A Novel Assay Reveals a Role for SolubleN-Ethylmaleimide-sensitive Fusion Attachment Protein in Mannose 6-Phosphate Receptor Transport from Endosomes to the Trans Golgi Network*

Soluble N-ethylmaleimide-sensitive fusion protein (NSF) attachment protein (α-SNAP) is a soluble protein that enables the NSF ATPase to associate with membranes and facilitate membrane trafficking events. Although NSF and α-SNAP have been shown to be required for many membrane transport processes, their role in the transport of mannose 6-phosphate receptors from endosomes to the trans Golgi network was not established. We present here a novel in vitro assay that monitors the transport of cation-dependent mannose 6-phosphate receptors between endosomes and the trans Golgi network. The assay relies on the trans Golgi network localization of tyrosine sulfotransferase and monitors transport of mannose 6-phosphate receptors engineered to contain a consensus sequence for modification by this enzyme. Using this new assay we show that α-SNAP strongly stimulates transport in reactions containing limiting amounts of cytosol. Together with α-SNAP, NSF can increase the extent of transport. These data show that α-SNAP, a soluble component of the SNAP receptor machinery, facilitates transport from endosomes to the trans Golgi network.

NSF 1 is a trimeric ATPase that is required for the transport of proteins between a variety of membrane-bound compartments in eukaryotic cells (1)(2)(3). NSF requires the so-called SNAP proteins to bind to membranes. Three SNAP isoforms have been identified: ␣, ␤, and ␥ (4). SNAPs are soluble proteins that bind saturably to Golgi membranes (5,6).
When NSF is mixed together with ␣-SNAP and a detergent extract of salt-washed Golgi membranes, NSF and ␣-SNAP become incorporated into a 20 S complex (7). Characterization of the proteins contained within this complex led to the identification of the SNAP receptors or SNAREs (8). Pairing of v-SNAREs on transport vesicles and t-SNAREs on their cog-nate membrane targets is thought to provide specificity for vesicle docking and fusion events (3). ATP hydrolysis by NSF induces a conformational change in the SNARE proteins (9) that triggers SNARE complex disassembly, prior to membrane fusion.
␣-SNAP and NSF have been implicated in most heterotypic and homotypic docking and fusion reactions (3, 10 -14). However in MDCK cells, transport from the trans Golgi network (TGN) to the apical plasma membrane was independent of NSF and could not be stimulated by ␣-SNAP (15). In yeast, a late step in the delivery of internalized ␣-factor to the vacuole requires an NSF homolog, the END13 gene product (16). Furthermore, delivery of carboxypeptidase Y from the TGN to the vacuole was shown to be independent of NSF (17). Carboxypeptidase Y is sorted by a receptor in the TGN for delivery to the endosomal/vacuolar system. The sorting receptor, Vps10p, has to return to the TGN for additional rounds of transport (18). Thus, it was possible that recycling from endosomes to the TGN might also be independent of NSF.
We study the transport of mannose 6-phosphate receptors (MPRs) between endosomes and the TGN, both in cultured cells (cf. Ref. 19) and in an in vitro assay that reconstitutes this transport process (20). In mammalian cells, MPRs carry lysosomal hydrolases from the TGN to endosomes, where they release the hydrolases and then return to the TGN for another round of transport (21). In previous work, we were unable to determine if NSF was required for endosome-to-TGN transport because this step requires a factor that is more sensitive to N-ethylmaleimide than NSF (22). Although NSF-depleted cytosol was active in in vitro transport, it was possible that the abundant, membrane-associated NSF present in those reactions was sufficient for the transport observed. Using a novel in vitro transport assay, we show here that ␣-SNAP stimulates the recycling of MPRs from endosomes to the TGN. Furthermore, NSF can stimulate transport in the presence of ␣-SNAP.

