The Coatomer Protein β′-COP, a Selective Binding Protein (RACK) for Protein Kinase Cε*

Distinct subcellular localization of activated protein kinase C (PKC) isozymes is mediated by their binding to isozyme-specific RACKs (receptors for activatedC-kinase). Our laboratory has previously isolated one such protein, RACK1, and demonstrated that this protein displays specificity for PKCβ. We have recently shown that at least part of the PKCε RACK-binding site on PKCε lies within the unique V1 region of this isozyme (Johnson, J. A., Gray, M. O., Chen, C.-H., and Mochly-Rosen, D. (1996) J. Biol. Chem. 271, 24962–24966). Here, we have used the PKCε V1 region to clone a PKCε-selective RACK, which was identified as the COPI coatomer protein, β′-COP. Similar to RACK1, β′-COP contains seven repeats of the WD40 motif and fulfills the criteria previously established for RACKs. Activated PKCε colocalizes with β′-COP in cardiac myocytes and binds to Golgi membranes in a β′-COP-dependent manner. A role for PKC in control of secretion has been previously suggested, but this is the first report of direct protein/protein interaction of PKCε with a protein involved in vesicular trafficking.

Anchoring proteins appear to be essential components of a number of signal transduction pathways (1). In the case of PKC, 1 activation of PKC isozymes is associated with their translocation to the particulate fraction, with each activated isozyme localized to distinct subcellular sites (1,2). This translocation appears to be mediated by the binding of each activated isozyme to specific anchoring proteins we have termed RACKs (receptors for activated C-kinase) (1,3).
In cultured cardiac myocytes, PKC⑀ translocates from the nucleus to cross-striated structures, the perinucleus, and cell/ cell contacts following stimulation with 4␤-phorbol 12-myristate 13-acetate or with ␣ 1 -adrenergic receptor agonists; other PKC isozymes translocate to other distinct intracellular sites (2,4). Our laboratory has recently demonstrated that the V1 region of PKC⑀ contains at least part of the RACK-binding site in this enzyme. A recombinant PKC⑀ V1 polypeptide (amino acids 2-144; which does not include the pseudosubstrate sequence) selectively inhibits stimulation-induced PKC⑀ translo-cation and regulation of contraction rate in cardiac myocytes. In contrast, translocation and function of other PKC isozymes are not altered by this fragment (5).
This study describes the use of the PKC⑀ V1 region in an overlay screen to clone a PKC⑀-selective RACK (6,7). Using this procedure, we have identified ␤Ј-COP, a COPI (coat protein I) coatomer complex protein essential for Golgi budding and vesicular trafficking (8), as a PKC⑀-selective RACK. Previous studies have linked PKC to control of constitutive membrane trafficking and Golgi function (9 -14). However, the mechanism by which PKC exerts this control has not yet been determined. Our finding that ␤Ј-COP can also serve as a selective anchoring protein (RACK) for activated PKC⑀ provides a possible mechanism by which PKC regulates Golgi function.
Column Overlay Assay-Recombinant ␤Ј-COP-MBP fusion proteins were immobilized on amylose affinity columns (New England Biolabs Inc.), and nonspecific proteins were eluted according to the manufacturer's protocol. Overlay buffer (200 mM NaCl, 50 mM Tris-HCl, pH 7.5, 12 mM 2-mercaptoethanol, 0.1% (w/v) bovine serum albumin, 1% (w/v) polyethylene glycol, 10 g/ml soybean trypsin inhibitor, and 10 g/ml leupeptin) was added with or without 1 unit of rat brain PKC, 60 g/ml PS, 2 g/ml DG, and 1 mM CaCl 2 as indicated, and the column was sealed and incubated at room temperature for 30 min while shaking.
Unbound material was eluted with 20 column volumes of wash buffer (200 mM NaCl, 20 mM Tris-HCl, pH 7.5, 1 mM EDTA, and 10 mM 2-mercaptoethanol). The recombinant protein and any associated proteins were eluted with wash buffer containing 20 mM maltose and subjected to SDS-PAGE followed by Western blot analysis using anti-PKC⑀ antibodies (Life Technologies, Inc.).
