Export of Cellubrevin from the Endoplasmic Reticulum Is Controlled by BAP31

To generate cellubrevin antibodies, a cDNA was constructed encoding the NH₂-terminal cytoplasmic part of cellubrevin (amino acids [aa] 1–81) devoid of its transmembrane anchor (ceb-cyt). The PCR product was ligated into pTrcHis (Invitrogen, Carlsbad, CA), resulting in a fusion protein containing the amino terminal His(6) tag. After expression in E. coli, the protein was extracted and purified on a nickel resin (Probond; Invitrogen) as described in Chapman et al. (1994). Bound proteins were eluted with a gradient of 0–500 mM imidazole. Fractions containing the fusion protein were pooled, dialyzed against PBS, and concentrated before immunization.

The same domain of cellubrevin was expressed as glutathione-S-transferase (GST)–fusion protein and purified on glutathione–Sepharose. 20 mg of purified fusion protein were coupled to 1 g (dry weight) CNBr– Sepharose 4B according to the manufacturer's instructions (Pharmacia Biotech., Piscataway, NJ), and used for affinity purification of the antibody. Bound IgGs were eluted with 0.1 M glycine (pH 2.7), neutralized with 1 M Tris (pH 9), dialyzed against PBS, and concentrated (Centricon 30; Amicon Inc., Bedford, MA) to yield a final protein concentration of 2–4 mg/ml. For some experiments, the affinity-purified antibodies were biotinylated using sulfo-NHS-Lc biotin (EZ-Link; Pierce Chemical Co., Rockford, IL) according to the manufacturer's instructions.

Rabbit antibodies against BAP31 were raised using the COOH-terminal cytoplasmic half of the protein (BAP31, aa 137–246) fused to GST as the antigen and then purified with Rivanol (Hoechst, Frankfurt, Germany; Franek, 1986; Adachi et al., 1996). To remove antibodies reacting with GST, the antibody (10 μl) was diluted (1:2,000) in TBS with Tween-20 [0.1%] and 5% dry milk and then incubated overnight with strips of nitrocellulose loaded with 125 μg of recombinant GST by means of SDS-PAGE and electrotransfer.

The following antibodies have been described previously: monoclonal antibodies against Rab3 (clone 42.1, Matteoli et al., 1991), and Rab 5 (clone 621.1–3, Fischer von Mollard et al., 1994). Monoclonal antibodies against the transferrin receptor, secretory carrier membrane proteins (SCAMP) (Brand et al., 1991), protein disulfide isomerase (PDI) (1D3, Vaux et al., 1990), and endoplasmic reticulum–Golgi intermediate compartment (ERGIC)-53 (Schweizer et al., 1988) were given by I. Trowbridge (Salk Institute, San Diego, CA), J.D. Castle (University of Virginia, Charlottesville, VA), S. Fuller (EMBL, Heidelberg, Germany), and H.-P. Hauri (Biocenter, University of Basel, Basel, Switzerland), respectively. The polyclonal antibody against p58 was a gift of J. Saraste (University of Bergen, Bergen, Norway). Antibodies to β-coat proteins (COP) were provided by T. Kreis (University of Geneva, Geneva, Switzerland). The polyclonal antibody against calnexin was a gift of A. Helenius (Yale University, New Haven, CT). Monoclonal anti-myc (9E10) ascites fluid was purchased from Berkeley Antibody Co. (Berkeley, CA). All donkey anti– rabbit or donkey anti–mouse secondary antibody– and streptavidin– conjugates were from Jackson ImmunoResearch Laboratories (West Grove, PA).

cDNAs encoding rat synaptobrevin I, II, and cellubrevin were provided by T.C. Südhof (University of Texas, Dallas, TX). Full-length or truncated (see above) coding regions were amplified using the PCR with oligonucleotides containing BamHI and EcoRI restriction sites. The PCR products were further cloned into the BamHI–EcoRI sites of the pGex-2T vector (Pharmacia Biotech. Inc.). Fusion proteins were expressed in E. coli strain JM109 and purified as described in Chapman et al. (1994). Immobilized proteins were analyzed by SDS-PAGE and Coomassie blue staining and then the concentration of the bound protein was determined by comparison with GST (3–4 μg/μl beads). Recombinant fusion proteins were always used in subsequent binding assays.

An expression vector coding for full-length cellubrevin in pCMV2 (McMahon et al., 1993) was provided by T.C. Südhof. cDNA encoding a myc-tagged full-length cellubrevin was constructed using a sense primer including a BamHI restriction site, the nucleotide sequence encoding myc, and 12 nucleotides of the cellubrevin sequence and then amplified using PCR. The PCR product was subcloned in pCDNA3 and used for transfection of BHK-21 cells.