EXPERIMENTAL PROCEDURES
GDI and Rab9⅐GDI complexes were prepared as described (23). Mouse cation-dependent mannose 6-phosphate receptor (MPR46) cDNA was a gift of Dr. Bernhard Hoflack (EMBL, Heidelberg, Germany). The bacterial expression plasmid pMW183 containing His 6 ::NodQ 2 was a gift of Dr. Sharon Long (Stanford University). Dr. James Rothman (Sloan-Kettering, New York), Dr. Sidney Whiteheart (Kentucky College of Medicine, Lexington), and Dr. Thomas Mayer (University of Muenster, Germany) kindly provided expression plasmids for His-tagged versions of NSF, ␣-SNAP, and Rab1b N121I, respectively. The following chemicals (Sigma) were kept as stocks: 100 ϫ protease inhibitor mix: 1 mg/ml leupeptin, 4 mg/ml aprotinin, 100 M pepstatin A, stored at Ϫ80°C; 20 mg/ml cycloheximide in ethanol and 20 mM primaquine in H 2 O, stored at Ϫ20°C. cDNA position 241 of mouse MPR46 by polymerase chain reaction and cloned via a primer-based 5Ј-NcoI site to the 3Ј-end of an influenza virus hemagglutinin leader sequence fused to a FLAG epitope tag that was derived from pSP65SF␤2 (26). Next, the sequence encoding the FLAG epitope tag was replaced by polymerase chain reaction with a cassette coding for a (His) 6 /myc-epitope tag (see Fig. 1B). The modified 5Ј-region was cloned back to the MPR46 cDNA via a ScaI site at position 346, replacing its 5Ј-region. This construct was designated MPR46HMY and confirmed by DNA sequencing. MPR46HMY was cloned as an XhoI/ XbaI fragment into pME18S (K. Maruyama, University of Tokyo). This construct and pSV2neo (ratio 10:1) were cotransfected by electroporation into CHO cells. Stable transformants were selected with geneticin (1 mg/ml, Life Technologies, Inc.) in ␣-MEM supplemented with 7.5% fetal calf serum. Of 23 isolated clones, 11 were positive by Western blot and indirect immunofluorescence (27) using monoclonal antibody 9E10 against the c-myc epitope (28). Clones were maintained in the above medium. All studies were carried out with clone 3.

Synthesis of 35 S-Labeled 3Ј-Phosphoadenosine 5Ј-Phosphosulfate
The synthesis of 35 S-labeled 3Ј-phosphoadenosine 5Ј-phosphosulfate (PAPS) was modified from Ehrhardt et al. (29). The reaction is outlined in Fig. 1C. A typical synthesis reaction, total volume of 200 l, consisted of 1 mCi of sodium [ 35 S]sulfate (ICN) in reaction buffer (10 mM Tris/ HCl, pH 8, 30 mM KCl, 5 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol) containing 20 mM ATP, 10 mM GTP, 10 units/ml ATPsulfurylase from Saccharomyces cerevisiae (Sigma), 100 units/ml inorganic pyrophosphatase (Sigma), and 10 g/ml recombinant His-tagged NodQ2 gene product from Rhizobium meliloti (NodQ2 has both ATPsulfurylase and APS kinase activity (49)). The reaction was incubated for 2 h at 30°C, and enzymes were inactivated by boiling for 1 min and aggregates pelleted for 2 min in a microcentrifuge. The reaction mix was diluted to 2 ml with distilled H 2 O yielding a specific activity of 0.5 Ci/l (corresponding to ϳ 0.3 mM PAPS). The extent of conversion of [ 35 S]sulfate into [ 35 S]PAPS was monitored by spotting 2 l of the reaction product onto a polyethyleneimine-cellulose F (Merck) TLC plate, previously washed for 5 min with H 2 O and dried. The reaction products were resolved by 0.9 M LiCl.