Immunofluorescence-Immunofluorescence studies were carried out as described previously (2). Anti-PKC⑀ antibodies were obtained from Research and Diagnostics Antibodies, and 23C anti-␤Ј-COP rat monoclonal antibodies were a generous gift from K. J. Harrison-Lavoie and Dr. K. R. Willison (15). Fluorescein-and rhodamine-conjugated secondary antibodies were from Organon Technika Corp. Adsorption of anti-PKC⑀ and anti-␤Ј-COP antibodies was carried out in solid phase. Rat brain PKC (0.1 g), recombinant ␤Ј-COP (5 g), or 3% normal goat serum was spotted on 0.5-cm 2 nitrocellulose membrane. After 5 min at room temperature, the membranes were washed and blocked with 3% normal goat serum in phosphate-buffered saline for 1 h. The membranes were then incubated with the appropriate primary antibody overnight at 4°C, and the unadsorbed material was used for immunostaining. Incubation of either antibody on nitrocellulose spotted with 3% normal goat serum did not reduce any immunostaining seen in Fig. 3 (A and B) (data not shown).
Recombinant PKC Isozyme Binding Experiments-Sf9-expressed purified recombinant PKC isozymes were obtained from PanVera Inc. Binding experiments were carried in the presence of 1 g of purified recombinant PKC as described above (see "Column Overlay Assay"). For the classical PKC isozymes (␣, ␤I, ␤II, and ␥), binding was carried out in overlay buffer containing PS, DG, and Ca 2ϩ as activators. For the novel isozymes (⑀ and ␦), binding was carried out in overlay buffer containing PS and DG as activators. PKC binding was determined by Western blot analysis of the column eluates using anti-PKC␣, -␤, and -␥ antibodies from Seikagaku America, Inc. and anti-PKC⑀ and -␦ antibodies from Life Technologies, Inc. For competition binding experiments, an excess of PKC␤I (1 g) was added to the columns in overlay buffer containing PS, DG, and Ca 2ϩ and incubated for 30 min at room temperature. After 30 min of preincubation, activated PKC⑀ was added and incubated for 30 min at room temperature, and then the columns were washed, eluted, and analyzed as described above.
Immunoprecipitations-Neonatal rat hearts were homogenized in 20 mM Tris-HCl, pH 7.5, containing 10 mM EGTA, 2 mM EDTA, 0.25 M sucrose, 10 M phenylmethylsulfonyl fluoride, 20 g/ml leupeptin, and 20 g/ml soybean trypsin inhibitor. The homogenate was centrifuged at 100,000 ϫ g for 40 min at 4°C, and the pellet was extracted with 20 mM Tris-HCl, pH 7.5, 1 mM EGTA, 1 mM EDTA, 12 mM 2-mercaptoethanol, and 10% (v/v) glycerol containing the above protease inhibitors and 0.5% Triton X-100 for 30 min on ice. The Triton-soluble extract (supernatant) was separated from the insoluble material following centrifugation at 100,000 ϫ g for 40 min at 4°C. For each immunoprecipitation reaction, 2-3 g of primary antibody was added to 50 l of recombinant protein G-agarose (Life Technologies, Inc.) and incubated for 2 h at 4°C. The protein G-agarose was washed three times with phosphate-buffered saline containing 0.1% Triton X-100. The Triton-soluble fraction from the equivalent of one neonatal heart was added to the protein G-agarose and incubated for 4 h at 4°C. The protein G-agarose was washed three times with phosphate-buffered saline containing 0.1% Triton X-100. Laemmli sample buffer was added, and the samples were subjected to SDS-PAGE followed by Western blot analysis using anti-PKC⑀, 23C anti-␤Ј-COP, and anti-PKC␤ antibodies.
PKC⑀ Binding to Rat Liver Golgi Membranes-Golgi membranes (100 g/ml) from rat liver (16,17) and cytosol from rat brain (1.5 mg/ml) were incubated in binding buffer (25 mM HEPES-KOH, pH 7, 0.2 M sucrose, 25 mM KCl, 2.5 mM MgCl 2 , 1 mM ATP, 25 mM creatine phosphate, 10 IU/ml creatine phosphokinase, and 1 mM dithiothreitol with or without 20 M GTP␥S for the last 10 min) for 25 min at 37°C. At the end of the incubations, samples were centrifuged at 13,000 rpm for 15 min at 4°C. The pellet was washed once with binding buffer and resuspended in Laemmli sample buffer. Samples were subjected to SDS-PAGE on 4 -15% gradient gels and Western blot analysis using anti-PKC⑀ and 23C anti-␤Ј-COP antibodies.

