SNAP-23 Requirement for Transferrin Recycling in StreptolysinO-permeabilized Madin-Darby Canine Kidney Cells*

Fusion of recycling and transcytotic vesicles with the apical and basolateral plasma membrane domains of Madin-Darby canine kidney (MDCK) cells requires theN-ethylmaleimide-sensitive factor and is sensitive to botulinum neurotoxin serotype E (BoNT/E). BoNT/E is thought to selectively proteolyze the 25,000-dalton synaptosomal associated protein (SNAP-25), a protein found in neurons or cells of neuroendocrine origin. However, SNAP-25 is not found in MDCK cells. One possible target for BoNT/E in MDCK cells is SNAP-23, a newly described SNAP-25 homolog that is found in several organs including kidney. Currently, the function of SNAP-23 is unknown. We have reconstituted transferrin recycling in permeabilized MDCK cells to assess the role of SNAP-23 in the endocytic traffic of this protein. We find that: (i) SNAP-23 is expressed in MDCK cells and is found both at the basolateral plasma membrane and associated with apical and basolateral vesicles, (ii) canine SNAP-23 is cleaved by BoNT/E, (iii) transferrin recycling is N-ethylmaleimide-sensitive factor-dependent and BoNT/E-sensitive, and (iv) addition of either exogenous SNAP-23 or anti-SNAP-23 antibodies inhibits ligand recycling. Our observations suggest that SNAP-23 may be required for fusion of recycling vesicles with the basolateral membrane of polarized MDCK cells.

Membrane trafficking is mediated by transport vesicles that bud from a donor membrane and then fuse with an acceptor membrane (1). An important regulatory step in this process is controlling the vesicle target specificity and subsequent fusion event. In many systems, the vesicle targeting/fusion machinery is composed of both cytosolic and membrane bound proteins (2). The cytosolic components are the N-ethylmaleimide-sensitive factor (NSF) 1 (3,4), and the soluble NSF attachment protein (SNAP) that binds NSF to the fusion complex (5,6). The membrane bound components, the SNAP receptors (SNAREs), are divided into two types: vesicle membrane-associated (v-SNAREs) and target membrane-associated (t-SNAREs). The v-SNAREs are thought to dock vesicles to the appropriate target membranes via interactions with cognate t-SNAREs (7). Recent evidence suggests that the v-SNARE⅐t-SNARE complex may also be the minimal machinery needed for vesicle fusion to occur with the target membrane (8). Interactions between v-SNAREs/t-SNAREs are regulated by Rab proteins, N-sec-1/ Munc-18 homologs (9 -12), as well as hydrolysis of ATP by NSF, which may act to dissociate v-SNARE⅐t-SNARE complexes priming them for another round of fusion (13)(14)(15).
In neurons, SNAREs are thought to control synaptic vesicle docking and fusion. One v-SNARE, vesicle-associated membrane protein (VAMP, also known as synaptobrevin), and two t-SNAREs, syntaxin and SNAP-25 (not related to the SNAP of the cytosolic component), have been isolated and characterized from synaptosomal membranes (reviewed in Ref. 16). Clostridial neurotoxins inhibit synaptic vesicle fusion by acting as specific proteinases of a single SNARE component without cleaving the others (reviewed in Ref. 17). For example, VAMP is cleaved by tetanus toxin, BoNT/B, D, F, and G, while syntaxin is cleaved by BoNT/C1, and SNAP-25 is cleaved by BoNT/A and E. It has now been shown that many trafficking reactions require SNARE-mediated targeting and fusion including endoplasmic reticulum to Golgi transport and trans-Golgi to basolateral plasma membrane transport (18 -20). In other cases, SNAREs have been identified but their exact role in trafficking is unknown (for example, see Refs. [21][22][23]. Previously, we analyzed the role of the vesicle fusion machinery in the exit of IgA from the apical recycling endosome (ARE) and fusion with the apical and basal plasma membrane in streptolysin-O (SLO)-permeabilized MDCK cells expressing the polymeric Ig receptor (pIgR) (24). The ARE is a tubulovesicular compartment that receives molecules transcytosing in the basolateral to apical direction and those recycling from the apical and basolateral poles of the cell (25)(26)(27). The pIgR protein transports polymeric Igs (IgA and IgM) across epithelial cells and into secretions. Both transcytosis in the basolateral to apical direction and recycling of pIgR ligand is dependent on NSF and inhibited by BoNT/E (which cleaves SNAP-25) (24). Surprisingly, BoNT/A has no apparent effect on either pathway. In contrast, transport of a newly synthesized protein from the trans-Golgi network to the apical surface is apparently independent of NSF/SNAP/SNAREs while transport from the trans-Golgi network to the basolateral surface requires these complexes (19). The above observations suggest that fusion with the apical plasma membrane of MDCK cells may occur by multiple mechanisms, one of which may require NSF/SNAP/ SNAREs (basal to apical transcytosis) and one which may not (trans-Golgi network to apical cell surface delivery). While studies have found several isoforms of VAMP and syntaxin are expressed within MDCK cells (19,23), attempts to identify SNAP-25 in MDCK cells have failed (24). This led us to hypothesize that a BoNT/E-sensitive SNAP-25 homolog(s) may exist in MDCK cells and may be required for transcytosis and recycling (24).
