|Home | About | Journals | Submit | Contact Us | Français|
The importance of soluble N-ethyl maleimide (NEM)-sensitive fusion protein (NSF) attachment protein (SNAP) receptors (SNAREs) in synaptic vesicle exocytosis is well established because it has been demonstrated that clostridial neurotoxins (NTs) proteolyze the vesicle SNAREs (v-SNAREs) vesicle-associated membrane protein (VAMP)/brevins and their partners, the target SNAREs (t-SNAREs) syntaxin 1 and SNAP25. Yet, several exocytotic events, including apical exocytosis in epithelial cells, are insensitive to numerous clostridial NTs, suggesting the presence of SNARE-independent mechanisms of exocytosis. In this study we found that syntaxin 3, SNAP23, and a newly identified VAMP/brevin, tetanus neurotoxin (TeNT)-insensitive VAMP (TI-VAMP), are insensitive to clostridial NTs. In epithelial cells, TI-VAMP–containing vesicles were concentrated in the apical domain, and the protein was detected at the apical plasma membrane by immunogold labeling on ultrathin cryosections. Syntaxin 3 and SNAP23 were codistributed at the apical plasma membrane where they formed NEM-dependent SNARE complexes with TI-VAMP and cellubrevin. We suggest that TI-VAMP, SNAP23, and syntaxin 3 can participate in exocytotic processes at the apical plasma membrane of epithelial cells and, more generally, domain-specific exocytosis in clostridial NT-resistant pathways.
The involvement of N-ethyl maleimide (NEM)-sensitive fusion protein (NSF), soluble NSF attachment proteins (SNAPs), and their receptors on membranes (so-called SNAREs) (Söllner et al., 1993a ,b ) in most membrane fusion events has been extensively documented in yeast and mammalian cells (for review, see Johannes and Galli, 1998 ). In particular, the neuronal vesicle SNAREs (v-SNAREs), vesicle-associated membrane protein (VAMP)/synaptobrevin 1 and 2 and their plasma membrane targets (t-SNAREs) SNAP25 and syntaxin 1 have been shown to be involved in calcium-dependent synaptic vesicle exocytosis (Schiavo et al., 1992 ; Blasi et al., 1993a ,b ). Interestingly, nonneuronal cells possess homologues of these neuronal SNAREs. These include the v-SNARE cellubrevin (McMahon et al., 1993 ) and the t-SNAREs syntaxins 2–7 (Bennett et al., 1993 ; Bock et al., 1996 ; Wang et al., 1997 ) and SNAP23 (Ravichandran et al., 1996 ; Wong et al., 1997 ). An understanding of the role of the brevins syntaxin 1 and SNAP25 in exocytosis has been greatly aided by the use of clostridial neurotoxins (NTs), i.e., tetanus NT (TeNT) and botulinum NT (BoNT), which specifically proteolyze these proteins (for review, see Niemann et al., 1994 ; Schiavo et al., 1994 ). These toxins also have effects on several exocytotic pathways in nonneuronal cells. For instance, TeNT, which cleaves cellubrevin, inhibits transferrin receptor exocytosis in fibroblasts (Galli et al., 1994 ). Strikingly, several exocytotic events are at least partially insensitive to some NTs (for review, see Johannes and Galli, 1998 ). For example, although TeNT and BoNT F can inhibit transport of vesicular stomatis virus (VSV) G protein to the basolateral plasma membrane in Madin–Darby canine kidney (MDCK) cells, apical transport of influenza hemagglutinin (HA) is insensitive to both NTs (Ikonen et al., 1995 ). This led the authors to propose a mechanism of membrane fusion at the apical plasma membrane of epithelial cells independent of v- and t-SNAREs, although they had found HA transport to be sensitive to NEM (Simons and Ikonen, 1997 ). On the contrary, several groups have localized syntaxin isoforms at the apical plasma membrane of epithelial cells (Gaisano et al., 1996 ; Low et al., 1996 ; Mandon et al., 1996 ; Delgrossi et al., 1997 ) and proposed a role for these SNAREs in apical docking and fusion (Weimbs et al., 1997 ). We have searched for a TeNT-insensitive v-SNARE that could be a partner of apical t-SNAREs in epithelial cells. Recently, D’Esposito et al. (1996) identified SYBL1, a synaptobrevin-like gene in Xq28 pseudoautosomal region, which undergoes X inactivation. We find that SYBL1p is TeNT insensitive and propose to name it TeNT-insensitive VAMP (TI-VAMP). Hence, we have tested the localization and biochemical properties of TI-VAMP in CaCo-2 epithelial cells.