RESULTS AND DISCUSSION
We have used the recombinant V1 fragment of PKC⑀ to screen a rat heart cDNA library for a PKC⑀-specific RACK. One clone, pRACK2, contained a partial open reading frame encoding a 481-residue protein homologous to the C-terminal half of human and bovine ␤Ј-COP (15,18). A full-length 906-amino acid open reading frame was obtained by RACE and identified as rat ␤Ј-COP (GenBank™ accession number AF002705) as it was found to be 97% identical to human ␤Ј-COP at the amino acid level. Therefore, ␤Ј-COP protein, one of a seven-protein coatomer complex that forms the coat of COPI-coated transport vesicles (8), binds the V1 region of PKC⑀ in the overlay cloning procedure. Similar to RACK1 (the RACK for PKC␤), ␤Ј-COP contains seven internal repeats of the WD40 motif ( Fig. 1, A  and B). This motif is also present in the G protein ␤-subunit, which is an anchoring protein for another translocating protein kinase, the G protein receptor kinase, and in other proteins involved in multimeric protein/protein interactions (19).
To determine if ␤Ј-COP is a RACK, we tested whether recombinant ␤Ј-COP and ␤Ј-COP fragments bind PKC⑀ in vitro. Fragments or full-length ␤Ј-COP fused to MBP were immobilized on an amylose column, and binding of rat brain PKC⑀ was measured using the column overlay assay (Fig. 1C). Recombinant ␤Ј-COP did not bind PKC when the ␤Ј-COP was subjected to SDS-PAGE followed by the overlay assay (3), indicating that the native conformation of ␤Ј-COP is important for PKC⑀ binding (data not shown). Because a RACK should bind only activated PKC (3), we determined the effect of PS and DG on PKC binding. As expected for a calcium-independent, novel PKC isozyme, PKC⑀ binding to ␤Ј-COP required PS and DG and was calcium-independent; in fact, calcium inhibited PKC⑀ binding (Fig. 1C). Furthermore, binding of PKC⑀ to ␤Ј-COP was saturable at ϳ10 nM PKC⑀ (Fig. 1D). Finally, binding of PKC⑀ to ␤Ј-COP was not inhibited by an excess of a pseudosubstrate peptide (Fig. 1D, lane 9 versus lane 8), and ␤Ј-COP was not a PKC substrate (data not shown). Rather, as in the case of RACK1 (7), phosphorylation of substrate increased in the presence of ␤Ј-COP. 2 Therefore, ␤Ј-COP fulfills the criteria previously established for RACKs (3).
The V1 region of PKC⑀ selectively inhibits PKC⑀ translocation and function in intact myocytes (5) and therefore should mediate the interaction of PKC⑀ with ␤Ј-COP. In vitro, the V1 fragment of PKC⑀ bound to the C-terminal fragment, but not the N-terminal fragment of ␤Ј-COP ( Fig. 2A). In contrast, PKC⑀ holoenzyme bound to both fragments (Fig. 1C). These data suggest that the PKC⑀/␤Ј-COP interaction involves more than one site on either or both molecules. However, the PKC⑀ V1 fragment inhibited PKC⑀ holoenzyme binding to ␤Ј-COP by 60% (Fig. 2B), indicating that the V1 region of PKC⑀ contains at least part of the ␤Ј-COP-binding site on PKC⑀. It is interesting to note that two binding sites on RACK1 for PKC␤ were also found (7), suggesting a common theme in PKC/RACK interactions.
To further demonstrate the specificity of PKC⑀ binding to ␤Ј-COP, we compared the binding of purified recombinant PKC isozymes to immobilized ␤Ј-COP (Fig. 2C). When using equal amounts of each isozyme, PKC⑀ binding to ␤Ј-COP was at least an order of magnitude better than the binding of the other isozymes. In addition, preincubation of ␤Ј-COP with PKC␤I ( Fig. 2D) or concomitant incubation with both isozymes (data not shown) did not inhibit subsequent binding of PKC⑀ to ␤Ј-COP. Together, these data show preferred binding of PKC⑀ by ␤Ј-COP in vitro.