A SNAP-25 homolog, called SNAP-23, has recently been cloned and characterized from human B lymphocytes (28). Moreover, a mouse isoform of SNAP-23 (also known as syndet) has been cloned from 3T3 L1 adipocytes and is 98.6% similar and 86.7% identical to human SNAP-23 (29,30). Unlike SNAP-25, which is distributed along axons, SNAP-23 is concentrated on nerve cell bodies with only limited staining of the axons and syntaptosomes. 2 In addition, SNAP-23 is expressed in several non-neuronal tissues including heart, lung, liver, muscle, pancreas, and kidney (where it is expressed at high levels) (28,31). It has been shown to bind various syntaxins (isoforms 1, 2, and 3 and with high affinity to 4) and VAMP (28,30). 2 While SNAP-23 is localized to the plasma membrane and in vesicular structures, little is known presently about its role in membrane trafficking. Our goal was to determine if SNAP-23 was required for endocytic traffic in polarized epithelial cells.
We find that SNAP-23 is expressed in MDCK cells, both at the basolateral plasma membrane and in apical and basolateral vesicles. Tf recycling is NSF-dependent and BoNT/E-sensitive, and addition of either exogenous SNAP-23 or anti-SNAP-23 antibodies inhibits ligand recycling. Our observations suggest that SNAP-23 may regulate fusion of recycling vesicles with the basolateral domain of polarized MDCK cells. These findings further our understanding of some of the underlying fusion mechanisms used in polarized MDCK cells and may ultimately lead to a better understanding of how polarized cells achieve and maintain their distinct plasma membrane domains.

EXPERIMENTAL PROCEDURES
Antibodies, Proteins, and Other Reagents-Canine apo-Tf (Sigma) was loaded with Fe 3ϩ as described (32). Rat monoclonal antibody ascites to ZO-1, a protein associated with tight junctions, was obtained from Chemicon (Temecula, CA) and was used at 1:100 dilution. Mouse monoclonal antibody supernatant to E-cadherin (hybridoma rr1) (33) was used at a 1:1 dilution. Horseradish peroxidase-, fluorescein isothiocyanate-, or Texas red-conjugated goat anti-rabbit IgG, goat anti-mouse IgG, or goat anti-rat IgG were obtained from Jackson ImmunoResearch Laboratories (WestGrove, PA) and fluorescent labeled secondary antibodies used at 5-10 g/ml.
Preparation of affinity purified antiserum to SNAP-23 was as described (28,30). 2 A human SNAP-23 cDNA was obtained by polymerase chain reaction from a human platelet cDNA library using primers (cg-ggatccATGGATAATCTGTCATCAG and cccaagcttTTAGCTGTCAATG-AGTTTC) based on the human B cell SNAP-23 sequence (28). The sequence was confirmed by dideoxyribose nucleic acid sequencing. The human platelet cDNA encoding SNAP-23 is identical to that reported (28). This human SNAP-23 cDNA was inserted into pQE-9 vector using BamHI and HindIII. Recombinant His 6 -SNAP-23 was prepared from Escherichia coli and purified by Ni 2ϩ -NTA affinity chromatography. Rabbits were immunized with purified His 6 -SNAP-23. The serum was first purified by affinity chromatography using glutathione S-transferase-SNAP-23 covalently coupled by BS 3 cross-linking to glutathione S-transferase-agarose beads. This antibody (Ab23) was further purified by passing it through an affinity column using glutathione S-transferase-SNAP-25 covalently coupled beads. The His 6 -SNAP-25 construct was a generous gift of Dr. Thomas Söllner. The flow-through antibody, termed affinity purified anti-SNAP-23, showed no detectable cross reactivity to His 6 -SNAP-25.
Cell Culture-MDCK strain II cells expressing the wild-type rabbit pIgR have been described (34). Cells were maintained in minimal essential medium (MEM; CellTech, Fisher, Pittsburgh, PA) supplemented with 10% fetal bovine serum (Hyclone, Logan, UT), 100 units/ml penicillin, and 100 g/ml streptomycin in 5% CO 2 , 95% air. In order to maintain a high level of receptor expression, and to obtain reproducible results, new cells were thawed every 1-2 weeks. For all experiments, cells were cultured on 12-mm diameter, 0.4-m pore size Transwells (Costar, Cambridge, MA) as described (34). Following the second day of plating, cells were fed everyday with fresh medium (1.0 ml apical and 1.5 ml basal) and were used 3-4 days after plating.
Fixation, Fluorescent Labeling, and Scanning Laser Confocal Analysis of Labeled Cells-Samples were fixed with paraformaldehyde using a pH-shift protocol, quenched, blocked with 5% (v/v) goat serum, stained, mounted, and stored as described previously (25). The samples were analyzed using an krypton-argon laser coupled to a Molecular Dynamics (Mountain View, CA) Multiprobe 2001 confocal, attached to a Diaphot microscope (Nikon, Melville, NY) with a Plan Apo 60X 1.4 N.A. objective lens (Nikon). The samples were scanned using the appropriate filter combinations. Collection parameters were as follows: laser output, 60 mW, PMT-1 and PMT-2 set to 700 -800 mV, laser attenuation at 10%, 100 m slit. The images (512 ϫ 512 pixels, 0.8 m pixel size) were acquired using ImageSpace software. The images were converted to tag-information-file-format and the contrast levels of the images adjusted in the Photoshop program (Adobe Co., Mountain View, CA) on a Power PC Macintosh 9500 (Apple, Cupertino, CA). The contrast-corrected images were imported into Freehand (Macromedia, San Francisco, CA) and printed from a Kodak (Rochester, NY) 8650PS dye sublimation printer.
