Membrane Recruitment of Coatomer and Binding to Dilysine Signals Are Separate Events*

It has previously been shown that transport of newly synthesized proteins and the structure of the Golgi complex are affected in the Chinese hamster ovary cell line ldlF, which bears a temperature-sensitive mutation in the Coat protein I (COPI) subunit e -COP (Guo, Q., Vasile, E., and Krieger, M. (1994) J. Cell Biol. 125, 1213–1224; Hobbie, L., Fisher, A. S., Lee, S., Flint, A., and Krieger, M. (1994) J. Biol. Chem. 269, 20958–20970). Here, we pin-point the site of the secretory block to an intermediate compartment between the endoplasmic reticulum (ER) and the Golgi complex and show that the distributions of ER-Golgi recycling proteins, such as KDEL receptor and p23, as well as resident Golgi proteins, such as mannosidase II, are accordingly affected. At the nonpermissive temperature, neither the stability of the COPI complex nor its recruitment to donor Golgi membranes is affected. However, the binding of coatomer to the dil-ysine-based ER-retrieval motif is impaired in the absence of e -COP, suggesting that dilysine signal binding is not the major means of COPI recruitment. Because expression of the exogenous chimera of e -COP and green fluorescent protein in ldlF cells at nonpermissive temperature rapidly restores the wild type properties, e -COP is likely to play an important role in the cargo selection events mediated homogenization.

of newly synthesized proteins and in the maintenance of Golgi structure at the nonpermissive temperature (33,34). At nonpermissive temperature, ⑀-COP is rapidly degraded without affecting the levels of the other subunits, except for an ϳ2-fold reduction in ␣-COP (35,36). These phenotypes are corrected by stably transfecting wild type ⑀-COP (33,34,37). We have used this ldlF cell line as a tool to examine the specific role of ⑀-COP in the COPI complex in the biosynthetic pathway. We find that at nonpermissive temperature, transport to the plasma membrane of the temperature-sensitive mutant of the glycoprotein of vesicular stomatitis virus (ts-045-G) is blocked in an ER/ Golgi intermediate compartment. In addition, we show that the distributions of recycling markers, such as KDEL receptor and p23, or resident Golgi proteins, such as mannosidase II, are also affected. At nonpermissive temperature, neither the stability of the COPI complex nor its recruitment to donor membranes is altered. However, the binding of coatomer to the dilysine ER-retrieval motif is impaired, indicating that coatomer recruitment to donor membranes and cargo selection are not identical but separate events. Because microinjection and expression of ⑀-COP-GFP cDNA in ldlF cells at nonpermissive temperature rapidly restores secretory function and Golgi structure, ⑀-COP is most likely involved in sorting/retrieval events in the early secretory pathway between the ER and Golgi and in the maintenance of Golgi structure.

EXPERIMENTAL PROCEDURES
Antibodies-Antibodies (Abs) used in this work were as follows: Abs against synthetic EAGE peptide of ␤-COP, which were prepared and affinity purified as described (38,39); murine monoclonal Ab against the peptide D1 (maD), which was prepared as described (39); monoclonal Ab P5D4 directed against the COOH terminus of VSV-G (40); rabbit anti-␣-, anti-␥-, and anti-␦-COPs (from Drs. C. Harter  COP cDNAs-The ⑀-COP-GFP construct has been previously described (37); the -COP-GFP construct was made by J. A. Whitney, 2 and the ts-045-G cDNA was obtained from C. Machamer. The Myc-␤-COP construct contains the SalI to SacII fragment of Myc-␤-COP from pGEM-Myc-␤-COP, subcloned into the blunted HindIII site of a modified version of pSG5; junctions were verified by sequencing (Amersham Pharmacia Biotech T7 kit).