Metabolic Labeling and Immunoprecipitation
Cells were grown to subconfluency, washed with TD (25 mM Tris/ HCl, pH 7.4, 5.4 mM KCl, 137 mM NaCl, 0.3 mM Na 2 HPO 4 ), and preincubated in labeling medium for 15 min (␣-MEM lacking methionine and cysteine but containing 7.5% dialyzed fetal calf serum). Cells were pulsed for 15 min with 1 mCi/ml [ 35 S]methionine and -cysteine (Tran 35 S-label, ICN), transferred to ice, washed twice with chase medium (␣-MEM, 7.5% fetal calf serum, 10 mM methionine, 10 mM cysteine), and chased for the indicated times. Cells were transferred to ice, washed with TD, harvested in RIPA (50 mM Tris/HCl, pH 7.8, 150 mM NaCl, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate (SDS), 1.5% Triton X-100) in the presence of 1 mM phenylmethylsulfonyl fluoride, and solubilized for 1 h. After centrifugation at 100,000 ϫ g for 10 min at 4°C, the supernatant was transferred to a new tube; 5 l of mAb 9E10 ascites fluid was added and incubated for 1 h on ice. Rabbit anti-mouse IgG (2.5 l, Calbiochem) was added for 30 min, and the immune complexes were harvested at 4°C overnight with 25 l of Staphylococcus aureus cells (Pansorbin, Calbiochem), preincubated for 1 h with RIPA containing 0.5% bovine serum albumin. Complexes were pelleted and washed 4 ϫ with RIPA and 1 ϫ with TD. Immune complexes were released by boiling for 5 min in 50 l of sample buffer. Proteins were resolved by 10% SDS-PAGE (30); the gel was treated with Entensify (ICN), dried, and exposed to film.

Cycloheximide Chase and Endo H Digestion
For the cycloheximide chase experiments, preincubation and labeling were as described above. The chase was initiated by adding 20 g/ml cycloheximide (Sigma) to the labeling medium; [ 35 S]methionine and -cysteine were also present throughout the chase. Cells were transferred to ice, washed 2 ϫ with TD, and solubilized for 2 h with RIPA, 25 mM imidazole. After centrifugation at 100,000 ϫ g at 4°C, the supernatant was added to 10 l of Ni-NTA agarose beads (Qiagen), prewashed with RIPA, 25 mM imidazole. For endo H reactions, MPR46HMY was allowed to bind to the beads at 4°C for 1 h, washed 4 ϫ with RIPA, 25 mM imidazole, 1 ϫ with TD, boiled for 3 min in 50 l of 200 mM sodium citrate, pH 5.5, 2% SDS, frozen at Ϫ20°C for 30 min, and boiled again for 3 min. Then the samples were diluted with an equal volume of distilled water containing 2 ϫ protease inhibitors and incubated with 4 milliunits of endo H (Boehringer Mannheim) for 22 h at 37°C. Proteins were resolved by 10% SDS-PAGE, and the gel was treated with Entensify (ICN), dried, and exposed to film.
Cell Breakage-Cell breakage was modified from Acharya et al. (31). Cells were washed once with KHM buffer (25 mM HEPES-KOH, pH 7.4, 125 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM dithiothreitol), drained well, and snap-frozen on a metal plate cooled on dry ice. After thawing the cells for 4 min at room temperature, they were transferred on ice and washed 2 ϫ with KHM to remove cytosol. The plates were drained extensively; 250 l of SEAT buffer (10 mM triethanolamine, 10 mM acetic acid, pH 7.4, 1 mM EDTA, 0.25 M sucrose) was added, and the cells were scraped with a rubber policeman and pooled.
Preincubation with Non-labeled PAPS, Collection, and Freezing Membranes-Broken cells were preincubated with unlabeled PAPS in the absence of added cytosol to reduce Golgi-derived background. Purification and Quantification of Tyrosine-sulfated MPR46HMY-The membranes were solubilized with 500 l of RIPA, 25 mM imidazole for 2 h on ice; insoluble material was pelleted in a microcentrifuge for 5 min, and the supernatant was transferred onto 10 l of prewashed Ni-NTA agarose beads. MPR46HMY were allowed to bind for 1 h at 4°C, followed by 4 washes with RIPA, 25 mM imidazole and eluted with 100 l of 25 mM EDTA in RIPA, 25 mM imidazole.