For expression, mouse BAP31 cDNA (full length) was cloned in the pA vector (Zhang et al., 1996). COOH-terminal–truncated or –mutated BAP31 coding sequences were obtained using the same sense primer with a SalI restriction site and the following antisense primers: 5′ GCGGGATCCTTAGACTGAGGGACCACGTAC 3′ for the BAP31 minus 4 last amino acids (BAP31-KKEE); 5′ GCGGGATCCTTATTCTTTGGTAAGGCCCTC 3′ for the BAP31 minus 24 last amino acids (BAP31-24aa); 5′ GCGGGATCCTTACTCCTCGCTGCTGACTGAGGGACCACGTAC 3′ for the BAP31 in which the two COOH-terminal lysines at -3 and -4 positions were changed in two serines (BAP31 KK/SS). Finally, a construct was made encoding only the NH₂-terminal half of BAP31(aa 1–137) that contained a myc epitope at the COOH-terminal end (residue 137; myc-BAP31TMR). This cDNA was obtained by PCR using the following antisense primer: 5′ GCGGGATCCTTAGTTCAGGTCCTCCTCGCTGATCAGCTTCTGCTCAAAGGCTTCATT 3′. All antisense primers contain a BamHI restriction site. PCR products were subcloned into the SalI–BamHI sites of the expression vector pA and used for transfection in BHK-21 cells.

Recombinant wild-type and mutant BAP31 proteins were expressed by in vitro translation in the presence of radiolabeled methionine. cDNAs encoding full-length BAP31 and the NH₂ terminal half of BAP31 (myc-BAP31TMR) were placed under control of the T7 promotor by subcloning into the SalI-BamHI restriction site of pBlueScript SK+. Radiolabeled protein was generated by coupling in vitro transcription translation using the TnT system (Promega Corp., Madison, WI) in the presence of [³⁵S]methionine and microsomes according to the manufacturer's instructions. Vector without insert was used as a control. Aliquots (5 μl) were diluted 40-fold in extraction buffer (1% Triton X-100 in 75 mM KCl, 2 mM EDTA, and 20 mM Hepes-KOH, pH 7.4 [unless indicated otherwise]) and incubated for 1 h at 4°C. Extracts were cleared by centrifugation (at 90,000 g_max), for 20 min in a rotor (model TL100.3; Beckman and Dickinson Co., Mountain View, CA) and then used for binding assays with recombinant synaptobrevin fusion proteins (see below).

For preparation of microsomal extracts, rat liver was homogenized in homogenization buffer (75 mM KCl [unless indicated otherwise] containing 20 mM Hepes-KOH, pH 7.3, and 2 mM EDTA) using a teflon–glass homogenizer and centrifuged at 15,000 g_max for 15 min. The low speed supernatant was spun at 230,000 g_max for 60 min in a Beckman TL100.3 rotor, and the resulting pellet was used for extraction. When BHK-21 cells were used, homogenization was done with a cell cracker (clearance 0.0010”). Pellets were stored at −70°C until use.

For the preparation of detergent extracts, the pellets were extracted for 1 h at 4°C in ice-cold homogenization buffer containing 1% Triton X-100 (1 mg/ml of protein) and centrifuged at 90,000 g_max for 20 min in a Beckman TL100.3 rotor to remove unsolubilized material. 40 μl of glutathione–Sepharose (60% suspension in extraction buffer) containing the immobilized fusion proteins were added to 1 ml extract and incubated overnight (4°C) under slow rotation. The beads were washed four times in extraction buffer and resuspended in a small volume of PBS. To elute bound proteins, the immobilized fusion proteins were cleaved from the beads using two NIH U/thrombin (Sigma Chemical Co., St Louis, MO) per 100 μl beads for 2 h at room temperature. The supernatant was adjusted to 1 mM phenylmethylsulfonyl fluoride, and 1:10 vol was analyzed by SDS-PAGE followed by silver staining or immunoblotting.

For immunoprecipitation, BHK-21 cells (control or transfected) were homogenized in homogenization buffer and centrifuged at 1,300 g_max for 15 min. The postnuclear supernatant (PNS) was extracted by adding Triton X-100 (final concentration 1%, vol/vol) for 1 h at 4°C and then centrifuging at 90,000 g_max for 20 min in a Beckman TL100.3 rotor. 10 μl of anti- myc ascites, 15 μl of affinity-purified anti-cellubrevin, or 25 μl of anti-BAP31 (whole IgG fraction) were added to 200–250 μl of extract (1 mg protein/ml), followed by overnight incubation (4°C). These amounts of antibody were sufficient for quantitative depletion of the antigen. Next, 30–40 μl of protein G–Sepharose slurry (Pharmacia Biotech Inc.) were added and then the incubation was continued for 1.5 h at 4°C. The beads were washed four times, resuspended in SDS-sample buffer free of reducing agents, and analyzed by SDS-PAGE and immunoblotting.

All fractionation steps occurred at 4°C and a cocktail of protease inhibitors (0.1 mM phenylmethylsulfonyl fluoride, 10 μg/ml trypsin inhibitor, 0.7 μg/ml pepstatin A) was added freshly to all homogenates.

Clathrin-coated vesicles were purified from rat liver as described previously (Maycox et al., 1992). For enrichment of early endosomes, BHK-21 cells were homogenized in 0.25 M sucrose containing 3 mM imidazole, pH 7.4 (HB) using a ball-bearing homogenizer (eight passages, 0.0009” clearance). Fractionation was carried out according to Gorvel et al. (1991).

Fractions enriched in intermediate compartment and ER were obtained from BHK-21 cells exactly as described by Schweizer et al. (1991), except that Nycodenz instead of Metrizamide (both from GIBCO BRL, Gaithersburg, MD) was used in the final gradient centrifugation step.

BHK-21 cells were routinely grown as monolayer cultures in Dulbecco's modified Eagle's medium supplemented with 5% fetal calf serum, 10% tryptose phosphate broth, 2 mM glutamine, Pen/Strep (1:100 final dilution; GIBCO BRL, Gruenberg et al., 1989). Tissue culture reagents were from GIBCO BRL. For immunocytochemistry, confluent cells were split (1:10), plated on glass coverslips, and allowed to grow for 18–20 h.

