KDEL Receptor (Erd2p)-mediated Retrograde Transport of the Cholera Toxin A Subunit from the Golgi Involves COPI, p23, and the COOH Terminus of Erd2p

Wild-type CTX and CTX-A were obtained from Calbiochem-Novabiochem (Bad Soden, Germany). CTX with a mutated A subunit was generated as previously described (Fontana et al., 1995). We have used a mutation in which serine₆₃ of the mature CTX-A had been replaced by a lysine (CTX–K63). The mutated protein is completely unable to ADP ribosylate polyarginine when tested according to Lai et al. (1981) and does not induce a rise of cAMP (results not shown here).

Antibodies were raised in rabbits against CTX-A and against the following peptides, which all contained an additional NH₂-terminal cysteine: COOH-LYITKVLKGKKLSLPA (COOH-terminal peptide of Erd2p), COOH-KKEAGELKPEEEITVGPVQK (residues 494–513 of β-COP) (Duden et al., 1997), COOH-RRFFKAKKLIE (COOH terminus of p23; Sohn et al., 1996), and COOH-YLKRFFEVRRVV (COOH terminus of p24; Stamnes et al., 1995). The peptides were coupled to keyhole limpet hemocyanin via the NH₂-terminal cysteines using the bifunctional reagent sulfo-SMCC (Pierce Chemical Co., Rockford, IL). The antibodies against CTX-A reacted also with the mutated CTX-A–K63. IgGs from the antisera were purified via protein A–Sepharose chromatography. The resulting IgGs were purified by affinity chromatography on the immobilized peptides or immobilized CTX-A (anti–CTX-A antibodies). The purity of the antibodies was checked by immunoblotting with extracts from Vero cells. All microinjested IgGs were equilibrated with “internal medium” (120 mM potassium glutamate, 5 mM NaCl, 5 mM MgCl₂) by gel filtration before microinjection. An mAb against p53 and antibodies against giantin came from H.P. Hauri (Biocenter, University of Basel, Basel, Switzerland) and an affinity-purified antibody against recombinant mSEC13 from B.L. Tang (Institute of Molecular and Cellular Biology, Singapore, Singapore).

Antibodies or BSA were labeled as recommended by the manufacturer with the disuccinimides of Cy2, Cy3, or Cy5 to a molar dye/protein ratio of 1:1 to 2:1. Cy dye–labeled primary antibodies allowed double- and triple-immunofluorescence studies without the need for secondary antibodies. Staining for p24 and mSEC13 was performed using the unlabeled primary antibodies and fluorescein- or TRITC-labeled goat anti–rabbit secondary antibodies.

F_ab fragments were prepared by digestion of the corresponding IgGs with agarose-immobilized papain (Sigma-Aldrich Chemie GmbH, Deisenhofen, Germany) followed by blocking of free SH groups with iodoacetamide and removal of agarose–papain by centrifugation. Complete digestion of IgGs was controlled by SDS-PAGE (Fig. 1). Iodoacetamide was removed and the F_ab fragments equilibrated with “internal medium” by gel filtration. Dye-labeled F_ab fragments were prepared in the same way from dye-labeled IgGs. F_ab fragments were used within 12 h after preparation.

Bacculo-Gold™ DNA was obtained from PharMingen (Hamburg, Germany).

All transport studies were performed with Vero cells, which had been grown on coverslips to ∼70% confluency. Binding of WT–CTX (0.5 μg/ml) or the CTX–K63 (0.5 μg/ml) was performed at 0°C. Following removal of the unbound toxin the uptake was initiated and the intracellular transport followed at 37°C and 5% CO₂ as described previously (Majoul et al., 1996). Unless otherwise mentioned, antibodies or GTP-γ-S were either injected 20 min before addition of CTX, or 15–20 min after start of CTX uptake, when some of the CTX-A had already reached the Golgi (Majoul et al., 1996).