Preparation of Recombinant ␣-SNAP, NSF, and His 6 ::NodQ 2
Histidine-tagged bacterially expressed ␣-SNAP and NSF were purified essentially as described by Whiteheart et al. (2,32). For NSF the only change is that the final step is an S-300 column instead of a Superose 6 Fast Flow (Pharmacia Biotech Inc.). ATPase assays were carried out as described in Tagaya et al. (33, data not shown). His 6 ::NodQ 2 , the ATP-sulfurylase/APS kinase from R. meliloti, was expressed in XL-1 blue cells grown to an A 660 of 0.8 at 37°C, induced for 4 h at 30°C with 20 M isopropyl-1-thio-␤-D-galactopyranoside. The cells were harvested by centrifugation, resuspended in 50 mM sodium phosphate, pH 8, 300 mM NaCl, 20 mM imidazole, 10% glycerol, and disrupted by French press. Homogenates were cleared by centrifugation, and the supernatant was incubated with 1 ml of Ni-NTA agarose for 30 min at 4°C. The column was packed, washed with 100 volumes of buffer, and eluted with 500 mM imidazole in buffer. Essentially pure fractions were pooled and dialyzed against 20 mM Tris/HCl, pH 8, 30 mM KCl, 5 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol, aliquoted, and stored at Ϫ80°C. Activity of His 6 ::NodQ 2 was verified measuring the conversion of radiolabeled sulfate to radiolabeled PAPS and separation of the reaction products by TLC as described above (see Fig. 3B).

RESULTS
In previous work, we were unable to determine if NSF was required for endosome-to-TGN transport because this step requires a factor that is more sensitive to N-ethylmaleimide than NSF (22). In these experiments, it was possible that the abundant, membrane-associated NSF was sufficient for the transport observed. Thus, we sought to establish a facile assay that reconstituted the same transport process but might show greater dependence upon added cytosolic factors.
Our previously established in vitro transport reaction makes use of the TGN localization of sialyltransferase and a CHO mutant cell line in which N-linked oligosaccharide sialylation does not occur (20). Transport requires the presence of purified, "wild type" rat liver Golgi membranes, as well as ATP and cytosolic proteins. Although this assay has enabled us to show a role for the Rab9 GTPase and Rab9⅐GDI complexes in MPR trafficking (23,34), it requires time-consuming metabolic labeling, tedious chromatography steps, SDS-PAGE, and fluorography.
Our new assay ( Fig. 1) takes advantage of another TGNspecific enzyme, tyrosine sulfotransferase (35,36). The N terminus of the cation-dependent, murine MPR (MPR46, 24) was engineered to contain a tyrosine sulfotransferase recognition sequence derived from the C terminus of the cholecystokinin precursor (37). Leitinger et al. (25) have shown that this nonapeptide can serve as a substrate for tyrosine sulfation when transplanted to the C termini of the asialoglycoprotein receptor H1 subunit or ␣1-proteinase inhibitor. In addition, we added an N-terminal histidine tag for purification (38), followed by a myc-tag for detection (Ref. 28; Fig. 1B). The modified receptor will be referred to as MPR46HMY.
MPR46HMY Is Properly Targeted to the MPR Recycling Pathway in Vivo-At steady state in cultured cells, endogenous MPRs are found predominantly in late endosomes; a smaller number are present in the TGN and early endosomes, and less than 5-10% are located at the cell surface (21). We first analyzed the biochemical properties of MPR46HMY in vivo. CHO cells stably expressing MPR46HMY were pulse-labeled for 15 min and chased for up to 29 h. MPR46HMY was immunoprecipitated with anti-myc tag antibodies and analyzed by SDS-PAGE and fluorography. Fig. 2A shows that MPR46HMY is stable for more than 29 h. The protein was first detected as a 46-kDa glycoprotein which then increased to ϳ50 kDa due to oligosaccharide maturation; it was converted to its mature form, as judged by endo H cleavage, after ϳ1 h (Fig. 2B). As shown previously, mature MPR46 contains both endo H-resistant and endo H-sensitive oligosaccharide chains (cf. Ref. 50).
These experiments indicate that transport of tagged MPRs through the endoplasmic reticulum and up to the medial Golgi is not delayed due to inefficient folding.