Antibodies against syntaxin 3 (TG1), SNAP23 (TG7), and TI-VAMP (TG11) were raised in rabbit against the cytoplasmic domain of rat syntaxin 3 fused to glutathione S-transferase (GST) (Calakos et al., 1994 ), recombinant human SNAP23, and the recombinant cytoplasmic domain of human TI-VAMP fused to GST, respectively, and purified on the corresponding recombinant proteins. Antibodies against human cellubrevin (TG2) were obtained in rabbit against the N-terminal peptide MSTGPTAATGSNC of cellubrevin and affinity purified on the cytoplasmic domains of human cellubrevin fused to GST. All of these antibodies reacted specifically with their corresponding antigen by Western blotting and immunoprecipitation (see Results). Monoclonal antibodies against E-cadherin/uvomorulin (DECMA), sec8 (clone 14), ZO-1, and transferrin receptor (H68.4) were from Sigma (St. Louis, MO), Transduction Laboratory (Lexington, KY), Dr. M. Mooseker (Yale University, New Haven, CT), and Dr. I. Trowbridge (Salk Institute, La Jolla, CA), respectively. Secondary Cy-2 or Texas Red–labeled goat anti-rat, anti-mouse, and anti-rabbit immunoglobulins were from Jackson ImmunoResearch (West Grove, PA).
Full-length cDNA of the human isoforms of cellubrevin, syntaxin 3, SNAP23, and TI-VAMP were obtained by reverse transcription-PCR on CaCo-2 mRNA using standard procedures and the following sets of oligonucleotides: 5′-ATGTCTACAGGTCCAACTGCTG-3′ (hcb5′) with 5′-GCTGGTTCTTCATGAAGAGACA-3′ (hcb3′), 5′-ATGAAGGACCGTCTGGAGCAG-3′ (hstx3-5′) with 5′-TTAATTCAGCCCAACGGAAAG-3′ (hstx3-3′), 5′-TGGGCTTCAGGATGAAGGA-3′ (hstx3-5′ups) with 5′-GCTAGATTGTTAGCTGAGTCA-3′ (hstx3-3′dos), 5′-ATGGATAATCTGTCATCAGAAGAA-3′ (SNAP23-5′) with 5′-TTAGCTGTCAATGAGTTTCTT-3′ (SNAP23-3′), and 5′-AGACTGAAGCCATGGCGATT-3′ (TI-VAMP 5′) with 5′-CTATTTCTTCACACAGCTTGGC-3′ (TI-VAMP3′), respectively. In the case of human syntaxin 3, all the clones that we obtained either with the oligonucleotides hstx3-5′ and hstx3-3′ or with hstx3-5′ups and hstx33-3′dos, which correspond to regions upstream of the starting ATG and downstream of the stop codon, had the following modifications compared with the human syntaxin 3 cDNA sequence deposited in GenBank (accession number U32315): position 519, CACGTC → CAGCTC, which translates into HV → QL; and position 816, AGC → ACG, which translates into S → T. The rat sequence is QL and T at the same positions. In the case of human SNAP23, all our clones had the following modification compared with the sequence deposited in GenBank (accession number U55936): 207GAA → GAG, which is conservative; and 462 GTC → GCC, which translates into V → A, as seen in SNAP23A (GenBank accession number Y09567). No modification was observed in our human clones of cellubrevin, identical to synaptobrevin 3 (GenBank accession number U64520), and in TI-VAMP (GenBank accession number M90418). Fragments of human cellubrevin predicted to contain only the first coiled coil domain (amino acids 1–52, hCB1) and both coiled coil domains (amino acids 1–71, hCB2) and of human TI-VAMP predicted to contain both coiled coils (amino acids 1–179) were obtained by cloning the corresponding cDNA into pGEX vectors (Pharmacia, Saclay, France). SNAP25 and syntaxin 1a clones were described previously (Hayashi et al., 1994 ).