For transfection experiments BHK-21 cells were split (1:10) 1 d before transfection. The next day, cells were cotransfected with the vectors, encoding for myc-tagged cellubrevin and the different BAP31 constructs, or for full-length cellubrevin and myc-BAP31TMR using Lipofectamine (GIBCO BRL) exactly as described by the manufacturer. After transfection, cells were grown for 24 h in normal medium, plated on glass coverslips, and grown for another 18–20 h. When indicated, cells were treated with 33 μM nocodazole (Sigma Chemical Co.) for 2 h before fixation. Nontransfected cells were grown on glass coverslips to 50–60% confluency. Cells were rinsed three times for five min with PBS supplemented with 1 mM CaCl₂ and 1 mM MgCl₂ (PBS²⁺), fixed for 5 min in methanol at −20°C and then followed by treatment with acetone for 30 s at −20°C. The coverslips were then rinsed several times in PBS²⁺ and processed immediately for double labeling immunocytochemistry. All subsequent blocking and incubation steps with primary and secondary antibodies were done in PBS²⁺ containing 8% goat serum. For double labeling with mouse monoclonal and rabbit polyclonal antibodies, the cells were incubated with a mixture of both primary antibodies followed by washing three times in PBS²⁺ and then subsequently incubated with both secondary antibodies (5-(4,6-dichlorotriazinyl) aminofluorescein (DTAF) in combination with CY3- or lissamine–rhodamine conjugates). When biotinylated rabbit anti-cellubrevin antibody was used for double labeling, the cells were first incubated with polyclonal anti-BAP31 and CY3-conjugated donkey anti–rabbit antibodies, followed by washing and blocking with PBS²⁺ supplemented with 10% rabbit serum for 1 h. After rinsing with PBS²⁺, the cells were incubated in biotinylated anti-cellubrevin followed by FITC–conjugated streptavidin. Controls were included where either one of the first or second antibodies was omitted. At the final step cells were rinsed in PBS and distilled water and mounted in VectaShield (Vector Labs, Inc., Burlingame, CA). Confocal laser scanning microscopy was performed on a MRC-600 system (Bio-Rad Laboratories, Hercules, CA) attached to a compound microscope (Axiovert; Carl Zeiss, Inc., Thornwood, NY). Image files were converted using the Confocal Assistant software, and finally processed and annotated using Adobe Photoshop 3.0 (Adobe Systems, Inc., Mountain View, CA) and PowerPoint 6.0 (Microsoft Corporation, WA).

SDS-PAGE (Laemmli, 1970) and immunoblotting (Towbin et al., 1979) was done according to established procedures. Immunoreactive bands were visualized using the enhanced chemiluminiscence (ECL) kit (Pierce Chemical Co.), DAB reaction in the case of HRP-conjugated second antibodies (Bio-Rad Laboratories) or by the alkaline phosphatase (AP) reaction when AP–conjugated second antibodies (Bio-Rad Laboratories) were used. The method of Heukeshoven and Dernick (1985) was used for silver staining of the minigels. Proteins were quantitated according to the method of Bradford (1976), following the manufacturer's instructions (Bio-Rad Laboratories).

To search for proteins interacting with cellubrevin, recombinant GST–cellubrevin fusion protein was immobilized on glutathione–Sepharose and incubated with Triton X-100 extracts of BHK-21 cells. After washing, cellubrevin was released by thrombin cleavage to recover bound proteins. The eluted proteins were analyzed by SDS-PAGE and silver staining. A band with an apparent molecular mass of ∼30,000 eluted from immobilized GST–cellubrevin (Fig. 1,a, arrow). This band specifically interacts with cellubrevin because it was not detected when beads containing bound GST–synaptobrevin II (Fig. 1,a) or bound GST (data not shown) were used. No binding was observed in 450 mM NaCl, similar to the synaptobrevin–synaptophysin interaction (Edelmann et al., 1995). To isolate larger quantities, we incubated a Triton X-100 extract of liver membranes with immobilized GST–cellubrevin and eluted bound proteins with 450 mM NaCl. SDS-PAGE of the concentrated eluate, followed by Coomassie blue staining, revealed that the apparent molecular mass 30,000 band was the only major protein eluted from the column under these conditions (Fig. 1 b). The band was excised, further concentrated by electrophoresis (Lombard-Platet and Jalinot, 1993), and digested by trypsin followed by peptide analysis using HPLC and microsequencing. Two peptide sequences were obtained, VNLQNNPGAMEHFHML and AENEVLAMRK. Database searches revealed that the sequences matched with BAP31, an ubiquitously expressed integral membrane protein that has previously been identified as a member of a group of proteins associated with B cell antigen receptor in β lymphocytes (Kim et al., 1994; Adachi et al., 1996).

The interaction between cellubrevin and BAP31 was further characterized using antibodies specific for BAP31. To determine the influence of ionic strength on the interaction more precisely, immobilized GST–cellubrevin was incubated with extracts of BHK-21 cells or rat liver containing increasing concentrations of KCl. Binding of BAP31 decreased at KCl concentrations >140 mM and was no longer detectable at 450 mM (Fig. 2 a). Next, we tested BAP31 binding to GST-fusion proteins of synaptobrevin I, synaptobrevin II, cellubrevin, and to a cellubrevin-deletion mutant lacking the transmembrane domain (ceb-cyt). The latter mutant was chosen since we found previously that the binding of synaptobrevin to synaptophysin requires the presence of its transmembrane region (Edelmann et al., 1995).