Internalization of [ 125 I]Tf-Canine Tf was iodinated using the ICl method to a specific activity of 1.0 -2.0 ϫ 10 7 cpm/g (34). Filter-grown MDCK cells expressing the pIgR were washed three times with MEM/ BSA (Hank's balanced salt solution, 0.6% (w/v) bovine serum albumin, 20 mM HEPES, pH 7.4), and incubated for 45 min at 37°C in MEM/BSA to deplete intracellular stores of Tf. The MEM/BSA was aspirated and 0.5 ml of MEM/BSA was placed in the apical chamber of the Transwell, and the unit was placed on a 40-l drop of MEM/BSA containing 5 g/ml [ 125 I]Tf. The cells were incubated 45 min at 37°C in a humid chamber. Under these conditions the Tf-containing compartments in these cells were maximally loaded with ligand. Free [ 125 I]Tf was then removed from the cell surface by washing the cells three times over a 15-min period at 4°C with MEM/BSA. The cells were then warmed up to 37°C in MEM/BSA for 2.5 min to allow for cell surface [ 125 I]Tf to be internalized as described previously (25,32). Typically, 8,000 -12,000 cpm of [ 125 I]Tf were cell associated following this wash period.
Preparation of SLO, SLO Permeabilization of MDCK Cells, LDH Release, and Assessment of Monolayer Integrity-SLO was prepared and MDCK cells permeabilized with this toxin as described previously (24), however, 35 units/ml of SLO was used instead of 40 units/ml to permeabilize the cells. Briefly, cells were washed twice with ice-cold KOAc ϩ buffer (115 mM potassium acetate, 2.5 mM magnesium acetate, 25 mM HEPES, pH 7.4, 0.9 mM CaCl 2 , and 0.5 mM MgCl 2 ), and once with ice-cold KTM buffer (115 mM potassium acetate, 2.5 mM magnesium acetate, 25 mM HEPES, pH 7.4, 2 mM K 2 CaEGTA (prepared as described in Ref. 35), pH 7.4, 2 mM EGTA-KOH, pH 7.4, and 1 mM DTT). The Transwell was placed on a 15-l drop of SLO, and the cells were incubated for 10 min on ice, and then washed twice quickly and once for 5 min with ice-cold KTM. To induce permeabilization of the cells and wash out cytosol, the cells were incubated 7.5 min at 37°C in KTM buffer. Subsequently, the KTM buffer was aspirated, replaced with ice-cold KTM, and the cells incubated for an additional 30 -45 min on ice to further wash out cellular cytosol.
These permeabilized MDCK monolayers remain functionally intact by several previously described criteria (24): Ͼ90% of the cellular LDH activity is released into the basolateral medium of permeabilized cells; when permeabilized from the basolateral surface, no LDH activity is released apically; antibodies against cytoplasmic epitopes do not stain cells unless their basolateral surfaces have been SLO-permeabilized; and when examined by electron microscopy, the monolayers are found to be intact and junctional complexes are readily observable. In addition, when [ 125 I]Tf was added directly to the buffer bathing the apical side of permeabilized cells Յ0.8% of the marker was found in the opposite chamber following a 90-min incubation at 37°C in cytosol-free 2 D. Chen and S. W. Whiteheart, submitted for publication. transport buffer (standard assay conditions/buffers are given below). All of these observations confirm that the apical membrane remained unpermeabilized in these preparations and there was little leakage or transcytosis of unbound [ 125 I]Tf from one chamber of the Transwell to the opposite in the time frame of the experiments presented in this work.

Preparation of Rat Liver Cytosol and Reconstitution of [ 125 I]Tf
Recycling in Permeabilized MDCK Cells-Preparation of rat liver cytosol has been described previously (24). Reconstituting traffic in the permeabilized cells requires an exogenous source of both energy and cytosol, hence 0.5 ml of KTM containing 8 -10 mg/ml rat liver cytosol, 1 mM ATP, 80 mM creatine phosphate, and 50 g/ml creatine kinase (transport buffer) was added to the basolateral surface of the cells as described previously (24), and 0.5 ml of KTM alone was added to the apical cell surface. The cells were then incubated for periods up to 90 min at 37°C. At the end of the incubation the amount of [ 125 I]Tf released basolaterally (recycled), released apically (transcytosed), or that remained cell associated was quantified in a Cobra II Auto-Gamma counter from Packard (Downers Grove, IL). Cytosol was not added in some experiments, and in other initial experiments the ATP regenerating system was replaced with 40 units/ml of the ATP depleting enzyme apyrase (grade VI, Sigma). However, we later found that in this permeabilized cell system there is no difference between adding apyrase or not adding it, therefore, the addition of this enzyme was omitted in later experiments.
BoNT/E Treatment of Permeabilized MDCK Cells-The 150-kDa single-chain BoNT/E was produced and purified as described (36).