CaCo-2 cells were seeded at confluency (3.105 cells/cm2) on Transwell-Clear filters (Costar, Cambridge, MA) and grown 7–10 d to full polarization (transepithelial resistance, 800 Ω·cm2). The cells were washed in PBS supplemented with 0.1 mM CaCl2, 0.1 mM MgCl2 (PBS+), fixed with methanol at −20°C for 3 min, and rehydrated in PBS+. The filter was then cut and incubated over a drop of PBS containing the primary antibodies, washed three times with PBS, incubated with Cy2-conjugated goat anti-rabbit antibody and Texas Red–conjugated goat anti-mouse or anti-rat antibody, washed three times, and mounted in 90% glycerol in PBS. Confocal laser scanning microscopy was performed using a Leica (Nussloch, Germany) TCS microscope. The images were assembled without modification using Adobe (Mountain View, CA) Photoshop.
CaCo-2 cells were fixed with 2% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, for 1 h at room temperature and processed for ultracryomicrotomy as described elsewhere (Raposo et al., 1997 ). Ultrathin cryosections were collected using a mixture of 2.3 M sucrose and methylcellulose (vol/vol) and were immunogold labeled with antibodies against syntaxin 3, SNAP23, or TI-VAMP and protein A-gold conjugates (PAG 10, Department of Cell Biology, Utrecht University, Utrecht, The Netherlands) (Raposo et al., 1997 ). No labeling was observed with protein A-gold alone or nonimmune serum and protein A-gold.
The effect of light chains of the clostridial NTs was tested according to the procedure of Hayashi et al. (1994) . Briefly, the cDNAs of syntaxin 3, cellubrevin, and TI-VAMP cloned in pCR3 (Invitrogen, NV Leek, The Netherlands), of SNAP23 cloned in pET15b (Novagen, R & D Systems, Abingdon, UK), and of SNAP25 and syntaxin 1a (Hayashi et al., 1994 ) were translated in vitro using T7 polymerase and rabbit reticulocyte lysate in the presence of [35S]methionine. The proteins were incubated with the light chains of TeNT (2 μM) and BoNT A (0.8 μM), B (2 μM), C (0.9 μM), D (2 μM), E (0.4 μM), F (2 μM), and G (2 μM) in 150 mM potassium glutamate, 10 mM HEPES-KOH, pH 7.2, for 1 h at 37°C. The cleavage was analyzed after the products were separated on 15% SDS-PAGE gels followed by visualization of bands by autoradiography.
For binding assays, full-length human syntaxin 3 and SNAP23 were translated in vitro using the Quick T7 kit (Promega, Lyon, France) according to the manufacturer’s procedure. The resulting extract (8 μl) was incubated with glutathione beads coated with GST, GST-hCB1, GST-hCB2, or GST–TI-VAMP (1 μM final concentration) in 100 μl of binding buffer (4 mM HEPES-NaOH, pH 7.4, 100 mM NaCl, 3.5 mM CaCl2, 3.5 mM MgCl2, 1 mM EDTA, 0.1% NP-40) for 12 h at 4°C as described previously (Hayashi et al., 1994 ). The beads were collected by centrifugation and washed six times with washing buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 2.5 mM Mg Cl2, 0.1% NP-40), and the final bead fraction was eluted with gel sample buffer consisting of 60 mM Tris, pH 6.8, 2% (wt/vol) SDS, 10% (wt/vol) glycerol, 0.07% bromophenol blue, and 5% β-mercaptoethanol. The resulting samples were either kept at room temperature for 30 min or boiled for 10 min and analyzed using SDS-PAGE (Schagger and von Jagow, 1987 ) followed by visualization of bands by autoradiography.