Fig. 2 b shows that BAP31 binds not only to cellubrevin but also to synaptobrevin I. No binding to synaptobrevin II (in agreement with the data shown above) or ceb-cyt was observed. The lack of binding to synaptobrevin II is not because of inactivation of the protein, since binding of synaptophysin as well as SNAP-25 and syntaxin was observed when incubated with brain extracts (data not shown; Edelmann et al., 1995). Also, less BAP31 bound to synaptobrevin I when BHK21 cell extract was used instead of rat liver extract, possibly indicating some species difference between rat and hamster BAP31. To confirm the specificity of the interaction, we tested for several other membrane-bound proteins including the transferrin receptor, SCAMP (Brand et al., 1991), the small GTPases Rab3 and Rab5, the ER residents calnexin, PDI, and the markers for the intermediate compartment, p58 and ERGIC-53. With exception of small quantities of the transferrin receptor, none of these proteins bound to the immobilized synaptobrevins.

To further study the binding of BAP31, recombinant [³⁵S]methionine-labeled BAP31 was generated by in vitro translation. As shown in Fig. 3,a, recombinant BAP31 bound to immobilized cellubrevin and this binding was salt dependent, very similar to the native protein. Furthermore, the recombinant protein showed the same preference for cellubrevin and synaptobrevin I, although in this case weak binding to GST–synaptobrevin II was detectable. Analogous to the native protein, recombinant BAP31 did not bind to cellubrevin lacking its transmembrane domain or to GST alone (Fig. 3,b, top). To investigate which domain of BAP31 is responsible for the interaction we constructed, a BAP31 mutant which lacks the COOH-terminal cytoplasmic portion of the protein (myc-BAP31TMR). The interactions of the mutant protein were very similar to full-length BAP31, suggesting that the transmembrane regions of BAP31 are required for binding (Fig. 3 b, middle).

The experiments described above demonstrate that the interaction between cellubrevin and BAP31 is very similar to that of synaptobrevin and synaptophysin suggesting that, analogous to the synaptophysin–synaptobrevin complex, BAP31 may serve as a companion for cellubrevin. However, cellubrevin is known to be concentrated on recycling vesicles distal of the Golgi complex, whereas BAP31 contains a KKXX motif and, in preliminary experiments, appeared to reside primarily in the ER. Interestingly, BAP31 also contains a sequence motif (YDRL) that is known to be responsible for binding the medium chain of AP-2, the adaptor complex involved in recruiting clathrin molecules at the plasma membrane (Ohno et al., 1995), raising the possibility that at least a pool of BAP31 may reach post-Golgi compartments and colocalize with cellubrevin. We investigated, therefore, to what extent the subcellular localization of BAP31 overlaps with that of cellubrevin by means of subcellular fractionation and immunocytochemistry.

Subcellular fractionation was used as enrichment for recycling organelles in communication with the plasma membrane which are known to contain cellubrevin (McMahon et al., 1993; Galli et al., 1994). First, clathrin-coated vesicles were purified from rat liver (Maycox et al., 1992). When the enrichment of cellubrevin and BAP31 during fractionation was monitored by immunoblotting, a clear dissociation between the two proteins was observed (Fig. 4,a). Cellubrevin is highly enriched in clathrin-coated vesicles, in agreement with earlier observations (McMahon et al., 1993). However, virtually no BAP31 was detected in the coated vesicle fraction. Second, fractions enriched in early endosomes were prepared by flotation density gradient centrifugation (Gorvel et al., 1991). As expected, cellubrevin and two additional constituents of early endosomes, the small GTPase Rab5 and the transferrin receptor, were enriched in the early endosomal fraction (Fig. 4 b). In contrast, most of the BAP31 was recovered in the low density interface of the gradient.

To enrich the ER and intermediate compartment, we used consecutive Percoll and Nycodenz gradient centrifugation steps (F3; Schweizer et al., 1991). As shown in Figure 4 c, BAP31 cofractionated with proteins p58 and ERGIC-53, two constituents of the intermediate compartment, and with ER proteins PDI and calnexin. In contrast, cellubrevin, SCAMP, and the transferrin receptor, proteins known to localize mainly in compartments at the proximal side of the Golgi complex, were enriched in the F₂ rather than in the F₃ fraction.

Together, these results suggest that despite their biochemical interaction, the majority of BAP31 and cellubrevin reside on different organelles. To identify these organelles more precisely, we used immunocytochemistry to compare the localization of BAP31 and cellubrevin in BHK-21 cells with each other, as well as with PDI and the transferrin receptor, established markers for the ER and recycling endosomes, respectively. PDI mediates the folding of newly synthetized proteins in the lumen of the ER. It is characterized by the COOH-terminal KDEL ER retention signal and, consequently, its distribution is restricted to the ER (Vaux et al., 1990).