If ␤Ј-COP is a PKC⑀-selective RACK, ␤Ј-COP should colocalize with activated PKC⑀ in cells. Indeed, immunofluorescence studies showed that ␤Ј-COP colocalized with activated PKC⑀ to cross-striated structures in cardiac myocytes (Fig. 3B). Immunostaining at the perinucleus and cell/cell contacts was also observed with both anti-␤Ј-COP and anti-PKC⑀ antibodies. None of the other PKC isozymes or RACKs localize to these structures in either control or 4␤-phorbol 12-myristate 13-ace- FIG. 1. A, a scheme indicating the approximate positions of the WD40 motifs in ␤Ј-COP and RACK1. B, alignment of the WD40 motifs from RACK1 (7) and rat ␤Ј-COP and the consensus sequence identified by Neer et al. (19). Previously published sequence analyses (15,18) have not identified WD40 motifs in the C-terminal half of ␤Ј-COP, the portion of ␤Ј-COP encoded by the original pRACK2 clone. Closer inspection of the sequence in this region revealed a partial WD40 sequence differing in 6 amino acids from the consensus sequence (19). The repeats numbers are indicated by rI through rVII. Numbers on the left indicate the position of the repeat in the sequence. Amino acid sequence is provided using a single-letter code, where X represents any amino acid. Numbers in braces indicate the minimal and maximal numbers of amino acids inserted at tate-treated cells (2). The specificity of the antibodies used for the immunostaining is shown in Fig. 3 (D and E); no immunostaining was observed after absorption with the corresponding protein. Image analysis indicated that fluorescence intensities for ␤Ј-COP and PKC⑀ correlated (p Ͻ 0.01). Moreover, using a defined visual threshold to isolate background from signal arising from brighter discrete cellular structures gave a highly significant overlap between ␤Ј-COP and PKC⑀ staining (p Ͻ that position; numbers in parentheses indicate the number of amino acids of the motif that deviate from the consensus sequence and are marked in the motif with a lower-case boldface letter. (Alignment was carried out by eye.) C, PKC⑀ binds to N-terminal, C-terminal, and full-length ␤Ј-COP in vitro. N-terminal (amino acids 1-449), C-terminal (amino acids 425-906), and full-length ␤Ј-COP-MBP fusion proteins were immobilized on amylose affinity columns and subjected to the column overlay assay (see "Experimental Procedures"). The recombinant protein and any associated proteins were eluted by wash buffer containing 20 mM maltose and subjected to SDS-PAGE followed by Western blot analysis using anti-PKC⑀ antibodies. These antibodies gave one major 92-kDa band that is immunoadsorbed by the immunizing peptide or recombinant PKC⑀ produced in Sf9 cells (5). PKC⑀ only binds to ␤Ј-COP in the presence of the PKC activator PS or PS/DG and is partially inhibited by Ca 2ϩ . PKC⑀ did not bind to MBP alone in the presence of activators (lane 16). The results of three independent experiments are summarized in the histogram. Binding to each fragment is expressed as percent optical density (OD) Ϯ S.E. correlated against the PS/DG/no Ca 2ϩ condition as 100%. D, binding of PKC⑀ to ␤Ј-COP is dose-dependent and saturable at ϳ10 nM. Full-length ␤Ј-COP-MBP fusion protein was immobilized on amylose columns, and column overlay assays were carried out with increasing amounts of rat brain PKC in the presence of PS and DG but no Ca 2ϩ , followed by SDS-PAGE and Western blot analysis using anti-PKC⑀ antibodies. Preincubation of PKC with 10 M pseudosubstrate peptide (PKC␤ amino acids 9 -31; a nonselective inhibitor of classical and novel PKC isozymes) for 15 min at room temperature prior to addition of PKC to the overlay assay did not inhibit PKC binding to ␤Ј-COP (lane 8) when compared with control PKC preincubated without peptide (lane 9). Averaged results of three independent experiments presented as optical density Ϯ S.E. compared with control are provided below the figure.   