CaCo-2 cells grown on 25-mm-diameter wells to confluency were washed twice briefly in PBS+ and were treated in PBS+ with 1 mM NEM or 1 mM NEM plus 2 mM DTT for 15 min on ice. In the first condition, NEM was quenched by 2 mM DTT for 15 min on ice. The cells were then washed in PBS+, further incubated in culture medium for 30 min at 37°C, lysed in solubilization buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl, 10 mM EDTA, 1 mM PMSF, 1 mM benzamidine, 1 μg/ml pepstatin, 1 μg/ml antipain, 1 μg/ml leupeptin, 1 μg/ml aprotinin, 1% Triton X-100), and centrifuged at 200,000 × g for 15 min. The resulting supernatant was incubated overnight at 4°C with rotation with anti-syntaxin 3, anti-GST, anti-cellubrevin, or anti–TI-VAMP immunobeads, which had previously been prepared as described (Chilcote et al., 1995 ) by covalent coupling of corresponding affinity-purified antibodies. The beads were then pelleted for 1 min at 1000 × g and washed with solubilization buffer containing 0.5% Triton X-100. Finally, the beads were eluted with 100 mM glycine, pH 2.7, 0.5% Triton X-100, and the corresponding eluate was neutralized with Tris, pH 8, supplemented with gel sample buffer, boiled for 5 min, and run on SDS-PAGE gels (Schagger and von Jagow, 1987 ).
Best-fit comparison (Genetics Computer Group [GCG], Madison, WI) of human TI-VAMP with the human brevins shows that the two putative coiled coils of this protein are 64 and 57%, respectively, similar to corresponding domains in the VAMP/brevins (Figure (Figure1A).1A). Overall, we found that the similarity is 59% with synaptobrevin 2 and 56% with SNC1p, a v-SNARE implicated in Golgi to the plasma membrane transport in yeast (Protopopov et al., 1993 ). The similarity between TI-VAMP and ERS-24, GOS28, human ykt6, or rsec22 v-SNAREs found in the endoplasmic reticulum and/or in the Golgi is <50% (46, 40, 49, and 43%, respectively). Two sets of four hydrophobic residues present in hepta repeats that overlap with the two coiled coils are found both in this new molecule and in the VAMP/brevins (Figure (Figure1A).1A). We found that TI-VAMP, translated in vitro, is insensitive to TeNT and to BoNTs B, D, F, and G, which all proteolyze cellubrevin (Figure (Figure2A),2A), and has structural properties of v-SNAREs (see below). Moreover, it is noteworthy that the first 90 amino acids of TI-VAMP define a domain that is not found in any other v-SNARE. In this domain, we have identified a region of 40 amino acids that has 47.5% similarity and 27.5% identity with the first annexin repeat of annexin XIII (Figure (Figure1B)1B) which forms part of an important calcium- and phospholipid-binding site (Raynal and Pollard, 1994 ; Smith and Moss, 1994 ). We also showed that, similar to the VAMP/brevins, TI-VAMP is a type II membrane protein, because the protein inserted into microsomes is sensitive to proteinase K (our unpublished results). We have generated in rabbit an antibody directed against the putative cytoplasmic domain of TI-VAMP (amino acids 1–179) fused to GST (serum referred to as TG11). This antibody reacts specifically with its corresponding antigen, recognizes a band of the expected size (25 kDa) in CaCo-2 cell extract by Western blotting, and specifically immunoprecipitates TI-VAMP (Figure (Figure1C).1C).
