that contain a mutated methionine respond with a clear translocation to the plasma membrane upon TPA treatment. We suggest, therefore, that the methionine residue is important for interaction with a Golgi-specific protein. For ␤2-chimaerin, a C1 domain-mediated interaction with the Golgi protein Tmp21-I has been identified (31), indicating that proteinprotein interaction is important for Golgi localization of C1 domain-containing proteins. It remains to be investigated whether this protein, or an as-yet-unidentified protein, also determines the Golgi localization of PKC.
As mentioned in the introduction, the C1 domain-mediated localization of proteins to the Golgi complex is important for their function (5,7,23,24). We have shown recently that the RD of PKC is targeted to the Golgi complex and induces apoptosis in neuroblastoma cells (9). The C1b domain is required for Golgi localization, and deletion of the C1b domain abolishes the apoptotic effect of PKCRD. In this study, we demonstrate that mutation of Met-267, which disrupts the Golgi localization of PKCRD, abrogates the apoptotic activity of the fusion protein. This further suggests that there is a direct correlation between Golgi localization and apoptotic activity of PKCRD. We speculate that PKCRD needs to be targeted to the Golgi complex and interact with a Golgi-specific structure to induce apoptosis, and that Met-267 is important for this association. However, in the presence of TPA, there was an elevated rate of apoptosis by the M267G mutant compared with unstimulated conditions. We currently have no certain explanation for this effect, but it might possibly be related to the minor amount of mutated PKCRD that is observed in the Golgi network of some cells (Fig. 5), which could become more apoptotic when exposed to TPA. It should be noted, however, that also in the presence of TPA, the wild-type regulatory domain had a significantly higher apoptosis-inducing effect than the mutated variant, demonstrating the importance of Met-267.
The Golgi apparatus has emerged as an important site for apoptosis signaling. Several pro-apoptotic proteins such as caspase-2 (37), Fas/CD95 (38), and GD3 synthase (39) are enriched in the Golgi complex. Moreover, the Golgi apparatus has FIG. 9. Mutation of the C1b domain makes full-length PKC⑀ less responsive to ceramide. SK-N-BE(2)C cells were transfected with expression vectors encoding wild-type full-length PKC⑀ (wt) or the double mutant D257G/M278G (DM) thereof, both fused to EGFP. After transfection, cells were treated with 16 nM TPA and/or 50 M C 2ceramide or vehicle for 16 h, fixed, and mounted on object slides. A, the localization of the fusion proteins was analyzed by confocal microscopy. B, transfected cells with neurites longer than two cell bodies were counted. Data (mean Ϯ S.E., n ϭ 3) are presented as percentage of transfected cells with neurites. *, statistically significant difference (analysis of variance followed by Duncan's multiple range test, p Ͻ 0.05). The formation of protein products was analyzed by Western blotting using an antibody toward EGFP (top) or actin as a loading control (bottom). The positions of the molecular mass markers 220 and 97 kDa are shown to the left of the blot. been shown to be a site of ceramide production (40), and another PKC isoform, PKC␦, was recently shown to depend on its Golgi localization for induction of apoptosis in HeLa cells (41).
However, the induction of apoptosis by PKCFL does not seem to depend on a localization to the Golgi, because no enrichment in the Golgi network was observed in the presence of TPA. Thus, PKCFL and the isolated PKCRD probably induce apoptosis via separate pathways in neuroblastoma cells. It is difficult to assess the physiological importance of these two effects. We could detect an increase in immunoreactivity toward the regulatory domain of PKC in perinuclear structures during Fas-induced apoptosis in Jurkat cells. We do not know whether this represents the isolated regulatory domain, which is formed upon Fas stimulation (42), or if it represents a translocation of endogenous PKCFL. Nevertheless, it illustrates a correlation between localization of the PKCRD to the Golgi network and induction of apoptosis in a setting resembling a physiological situation.
Mutation of analogous residues in PKC⑀C1b (i.e. Asp-257 and Met-278) abolished the Golgi localization of the protein. However, PKC⑀C1ab Asp-257/Met-278 retained the ability to induce neurites. Thus, Golgi localization is not important for PKC⑀-mediated neurite outgrowth. This is in line with previous results from our group, indicating that localization of PKC⑀ to the plasma membrane and/or the cortical cytoskeleton is important for its ability to induce neurites (43). Furthermore, mutation of Asp-257 and Met-278 in PKC⑀FL abolished its translocation to the Golgi network induced by ceramide exposure. These C1b residues are thus crucial for a proper localization to the Golgi apparatus, not only for isolated PKC domains, but also for PKC⑀FL. We have also seen that ceramide suppresses the neurite-inducing capacity of PKC⑀ (29), but PKC⑀ with the D257G and M278G mutations is more resistant to this effect of ceramide. Thus, the residues are important both for the localization and function of PKC⑀.
In conclusion, this study identifies one amino acid residue, conserved among Golgi-localizing C1 domains, that is critical for the localization of the PKCC1b domain to the Golgi complex. Thus, this residue is a prominent candidate as a common mediator of Golgi-targeting of C1 domains. The data also further support the hypothesis that localization to the Golgi network is critical for apoptosis induction by PKCRD. The corresponding residue in PKC⑀ is, together with Asp-257, important for ceramide-influenced localization and effects of PKC⑀FL, illustrating that the identified residues are important also for the holoenzyme.