In methanol-fixed BHK-21 cells, BAP31 immunolabeling resulted in a reticular staining pattern that overlapped perfectly with the staining obtained for PDI (Fig. 5, first row). However, BAP31 staining was extended to a dense cluster of dots in a paranuclear region devoid of PDI staining. Treatment with 33 μM nocodazole, a drug that depolymerizes microtubules, for 2 h at 37°C led to the disappearance of the paranuclear cluster and an almost complete colocalization of BAP31 with PDI (Fig. 5, second row). The transferrin receptor did not colocalize with BAP31, with exception of the paranuclear cluster where extensive overlap was observed. (Fig. 5, third row). This cluster probably corresponds to vesicles accumulating around the microtubule organizing center (MTOC) where the transferrin receptor is known to be concentrated (Trowbridge et al., 1993; Daro et al., 1996). To establish whether BAP31 and transferrin receptor are indeed overlapping in the same organelle population, cells were again treated with nocodazole to disrupt the MTOC. Clearly different patterns were obtained: a punctate reticular staining for BAP31, and dispersed dots positive for the transferrin receptor. Taken together, we assume that BAP31 is an ER resident that probably shuttles between the ER and the intermediate compartment/cis-Golgi complex. The accumulation of BAP31-containing organelles may reflect accumulation of vesicles en route from the ER to the Golgi apparatus, a pathway that is disrupted by nocodazole. Indeed, we observed some overlap between the staining patterns of BAP31 and β-coat protein (COP), a component of COP-coated vesicles mainly accumulating in the cis-Golgi network (Oprins et al., 1993; data not shown).

To perform double labeling for cellubrevin and BAP31, rabbit antibodies for cellubrevin were affinity purified and biotinylated. Cellubrevin immunoreactivity was concentrated in the area of the MTOC. Here it overlaps with BAP31, but in peripheral areas of the cell the staining patterns were different (Fig. 6, top), even though a weakly stained reticulum positive for cellubrevin with the reticular ER pattern typical for BAP31 staining was noticeable (Fig. 6, top row, arrowheads). Double labeling for cellubrevin and PDI (Fig. 6, middle row) confirms that the proteins have different distributions, with PDI being virtually excluded from the region of the MTOC. Only a minor overlap was observed between cellubrevin and β-COP in control and nocodazole-treated cells (data not shown). All spots positive for the transferrin receptor were also positive for cellubrevin (Fig. 6, bottom), in agreement with earlier results (McMahon et al., 1993; Galli et al., 1994; Daro et al., 1996).

The experiments described so far provided evidence that native BAP31 binds with high specificity and efficiency to recombinant cellubrevin, but also showed that the major pools of the two proteins are localized to different compartments. Therefore, we investigated whether complexes of native cellubrevin and BAP31 can be directly isolated from cell extracts using immunoprecipitation with affinity-purified antibodies specific for either cellubrevin or BAP31. Equal amounts of detergent extracts derived from untreated and nocodazole-treated BHK-21 cells were incubated with excess amounts of antibodies, resulting in the disappearance of cellubrevin and BAP31, respectively, from the extract (Fig. 7,b; data not shown). As shown in Fig. 7,a (right lanes), a considerable amount of cellubrevin coimmunoprecipitated with the anti-BAP31 antibody. This amount, however, constituted only a relatively small fraction of the cellubrevin pool in these extracts (compare the amount of cellubrevin precipitated with the anti-cellubrevin and anti-BAP31 antibodies, respectively) which probably reflects cellubrevin pools in the ER that are freshly synthesized and en route to the Golgi apparatus. Coprecipitation was specific since neither PDI nor TfRn or SCAMP, the latter two being integral membrane proteins colocalizing with cellubrevin, were detectable on the beads (Fig. 7,b). Also, no differences were observed when the cells were treated with the microtubule-disrupting agent nocodazole before extraction (Fig. 7 a), demonstrating that disruption of the MTOC does not affect the amount of the cellubrevin– BAP31 complex.

The picture emerging from the data presented above suggests that cellubrevin interacts with BAP31 only during the early phases of its life cycle, i.e., during export out of the ER. We hypothesized, therefore, that BAP31 may regulate the export of cellubrevin from the ER, by serving as a sorting companion for cellubrevin which delivers cellubrevin to the Golgi and then returns to the ER by retrograde vesicular transport, for instance. To test this idea further, we generated mutants of BAP31 carrying alterations in the cytoplasmic tail, hoping that at least some of them would lead to missorting of the protein and, in turn, affect the distribution of cellubrevin.

As discussed above, the COOH terminus of BAP31 contains two different sorting motifs, a YDRL motif frequently involved in adaptin binding, and a KKXX motif. To disrupt these signals, the following mutants were constructed (Fig. 8, top): BAP31 KK/SS in which the penultimate lysines were replaced with serines; BAP31-KKEE in which the last four amino acids were deleted; BAP31-24 aa in which the last 24 amino acids (including both the YDRL and the KKXX motif) were deleted; and myc-BAP31TMR in which the entire cytoplasmic domain was deleted. Instead, a myc-epitope tag was added for detection since our BAP antibody binds to the cytoplasmic tail.