BoNT/E was treated with trypsin to convert the single-chain protein into the 150-kDa di-chain BoNT/E as described previously (37). Neurotoxins (1 mg/ml stocks, stored on ice) were diluted 1:10 in KTM buffer containing 10 mM DTT (freshly dissolved in degassed buffer), and incubated at room temperature for 60 min. The neurotoxins were further diluted in KTM containing 1 mM DTT and then added to the cells during their permeabilization at 37°C. One difference from the normal assay condition described above was that following SLO binding the permeabilization step was extended to 10 min at 37°C to allow for maximal proteolytic action of the neurotoxins on SNAREs. No additional inhibition was observed in this permeabilized cell system when neurotoxin was included in the 90-min transport reaction at 37°C so this step was omitted.
In one set of experiments (see Fig. 7) single chain or di-chain BoNT/E was reduced with DTT as described above or sham treated. Following SLO binding, cells were washed with DTT-free KTM and then warmed up for 10 min at 37°C in the presence of neurotoxin reduced or unreduced with DTT. The cells were then washed once with ice-cold DTTfree KTM and then incubated 30 min in ice-cold DTT-free KTM. Transcytosis and recycling of ligand in these neurotoxin-treated cells was assayed as described above except the transport buffer did not contain DTT.
Western Blot Analysis-Immunoblotting was performed as follows. Transwell filters with BoNT/E-treated or sham-treated cells were boiled 5 min in 100 mM triethanolamine, pH 8.6, 5 mM EDTA, 0.5% SDS, vortex-shaken for 15 min at 4°C in a vortex mixer (Model 5432 Eppendorf), and aliquots resolved by sodium dodecyl sulfate (SDS)polyacrylamide gel electrophoresis (15% gels). Rainbow markers (Amersham, Arlington Heights, IL) were loaded in adjacent lanes as molecular weight standards. The gels were then equilibrated in 10 mM CAPS-NaOH buffer, pH 11.0 (CAPS buffer), for 10 min at room temperature before the proteins were transferred to Immobilon-P membrane (Millipore, Bedford, MA), for 75 min at 375 mA constant current in CAPS buffer. The Immobilon-P membrane was blocked overnight in 5% BSA dissolved in Dulbecco's phosphate-buffered saline, and then incubated with a 1:1000 dilution of affinity purified anti-SNAP-23 antibody in 1% dry non-fat milk/phosphate-buffered saline for 2 h at room temperature. Unbound primary antibody was removed by three 15-min washes in TBST (25 mM Tris, pH 8.0, 500 mM NaCl, 25 mM KCl, 0.05% (w/v) Tween 20), and three washes with phosphate-buffered saline, and the Immobilon-P membrane incubated 60 min at room temperature with goat anti-rabbit horseradish peroxidase diluted 1:25,000 in 1% dry non-fat milk/phosphate-buffered saline. Following three 15-min washes in TBST, SNAP-23 was detected using the SuperSignal chemiluminescent detection system (Pierce, Rockford, IL) following the protocol described by the manufacturer. The resulting bands on the exposed film were quantified by densitometry.
NEM Treatment of Permeabilized MDCK Cells and NSF, SNAP-23, and Anti-SNAP-23 Addition-NEM treatment and addition of recombinant NSF were performed as described (24). In some experiments 0.5-0.05 M recombinant SNAP-23 was added to the transport buffer during the 90-min reaction at 37°C. In other experiments affinity purified rabbit anti-SNAP-23 or affinity purified anti-human IgA antibodies were added (following the SLO binding step) in the warm-up step at 37°C, during the cytosol wash-out at 4°C and in the transport buffer during the 90-min reaction at 37°C.

MDCK Cells
Express SNAP-23-We previously hypothesized that a SNAP-25 homolog exists in polarized MDCK cells (24). One possible candidate is SNAP-23, a widely distributed SNAP-25 homolog that has recently been found in several organs including kidney (28 -31). Initially, we determined if SNAP-23 was expressed in MDCK cell by probing Western blots of lysates from SLO-permeabilized MDCK cells with an affinity-purified rabbit polyclonal antibody against SNAP-23. A major protein band was observed at approximately 28 kDa (Fig. 1, left lane), and is slightly larger than the predicted molecular mass of 23 kDa. The slower than expected mobility has been reported previously (31). There were additional protein bands at 60 and 90 kDa. The 60-kDa protein was susceptible to BoNT/E treatment (see below; Fig. 1, right lane), and may represent dimers of SNAP-23. The faint 90-kDa band was not susceptible to toxin treatment, and may be nonspecific.
SNAP-25 is specifically cleaved by BoNT/E at the Arg 180 -Ile 181 peptide bond (38), and SNAP-23 was predicted to be susceptible to BoNT/E since it retains the identical cleavage sequence (Arg 186 -Ile 187 ), although, at a different position. Surprisingly, recent reports determined that rat and human SNAP-23 are not cleaved by BoNT/E (39,40). To determine whether canine SNAP-23 is susceptible to BoNT/E, we treated SLO-permeabilized MDCK cells with BoNT/E and analyzed the effects of toxin treatment on SNAP-23 immunoreactivity. In cells treated with 650 nM of reduced di-chain BoNT/E there was a large reduction (approximately a 70% loss) in the amount of the 28-kDa band (Fig. 1, right lane) and the appearance of an additional faint band at 25 kDa (marked with an asterisk in Fig. 1). These results are consistent with earlier observations that BoNT/E cleavage of SNAP-25 removes an approximately 3-kDa fragment from the C terminus of this protein (38) and is predicted to remove a similar size fragment from SNAP-23 (31). This cleavage is consistent with a recent report that canine SNAP-23, unlike human SNAP-23, is susceptible to BoNT/E cleavage (39).