As none of the microinjected IgGs affected the plasma membrane– Golgi transport of CTX-A–K63, F_ab fragments were microinjected immediately before addition of CTX–K63 to the cells. For microinjection coverslips were transferred to DME, 10% FCS, 10 mM Hepes, pH 7.4, in a 3.5-cm Petri dish with a central hole (1-cm-diam) that had been closed from the bottom side by a glued glass coverslip. Microinjection was then performed over a period of ≤5 min using a semi-automatic micromanipulation and injection unit and Eppendorf femtotips (Eppendorf-Netheler-Hinz GmbH, Hamburg, Germany). After microinjection, the cells were incubated for the indicated times at 37°C and 5% CO₂. Cells injected with GTP-γ-S were identified by coinjecting Cy2-labeled BSA. Cells microinjected with antibodies or F_ab fragments directed against β-COP, p23, or Erd2p were identified by coinjection of the same antibodies or F_ab fragments labeled with Cy2. The ratio of unlabeled IgGs or F_ab fragments to Cy-labeled IgGs or F_ab fragments was ∼3:1. At the appropriate time points, cells were fixed with 3.5% PFA, permeabilized with 0.1% (wt/vol) saponine, and then immunostained as previously described (Majoul et al., 1996).

Standard immunofluorescence was performed with a Zeiss Axioplan microscope (Carl Zeiss, Oberkochen, Germany) equipped with a Zeiss Plan Neofluar 40×/0.75 objective and a Plan Neofluar 100×/1.30 oil objective. Cy2 and Cy3 were exited at 488 and 514 nm, respectively. Images were collected with a digital CF8/1DX camera (Kappa, Reinhausen, Germany).

Confocal laser scanning microscopy was performed with a Zeiss LSM410 microscope with a 40× 0.9 Plan Neofluar objective and a 63× 1.4 Plan Neofluar objective. Excitation was performed at 488 nm (argon laser, Cy2), 543 nm (Cy3), and 633 nm (Cy5) (both Helium/Neodym laser). The following emission filters were used: LP 515 (Schott, Mainz, Germany) or BP530 (Omega Optical, Brattleboro, VT) for Cy2, LP 570 (Schott) or 543BP12 (Omega Optical) for Cy3, and LP665 (Schott) for Cy5. Reconstruction of images was performed using the NIH-Image software and Photoshop software (version 4 for Macintosh; Adobe Systems, Inc., Mountain View, CA). Background intensity quantitation was performed using the Scil Image software (Technical University, Delft, The Netherlands).

The cDNA coding for human Erd2p was cloned into the Bacculovirus transfer vector and Sf9 cells were co-transfected with this vector together with BacculoGold DNA (PharMingen). The transfected cells were cultivated for 4–5 d in TNM FH medium at 27°C. Virus particles were isolated by centrifugation and stored at −80°C. Sf9 cells grown to 80–90% confluency were harvested and seeded at a density of 1.9 × 10⁷ cells/plate (15 cm). After attachment of the cells they were infected with virus stock solution, cultivated for 3 d, and then harvested. The Sf9 cells overexpressing Erd2p were homogenized with a glass-glass Potter homogenizer in 0.25 M sucrose, 10 mM Tris-Cl, pH 7.4 (five times the volume of the sedimented cells). The homogenate (800 μl) was mixed with 900 μl of 2.3 M sucrose, 10 mM Tris-Cl, pH 7.4, and 100 μl 100 mM Na₂EDTA and pipetted onto the bottom of a Beckman SW50 centrifuge tube (Beckman Instruments GmbH, München, Germany), followed by overlays with 1.2 M sucrose, 10 mM Tris-Cl, pH 7.4, 0.8 M sucrose, 10 mM Tris-Cl, pH 7.4. After centrifugation for 3.5 h at 90,000 g, the Golgi-enriched fraction was recovered from the 1.2 M sucrose/0.8 M sucrose interphase. Golgi-enriched fractions were mixed with an equal volume of 10 mM Hepes-KOH, pH 7.5, and spun for 1 h at 100,000 g. The pellet from 600 μl of a Golgi-enriched fraction was suspended in 200 μl of 0.1 M Na₂CO₃ for 10 min and spun again for 30 min at 100,000 g. The sediment was resuspended in 200 μl 10 mM Hepes-KOH, pH 7.5, and used for the binding assays.