Cell Culture-Wild type CHO cells and the temperature-sensitive mutant ldlF cells were grown in ␣-modified minimum Eagle's medium (Sigma) supplemented with 1% L-glutamine, 1% nonessential amino acids and 10% FCS (all from Life Technologies, Inc.) at 34°C. For BFA treatment, BFA (Roche Molecular Biochemicals) was used at 10 M from a 1 mg/ml stock in methanol in low carbonate cell medium (Life Technologies, Inc.).
Preparation of Cytosol from Cells-Cells from twenty 10-cm dishes were washed twice with cold homogenization buffer (0.15 M NaCl, 20 mM Tris, pH 7.5, 2 mM EDTA), scraped into 1 ml each of homogenization buffer containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 0.5 mM 1:10 phenanthroline, 2 M pepstatin A, 2 g/ml aprotinin, 0.5 g/ml leupeptin), and passed 20 times through a 25 gauge needle. The resulting homogenate was centrifuged at 15,000 rpm for 15 min at 4°C in a microcentrifuge, and the supernatant was collected and centrifuged again at 100,000 ϫ g for 1 h at 4°C. The resulting high speed supernatant (cytosol) was used for experiments. For preparation of ldlF mutant cytosol, cells were incubated at 40°C for 10 h prior to homogenization.
Immunoprecipitation Experiments-Cells were incubated in labeling medium (9 parts of methionine/cysteine-free minimal essential medium to 1 part of normal minimal essential medium) containing 30 Ci/ml [ 35 S]methionine/cysteine (Pro-mix, Amersham Pharmacia Biotech) for 20 h at 34°C. For labeling at nonpermissive temperature (40°C), cells were incubated at 40°C during the last 10 h. Cells were washed twice with ice-cold homogenization buffer. Cells were then lysed in 800 l of immunoprecipitation buffer (homogenization buffer containing 0.5% Triton X-100 containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 0.5 mM 1:10 phenanthroline, 2 M pepstatin A, 2 g/ml aprotinin, 0.5 g/ml leupeptin) for 15 min on ice. The lysates were cleared for 10 min at 15,000 rpm in a microcentrifuge.
Lysates were incubated with 20 l of protein A-Sepharose (Amersham Pharmacia Biotech; 50% slurry in PBS) for 1 h at 4°C on a rotating wheel. The beads were removed by centrifugation, and 2 g of affinity-purified anti-␤Ј-COP (CM1A10) was added to the supernatant. After incubation for 1 h at 4°C, 20 l of protein A-Sepharose was added, and the mixture was incubated for a further 2 h at 4°C. The beads were washed five times with 1 ml of immunoprecipitation buffer and then once with 1 ml of PBS. Proteins were eluted from the beads by boiling in SDS sample buffer and subjected to SDS-PAGE analysis.
In Vitro Binding Experiments-Enriched Golgi membranes were prepared as described previously (42) and were enriched 180ϫ over the homogenate, as assayed by galactosyl transferase activity.
In vitro binding experiments were performed essentially as described previously (29). Briefly, incubations (120 l) were carried out for 10 min at 37°C in the presence of 25 mM Hepes-KOH, pH 7.0, 25 mM KCl, 2.5 mM MgCl 2 , 25 M guanosine 5Ј-3-O-(thio)triphosphate and an ATP regenerating system (50 M ATP, 2 mM creatine phosphate 12.5 units/ml creatine kinase). 13 g of Golgi-enriched membranes and 160 g (3 mg/ml) of cytosol were added as indicated in the lenged to Fig. 5. The same results were obtained whether the cytosol was freshly prepared or frozen before the binding assay. Reactions were layered onto 15% sucrose and centrifuged at 4°C in a microcentrifuge for 30 min at 15000 rpm. The membrane pellets were resuspended in 15 l of SDS sample buffer and subjected to SDS-PAGE analysis.