The amount of cleaved fragment and uncleaved SNAP-23 following toxin treatment did not add up to the amount of SNAP-23 found in untreated cells. One possible explanation is that the cleaved fragment is only inefficiently recognized by our polyclonal antibody (i.e. as a result of loss of epitopes). Another possibility is the cleaved SNAP-23 fragment is susceptible to other proteinases that rapidly degrade it into small peptides. The significant fraction of uncleaved SNAP-23, following BoNT/E treatment most likely represents SNAP-23 bound in SNARE complexes since, as previously demonstrated, BoNT/E is only weakly effective on assembled fusion complexes (41). 650 nM BoNT/E causes a significant inhibition of recycling in our permeabilized cell system (see below), and we have previously shown that this concentration of toxin gives maximal inhibition of IgA transcytosis and recycling (24).

SNAP-23 Is Found at the Basolateral Cell Surface and in Small Intracellular
Vesicles-The distribution of SNAP-23 has been analyzed in several non-polarized cell lines and is found at the plasma membrane and in intracellular vesicles (29,31). In addition, SNAP-23 has recently been localized to the basolateral surface of pancreatic acinar cells and when human SNAP-23 is overexpressed in MDCK cells it is found at both the apical and basolateral cell surfaces (39,42). We localized endogenous canine SNAP-23 within MDCK cells using confocal microscopy which allows visualization of individual optical planes through the three-dimensional MDCK monolayer. Our goal was to identify if endogenous SNAP-23 was associated with the plasma membrane and might therefore act (in a manner analogous to SNAP-25) as a t-SNARE for endocytic traffic. The cells were co-stained with a monoclonal antibody that recognizes ZO-1, a marker for the tight junction which separates the apical and basolateral plasma membrane (Fig. 2B). At the level of the brightest ZO-1 staining, SNAP-23 was found in small vesicles across the apical pole of the cell (Fig. 2D). It was not possible to determine whether SNAP-23 was present at the apical plasma membrane because the signal was low and at the light level it is hard to discriminate microvillar staining (which also gives a punctate staining pattern) from SNAP-23 found in subapical vesicles. The brightest SNAP-23 staining was observed in small vesicles along the lateral and basal surfaces and also on what appeared to be the basolateral plasma membrane (Fig. 2, F and H). The specificity of the antibody labeling was confirmed by the addition of exogenous SNAP-23 protein which competitively inhibited binding of SNAP-23 antibodies to the cells and completely eliminated the signal (data not shown).
In order to confirm the localization of SNAP-23 at the basolateral plasma membrane, we performed double labeling with an antibody against E-cadherin (Fig. 3, A, D, G, and J) and anti-SNAP-23 (Fig. 3, B, E, H, and K). At steady-state Ecadherin was found predominantly at the lateral plasma membrane. Fig. 3, G, H, and I, demonstrate that SNAP-23 and E-cadherin were colocalized along the lateral membrane surfaces, but there were distinct, SNAP-23-containing vesicles that lacked E-cadherin. The presence of SNAP-23 on the plasma membrane and intracellularly is consistent with previous observations on the localization of this protein in other cell types (29, 31); however, in MDCK cells there seemed to be an extensive amount of vesicular staining.
Tf Recycling in Permeabilized MDCK Cells Requires ATP, Cytosol, and NSF-The exit of pIgR⅐IgA complexes from the ARE and fusion with the apical and basolateral plasma membrane were previously shown to require NSF and were BoNT/E sensitive (24). However, basolateral recycling of IgA ligand was more sensitive to BoNT/E than transcytosis (50% inhibition of recycling versus 30% inhibition of transcytosis). Coupled with the above observations that SNAP-23 was found predominantly on the basolateral plasma membrane and in basolateral vesicles, we determined whether basolateral recycling of Tf (the classical basolateral recycling marker) could be reconstituted

SNAP-23 Required for Transferrin Recycling
in a permeabilized cell system. Our objective was to use this system to determine if Tf recycling was susceptible to BoNT/E and required SNAP-23. In addition, we hoped to confirm and extend our original observations and gain insight into the molecular requirements for the basolateral recycling pathway.
Because MDCK cells do not express the appropriate receptors for BoNT/E binding, we have reconstituted Tf recycling traffic in SLO-permeabilized MDCK cells. SLO forms pores in the plasma membrane which are sufficiently large enough to allow BoNT/E to enter the cells. In this assay [ 125 I]Tf was internalized from the basolateral surface for 45 min at 37°C (upon which all the Tf containing compartments were maximally loaded with ligand). The cells were washed at 4°C to remove nonspecifically bound ligand and then they were incubated for 2.5 min at 37°C to allow for cell surface-bound Tf to be internalized, as described previously (25,32). Under these internalization conditions Tf was found in small vesicles in the basal region of the cell and underlying the lateral borders (data not shown). In addition, Tf was found at the apical pole of the cell, both at and above the level of the tight junctions, in a vesicular compartment that has previously been characterized as a subdomain of the ARE (25).