Potential partners of TI-VAMP, SNAP23, and syntaxin 3 are produced in most cell types, including epithelial cells (see Figure Figure7;7; our unpublished observations) and are also insensitive to NTs (Figure (Figure2,2, B and C). In vitro–translated SNAP23 is not cleaved by BoNTs A, C1, or E, whereas these toxins very efficiently cleaved SNAP25 (Niemann et al., 1994 ; Schiavo et al., 1994 ; Osen-Sand et al., 1996 ) (Figure (Figure2B),2B), a neuronal isoform of SNAP23. We have reinvestigated the effect of BoNT C1 on syntaxin 3. We found that contrary to syntaxin 1a, human syntaxin 3 is insensitive to BoNT C1 at a concentration as high as 0.9 μM (Figure (Figure2C),2C), a 10-fold excess over the concentration required for maximal cleavage of syntaxin 1a. A study with increasing concentrations of BoNT C1 (ranging from 1 nM up to 3.8 μM) showed that rat syntaxin 3 is not cleaved by this toxin even at the highest concentration tested (our unpublished observations).
Because it was shown that apical transport of HA is insensitive to clostridial NTs (Ikonen et al., 1995 ), we investigated the subcellular distribution of clostridial NT-insensitive SNAREs in the human epithelial cell line CaCo-2, a well-documented model of polarized intestinal cells (Weimbs et al., 1997 ). By confocal microscopy, we found that syntaxin 3 is present at high concentration at the apical plasma membrane and is not detected in the lateral plasma membrane, a region where E-cadherin is specifically found (Figures (Figures3A3A and 4). In rat kidney and intestine sections, syntaxin 3 is strictly localized in the microvilli structures of the apical domain (our unpublished results). The bulk of SNAP23 is found at the apical plasma membrane by confocal microscopy (Figures (Figures3B3B and and4).4). Small amounts of SNAP23 were found in tight junctions, where it colocalizes with ZO-1 (Figure (Figure4),4), and along the lateral plasma membrane (Figures (Figures3B3B and and4),4), by confocal microscopy. Immunoelectron microscopy showed that syntaxin 3 (Figure (Figure5A)5A) and SNAP23 (Figure (Figure5B)5B) are present in the apical plasma membrane rather than in intracellular structures underneath and that syntaxin 3 labeling is strictly restricted to the apical plasma membrane. In CaCo-2 cells, cellubrevin and TI-VAMP localize to small vesicular structures dispersed throughout the peripheral cytoplasm, along the longitudinal axis with a pronounced enrichment of both v-SNAREs in the apical domain (Figure (Figure3,3, C–F). We did not observe any significant codistribution of these v-SNAREs either with CTR433, a medial Golgi marker (Jasmin et al., 1989 ), or with lamp-2, a lysosomal marker (Chen et al., 1985 ). Strikingly, we noticed that neither cellubrevin nor TI-VAMP codistributed with Tf receptor (Figure (Figure3,3, C–F) or with Tf-containing organelles (our unpublished observations) as seen by confocal microscopy. Because Tf receptor is known to recycle mainly between basolateral endosomes and plasma membrane (Hughson and Hopkins, 1990 ), we investigated whether TI-VAMP could be detected in the apical plasma membrane as expected if TI-VAMP–containing vesicles dock and fuse at this site. For this purpose, we have analyzed the localization of TI-VAMP by immunoelectron microscopy with gold-labeled antibodies in CaCo-2 cells, and we observed labeling of the apical plasma membrane, suggesting the presence of a pool of this protein (Figure (Figure5C).5C).
We then studied the capacity of TI-VAMP and cellubrevin to interact with t-SNAREs in vitro. Using recombinant fragments of the cytoplasmic domains of these v-SNAREs fused to GST, we found that cellubrevin and TI-VAMP were both able to form SDS-resistant SNARE complexes (Hayashi et al., 1994 ) with syntaxin 3 and SNAP23 translated in vitro in the presence of [35S]methionine (Figure (Figure6).6). SDS resistance of the brain syntaxin 1/SNAP25/synaptobrevin 2 SNARE complex is thought to be a key property of SNARE complexes that is probably attributable to the high stability of the coiled–coiled interactions (Hayashi et al., 1994 ). We also found that both cellubrevin and TI-VAMP interacted poorly with syntaxin 3 alone (Figure (Figure6)6) or SNAP23 alone (our unpublished results), in agreement with a previous report showing the lack of interaction between syntaxin 3 and the synaptobrevins (Calakos et al., 1994 ). Interestingly, a shorter fragment of the cytoplasmic domain of cellubrevin, excluding the second coiled coil fused to GST (GST-hCB1), was only slightly less efficient in this assay (Figure (Figure6),6), as was expected, because deletion of the second putative α helix in synaptobrevin 2 did not prevent association of SNAP25 (Regazzi et al., 1996 ). These data clearly indicate that cellubrevin and TI-VAMP have the capacity to form SNARE complexes with the same t-SNAREs, including the NT-insensitive syntaxin 3 and SNAP23.