Coatomer Binding to Glutathione S-Transferase (GST)-Cytoplasmic Tail Fusion Proteins-cDNAs corresponding to the cytoplasmic domains of WBP1 and WBP1-SS fused to the COOH terminus of GST, kindly provided by Dr. P. Cosson (University of Geneva), were expressed essentially as described previously (10). 30 g of GST-cytoplasmic tail fusion proteins was preincubated with 50 l of glutathioneagarose beads in GST incubation buffer (50 mM Hepes-KOH, pH 7.0, 90 mM KCl, 0.5% Triton X-100) for 1 h at 4°C and washed twice with GST incubation buffer. 160 g of CHO cytosol was then added to the beads in GST incubation buffer and incubated for a further 3 h at 4°C. Beads were then washed four times with PBS 0.5%/Triton X-100 and once with PBS. The bound CHO cytosol proteins were eluted by boiling in 30 l of SDS sample buffer and subjected to SDS-PAGE analysis.
Electrophoresis and Immunoblotting-Reduced proteins were separated on SDS-polyacrylamide gels (43). For the analysis of 35 S-labeled proteins, gels were stained with Coomassie Blue, destained, washed in water for 30 min, incubated in 1 M salicylate for 20 min, and vacuum dried, and fluorographs were recorded on x-ray film (X-Omat-AR films, Eastman Kodak) at Ϫ70°C. For immunoblotting, proteins separated by SDS-PAGE were transferred onto nitrocellulose filters (Schleicher and Schuell). Filters were then incubated with primary antibodies followed by peroxidase-conjugated secondary antibodies (Cappel) and detected by ECL (Amersham Pharmacia Biotech).
Indirect Immunofluorescence-Wild type and ldlF CHO cells were plated on serum-coated coverslips 2 days before the experiment and incubated at nonpermissive temperature for 10 h to allow degradation of ⑀-COP as indicated. Cells were fixed with 3% paraformaldehyde at room temperature for 20 min, extracted with 0.1% saponin, and subsequently labeled with specific antibodies in PBS containing 0.1% saponin. The primary antibodies were visualized with fluorescein-and rhodamine-conjugated secondary antibodies. Immunolabeled cells were visualized on a Zeiss inverted fluorescence microscope (Axiovert TV135), and images were recorded with a cooled CCD camera (Photometrics CH250, Tucson, AZ). Images were further processed using Adobe Photoshop V4.0 (Adobe Systems Inc., Mountain View, CA).
Microinjection-LdlF cells were incubated at nonpermissive temperature for 8 h and then microinjected at nonpermissive temperature with either ⑀-COP-GFP (40 g/ml in PBS), -COP-GFP, Myc-␤-COP, or ts-045-G cDNAs (80 g/ml in PBS). In some experiments, co-microinjection of both ⑀-COP-GFP and ts-045-G cDNAs was performed. cDNAs were microinjected into nuclei using an automated microinjection system (Zeiss AIS) as described previously (55). Cells were then incubated for 2.5 h at 40°C. For experiments involving ts-045-G cDNA microin-2 J. A. Whitney, manuscript in preparation. jection, cells were chased at the permissive temperature of 31°C for 1 h in low carbonate medium containing 100 g/ml cycloheximide (Sigma) to prevent further protein synthesis. After fixation and extraction, cells were labeled with antibodies described under "Results" or in the Figure legends.

Transport of ts-045-G Protein Is Blocked in Transport Complex (TC)-like Structures in the Absence of ⑀-COP-To investi-
gate whether the transport defect of newly synthesized proteins in ldlF cells at nonpermissive temperature (34) was due to arrest of the proteins in an ER-to-Golgi intermediate compartment, we studied the transport of a well characterized transport marker, ts-045-G. ts-045-G misfolds and accumulates in the ER at the nonpermissive temperature (39.5°C), is rapidly folded at 31°C, and is transported to the plasma membrane (44). ldlF cells grown for 8 h at the nonpermissive temperature of 40°C to provoke the degradation of ⑀-COP were microinjected with ts-045-G cDNA, and after 3 h of incubation at the nonpermissive temperature to allow ts-045-G expression and retention in the ER, they were shifted to permissive temperature (31°C) for 1 h and processed for immunofluorescence. ts-045-G protein was transported normally to the plasma membrane upon shifting to 31°C in wild type CHO cells (detected as expected in the Golgi complex and at the cell surface, Fig. 1A, a), whereas in ldlF cells previously held at nonpermissive temperature, it accumulated into patch-like structures positive for ␤-COP (Fig. 1A, c and d). Interestingly, these structures appeared similar to the TCs described previously in Vero cells (5), into which a chimera of the temperature-sensitive glycoprotein of vesicular stomatitis virus and green-fluorescent protein (ts-G-GFP ct ) concentrates upon shifting to 15°C and which have been proposed to act as ER-to-Golgi transport intermediates.