The binding assay was performed in cacodylate buffer as described by Wilson et al. (1993). Although the maximum binding of CTX-A and of KDEL peptides to Erd2p occurred at pH 5 under our conditions (confirming the data presented first by Wilson et al. [1993]), we have observed (unpublished results) that the decrease in binding upon raising the pH is much less for ¹²⁵I-KDEL proteins than for various ¹²⁵I-KDEL peptides. To study the effects of anti-Erd2p antibodies on the binding of ¹²⁵I–CTX-A, the permeabilized Golgi vesicles were preincubated at pH 7.2 with anti-Erd2p IgGs at room temperature for 30 min. Control vesicles were either incubated in buffer or with the same dilution of IgGs from the preimmune serum. After preincubation, the samples were brought to pH 6 with cacodylate, and the binding assays performed with different concentrations of ¹²⁵I–CTX-A. Controls with ¹²⁵I–CTX-A but without membranes were subtracted. Using Iodobeads CTX-A was ¹²⁵I-labeled in PBS at pH 7.4 to a specific radioactivity of 2.2 μCi/μg.

After binding to the ganglioside GM1 via its B subunits, CTX undergoes cellular uptake (Janicot and Debuquois, 1987; Sofer and Futerman, 1987; Tran et al., 1987; Nambiar et al., 1993; Orlandi et al., 1993) via caveolae (Orlandi and Fishman, 1998; Wolf et al., 1998). In a late endosomal or a Golgi compartment, CTX-A and CTX-B dissociate from each other (Bastiaens et al., 1996). Only CTX-A is transported further by retrograde transport to the intermediate compartment and the ER (Lencer et al., 1995; Majoul et al., 1996). CTX induces an increase in cAMP. As increase in cAMP had been reported to stimulate transport of CTX-A to the ER (Sandvig et al., 1996), we have compared transport of WT–CTX and a mutated form of CTX–K63 that lacks any ADP-ribosylating activity and therefore does not induce an increase of cellular cAMP.

Transport of both A-subunits, CTX-A–K63 and WT– CTX-A from the Golgi to ER-like structures occurred with a similar time course (Fig. 2). This indicates that under our conditions the stimulation of adenylate cyclase by CTX (Majoul et al., 1996) does not have a significant effect on the retrograde transport of the toxin. Unless otherwise mentioned, all subsequent experiments were carried out with the CTX–K63 mutant.

GTP-γ-S has been shown to inhibit intracellular vesicular transport at various steps (Melancon et al., 1987; Schwaninger et al., 1992; Tan et al., 1992; Taylor at al., 1992; Barlowe et al., 1993; Oka and Nakano, 1994; Barthel et al., 1995; Hidalgo et al., 1995). Microinjection of GTP-γ-S immediately before addition of CTX–K63 to the medium led to a retention of CTX–K63 at the plasma membrane and in endosomal structures under conditions where control cells showed an appearance of CTX-A–K63 in the Golgi (Table I).

When GTP-γ-S was microinjected 15–20 min after start of CTX uptake, i.e., when most of CTX-A had appeared in the Golgi (Majoul et al., 1996), CTX-A–K63 remained in perinuclear compartments or in perinuclear tubular-vesicular structures over the following 85 min, whereas under control conditions most of the CTX-A–K63 had arrived in the ER (Table II). Thus, GTP-γ-S inhibits internalization of CTX as well as transport of the A subunit of CTX from the Golgi to the ER.

Arrival of CTX-A–K63 in the Golgi led to a dispersed distribution of Erd2p, which starts already 15 min after start of CTX–K63 uptake (Fig. 3). Giantin, a resident membrane protein of the Golgi (Linstedt and Hauri, 1993) remained in the more condensed perinuclear structures characteristic for Golgi cisternae under conditions where Erd2p showed a CTX-induced dispersion (Fig. 3). Apparently the increased occupancy of Erd2p resulting from the “wave” of CTX-A–K63 had led to a transient change of the steady-state distribution of Erd2p, most likely by an enhanced budding from the Golgi of vesicles containing occupied Erd2p. The dispersion of Erd2p was accompanied by a reproducible shift in the distribution of the ERGIC marker p53 (Fig. 4).