To show that exocytic organelles can dock at the apical plasma membrane, we set up an immunoprecipitation assay based on the observation that syntaxin 3 is found strictly at the apical plasma membrane (see above). NEM treatment was used to block NEM-dependent ATPases, including NEM-sensitive NSF (Beckers et al., 1989 ), and therefore should elicit accumulation of assembled SNARE complexes as seen in PC12 cells (Banerjee et al., 1996 ). NEM has also been shown to inhibit the binding of nSec1 to syntaxin 1 in vitro (Meffert et al., 1996 ), an effect that is also expected to enhance SNARE complex formation, because nSec1 prevents syntaxin 1 from forming SNARE complexes (Pevsner et al., 1994 ). To compare a control condition with a condition in which SNARE complexes are accumulated, we pretreated CaCo-2 cells grown on filters at 4°C with either 1) 1 mM NEM plus 2 mM DTT for 30 min (control conditions) or 2) 1 mM NEM (15 min) followed by 2 mM DTT (15 min) and allowed vesicles to accumulate at their target membrane by returning the cells to culture medium at 37°C as reported previously in the case of PC12 cells (Banerjee et al., 1996 ). First, we observed by immunofluorescence that syntaxin 3 and SNAP23 are still localized at the apical plasma membrane after treatment with NEM. Second, using this experimental approach, we could immunoprecipitate virtually all of the syntaxin 3 present in the cell lysate using anti-syntaxin 3 immunobeads (Figure (Figure7).7). Western-blotting analysis of the fractions bound to the syntaxin 3 immunobeads showed a dramatic simultaneous increase of the amounts of SNAP23 and cellubrevin coimmunoprecipitated with syntaxin 3 when the cells were pretreated with NEM (Figure (Figure7).7). Third, TI-VAMP was found to coimmunoprecipitate with syntaxin 3 only from NEM-pretreated cell extracts (Figure (Figure7).7). None of the SNAREs were detected in pellet fractions recovered with anti-GST immunobeads, and Sec8, a protein that can be found in association with syntaxin 1 to a small extent (Hsu et al., 1996 ), was also absent from all bead fractions (Figure (Figure7),7), therefore showing the high specificity of the experimental procedure. This result clearly indicates that the t-SNAREs present at the apical plasma membrane, syntaxin 3 and SNAP23, allow NEM-dependent SNARE complex formation, which includes cellubrevin or TI-VAMP (Figure (Figure7).7).
Our results suggest the involvement of the clostridial NT-resistant t-SNAREs syntaxin 3 and SNAP23 in exocytosis at the apical plasma membrane of epithelial cells.
First, we have found that in contrast to their neuronal counterparts, SNAP23 and syntaxin 3 were not proteolyzed by clostridial NTs. Our finding that syntaxin 3 was insensitive to BoNT C1 is in contrast to a report showing the cleavage of a synaptosomal protein recognized by an anti-syntaxin 3 antibody by BoNT C1 (Schiavo et al., 1995 ). It should be mentioned that the antibody used in the latter study also weakly recognizes syntaxin 1 (Gaisano et al., 1996 ). In the present article, we report the lack of effect of the toxins on full-length in vitro–translated human SNARE proteins. This method avoids the kind of technical problems described above. The resistance of SNAP23 and syntaxin 3 to clostridial NTs is surprising given the high degree of similarity between SNAP25 and SNAP23 (72%) and between syntaxin 1 and syntaxin 3 (77%). The lack of cleavage by NTs of SNAP23 and syntaxin 3 might be the result of the inability of the toxins either to bind to the putative coiled coils of these SNAREs or to recognize them as substrates because of key amino acid substitutions within the cleavage sites (Pellizzari et al., 1996 ).