To further investigate the nature of these TC-like structures, we first checked whether they contained Golgi-resident enzymes. At nonpermissive temperature, mannosidase II partially redistributed from the Golgi complex to ER-like structures in a BFA-like effect, as previously reported (33) (Fig. 2, c and d). However, TC-like structures were also observed to contain mannosidase II (Fig. 2c, arrows), and these were positive for ␤Ј-COP (Fig. 2d). These structures disappeared when cells were treated with BFA ( Fig. 2, e and f), further indicating that depletion of ⑀-COP has a different effect than BFA-dependent COPI inhibition.
We then examined the distribution of the dilysine-containing protein p23, which cycles between early biosynthetic compartments (16), and the KDEL receptor, known to be involved in the retrieval of ER-resident proteins from the intermediate compartment and the Golgi complex to the ER (45,46). As shown in Fig. 3, at nonpermissive temperature, KDEL receptor and p23 were redistributed from compact perinuclear structures into TC-like structures, which still mostly colocalized with ␤Ј-COP.
These data show that in conditions where ⑀-COP is absent (Fig. 1B, c), the coatomer localizes on structures that are at the same time positive for secretory proteins, Golgi enzymes, and ER-Golgi recycling proteins, suggesting that protein sorting did not proceed normally. In addition, this experiment shows that coatomer still binds to these TC-like membrane structures at nonpermissive temperature, which suggests that ⑀-COP is not required for membrane binding of coatomer in vivo (Fig. 1B, d).
The Stability of Coatomer and Its Recruitment to Membranes Are Not Affected by the Absence of ⑀-COP-Because coatomer exists as a preassembled complex in the cytosol and is recruited as such to the donor membrane (47), we asked whether the ldlF cell phenotypes observed at nonpermissive temperature might be explained by an effect on the stability of coatomer and/or its recruitment to donor membranes. All of the subunits of coatomer, except ⑀-COP, could be immunoprecipitated with an anti-␤Ј-COP antibody from lysates of ldlF cells grown at nonpermissive temperature (Fig. 4, lane 5) in amounts comparable to those from wild type CHO and ldlF cells at permissive temperature (Fig. 4, lanes 2-4), showing that coatomer does not disassemble in the absence of ⑀-COP.
Because a complex composed of the subunits ␣-, ␤-, ␤Ј-, ␥-, ␦-, and -COP was immunoprecipitable from lysates of ldlF cells at nonpermissive temperature, we examined whether this ⑀-COPdeficient complex could still be recruited to membranes by performing in vitro binding experiments. This was particularly interesting because it has been shown that in wild type cells, the ⑀-COP subunit belongs to the COPI complex when the coatomer is recruited en bloc to membranes (47). Cytosols prepared from either wild type CHO or ldlF cells grown at permissive or nonpermissive temperature were incubated with enriched Golgi membranes in the presence of guanosine 5Ј-3-O-(thio)triphosphate. Coatomer from cytosol prepared from ldlF ⑀-COP-depleted COPI Binding to Membranes and Sorting Signal cells grown at nonpermissive temperature was still recruited to membranes, although slightly less efficiently than that from wild type CHO cells and ldlF cells at permissive temperature (Fig. 5, compare lane 4 with lanes 2 and 3). As expected, this recruitment was GTP-dependent and was increased by guanosine 5Ј-3-O-(thio)triphosphate (data not shown; see Ref. 36). Thus, ⑀-COP appears not to be required for the stability of coatomer and, in agreement with the immunofluorescence data ( Fig. 1), for coatomer recruitment to membranes in vitro.