Following [ 125 I]Tf internalization, the cells were rapidly cooled and SLO bound to the basolateral surface. Free SLO was washed from the cell surface, the cells were warmed up to 37°C to allow permeabilization to occur and then re-cooled to 4°C for extended cytosol washout. When permeabilized cells were incubated in the presence of an ATP regenerating system and exogenous rat liver cytosol, approximately 45-50% of the preinternalized ligand was released basolaterally (recycled) over a 90-min incubation period at 37°C (Fig. 4). This is an efficiency of 50 -55% (approximately 90% of ligand is recycled basolaterally in non-permeabilized cells during the same time course). (Fig. 4). These observations are consistent with a previous analysis of the energy and cytosol requirements for Tf recycling in mechanically perforated MDCK cells (32). In the absence of ATP there was a 4 -5-fold reduction in recycling and cytosol depletion resulted in a 3-fold reduction. In this cell system, there was a fraction of Tf (approximately 18 -19%) that was released apically (transcytosed). This is more than has been observed in nonpermeabilized cells (typically 5-8%) and may reflect some loss of cytosolic machinery that controls the amount of Tf transcytosis. Because this signal is relatively small, and because Tf is generally considered a marker of the basolateral recycling pathway we have not included the apical release of this marker in the subsequent figures. However, for the sake of completeness the amount of Tf released apically is included in each of the subsequent figure legends. In general, all treatments affected Tf recycling and transcytosis in a similar manner.

Basolateral recycling of [ 125 I]Tf required both ATP and cytosol
As an additional control, we also determined if Tf recycling requires NSF. If so, it would suggest that a bona fide membrane fusion event(s) was occurring. In fact, NEM treatment (which inactivates NSF) reduced recycling by 82% (Fig. 5). No inhibition was observed if both NEM and DTT were added simultaneously, confirming that NEM treatment is the cause of the inhibition. More importantly, exogenous NSF added subsequent to NEM treatment restored much of the recycling activity (72% of the control value) (Fig. 5). This observation confirmed that Tf recycling is NSF-dependent and likely requires use of an NSF⅐SNAP⅐SNARE complex. The inability of NSF to completely restore activity has been reported previously and probably reflects the inactivation of other NEM-sensitive proteins other than NSF (24,43).  The primarily basolateral release, high efficiency, and the requirements for energy, cytosol, and NSF confirm that we have faithfully reconstituted the Tf basolateral recycling pathway in MDCK cells. Previous investigators have permeabilized the plasma membrane domain opposite of the cell surface from which trafficking will be explored (19,44). However, we find that basolateral recycling in apically permeabilized cells is inefficient and the large amount of SLO toxin required is cost prohibitive. Permeabilizing the same membrane that is the chief target for fusion is not inherently different from other SLO-permeabilized cell systems involving nonpolarized cells in which there is only one plasma membrane domain that is both permeabilized and the target of fusion (45)(46)(47).
Basolateral Recycling of Tf Is BoNT/EϪsensitive-The dependence of Tf recycling on NSF and the expression of SNAP-23 at the basolateral plasma membrane suggests that SNAP-23 may be a t-SNARE involved in Tf delivery to the basolateral cell surface. In order to determine the involvement of SNAP-23 in Tf traffic, SLO-permeabilized cells were treated with reduced di-chain BoNT/E. Tf recycling was inhibited by reduced di-chain BoNT/E in a concentration-dependent fashion (Fig. 6). Maximal inhibition was observed at 650 nM which inhibited ϳ80% of the cytosol-dependent basolateral release. The cytosol-dependent release was calculated by subtracting the percent release that occurred in the absence of cytosol from the percent release that occurred in the presence of cytosol. This is similar to the reporting technique used by Simons and co-workers (19,44). BoNT/E had little effect on the fraction of basolateral recycling that occurred in the absence of cytosol (Fig. 6). These cytosol-independent reactions may represent SNAP-23 that is already bound in fusion complexes. As described above, these complexes are thought to be toxin-insensitive. These results support the conclusion that the majority of the cytosol-dependent reaction is sensitive to BoNT/E and therefore may require SNAP-23.
The neurotoxicity of single chain neurotoxins is enhanced by proteolytic cleavage (nicking) into a di-chain form (48), and the inhibitory activity of the di-chain form is further enhanced after reduction of its disulfide bond (49,50). For example, proteolysis of synthetic peptides of SNAP-25 and VAMP-2 by BoNT/E, BoNT/A, and BoNT/B are stimulated by reduction with DTT (51,52). We therefore examined the inhibitory activity of either single chain or di-chain BoNT/E (Ϯ reduction) (Fig.  7). As expected, only reduced di-chain toxin inhibited basolateral release of Tf (Fig. 7A); single chain toxin had little effect both before and after reduction (Fig. 7B). Note that only 300 nM concentration of toxin was used in the experiment described in Fig. 7A as our toxin supply was limited. These results demonstrate that inhibition of Tf recycling by BoNT/E is only observed following nicking and reduction of the toxin, and confirms that the inhibition we have observed is not due to some contaminating inhibitory activity present in the toxin preparation.