In the first series of experiments, we examined whether any of these mutations affects the interaction between BAP31 and cellubrevin. For this purpose, BHK-21 cells were cotransfected with cDNAs encoding the NH₂-terminal myc-tagged cellubrevin together with either full-length or mutant BAP31 constructs. Immunoblotting of cell extracts (Fig. 8,b) showed that all mutant proteins were recognized by the anti-BAP31 antibody and that expression levels were comparable. myc-BAP31TMR was also expressed as detected by the myc-antibody (data not shown). Equal amounts of each cell extract were immunodepleted using excess amounts of anti–myc monoclonal antibodies and then the immunoprecipitates were analyzed by immunoblotting (Fig. 8,c). All BAP31 mutants coprecipitated with myc–cellubrevin with no difference being observed between the various mutants (Fig. 8,c). This binding was disrupted upon SDS pretreatment of the cell extracts (data not shown), and likewise for the synaptobrevin–synaptophysin interaction (Edelmann et al., 1995). To test binding of myc-BAP31TMR to cellubrevin, cells were cotransfected with cDNAs coding for untagged cellubrevin and myc-BAP31TMR. Again, equal quantitative immunoprecipitation from equal amounts of cell extracts were performed with either anti-cellubrevin or anti-myc and the immune complexes were analyzed by immunoblotting (Fig. 8,d). When anti-myc was used as the depleting antibody, a fraction of cellubrevin was found in the immunoprecipitate (Fig. 8,d, left panel, beads) but the supernatant was not depleted. Similarly, a fraction of myc-BAP31TMR coprecipitated with cellubrevin (Fig. 8 d, right panel, beads) but again, the majority stayed in the supernatant. These data show that all mutants interact with cellubrevin in a similar manner, but that large pools of both proteins exist in the transfected cells that do not interact with each other.

Next, we analyzed the distribution of the mutant proteins by immunocytochemistry using confocal laser scanning microscopy. To discriminate between endogenous and transfected BAP31, we used a less sensitive detection method than in the experiment shown in Figs. 5 and 6. Control stainings of nontransfected cells with this procedure confirmed that endogenous BAP31 was barely detectable and that BAP31 staining was dominated by the mutant protein. Transfections with wild-type BAP31 (BAP31wt) and myc–cellubrevin resulted in staining patterns indistinguishable from the respective endogenous proteins (Fig. 9, a–c, compare with Figs. 5 and 6). Thus, BAP31wt had a marked reticular distribution in addition to the accumulation around the MTOC (Fig. 9, a and c). myc–Cellubrevin colocalized with BAP31wt in this paranuclear region but was differentially distributed in the cell periphery (see above; Fig. 6). As described earlier, nocodazole treatment resulted in the disappearance of the MTOC and a dispersed punctate staining pattern for cellubrevin (data not shown). We then studied the distribution of the mutants of BAP31, in which the last four amino acids were modified or deleted (BAP31KK/SS and BAP31-KKEE). No differences in the staining pattern for BAP31 and cellubrevin were observed (Fig. 9, d and e; data not shown). Similarly, loss of the COOH-terminal 24 amino acids including both sorting motifs did not alter the distribution of the proteins (data not shown), suggesting that signals upstream in the BAP31 sequence contribute to ER retention.

In contrast, striking alterations were observed in cells transfected with myc-BAP31TMR and cellubrevin. When cells were stained for the mutant using myc-antibodies, the staining in the cell periphery was still mostly reticular but occasionally larger immunopositive blebs were observed, and the vesicle cluster around the MTOC was less conspicuous with a more fragmented appearance (Fig. 9, g and j). The staining pattern of endogenous BAP31 was identical (Fig. 9, h and i), demonstrating that the mutant-induced changes are dominant. Double labeling for myc-BAP31TMR and calnexin, a resident ER protein, resulted in virtually complete colocalization, confirming that myc-BAP31TMR is now restricted to the ER (data not shown). When these cells were analyzed for the distribution of cellubrevin, we noticed a dramatic change in the intracellular distribution of the protein (Fig. 9, j–l): cellubrevin fully colocalized with myc-BAP31TMR. In these cells, blebs positive for myc- BAP31TMR seemed to be distinguished by cellubrevin staining (Fig. 9, compare arrowheads in j and k with color overlay in l). These immunoreactive structures probably mark the boundary between the ER and the intermediate compartment (ER export complexes; see Discussion), since they are also positive for p58, a marker protein of the intermediate compartment (Fig. 9, arrowheads in m–o).

These experiments suggest that intact BAP31 is required for exporting cellubrevin from the ER. However, they do not distinguish whether the retention of cellubrevin by myc-BAP31TMR is because of a general defect in ER to Golgi traffic, or whether the effect is specific for cellubrevin. To distinguish between these alternatives, we compared the distribution of cellubrevin with that of transferrin receptor in the cells transfected with myc-BAP31TMR and cellubrevin. In these cells (Fig. 9, p and q), the distribution of the transferrin receptor was strikingly different from the reticular pattern of cellubrevin (compare with Fig. 6, bottom). The transferrin receptor displayed a typical pattern of fine dots concentrated in the perinuclear region, documenting that it is not retained in the ER as is the case for cellubrevin.

If temporary association with BAP31 is needed for cellubrevin to be exported, it is possible that export becomes rate limiting when excess cellubrevin is produced, resulting in accumulation of cellubrevin in the ER because of saturation of BAP31. Using BHK-21 cells overexpressing myc-tagged cellubrevin, we examined the localization of the protein with respect to ER residents. As shown in Fig. 10, overexpression of cellubrevin resulted in accumulations of the protein in calnexin-positive reticular structures. These accumulations were significantly more pronounced than those observed with the endogenous protein (compare Fig. 6), although not all of the calnexin-positive structures were labeled. Taken together, we conclude that although both cellubrevin and the transferrin receptor are targeted to the same post-Golgi compartments, cellubrevin, but not the transferrin receptor, interacts with BAP31 before or while exiting the ER.