Coatomer Binding to GST-KKXX Fusion Proteins Is Inhibited in the Absence of Normal Levels of ⑀-COP-It has been shown that the ␣/␤Ј/⑀ trimer is capable of binding to KKXX fusion proteins and to Golgi membranes in vitro (10,29), suggesting that coatomer may bind membranes at least partly via dilysine-containing cargo. We investigated whether, in the absence of ⑀-COP, coatomer could still bind to KKXX-like motifs. In vitro binding experiments were performed by incubating bacterially expressed GST fused to the cytoplasmic domain of WBP1 (which contains a dilysine motif) (GST-WBP1; Ref. 10) with cytosol from wild type CHO and ldlF cells grown at permissive or nonpermissive temperature. Strikingly, the clear binding observed when using wild type CHO cytosol (Fig. 6, lane 1) was almost completely abolished with cytosol prepared from ldlF cells grown at nonpermissive temperature (Fig. 6,  lane 5). When GST-WBP1 was incubated with cytosol from ldlF cells grown at permissive temperature, binding to the GST-WBP1 was only slightly reduced (Fig. 6, lane 3) compared with binding with wild type CHO cytosol, consistent with the fact that in these cells the level of ⑀-COP is also reduced (35). As expected, there was no coatomer binding from any of the cytosols to the fusion protein GST-WBP1-SS, in which the two critical lysines residues have been mutated to serines (Fig. 6,   lanes 2, 4, and 6). Similar results were obtained using a dilysine-containing p24 family member (data not shown). These results indicate that ⑀-COP is necessary for the interaction of the COPI complex with the cytoplasmic tail of dilysine signalbearing proteins. This suggest that ⑀-COP may play a role in protein sorting in vivo and that this function may be important to the maintenance of the structure of membranes of the biosynthetic pathway. Additionally, because coatomer can still be recruited to membranes in the absence of ⑀-COP (Fig. 4), this further suggests that KKXX-protein binding is not the only means of coatomer recruitment.
ldlF Cells at Nonpermissive Temperature Rapidly Recover a Wild Type Phenotype upon Expression of Exogenous ⑀-COP-GFP-Coatomer, once formed, is very stable, with a half-life of ϳ28 h for all of the subunits (␣-to -COP), suggesting, as described previously, that little or no subunit exchange occurs between coatomer and newly synthesized COPs (28). Accordingly, short term expression of a GFP-tagged version of ⑀-COP (⑀-COP-GFP) did not lead to incorporation into endogenous coatomer in wild type CHO cells (Fig. 7, a and b), and cytosolic staining was observed in all of the cells analyzed (150 -200 cells). In ldlF cells grown at nonpermissive temperature, however, exogenously expressed ⑀-COP-GFP colocalized rapidly (within 2.5 h) with endogenous ␤Ј-COP (Fig. 7, c and d). These results suggest that ⑀-COP was able to incorporate into an "incomplete" coatomer complex devoid of ⑀-COP, consistent with the observation that ⑀-COP is the last subunit to be incorporated into nascent coatomer (28). This was specific for ⑀-COP, because exogenously expressed Myc-tagged-␤-COP and -COP-GFP were not incorporated into preexisting coatomer within this time period (Fig. 7, e and f, respectively), confirming that probably no subunit exchange occurs between assembled coatomer and newly synthesized proteins.