While the distribution of Erd2p and mSEC13 (a member of the COPII complex), was clearly different in the absence of CTX (Fig. 5), Erd2p assumed an increased colocalization with mSEC13 between 45 and 100 min after start of CTX–K63 uptake (Fig. 6).

To analyze for a possible contribution of the cytoplasmically oriented COOH terminus of Erd2p to the sorting and the retrograde transport of the occupied Erd2p we analyzed the effect of antibodies raised against the COOH terminus of Erd2p on the retrograde transport of CTX-A–K63. Anti-Erd2p IgGs injected before start of CTX– K63 uptake, did not significantly affect the retrograde transport from the plasma membrane to the Golgi (Table I). However, when the IgGs were injected 10–20 min after start of CTX–K63 uptake, we observed in ∼60% of the cells an accumulation of CTX-A–K63 in perinuclear structures and an inhibition of translocation of CTX-A–K63 from the Golgi to the ER (Table I). When anti-Erd2p F_ab fragments were injected 5 min before start of CTX uptake, CTX-A–K63 became also retained in Golgi-like structures (Fig. 7) in 12 out of 15 randomly chosen intact microinjected cells. Microinjection of BSA or preimmune IgGs did not have a significant effect under the same conditions (Table II). The CTX-K63–induced translocation of Erd2p as well as that of p53 was also strongly inhibited by microinjected anti-Erd2p F_ab fragments (Fig. 8).

To exclude the possibility that the anti-Erd2p–induced inhibition of retrograde transport of CTX-A–K63 from the Golgi to the ER had resulted from a decreased binding of CTX-A–K63 to Erd2p, we measured the effect of the anti-Erd2p antibodies on the binding of ¹²⁵I-labeled CTX-A to permeabilized Golgi vesicles from Sf9 cells overexpressing human Erd2p. The antibody binding was performed at pH 7.2, the binding of CTX-A to Erd2p at pH 6. At pH 6 the binding of CTX-A to Erd2p was still sufficiently high (∼60% of the binding observed at pH 5, unpublished results). Preincubation of the permeabilized Golgi membranes with anti-Erd2p antibodies did not affect binding of CTX-A to Erd2p (Fig. 9). Binding of CTX-A was specific as CTX-A could be displaced from the membranes by low concentrations of KDEL peptides but not by peptides without a COOH-terminal KDEL sequence (results not shown here). Therefore, inhibition of retrograde transport of CTX-A from the Golgi by microinjected anti-Erd2p antibodies results from an impaired sorting of the occupied receptor into COPI-coated vesicles or from an impaired vesicle budding rather than from a decreased binding of CTX-A to Erd2p.

It has been shown previously (Pepperkok et al., 1993) that a specific anti–β-COP antibody (anti-EAGE; Duden et al., 1991; Pepperkok et al., 1993) inhibits the anterograde transport of the VSV-G protein from the ER to the Golgi. If the occupied Erd2p would undergo retrograde transport from the Golgi to the ER via COPI-coated vesicles, anti-EAGE antibodies should also inhibit this transport step. When affinity-purified antibodies were injected before CTX–K63 uptake the transport of CTX-A–K63 to the Golgi as measured 30 min after start of CTX uptake was not impaired (Table I). When antibodies were injected ∼20 min after start of CTX–K63 uptake and subsequently incubated for another 80 min before fixation and analysis by immunofluorescence CTX-A–K63 was mostly retained in compact perinuclear structures similar to those stained with antibodies against Erd2p, indicating that CTX-A–K63 was retained in the Golgi (Table II) in contrast to non-injected or mock-injected cells where CTX-A–K63 at this time point was observed in reticular structures, sometimes also in structures resembling the intermediate compartment (p53 compartment) (Table II). When the antibodies were injected later, namely 45–60 min after start of CTX uptake, only very little or no inhibition of transport of CTX-A–K63 from the Golgi to the IC and the ER was observed (results not shown here), indicating that once coatomer recruitment and vesicle budding had occurred, the anti–β-COP antibody was without effect.