FIG. 2. A, the PKC⑀ V1 fragment binds N-terminal and full-length ␤Ј-COP. N-terminal (amino acids 1-449), C-terminal (amino acids 425-906), or full-length ␤Ј-COP-MBP fusion proteins were immobilized on amylose columns, and column overlay assays were carried out with 1 M PKC⑀ V1-FLAG fragment in the absence of activators. The recombinant protein and any associated proteins were eluted with wash buffer containing 20 mM maltose and subjected to SDS-PAGE followed by Western blot analysis using anti-FLAG antibodies (IBI). These antibodies gave one major band of 21 kDa that disappeared when FLAG antibodies were immunoabsorbed with FLAG peptide or another FLAG-containing recombinant protein (data not shown). The PKC⑀ V1 fragment does not appear to bind C-terminal ␤Ј-COP. Results are representative of three independent experiments, and averaged data from these three experiments are provided below the figure. B, the PKC⑀ V1 fragment inhibits binding of rat brain PKC to full-length ␤Ј-COP. Full-length ␤Ј-COP-MBP fusion protein was immobilized on amylose columns. Overlay buffer with or without 1 M PKC⑀ V1 fragment was added, and the columns were incubated for 15 min at room temperature prior to addition of rat brain PKC and PS/DG as described under "Experimental Procedures." The recombinant protein and associated proteins were eluted with wash buffer containing 20 mM maltose and subjected to SDS-PAGE followed by Western blot analysis using anti-PKC⑀ antibodies. Results are representative of three independent experiments, and averaged data from these three experiments are provided below the figure. C, activated PKC⑀ binds to ␤Ј-COP better than other PKC isozymes. Purified recombinant PKC isozymes (1 g each) expressed in Sf9 cells were activated as described under "Experimental Procedures" and applied to full-length ␤Ј-COP-MBP fusion protein immobilized on amylose columns. Binding was determined with isozyme-specific antibodies as described for B, except that only one-fourth of the eluate was loaded on SDS-polyacrylamide gel for analysis. As standard for comparison, increasing amounts of recombinant isozymes (indicated in ng above each lane) were also included in the Western blot analysis (left lanes in each panel). Using these standards, we found that Ͼ250 ng of PKC⑀ bound to ␤Ј-COP as compared with 25 ng of PKC␤I, 20 ng of PKC␦, and Ͻ1 ng of the other isozymes. Hence, PKC⑀ binding to ␤Ј-COP was at least an order of magnitude better than the binding of the other isozymes. D, activated PKC␤ does not compete for PKC⑀ binding to ␤Ј-COP. The experiment was carried out as described for C, except that activated PKC⑀ was added after preincubation of ␤Ј-COP with activated PKC␤ (1 g). 0.001). Therefore, activated PKC⑀ is in close proximity to ␤Ј-COP in cardiac myocytes, suggesting a direct association between the two molecules. Using commercially available antibodies to another coatomer protein, ␤-COP (Sigma), we observed a faint perinuclear staining, but no cross-striated staining. These data may suggest that ␤Ј-COP, but not the full coatomer complex, localizes to cross-striated structures. However, we cannot rule out the possibility that the antibody recognition site on ␤-COP is hidden by an associated protein when the coatomer complex is present on the cross-striated structures. Future studies with multiple anti-coat protein antibodies will elucidate this question.
In addition, if ␤Ј-COP is a PKC⑀-selective RACK, ␤Ј-COP should co-immunoprecipitate with PKC⑀-specific antibodies and PKC⑀ should co-immunoprecipitate with anti-␤Ј-COP antibodies. This is shown in Fig. 3F. ␤Ј-COP was co-immunoprecipitated with anti-PKC⑀ antibodies, but not with control immunoglobulin. (Immunoprecipitating anti-PKC␤ antibodies are currently not available and therefore could not be used in the assay.) In addition, although there is a similar amount of PKC␤ immunoreactivity compared with PKC⑀ immunoreactivity in this preparation, some of PKC⑀, but no PKC␤, co-immunoprecipitated with anti-␤Ј-COP antibodies. This is not seen with control immunoglobulin. These data suggest that at least some of the cellular ␤Ј-COP associates with PKC⑀ even after extensive dilution, homogenization, and cell fractionation.