We then investigated whether expression of ⑀-COP-GFP in ldlF cells at nonpermissive temperature rapidly allowed recovery of the wild type phenotype. Immunofluorescence showed that mannosidase II, KDEL receptor, and p23 all recovered a normal distribution in cells microinjected with ⑀-COP-GFP cDNA (Fig. 8, a, c, and e). In addition, transport to the plasma membrane of ts-045-G was restored (Fig. 8g), indicating that ⑀-COP is essential for the proper functioning of coatomer. DISCUSSION We have taken advantage of the existence of a temperaturesensitive mammalian cell line (ldlF CHO cells) in which one of the coatomer subunits, ⑀-COP, is degraded at nonpermissive temperature to study the possible role played by this subunit in coatomer function. Our results suggest that ⑀-COP is not directly involved in the stability of the coatomer complex or in its recruitment to donor membranes. However, although coatomer still exists in the cells as a membrane-bound complex in the absence of ⑀-COP, transport from the ER to the plasma membrane of ts-045-G protein was blocked in an ER/Golgi intermediate compartment or in transport complexes. These results show that coatomer stability and recruitment to donor membranes are not sufficient for its proper function. In addition, we could show that, in the absence of ⑀-COP, binding of coatomer to the dilysine ER-retrieval motif was inhibited in vitro and that the distributions of the KDEL receptor (known to be involved in the retrieval of ER-resident proteins from the intermediate compartment and the Golgi complex) and the dilysinecontaining protein p23 were altered in vivo. At nonpermissive temperature, these proteins were redistributed to membranous structures to which the other non-⑀-COP subunits of coatomer were bound. Because exogenously expressed ⑀-COP-GFP cDNA in ldlF cells at nonpermissive temperature rapidly colocalized with endogenous coatomer and restored all of the wild type

⑀-COP-depleted COPI Binding to Membranes and Sorting Signal
properties, we conclude that ⑀-COP plays an important role in coatomer function, probably directly participating in sorting/ retrieval events.
Exactly how coat proteins are recruited to specific sites on the donor membrane to initiate budding is still unclear. Several lines of evidence suggest that cargo itself could stimulate coated vesicle formation. For example, the knockout in mice of the two mannose-6-phosphate receptors, the major cargoes of trans-Golgi network-derived clathrin-coated vesicles, results in decreased recruitment of AP-1 adaptors and clathrin to trans-Golgi network membranes (48). Furthermore, the binding of the plasma membrane AP-2 adaptors to tyrosine-based motifs was shown to be strengthened upon clathrin coat formation, suggesting that cargo recognition and coat assembly are indeed coupled (49). By analogy, it has been proposed that some members of the p24 family, which can bind via their dilysine motifs to coatomer in vitro and are found in COPI-coated vesicles (13,18,27), may act in the donor membrane as cargo receptors and coatomer and/or ARF receptors. In fact, p23 and p24 cytoplasmic tails attached to artificial liposomes of physiological lipid composition appear sufficient to recruit coatomer in the presence of ARF-GTP to these membranes (21). However, our results suggest that in ldlF cells at nonpermissive temperature, coatomer can still be recruited to membranes in vitro and in vivo, despite impaired binding to dilysine motifs, consistent with the fact that coatomer can also bind to Golgi membranes via ARF-GTP, probably through the ␤/␦ subunits (20). It seems unlikely, then, that the members of the p24 family act as the sole receptors for coatomer. It has actually been shown recently that p24 family proteins may be involved in protein sorting by modulating the COPI stimulation of the ARF GTPase