Tf Recycling Is SNAP-23-dependent-Although BoNT/E is believed to specifically proteolyze SNAP-25 and SNAP-25 homologs (in our case SNAP-23), it cannot be excluded that it may be proteolyzing some unknown substrate that is required for Tf recycling. Since endogenous SNAP-23 is believed to be membrane-bound (probably via palmitoylation of one of its internal cysteine residues) (28,29) and part of the SNARE complex for vesicle fusion, addition of exogenous SNAP-23 might act as a competitive inhibitor of basolateral recycling by forming nonproductive interactions between exogenous SNAP-23 and endogenous syntaxin or VAMP. Earlier studies have shown that a 20-amino acid peptide derived from the C terminus of SNAP-25 can inhibit SNAP-25-dependent exocytosis in chromaffin cells (53). In this system, addition of recombinant SNAP-23 inhibited approximately 45% of cytosol-dependent Tf recycling (Fig. 8A), but had no effect on the cytosol-independent reactions.
As additional evidence, we tested the effect of anti-SNAP-23 antibody addition on the assay (Fig. 8B). This is expected to inhibit endocytic traffic by binding SNAP-23 and sterically hindering its ability to interact with other proteins. Antibody treatment resulted in an approximately 35% inhibition of the cytosol-dependent recycling reaction (Fig. 8B). A nonspecific antibody against human IgA (used as a control) had no effect on Tf recycling. These results confirm that SNAP-23 is required for a fraction of Tf recycling.

DISCUSSION
The SNARE hypothesis predicts that the specificity of vesicle targeting and fusion is mediated through specific interactions between v-SNAREs and t-SNAREs (7). If true, then each membrane trafficking event would require distinct isoforms or homologs of the various SNARE components. Until recently, however, only isoforms of syntaxin and VAMP, but no mammalian isoforms of SNAP-25 have been identified (54,55). In the last 2 years, Ravichandran et al. (28) identified a SNAP-25 homolog with a predicted molecular mass of 23 kDa called SNAP-23. This protein, expressed in several tissues, is able to bind various syntaxins and VAMPs. A mouse isoform of SNAP-23, also known as syndet, has recently been cloned from an adipocyte 3T3 L1 cDNA library (29,30), and has a similar cellular distribution to human SNAP-23. Munc 18c, an N-sec-1 homolog and putative regulator of the NSF⅐SNAP⅐SNARE complex, has been shown to inhibit SNAP-23 and syntaxin 4 binding in vitro (30). While these observations identify some of the potential binding partners of SNAP-23, the role of SNAP-23 in membrane trafficking remains largely uncharacterized.
Previous work is consistent with the hypothesis that SNAP-23 plays a role in exocytosis. Banerjee et al. (56,57) have shown that exocytosis of large core granules in permeabilized PC12 cells is inhibited completely by BoNT/E but not BoNT/A, even though SNAP-25 is efficiently cleaved by both toxins. However, when PC12 cells are treated with nerve growth factor, release of granules became BoNT/A-sensitive. It is now known that these cells express both SNAP-25 and SNAP-23 (31). It is therefore possible that SNAP-23, which is sensitive to E-toxin and predicted to be insensitive to A-toxin, may be required for release of PC12 storage granules in untreated cells, while in nerve growth factor-treated cells SNAP-25 (and possibly SNAP-23) is required. These observations also suggest that usage and expression of SNARE components may change with the development of any particular cell or tissue.
In the present study we have demonstrated that SNAP-23 is not only present in MDCK cells, and cleaved by BoNT/E, but is required for basolateral recycling of Tf. Our observation that canine SNAP-23 is susceptible to toxin cleavage is consistent with work by Low et al. (39) which have shown that the canine isoform of SNAP-23 is sensitive to BoNT/E cleavage while the human isoform of SNAP-23 is insensitive. Even though BoNT/E is thought to specifically cleave SNAP-25 and SNAP-25 homologs, other proteins in the cell may be susceptible. In order to examine SNAP-23 involvement more directly, we added either exogenous SNAP-23 or anti-SNAP-23 antibodies to the assay. Addition of either resulted in an inhibition of recycling; however, the level of inhibition was less than that attained with BoNT/E treatment. The added SNAP-23 and anti-SNAP-23 antibodies might only compete at steps prior to SNARE complex formation and only ineffectively compete or was included during the 90-min transport reaction at 37°C. The percent of total ligand released basolaterally in the presence of an ATPregenerating system (ATPRS) and in the presence or absence of rat liver cytosol (RLC) was quantified. The percent of total ligand released apically was as follows: ATPRS ϩ RLC, 19.9 Ϯ 1.9%; ATPRS ϩ RLC ϩ 0.05 M SNAP-23, 18.9 Ϯ 1.0%; ATPRS ϩ RLC ϩ 0.125 M SNAP-23, 13.6 Ϯ 1.8%; ATPRS ϩ RLC ϩ 0.5 M SNAP-23, 11.8 Ϯ 3.3%; ATPRS Ϫ RLC, 6.9 Ϯ 1.3%; ATPRS Ϫ RLC ϩ 0.5 M SNAP-23, 7.5 Ϯ 1.4%. Mean and S.D. of values from a representative experiment are given (n ϭ 3). B, ligand was internalized, SLO bound to the basolateral cell surfaces, and cells permeabilized for 10 min at 37°C in the absence (control) or presence of anti-SNAP-23 or anti-human IgA (hIGA) antibodies. Antibodies were included during the cytosol wash-out and during the 90min transport reaction at 37°C. The percent of ligand released basolaterally during a 90-min incubation at 37°C