An inhibitory effect of anti–β-COP IgGs due to unspecific “cross-linking” could be excluded as anti-EAGE F_ab fragments injected 5 min before the start of CTX uptake, inhibited also the translocation of CTX-A–K63 from the Golgi to the ER in 16 out of 20 randomly chosen intact microinjected cells (Fig. 10).

Recruitment of COPI proteins to the Golgi membrane seems to require not only ADP ribosylating factor (ARF) GTP, but also their direct interaction with the cytoplasmically oriented COOH termini of either transmembrane cargo proteins or of proteins of the p24 protein family (Stamnes et al., 1995; Fiedler et al., 1996; Sohn et al., 1996). It appears that members of the p24 protein family are particularly involved in the recruitment of COP proteins and thus in the budding of transport vesicles (for a review see Gaynor et al., 1998). Recently, Blum et al. (1996), Sohn et al. (1996), and Rojo et al. (1997) have described a new member of this family, designated p23. Sohn et al. (1996) have demonstrated that the interaction of p23 with coatomers does not only depend on its COOH-terminal KKLIE sequence but also on two sequential phenylalanines positioned 8 and 9 amino acids upstream from the COOH-terminal glutamic acid. They have proposed that p23 might be particularly involved in vesicular transport from the Golgi back to the ER (Sohn et al., 1996). This was supported by the observation that p23 was enriched in Golgi-derived COPI-coated vesicles (Sohn et al., 1996). As the role of p23 in retrograde transport of KDEL proteins from the Golgi to the ER has not been studied so far, we have analyzed the effects of microinjection of antibodies directed against the COOH terminus of p23 on the transport of CTX–K63. When injected before uptake of CTX– K63, no significant effect of the anti-p23 IgGs on transport of CTX-A–K63 from the plasma membrane to Golgi-like structures was observed (Table I), a result similar to that observed with anti–β-COP antibodies. This indicates that neither β-COP nor p23 are involved in a rate-limiting step of transport of CTX–K63 to the Golgi compartment.

In further experiments, we microinjected anti-p23 IgGs ∼15–20 min after start of CTX–K63 uptake, when a considerable amount of CTX-A–K63 had already reached the Golgi. Cells were then incubated for another 80 min and subsequently analyzed for the distribution of CTX-A–K63. As a result, a significant inhibition of the translocation of CTX-A from the Golgi to the intermediate compartment and to the ER occurred (Table II). CTX-A–K63 accumulated in structures overlapping with the Golgi staining by anti-p23 antibodies. In view of the high concentration of p23 in Golgi or IC membranes (Sohn et al., 1996; Rojo et al., 1997) cross-linking of p23 by the anti-p23 IgGs had to be considered as the reason for an unspecific effect on the retrograde transport of CTX–K63. However, microinjection of anti–p23 F_ab fragments led also to an inhibition of the translocation of CTX-A–K63 from the Golgi to the ER in 13 out of 15 randomly selected intact microinjected cells (Fig. 11). The distribution of CTX-A–K63 in the presence of anti-p23 IgGs or F_ab fragments resembled that observed with anti–β-COP antibodies.

p24 shares with p23 the di-phenylalanine motif in the COOH terminus. It does not possess a COOH-terminal KKXX or KKXXX motif (Stamnes et al., 1995), but appears to bind coatomer (Fiedler et al., 1996). The cytoplasmic tail of p23 (Nickel et al., 1997) but not that of p24 (Fiedler et al., 1996) is able to retrieve the corresponding fusion proteins with CD8 (CD8-p23, CD8-p24) from post-ER compartments to the ER. Therefore, we have compared the effects of anti-p23 antibodies with those of anti-p24 antibodies on the retrograde transport of CTX-A–K63 from the Golgi to the ER. When antibodies raised against a peptide comprising the 12 COOH-terminal amino acids of p24 were injected into Vero cells about 20 min after the start of CTX–K63 uptake, transport of CTX-A–K63 from the Golgi to the ER was not significantly impaired (result from 20 randomly evaluated microinjected cells) although the binding of the antibodies to Golgi-like structures occurred (Fig. 12). Injection of anti-p23 into cells from the same coverslips was clearly inhibitory (Fig. 12).