In the present study, we have demonstrated that cellubrevin specifically interacts with BAP31. BAP31 is a resident of the ER that probably shuttles between the ER and the intermediate compartment and/or cis-Golgi complex. Our data suggest that BAP31 binds newly synthesized cellubrevin, and perhaps other proteins, to control their export to the Golgi apparatus where these transported proteins reach their final destinations.

As a resident of the ER, BAP31 does not colocalize with the major pools of cellubrevin. In addition, we found BAP31-positive membranes concentrated in a paranuclear region close to the Golgi apparatus and the MTOC, an area that is devoid of lumenal ER proteins. Proteins involved in endosomal recycling, such as cellubrevin and the transferrin receptor, are also concentrated but, as our analysis indicates, they reside on different vesicle populations. This area apparently serves as a central relay station for trafficking vesicles of different origins and destinations.

Export from the ER, the first step in the vectorial transport of proteins, commences in specialized regions of the ER. In some cells, e.g., the pancreatic acinar cells, these regions are juxtaposed to the cis-Golgi network and were originally referred to as transitional elements (Palade, 1975). In other cells, they appear to be distributed throughout the cytoplasm (Bannykh et al., 1996, Presley et al., 1997). They represent regions of the ER with many budding vesicles that are often adjacent to vesiculo-tubular clusters (Saraste and Svensson, 1991; Balch et al., 1994). Budding from the ER involves COPII coat proteins and results in the formation of COPII-coated transport vesicles (Barlowe et al., 1994). Before reaching the cis-Golgi, these transport vesicles pass through vesiculo-tubular clusters that may represent the intermediate compartment, functionally defined as the sorting compartment between the ER and the Golgi complex (Aridor and Balch, 1996). Here, ER resident proteins are probably sorted out and transported retrogradely to the ER, presumably involving COPI-coated transport vesicles (Aridor and Balch, 1996; Bannykh et al., 1996; Schekman and Orci, 1996). The accumulation of BAP31-containing vesicles around the MTOC, an area devoid of lumenal ER proteins, demonstrates clearly that the protein exits the ER during its life cycle. However, it remains to be established whether it is transported all the way to the cis-Golgi. Since we found only minor colocalization with the cis-Golgi marker β-COP, its steady-state concentration in that compartment must be low and the time BAP31 resides in these cisternae very short. Additionally, the protein may be sorted out earlier and shipped back to the ER by retrograde transport. It should be emphasized, however, that the evidence that supports recycling of BAP31 from these compartments to the ER is indirect. Thus, we cannot exclude that BAP31 is directed to lysosomes where it is degraded instead of returning to the ER. We regard this as less likely because BAP31 does not exhibit a lysosomal staining pattern and it is completely absent from purified clathrin-coated vesicles (Fig. 4).

Upon nocodazole treatment, the BAP31 staining in the region of the MTOC disappeared and BAP31 was almost exclusively retained in the ER with no overlap with cis-Golgi markers (our unpublished observations). Thus, nocodazole blocks forward transport of BAP31-containing vesicles from the ER to the MTOC, in addition to its inhibition of retrograde transport that is known to be microtubule dependent (Lippincott-Schwartz et al., 1990). If only retrograde transport was inhibited by the drug, BAP31 would be expected to accumulate in the cis-Golgi area. Forward transport between the ER and the Golgi dependent on microtubules is in agreement with recent observations (Bannykh et al., 1996; Rowe et al., 1996; Presley et al., 1997). Nocodazole also disrupted the accumulation of vesicles containing transferrin receptor and cellubrevin in this area (Daro et al., 1996), making the differential distribution of BAP31 and cellubrevin more obvious. These findings highlight the role of the MTOC as a central relay station for microtubule-based intracellular vesicle traffic. Apparently, both ER-derived forward trafficking vesicles and plasmalemma- or endosome-derived endocytic vesicles are collected by microtubular transport from the cell periphery and then passage through the area of the MTOC before reaching their destinations at the cis- and trans-side of the Golgi complex, respectively.

It remains to be established which domains of BAP31 are responsible for its intracellular sorting. Deletion of the KKXX motif (which functions in the recruitment of COPI proteins), as well as deletion of a longer stretch (including the YDRL motif), had no obvious effects on the localization of the protein. Apparently, the KKXX motif is redundant to another as yet unknown sorting signal, and probably has a secondary signal function in assisting other proteins in the recruitment of COPI. It is possible, however, that mutant BAP31 is associated with endogenous wild-type BAP31 that still contains intact sorting signals. Interestingly, BAP31 remained in the ER even when the entire cytoplasmic tail was deleted, although upon extended culturing, abnormal vesicles were observed and cell viability decreased (see below).

Although cellubrevin and BAP31 are localized to different subcellular membranes, the interaction between these two proteins is highly specific. Like synaptophysin for synaptobrevin, BAP31 appears to be the dominant binding protein for cellubrevin, clearly exceeding the still elusive putative SNARE partners of the protein. Binding was observed with native as well as with recombinant proteins, suggesting that the interaction is direct and does not require intermediate proteins. Furthermore, binding appears to require the transmembrane domains of both BAP31 and cellubrevin, although electrostatic interactions must also be involved that can be shielded by high ion concentrations. Despite the specificity and affinity of the interaction, only a relatively small proportion of the proteins are complexed in cellular detergent extracts. This finding agrees well with the differential localization of the proteins and indicates they are associated with each other only during the early phase of the intracellular traffic of cellubrevin. Since cellubrevin, like synaptobrevin, is probably synthesized on free ribosomes (Kutay et al., 1995), we assume that it interacts with BAP31 after posttranslational insertion into the ER membrane, although an additional role of BAP31 in membrane insertion of cellubrevin cannot be excluded at present.