in the presence of an ATP-regenerating system (ATPRS) and in the presence or absence of rat liver cytosol (RLC) was quantified. The percent of total ligand released apically was as follows: ATPRS ϩ RLC, 13.7 Ϯ 0.6%; ATPRS ϩ RLC ϩ 20 g/ml anti-SNAP-23, 10.4 Ϯ 1.0%; ATPRS ϩ RLC ϩ 4 g/ml anti-SNAP-23, 10.5 Ϯ 0.6%; ATPRS ϩ RLC ϩ 20 g/ml anti-hIgA, 15.1 Ϯ 1.6%; ATPRS Ϫ RLC, 1.0 Ϯ 0.5%; ATPRS Ϫ RLC ϩ20 g/ml anti-SNAP-23, 0.5 Ϯ 0.9%. Mean and S.D. of values from a representative experiment are given (n ϭ 3). sterically hinder already formed SNARE complexes. This possibility would be consistent with recent observations by Weber et al. (8) who have reconstituted v-and t-SNAREs into separate lipid bilayer vesicles which fuse with one another in a SNAREdependent reaction. This fusion reaction can be inhibited when soluble VAMP or t-SNAREs are included in a preincubation reaction (during which fusion complex formation is occurring). However, they have no effect when added to reactions in which fusion complexes have already formed. Another possibility is that there are additional BoNT/E-sensitive SNAP-23/SNAP-25 homologs in MDCK cells. These unidentified homologs could account for the higher level of BoNT/E inhibition observed than when SNAP-23 or anti-SNAP-23 antibodies were added exogenously to the permeabilized cell system.
Localization of SNAP-23 reveals that it is present along the basolateral plasma membrane. It was not possible to determine unequivocally if SNAP-23 was present at the apical cell surface. Ultrastructural localization of SNAP-23 is currently under way. There was extensive localization of endogenous SNAP-23 within intracellular vesicular structures. This is contrary to findings by Low et al. (39) who found little intracellular SNAP-23 in MDCK cells. However, this group examined the distribution of overexpressed human SNAP-23, whereas, we have examined the endogenous distribution of this protein. The amount of intracellular SNAP-23 is not unexpected since SNAP-23 localization in non-polarized cells is both at the plasma membrane and in intracellular vesicles (29,31). Some of the intracellular vesicles we observed could be endocytic in nature, a possibility we are actively exploring. In neurons, SNAP-25 and syntaxin 1 are found not only at the plasma membrane but are also abundant constituents of both synaptic vesicles and clathrin-coated vesicles (58). While the function of these proteins in this setting is unknown, it has been suggested that SNAP-25 and syntaxin 1 may reside in these structures as a result of incomplete dissociation from the v-SNARE VAMP, or that t-SNAREs may be reactivated by recycling (58). Alternatively, these t-SNAREs may be involved in synaptic vesicle fusion with already docked synaptic vesicles in a process termed compound exocytosis (58).
In MDCK cells it is thought that the final step of recycling and transcytosis is direct transport from endosomes to the cell surface. An alternative possibility is that several vesicles may fuse with one another prior to fusion with the cell surface. SNAP-23 could function in these intermediate steps of recycling or transcytotic vesicle fusion, and therefore may be acting at steps prior to or in addition to vesicle fusion with the plasma membrane. In fact, the differential effects we observe between BoNT/E treatment and either SNAP-23 or anti-SNAP-23 addition may reflect the differential sensitivity of these agents on homotypic fusion of recycling vesicles or fusion of these vesicles with the plasma membrane.
At present it is not clear if SNAP-23 is also required for fusion of vesicles with the apical plasma membrane domain of MDCK cells. The apical release of Tf has a similar sensitivity to BoNT/E and requirement for SNAP-23, and as such basolateral to apical transcytosis of Tf may also require SNAP-23. In addition, our previous observations that transcytosis of IgA in MDCK cells is inhibited by BoNT/E is consistent with there being a SNAP-23 requirement for fusion with the apical plasma membrane (24). If we assume that SNAP-23 is a SNARE required for fusion at both cell surfaces, then the specificity of fusion must be regulated by additional factors. One possibility is that distinct v-SNAREs are present on basolateral recycling versus transcytosing vesicles or that additional t-SNAREs, present on each of the plasma membrane domains, are required for fusion. VAMP-2, for example, interacts with syntaxin 1 and 4 but not with syntaxin 2 and 3 (59). Syntaxin 4 is found primarily at the basolateral surface of MDCK cells (23) and is known to bind with high affinity to SNAP-23 (30). It is therefore possible that VAMP-2 will be found on basolateral recycling vesicles and will be required for fusion with the basolateral cell surface. Syntaxin 3, in contrast, is found at both the apical membrane and in apical vesicles and may be involved in fusion of vesicles with the apical pole of the cell (22,23). Additional work is required to identify the binding partners of SNAP-23 and which components are required for the targeting specificity of recycling and transcytotic vesicles. The existence of fusion events that occur in the absence of NSF⅐SNAP⅐SNARE complexes further extends the possible types of fusion events possible in the cell (19,60). The present work suggests that SNAP-23 may be one component of a complex required for vesicle fusion with at least the basolateral plasma membrane domain of MDCK cells.