To test the functionality of the tyrosine sulfation tag, untransfected CHO cells (Fig. 2C, lane 1) or cells expressing MPR46HMY (Fig. 2C, lane 2) were labeled with [ 35 S]sulfate for 2 h, and MPR46HMY was purified by detergent solubilization and nickel agarose binding. Only the transfected cells expressed a sulfate-labeled polypeptide that bound to the nickel resin; only a single prominent polypeptide was detected that had the mobility expected for MPR46HMY. This confirmed that the tyrosine sulfation tag was utilized by tyrosine sulfotransferase in vivo and also that [ 35 S]sulfate-labeled MPR46HMY can be purified in one step by virtue of its histidine tag.
The presence of the TGN-specific, tyrosine sulfate modification showed that MPR46HMY was localized at or beyond the TGN. Moreover, the stability of MPR46HMY is a strong indication that it was not mis-sorted to lysosomes, as trafficking mutants of MPR46 are delivered to lysosomes and rapidly degraded there (39). Correct localization was confirmed by indirect immunofluorescence microscopy (Fig. 3). Upon transient transfection in COS cells, MPR46HMY showed excellent co-localization with the 300-kDa, cation-independent MPR  (Fig. 3, A and B). Unfortunately, the anti-300-kDa MPR antibodies do not recognize the CHO cell form of this protein.
Resialylation experiments were carried out to test recycling of MPR46HMY from the plasma membrane to the TGN in cultured cells. Cells were pulse-labeled with [ 35 S]sulfate, and glycoproteins reaching the plasma membrane were desialylated with neuraminidase. The cells were then recultured for several hours in the absence of neuraminidase to permit glycoproteins recycling to the TGN to re-acquire sialic acid. The rate of resialylation, determined by slug lectin binding (20), was identical to that reported for the cation-independent MPR in CHO wild type cells (data not shown; Ref. 19). We therefore conclude that MPR46HMY is functional in terms of its ability to recycle to the TGN in living cells.
In Vitro Sulfation as a Measure of Transport-The new in vitro reaction utilizes [ 35 S]3Ј-phosphoadenosine 5Ј-phosphosulfate (PAPS) as the precursor for tyrosine sulfation. PAPS is normally synthesized in the cytosol from sulfate and ATP in a two-step process (Fig. 4A). Sulfate and ATP are first converted to phosphoadenosine 5Ј-phosphosulfate (APS) by ATP-sulfurylase; APS is then converted to PAPS by APS kinase. ATPsulfurylase is reversibly inhibited by chlorate at millimolar concentrations, both in vivo and in vitro (40).
[ 35  ATP with an efficiency of more than 95% (Fig. 4B), using commercially available ATP-sulfurylase and inorganic pyrophosphatase together with recombinant APS kinase from R. meliloti (29). [ 35 S]PAPS synthesis was monitored by thin layer chromatography (Fig. 4B) and could be used in in vitro transport reactions without further purification. [ 35 S]PAPS was unstable at 37°C in the presence of cytosolic proteins. However, inclusion of 10 mM chlorate in reaction mixtures provided significant stabilization of the PAPS substrate (Fig. 4B).
To use tyrosine sulfation as a measure of endosome-to-TGN transport, the rate of tyrosine sulfation must be significantly faster than the rate of the overall transport process. We therefore examined the kinetics of [ 35 S]PAPS translocation and its subsequent incorporation into trichloroacetic acid-precipitable material using purified rat liver Golgi membranes. In the presence of 20 M Tyrosine Sulfation-based Endosome-to-TGN Transport Assay-The new assay utilizes donor membranes containing unsulfated MPR46HMY receptors. To accumulate unsulfated receptors, cultured cells are grown for 3 days in sulfate-free medium containing 10 mM chlorate. Four hours prior to cell harvest, cycloheximide is added to chase newly synthesized MPR46HMY to their steady state localization. After as little as 1 h of cycloheximide treatment, all MPR46HMY receptors are at or beyond the medial Golgi in these cells (Fig. 2B). Moreover, the protein synthesis block is complete because the intensity of the MPR46HMY band did not increase with time, despite the continued presence of radiolabeled methionine and cysteine (Fig. 2B). Cells are permeabilized by freeze-thaw; cytosol is washed away, and then the cells are further broken by scraping off the plate.