Second, by confocal and immunoelectron microscopy, we have found that endogenous syntaxin 3 and SNAP23 were both localized at the apical plasma membrane of CaCo-2 cells. In the case of syntaxin 3, we were unable to detect any intracellular or basolateral plasma membrane staining either by confocal microscopy or by immunoelectron microscopy on fully polarized and differentiated cells. This is in contrast to two studies on MDCK (Low et al., 1996 ) and CaCo-2 cells (Delgrossi et al., 1997 ) overexpressing syntaxin 3 in which significant intracellular labeling was detected, including lysosome labeling (Low et al., 1996 ). In our case, we observed syntaxin 3-positive vesicular structures only in CaCo-2 cells that were not confluent and not yet fully polarized (our unpublished observations). We believe therefore that intracellular syntaxin 3 labeling might represent the newly synthesized pool or a pool resulting from overexpression of the protein or both. In the case of the partner t-SNARE SNAP23, we found that this protein is localized mainly at the apical plasma membrane on fully polarized CaCo-2 cells, with a minor labeling of the lateral plasma membrane as shown by longitudinal confocal sections. We found a major association of SNAP23 with syntaxin 3 by coimmunoprecipitation. This demonstrates that syntaxin 3 and SNAP23 act in concert as apical plasma membrane t-SNAREs. The lateral pool of SNAP23 is probably associated with another syntaxin isoform, because SNAP23 was already found to associate with syntaxin 4 (Ravichandran et al., 1996 ), a basolateral t-SNARE in MDCK cells (Low et al., 1996 ). Because we could not detect syntaxin 4, 2, or 1 by immunofluorescence microscopy in CaCo-2 cells (our unpublished observations), we have not investigated this point in the present study.
Third, we show that syntaxin 3 and SNAP23 form apical SNARE complexes and provide the first evidence for the involvement of cellubrevin and a new v-SNARE, TI-VAMP, in apical SNARE complex formation. We have taken advantage of the specific localization of syntaxin 3 at the apical plasma membrane of polarized CaCo-2 cells to immunoprecipitate apical SNARE complexes containing syntaxin 3. We found that NEM pretreatment of CaCo-2 cells increased the recovery of cellubrevin and SNAP23 associated with syntaxin 3. We have also found that in NEM-pretreated cell extracts, SNAP23 and syntaxin 3 form SNARE complexes with TI-VAMP, a new member of the VAMP/brevin family that is the first to be shown to be insensitive to all clostridial NTs that proteolyze the brevins. Because of the restricted localization of syntaxin 3 to the apical plasma membrane and cellubrevin and TI-VAMP on intracellular vesicles, we believe that the most likely explanation of our immunoprecipitation experiments is that cellubrevin- and TI-VAMP–containing vesicles dock at the apical plasma membrane through the NEM-dependent formation of SNARE complexes, which include SNAP23 and syntaxin 3. Further evidence was obtained in the case of TI-VAMP by our observation of a pool of the protein at the apical plasma membrane revealed by immunogold labeling on ultrathin cryosections. In CaCo-2 cells, the syntaxin 3/SNAP23/cellubrevin and syntaxin 3/SNAP23/TI-VAMP SNARE complexes appear as the analogues of the syntaxin 1/SNAP25/synaptobrevin complex found in the nerve terminal.