activity (50), which is in agreement with the observed binding of ARF to coatomer in close proximity of the dilysine-binding domain (51). Interestingly, Zhu et al. (52) recently reported that, similarly, mannose 6-phosphate receptor tails are not essential determi-  35 S-labeled wild type CHO and ldlF lysates with CM1A10 antibodies. Proteins remaining bound to the protein A-Sepharose beads were eluted in sample buffer, fractionated by 10% SDS-PAGE, and detected by fluorography. The upper portion, containing ␣-, ␤-, ␤Ј-, ␥-, ␦-, and ⑀-COP, was exposed for 24 h, and the lower portion, containing -COP, was exposed for 5 days. Lane 1 shows that there is no significant background in wild type CHO lysate in the absence of CM1A10 antibodies. Note that in the absence of ⑀-COP, a coatomer complex consisting of ␣-, ␤-, ␤Ј-, ␥-, ␦-, and -COP was immunoprecipitated in comparable amounts to the one immunoprecipitated from lysates of wild type CHO and ldlF cells at permissive temperature (compare lane 5 to lanes 2, 3, and 4). The identities of the proteins were confirmed by Western blotting (not shown). The nature of the protein that migrates slightly faster than ⑀-COP (indicated by the dots in lane 5) is not known, but it may be a degradation product of ⑀-COP, although it does not cross-react with our anti-⑀-COP antibody. Molecular masses are indicated in kDa. nants in the initial steps of AP-1 binding to the trans-Golgi network but are most likely involved in sorting steps by modulating ARF GTPase activity. COPI complex may thus be re-cruited on Golgi membranes via ARF in its GTP bound form, and p24 family proteins may then contribute to stabilize or destabilize this primary complex due to particular sequences in their dilysine-containing (or diphenylalanine-containing) tail.
Although we showed that ⑀-COP plays a major role in sorting, we did not demonstrate a direct interaction between ⑀-COP and KKXX motifs, and we thus do not exclude the possibility that in the absence of ⑀-COP, binding to KKXX motifs was altered through the other subunits of the ␣/␤Ј/⑀ subcomplex that is thought to mediate binding to KKXX motifs. Indeed, yeast genetic studies have shown that coatomer from ␣-COP and ␤Ј-COP mutants has lost the ability to bind dilysine motifs in vitro. (11). Also, the KDEL receptor, although it is a well established cargo of COPI-coated vesicles, has not been shown to interact directly with coatomer (18,53). However, our observation that in the absence of ⑀-COP KDEL receptor redistributes into similar structures to the dilysine-containing p23 suggests that they use similar recycling mechanisms, perhaps implicating ⑀-COP. This is consistent with genetic experiments performed in yeast, implicating COPI in retrieval of both KKXX-and KDEL-containing proteins (11,12,54).  8. Exogenously expressed ⑀-COP-GFP cDNA in ldlF cells at nonpermissive temperature rapidly restores secretory function. ldlF cells at nonpermissive temperature were microinjected in their nuclei with ⑀-COP-GFP cDNA or co-microinjected with ⑀-COP-GFP and ts-045-G cDNAs and incubated for a further 2.5 h (4 h when co-microinjected with both constructs) at nonpermissive temperature. For co-microinjection experiments, cells were subsequently shifted to 31°C for 1 h in the presence of 100 M cycloheximide to prevent both neosynthesis of active ⑀-COP and further ts-045-G production. After fixation, cells were labeled with specific antibodies as indicated. Note that ldlF cells at nonpermissive temperature expressing exogenous ⑀-COP-GFP cDNA (b, d, f, and h; asterisks in a, c, e, and g) recovered a "wild type" Golgi structure (a) and proper localization of the markers p23 and KDEL receptor (c and e, respectively). Transport of newly synthesized ts-045-G to the plasma membrane was also restored in ldlF cells at nonpermissive temperature (g). Bar, 10 m.