␤Ј-COP assembles as part of the coatomer complex and binds to Golgi membranes in a GTP-dependent manner. To further demonstrate the interaction between ␤Ј-COP and PKC⑀, we used cell systems where ␤Ј-COP/Golgi interaction has been characterized. Using purified rat liver Golgi membranes and rat brain cytosol in binding experiments, we tested if PKC⑀ bound to Golgi membranes and determined whether association of PKC⑀ with Golgi membranes was dependent on ␤Ј-COP binding to this structure (Fig. 4). GTP␥S stimulates binding of ADP-ribosylation factor (ARF), and thus COPI coatomer binding, to Golgi membranes (10). Here, we show that the increased binding of ␤Ј-COP to the Golgi membranes seen in the presence of GTP␥S was associated with a corresponding increase in PKC⑀ binding (Fig. 4). Furthermore, treatment with brefeldin A (BFA), a fungal metabolite that inhibits GTP␥S-stimulated coatomer binding to Golgi membranes, resulted in reduced ␤Ј-COP association and a corresponding reduction of PKC⑀ association with the Golgi membranes (Fig. 4). Although 4␤phorbol 12-myristate 13-acetate treatment augments PKC⑀ binding to Golgi membranes (data not shown), PKC⑀ binding was seen without the addition of PKC activators. A plausible explanation for this is that DG or other fatty acid activators FIG. 3. ␤-COP and PKC⑀ association in neonatal cardiac myocytes. Primary cultures of neonatal cardiac myocytes were treated with 100 nM 4␤-phorbol 12-myristate 13-acetate for 5 min, stained for indirect immunofluorescence, and viewed using an Applied Biosystems confocal microscope. The white boxes indicate 5-m scale, and pseudo-color images show the following. A, anti-␤Ј-COP binding is detected with fluorescein isothiocyanate-conjugated anti-rat secondary antibodies. The panel demonstrates that ␤Ј-COP is also located at cross-striated structures, the perinucleus, and cell/cell contacts (white arrows). B, anti-PKC⑀ binding is detected with rhodamine-conjugated anti-rabbit secondary antibodies. The panel demonstrates that PKC⑀ is located at cross-striated structures, the perinucleus, and cell/cell contacts (white arrows). C, A and B are combined. Colocalization of the ␤Ј-COP and PKC⑀ antibodies is indicated in yellow. Results are representative of three independent experiments. D, no specific staining is obtained after incubation of the anti-␤Ј-COP antibody with 5 g of ␤Ј-COP. E, no specific staining is obtained after incubation of the anti-PKC⑀ antibody with 0.1 g of rat brain PKC (2,5). F, anti-PKC⑀ antibody (Ab; but not control immunoglobulins) immunoprecipitates (IP) ␤Ј-COP along with PKC⑀ from neonatal heart. Similarly, PKC⑀ (but not PKC␤) co-immunoprecipitates with ␤Ј-COP using anti-␤Ј-COP antibodies. A representative result of three independent experiments is shown.
were generated on the Golgi membrane as a result of GTP␥S activation of phospholipase D, for example. Nevertheless, these results show that the level of PKC⑀ bound to Golgi membranes is dependent on factors that affect ␤Ј-COP/Golgi association. Furthermore, these data are in agreement with our hypothesis that PKC⑀ associates with ␤Ј-COP and suggest that PKC⑀ may have a role in coatomer function and vesicular transport.
There are substantial data indicating that some PKC isozymes (including PKC⑀) are localized on the Golgi complex and that PKC plays a role in constitutive membrane trafficking in general and in Golgi function in particular (10 -14). The mechanisms responsible for such localization and for the regulation of secretory function by PKC are unknown. We propose that a direct interaction between PKC⑀ and ␤Ј-COP is one of the mechanisms sustaining the Golgi-specific localization of some PKC isoforms. In NIH 3T3 cells overexpressing PKC⑀, PKC⑀ binds to the Golgi apparatus, at least in part, via the C1 region of PKC⑀ and appears to regulate Golgi function (20,21). It has not yet been determined whether the C1 region of PKC⑀ also binds ␤Ј-COP. However, these data are in agreement with our findings in Fig. 1, which demonstrate that the V1 fragment contains only part of the ␤Ј-COP (RACK)-binding site on PKC⑀.