The properties of TI-VAMP, a new member of the VAMP/brevin family, have not been examined previously. It is insensitive to NTs, binds to plasma membrane t-SNAREs, and localizes to apical organelles in epithelial cells. This v-SNARE has a unique N-terminal domain that resembles the lipid/Ca2+ binding domain of annexin XIII and therefore may be a v-SNARE with additional lipid/Ca2+ binding properties. In any case, the N-terminal domain of TI-VAMP probably confers to this molecule specific features that would distinguish it from other members of the VAMP/brevin family. It is intriguing that the weak but nonetheless significant similarity we found in the N-terminal part of TI-VAMP establishes a relationship between this new v-SNARE and annexin XIII, an intestine-specific annexin (Wice and Gordon, 1992 ), which has been proposed to play a role in apical transport of HA in MDCK cells (Fiedler et al., 1995 ). Future studies should explore the functional relevance of the similarity between TI-VAMP and annexin XIII. Another remarkable property of TI-VAMP is that it is resistant to TeNT and to BoNTs B, D, F, and G. As in the case of SNAP23 and syntaxin 3, this might result from lack of binding of the NTs or from the differences in amino acid sequence observed in the NT cleavage sites (see the legend of Figure Figure1A).1A). In this study in epithelial cells, despite the biochemical specificities of TI-VAMP depicted above, we have not found major localization or biochemical differences regarding the t-SNARE partners between TI-VAMP and cellubrevin. This may be because these proteins have overlapping functions in the same apical exocytotic pathway or because our level of analysis could not distinguish different apical exocytotic pathways. Our observations 1) that cellubrevin and TI-VAMP could both form SNARE complexes with the same apical t-SNAREs but that these v-SNAREs were not complexed together and 2) that TI-VAMP was found associated with syntaxin 3 only in NEM-pretreated cells could be seen to favor the second hypothesis. In CaCo-2 cells, sucrase isomaltase was shown to be transported directly from the trans-Golgi network to the apical plasma membrane, whereas dipeptidyl-peptidase IV is first transported to the lateral membrane and then sorted to the apical plasma membrane (Matter et al., 1990 ). Given the role of cellubrevin in transferrin receptor recycling in fibroblasts (Galli et al., 1994 ) and the lack of effect of TeNT on transport of HA from the trans-Golgi network to the apical plasma membrane in MDCK cells (Ikonen et al., 1995 ), an attractive but speculative hypothesis could be that cellubrevin is involved in recycling of apical proteins and TI-VAMP in the direct trans-Golgi network to the apical plasma membrane pathway. Obviously, it will be important to characterize by electron microscopy the vesicular pools of cellubrevin and TI-VAMP in epithelial cells, to test whether TI-VAMP distributes on HA-containing vesicles and whether it is involved in apical docking/fusion of these vesicles with the apical plasma membrane in MDCK cells.
In conclusion, our data are in favor of the existence of SNARE-dependent exocytotic events at the apical plasma membrane of epithelial cells. They also imply that NTs should be unable to impair multiple exocytotic events in different cell types. The NT-resistant SNAREs described here have a very broad tissue and cell distribution. Our work opens the way for studying NT-resistant secretory pathways in nonneuronal cells and in neurons (Osen-Sand et al., 1996 ) and in particular the role played by TI-VAMP in these cell types.
We are indebted to Mark Bennett for the rat syntaxin 3 clone; to Michel Bornens for the CTR433 antibody; to Ian Trowbridge for the H68.4 antibody; to Monique Arpin for helpful discussions; to Dominique Morineau, Ahmed El Marjou, and Lucien Cabanié for excellent technical assistance; and to Margaret Butler and Roy Golsteyn for critical reading of this manuscript. This work was supported in part by an Association pour la Recherche contre le Cancer grant to A.Z., by an Alexander-von-Humboldt Foundation fellowship to V.V.V., by a Human Frontier Science Program Postdoctoral fellowship to J.T., by a Fonds der Chemischen Industrie grant to H.N., and by National Institutes of Health grants to M.K.
1Abbreviations used: BoNTs, botulinum NTs; HA, influenza hemagglutinin; NEM, N-ethyl maleimide; NSF, NEM-sensitive fusion protein; NTs, neurotoxins; SNARE, SNAP receptor; SNAPs, soluble NSF attachment proteins; t-SNARE, target SNARE; TeNT, tetanus NT; TI-VAMP, tetanus neurotoxin-insensitive VAMP; VAMP, vesicle-associated membrane protein; v-SNARE, vesicle SNARE.