How does BAP31 influence the export of cellubrevin from the ER? Two explanations are possible. First, BAP31 may function as a negative regulator that retains (or even recruits) newly synthesized cellubrevin in the ER until it is released by an unknown regulatory mechanism. Second, BAP31 may function as a positive regulator to which cellubrevin needs to bind to reach the Golgi compartment. According to this scenario, cellubrevin would be unable to leave the ER unless it is recruited by BAP31 into export vesicles, i.e., being actively transported rather than passively sorted by bulk flow.

According to the first view, BAP31 would bind to cellubrevin and other to be exported proteins and keep them in the ER membrane until they are either assembled with other membrane components or properly folded. However, we found that under all experimental conditions, only a fraction of cellubrevin is associated with BAP31 in detergent extracts. Neither treatment with nocodazole (resulting in a minor increase of cellubrevin in the ER) nor expression of a BAP31 mutant lacking the cytoplasmic domain (resulting in retention of cellubrevin in the ER), led to a noticeable increase of cellubrevin–myc–BAP31-TMR complexes relative to the uncomplexed protein pools. Although artefacts can never be excluded when assessing membrane protein complexes in detergent extracts, association of the proteins appears to be low even if they are both confined to the ER. This finding is difficult to reconcile with the negative regulator model. Rather, it suggests that BAP31 may serve as a sorting chaperone that recruits certain classes of membrane proteins into transport vesicles. BAP31 may directly interact with COPII budding components. Alternatively, it may bind (perhaps via its coiled coil regions) to integral membrane proteins that facilitate COPII-mediated export. In fact, there are precedents for protein-assisted export from the ER in yeast. For instance, the protein Erp25p forms a complex with Emp24p that is required for selective export of certain cargo molecules from the ER (Schimmöller et al., 1995; Belden and Barlowe, 1996). Another well-documented case is the yeast gene product Shr3p that, similar to BAP31, resides primarily in the ER and recruits specific amino acid permeases into transport vesicles (Kuehn et al., 1996).

Deletion of the cytoplasmic tail of BAP31 results in the retention of both BAP31 and cellubrevin in the ER, whereas other proteins such as the transferrin receptor still reach their normal destination beyond the Golgi complex. Moreover, the differences between transferrin receptor and cellubrevin localization are not caused by differences in the turnover rates of the proteins because preliminary observations suggest that their half lives are similar. Thus, these findings document that once the function of BAP31 is impaired, cellubrevin cannot reach its final destination and accumulates in the ER. Precisely how the function of BAP31 is affected by this deletion remains unclear. The deletion mutant still appears to be able to form a complex with cellubrevin, exhibiting properties that are not obviously different from the wild-type complex. We observed, however, that upon extended culturing cells developed abnormal membrane blebs, suggesting a delayed noxious effect of the mutant protein (Fig. 9, j–l). These blebs may represent membrane accumulations at the exit site of the ER (Bannykh et al., 1996), as suggested by the colocalization of myc-BAPTMR with p58 in these blebs (Fig. 9, m–o) or, alternatively, accumulation of membrane destined for degradation. It is possible that vesicle traffic out of the ER is affected to some extent, which may contribute to the phenotype.

We conclude that BAP31 is a representative of a novel class of proteins that regulates trafficking of certain membrane proteins out of the ER, either by retaining newly synthetized membrane proteins in the ER or by functioning as a conveyor belt for actively transporting these proteins from the ER to the Golgi complex. This class may include additional proteins such as BAP29 (Adachi et al., 1996) and other members of the BAP family which are specific for different membrane immunoglobulins (Kim et al., 1994; Terashima et al., 1994). Thus, trafficking of membrane proteins by means of such control proteins may be a common mechanism in eukaryotic cells.

We wish to thank A. Helenius, J.D. Castle, S. Fuller, J. Saraste, T. Kreis, H.-P. Hauri, and I. Trowbridge for their generous gifts of polyclonal and monoclonal antibodies and I. Mellman (all from Yale University) for critical comments to the manuscript. We are indebted to the W.M. Keck Foundation Biotechnology Resource laboratory at Yale University for their help in peptide purification and sequencing. We are especially grateful to L. Caron (Yale University) for invaluable advice in the use of the confocal microscope. We want to thank all the members of the Jahn lab and the Department of Cell Biology at Yale University for many helpful discussions and technical assistance.

aa

amino acids

AP

alkaline phosphatase

COP

coat proteins

DTAF

5-(4,6-dichlorotrianzinyl) aminofluorescein

ECL

enhanced chemiluminescence

ERGIC

endoplasmic reticulum–Golgi intermediate compartment

MTOC

microtubule organizing center

PDI

protein disulfide isomerase

PNS

postnuclear supernatant

SCAMP

secretory carrier membrane proteins

SNAP

soluble NSF attachment protein

SNAP-25

synaptosome-associated protein of 25,000 kD