The mechanism by which PKC⑀ exerts its action on secretion remains to be resolved, but a number of possibilities are raised by the direct interaction of PKC with a COPI protein. Activation of PKC enhances binding of ARF and ␤-COP to Golgi membranes and stimulates vesicle formation (10). Furthermore, it has been suggested that Golgi-associated PKC may control the activation of ARF (12). However, the PKC isozyme(s) responsible for the effects seen in these studies was not determined. This may suggest a model in which ARF, by promoting COPI binding to the Golgi complex and hence sustaining PKC⑀ localization and activity on it, initiates a positive feedback loop that reinforces its own binding to these mem-branes. In addition, the COPI coatomer subunits ␤-COP and ␦-COP appear to undergo regulated phosphorylation (22). It is therefore possible that phosphorylation of these COPI components by PKC⑀ may result in different COPI conformation. Such COPI conformation changes could result in a shift in the bimodal interaction, allowing it to engage with its receptors on Golgi membranes, and produce a shift in the anterograde versus retrograde membrane flux along the secretory pathway (23). The essential role of DG in vesicle formation (9) further supports the involvement of PKC in control of budding. A further possible target of PKC relevant to its role in regulating Golgi function could be phospholipase D, which is enriched on Golgi membranes. Phospholipase D activity is stimulated by ARF and appears to be involved in vesicle formation (24). Moreover, it has been known for some time that phospholipase D is also activated by PKC (25), although the isozyme responsible is a matter of some debate (26,27). A noncatalytic role for PKC has been proposed (28), and noncatalytic involvement of PKC␣ in porcine brain phospholipase D activation was recently suggested by Sternweis and co-workers (29). Finally, noncatalytic PKC activation of phospholipase D has been proposed to mediate vesicle coat assembly in an in vitro model system, but the isozyme required was not identified (12). Although PKC␣ and PKC␤ have been linked to phospholipase D activation, PKC⑀ has also been implicated in a number of studies (26) including one using cardiac myocytes (30).
As shown in Fig. 3, ␤Ј-COP was also found on cross-striated structures in cardiac myocytes. Since these structures are not characteristic Golgi structures, we wanted to determine the significance of this finding. Several cell types, including cardiac myocytes, appear to be resistant to BFA treatment, and hence, ␤Ј-COP association with Golgi and cross-striated structures could not be dissected. This still leaves the question of ␤Ј-COP/ PKC⑀ complex function at these structures. We previously found that PKC⑀ mediates inhibition of contraction rate (negative chronotropy) (5) and resistance to ischemia-induced cell death (31). Together with the finding that ␤Ј-COP is the anchoring RACK for activated PKC⑀, the data suggest that regulation of these functions may be mediated by secretion of autocrine factors. Alternatively, localization in cardiac myocytes of ␤Ј-COP to cross-striations in addition to Golgi membranes may suggest a novel role for ␤Ј-COP in this cell type, i.e. anchoring PKC⑀ close to substrates that are not involved in COPI function. Localization of ␤Ј-COP and PKC⑀ at crossstriated elements may bring PKC⑀ into contact with contractile proteins such as troponin (32) and/or with sarcoplasmic reticulum PKC substrates such as phospholamban (33). Finally, the complex may actually lie on T-tubules, plasma membrane invaginations that penetrate deep into the myofibrils close to the Z-lines (34).
In summary, we have shown that ␤Ј-COP fulfills the criteria for a PKC⑀-selective RACK: it binds activated PKC⑀ better than any other isozyme, and PKC binding is dose-dependent and saturable. ␤Ј-COP co-immunoprecipitates and colocalizes with activated PKC⑀ in cardiac myocytes. Finally, ␤Ј-COP mediates PKC⑀ binding to Golgi membranes. The significance of ␤Ј-COP as a PKC⑀-selective RACK remains to be fully determined, but the study of PKC⑀/␤Ј-COP interaction should lead to new insights into PKC⑀-specific functions. FIG. 4. A, shown is the binding of ␤Ј-COP and PKC⑀ to Golgi membranes. Similar to ␤Ј-COP, binding of PKC⑀ to rat liver Golgi membranes is GTP␥S-dependent. Binding was determined as described under "Experimental Procedures." Lane 1, in the absence of GTP␥S, control; lane 2, with 20 M GTP␥S; lane 3, with 40 g/ml BFA and 20 M GTP␥S. B, quantitative analysis of four experiments run in duplicate gave the following results for PKC⑀ binding when PKC⑀ binding in the presence of GTP␥S was normalized as 100%: control (without GTP␥S), 10 Ϯ 5%; and BFA and GTP␥S, 40 Ϯ 9%. Quantitative analysis gave the following results for ␤Ј-COP binding to Golgi membranes when ␤Ј-COP binding in the presence of GTP␥S was normalized at 100%: control (without GTP␥S), 15 Ϯ 7%; and BFA and GTP␥S, 35 Ϯ 10%.