In ldlF cells at nonpermissive temperature, not only the sorting of recycling proteins but also that of secretory proteins was perturbed. Transport to the plasma membrane of newly synthesized ts-045-G protein was indeed blocked in structures that were positive for ␤Ј-COP, and most likely also for KDEL receptor and p23. Similarly, we have previously shown that microinjection of antibodies against a synthetic peptide of ␤-COP, EAGE, blocks transport of ts-045-G to the cell surface, at the interface of the ER and the Golgi, in tubular structures containing ERGIC-53 (55), suggesting that ts-045-G protein may be arrested in similar compartments in ldlF cells at nonpermissive temperature and in EAGE-microinjected cells. These structures might represent the TCs recently described in Vero cells (5), although we cannot exclude the possibility that ts-045-G is arrested in the Golgi, which is also scattered at nonpermissive temperature in ldlF cells (33). A model was proposed (5,56) in which COPII-coated vesicles form from the ER, cluster into vesicular tubular structures, as proposed by Balch and co-workers (57,58), and fuse to form TCs. Subsequently, COPI replaces COPII, and the TCs to interact with microtubules and move toward the Golgi complex (5,37). In this model, COPI is required to retrieve material to the ER from the TC and the Golgi complex. Our observations of the phenotypes of ldlF cells are consistent with this model: we showed that coatomer in ldlF cells at nonpermissive temperature loses the ability to bind dilysine motifs in vitro, which could explain several of the defects we observed in these cells. If the steady state localizations of KDEL receptor and p23 require proper recycling between the Golgi complex and/or between the intermediate compartment and the ER, the inhibition of coatomer binding to motifs required for sorting and retrieval of these proteins would alter their distributions, as we observed. It could also explain the arrest of ts-045-G protein in an intermediate compartment between the ER and the Golgi (perhaps the TC); if the retrieval function of COPI is inhibited in ldlF cells at nonpermissive temperature, factors required for earlier steps in the secretory pathway (e.g. v-SNARES, ERGIC-53, and COPII-associated proteins) would not be removed, and transfer of cargo to the Golgi would be impaired. One of the proposed roles for the abundant p23 protein is that it contributes to the structure of the intermediate compartment (16,22); therefore, missorting of p23 induced by degradation of ⑀-COP could potentially explain the structural changes observed in the membranous compartments of the early biosynthetic pathway.
The reasons for the observed distribution of the Golgi enzyme mannosidase II in ldlF cells at nonpermissive temperature are less clear. To some extent, it is reminiscent of the effect observed with the drug BFA, the earliest detected effect of which is the removal of ␤-COP from Golgi membranes, resulting in the extension of tubules from the Golgi to the ER and the subsequent disappearance of the Golgi, with the redistribution of Golgi membrane proteins to the ER (59,60). In ldlF cells at nonpermissive temperature, Golgi structure has been shown to be dramatically altered, dissociating into vesicles and tubules (33). However, we observed that in ldlF cells at nonpermissive temperature, mannosidase II was still associated with patches positive for ␤Ј-COP that were not observed in ldlF cells at permissive or nonpermissive temperature after BFA treatment. This is consistent with the observation that dissociation of coatomer from membranes is required for BFA-induced transfer of Golgi enzymes to the endoplasmic reticulum (61). It is thus possible that the membrane-bound ⑀-COP-depleted coatomer, although it lost at least part of its sorting capacity, conserved its ability to prevent premature fusion of transport intermediate. An extension of the transport model mentioned above predicts not only that TCs mature as they travel toward the Golgi complex but also that they could potentially fuse with each other to form the cis-Golgi network (5,62). Because we have shown that TC maturation is blocked by the inactivation of ⑀-COP (this work and Ref. 5), this could also explain the observed changes in Golgi structure. This is consistent with recent data showing COPI to be necessary for the in vitro formation of vesiculo-tubular clusters (TC precursors) from transitional ER microsomes (63).
In conclusion, although a coatomer complex is still able to form and be recruited to membranes in ldlF cells at nonpermissive temperature, secretory function, protein sorting, and Golgi structure are affected. Exogenously expressed ⑀-COP-GFP, which rapidly incorporates into a nonfunctional coatomer devoid of ⑀-COP, restores all of the wild type properties, suggesting that ⑀-COP is essential for the proper functioning of coatomer. In addition, the ability of ⑀-COP-deficient coatomer to bind membranes without binding to KKXX motifs suggests that coatomer recruitment occurs prior to cargo selection rather than being directly coupled to it.