After uptake of CTX, its A subunit is dissociated and transported from the Golgi to the ER as an individual protein (Majoul et al., 1996). Specificity of this transport is conferred by the COOH-terminal sequence KDEL that is recognized within the lumen of the Golgi by Erd2p. The demonstration of Erd2p in Golgi-derived COPI vesicles (Sönnichsen et al., 1996; Orci et al., 1997) has indicated that the retrograde transport of Erd2p could occur via COPI-coated vesicles. However, the direct involvement of COPI vesicles in the retrograde transport of a KDEL protein from the Golgi to the ER has shown here for the first time in a functional assay using the retrograde transport of a KDEL protein. Microinjected GTP-γ-S inhibited the transport of CTX-A from the plasma membrane to the Golgi as well as the transport from the Golgi to the ER.

To date it is unclear whether transport steps of the early secretory pathway are constitutive or regulated, e.g., stimulated by an increased occupancy of receptor molecules involved. The transient translocation of Erd2p after uptake of CTX–K63 indicates that the increased occupancy of Erd2p enhances the retrograde transport of the occupied receptor. The fact that giantin did not participate in this translocation but kept its Golgi-like distribution underlines the specificity of the sorting of Erd2p into retrograde vesicles under the conditions of our experiment.

Whereas before internalization of CTX–K63, Erd2p was located mainly in Golgi-like structures that clearly differed from the punctated structures characterized by mSEC13, the arrival of CTX-A–K63 in the Golgi induced a preferential colocalization of Erd2p with mSEC13 (Fig. 6). Some of the punctated structures harboring Erd2p as well as mSEC13 may represent COPII vesicles transporting Erd2p from the IC back to the cis-Golgi.

Townsley et al. (1993) have shown previously that mutations in very different positions of Erd2p inhibit or abolish the translocation of the overexpressed Erd2p from the Golgi to the ER. Here we show for the first time that IgGs and the corresponding F_ab fragments directed against the COOH-terminal part of the cytoplasmic tail of Erd2p inhibit the retrograde transport of a KDEL ligand. Interestingly, anti–Erd2p F_ab fragments inhibited or abolished not only the CTX-K63–induced dispersion of Erd2p itself, but also that of p53, a protein known to participate in the binding of COPI proteins (Tisdale et al., 1997). This indicates that the increased occupancy of Erd2p does not simply induce an increased sorting of the occupied Erd2p into COPI-coated vesicles but leads also to an increased formation of these vesicles. How Erd2p might participate in the recruitment of coatomer to the Golgi membrane or in vesicle budding remains to be established. It may be that the interaction of the antibodies with the COOH terminus of Erd2p interferes with the homo-oligomerization of the occupied Erd2p. Such a homo-oligomerization has been proposed by Townsley et al. (1993), but has not yet been demonstrated. Our binding studies with ¹²⁵I–CTX-A make it unlikely that the inhibition could have resulted from a decreased binding of our cargo KDEL protein (CTX-A–K63) to Erd2p.

On the basis of microinjection experiments with IgGs against β-COP it had been concluded, that COPI-coated vesicles were involved in anterograde transport at some step between ER and Golgi (Pepperkok et al., 1993). On the other hand, inhibition of COPI-directed ER–cis-Golgi transport in yeast is cargo selective, whereas inhibition of COPII-directed transport is not (Gaynor and Emr, 1997). This could indicate that the inhibition of anterograde ER– Golgi transport had resulted indirectly from an inhibition of COPI-dependent retrograde transport from the cis-Golgi to the ER leading to a depletion in the ER of factors necessary for sorting and/or packaging of cargo into COPII-coated vesicles. Here we show for the first time that antibodies or F_ab fragments directed against β-COP inhibit indeed a retrograde transport step from the Golgi towards the ER. The fact that anti–β-COP antibodies had little or no inhibitory effect on CTX-A–K63 transport when applied after most of CTX-A–K63 had left the Golgi (45–60 min after start of CTX–K63 uptake), indicates that the antibodies act most likely by inhibition of either coatomer recruitment or the budding process itself.

Although a function of COPI proteins, notably β-COP in early endocytosis, has been described (Guo et al., 1994; Daro et al., 1997), we did not observe an inhibition of transport of CTX-A–K63 from the plasma membrane to the Golgi by anti-EAGE antibodies. This would be in line with CTX uptake occurring via caveolae (Orlando and Fishman, 1998; Wolf et al., 1998) rather than via early endosomes.

p23, a member of the p24 protein family (Sohn et al., 1996) was proposed to function as a coatomer receptor in the Golgi, because the protein is strikingly enriched in COPI vesicles when compared with their donor Golgi membranes and its cytoplasmic domain binds to coatomer (Sohn et al., 1996). Therefore, inhibition of retrograde Golgi–ER transport of CTX-A–K63 by microinjected antibodies directed against the cytoplasmic tail of p23 results most likely from inhibition of coatomer recruitment or of budding of COPI vesicles. The role of p23 in the recruitment of coatomers has been questioned by Rojo et al. (1997) who describe a de-enrichment of p23 in COPI-coated buds and vesicles and no inhibition of coatomer binding to p23-enriched membranes by antibodies directed against the COOH terminus of p23. However, under the conditions applied by Rojo et al. (1997) one might expect a decreased binding of anti-p23 antibodies to buds and COPI-coated vesicles when analyzed by immuno-EM as the binding of coatomer to p23 most likely competes with the binding of antibodies to p23. Nevertheless, these authors observed an inhibition of the anterograde transport of VSV-G by this anti-p23 antibody.

In contrast to the inhibition of VSV-G transport by anti-p23 IgGs, Rojo et al. (1997) could not observe a transport inhibition by F_ab fragments of the same antibodies. From that they concluded that part of the transport inhibition by anti-p23 IgGs might have resulted from a physical change in the IC membranes because of a cross-linking of p23 molecules by the bivalent antibodies. In our experiments, however, we observed a clear-cut inhibitory effect of anti– p23 F_ab fragments on CTX-A–K63 transport. The discrepancy may have technical reasons.

p24, another member of the p24 protein family which is found under steady-state conditions mainly associated with cis-Golgi structures but also with COPI vesicles, seems likewise to be able to bind coatomer (Stamnes et al., 1995; Fiedler et al., 1996; Dominguez et al., 1998) although it possesses a COOH-terminal RRVV instead of a KKXX motif. In our experiments, microinjection of antibodies directed against the cytoplasmic domain of p24, in contrast to p23, did not inhibit translocation of the retrograde KDEL cargo protein, although the antibodies bound efficiently to p24 (Fig. 12). We raised these antibodies against the 10 COOH-terminal amino acids of the cytoplasmic tail of p24 that consists of only 12 amino acid residues, expecting that their binding in vivo to p24 should compete with any physiological function associated with the COOH-terminal tail of p24. However, we have to consider the possibility that the discrepancies between the actions of anti-p23 and anti-p24 antibodies had simply resulted from an insufficient binding or neutralizing action of the anti-p24 antibodies. Further experiments are needed before it can be concluded that p23, but not p24 is involved in COPI-mediated retrograde transport from the Golgi.

The authors wish to thank R. Duden, Cambridge, UK for extensive and helpful discussions; and H.P. Hauri, Basel, Switzerland and B.L. Tang, Singapore for the generous gift of antibodies against giantin and p53, and against mSEC13, respectively.

This work was supported by grants by of the Deutsche Forschungsgemeinschaft (grant No. 43/56-1) and the Fonds der Chemischen Industrie given to H.-D. Söling.

BFA

brefeldin A

COPI and COPII

coat protein I and II

CTX

cholera toxin

CTX-A

cholera toxin A subunit

IC

intermediate compartment