In the new assay, the endogenous TGN is used as the acceptor compartment. Thus, any MPR46HMY receptors present in the TGN at the time of cell breakage must be masked by labeling with non-radioactive PAPS prior to the transport reaction. Cell extracts are incubated with 20 M non-radiolabeled PAPS for 15 min at 37°C in the presence of an ATP-regenerating system. This reaction is carried out in the absence of cytosol to prevent significant transport. Membranes are then collected on a sucrose cushion to reduce the amount of unla-beled PAPS carried over into the labeling reaction and to further wash membranes of cytosolic proteins. Membranes for ϳ500 assays are pooled, aliquoted, and stored at Ϫ80°C.
The transport reaction is carried out by mixing freshly thawed, pretreated membranes, desalted cytosol, 10 mM chlorate, an ATP-regenerating system, and 20 M [ 35 S]PAPS. Reactions are incubated for up to 2 h at 37°C, stopped on ice, and detergent-solubilized. MPR46HMY receptors are collected by binding to nickel agarose, which is washed and then eluted with EDTA. The reaction product is essentially pure, as shown by SDS-PAGE and fluorography (Fig. 2C, lane 3). The purity of the signal permitted direct quantification of the extent of MPR46HMY sulfation by liquid scintillation counting. Fig. 5A shows the general features of the in vitro transport process, monitored by the sulfation of MPR46HMY. Reactions were carried out for 2 h at 37°C. Under these conditions, sulfation was stimulated significantly by cytosol; we routinely observed 5-7-fold stimulation by cytosol addition (Fig. 5, A and  B). No transport was observed at 4°C or in the presence of an ATP depletion system. Moreover, Rab GTPases were required because addition of Rab GDP dissociation inhibitor-␣ (GDI) inhibited transport significantly (42). Thus, Rab proteins are required for efficient transport-coupled sulfation. These control experiments confirmed that the sulfate incorporation observed was not due to MPR46HMY that might have been localized to the TGN at steady state.
To rule out the possibility that we were monitoring intra-Golgi transport, we added primaquine, a reagent shown to completely block intra-Golgi transport without affecting endosome-to-TGN transport (20,43). As shown in Fig. 5A, primaquine had no effect on the extent of MPR46HMY sulfation. In addition, a recombinant dominant negative form of Rab1b (mutation N121I) added at final concentrations of up to 2.7 g/ml had no effect (not shown), confirming that endoplasmic reticulum-to-Golgi and intra-Golgi transport processes were not being scored (44 -46). Together, these data show that we are not monitoring intra-Golgi transport.
In vitro transport was maximal at 0.5 mg/ml cytosol (Fig.  5B). In addition, the reaction was complete after 60 min (Fig.  5C). The slight lag seen between 0 and 5 min may represent the time required for the formation of transport intermediates; alternatively, it may simply be a consequence of the presence of cold PAPS in the lumen of the TGN at the start of the reaction. The reaction process is somewhat faster than that reported previously (20). This difference is likely due to the fact that receptor sialylation is less efficiently detected than receptor tyrosine sulfation.
In summary, these data suggest that MPR46HMY sulfation can be used to measure the extent of transport of this receptor from endosomes to the TGN in vitro. Confirmation of this conclusion comes from our observation that Rab9 participates in this process (see below).
␣-SNAP, Rab9, and NSF Stimulate Endosome-to-TGN Transport in Vitro-When recombinant ␣-SNAP was added to reactions containing limiting amounts of cytosol (Fig. 6A), transport could be stimulated to the same level observed in the presence of saturating cytosol concentrations (Fig. 6B). To verify that this stimulation was not due to a potential enhancement of intra-Golgi transport, we showed that stimulation of transport by ␣-SNAP was unaffected by primaquine (Fig. 6C). These data suggest that ␣-SNAP is a very limiting component under the conditions of this assay and can stimulate endosometo-TGN transport.
We have previously shown that prenylated Rab9 (in complex with GDI) is required for endosome-to-TGN transport, both in vivo and in vitro (19, 23, 34). Here, we tested whether Rab9⅐GDI complexes could stimulate transport in the presence of purified recombinant ␣-SNAP. Fig. 7A shows that in the presence of limiting cytosol, Rab9⅐GDI complexes were only slightly stimulatory at the concentration employed. However, in the presence of 50 or 150 ng/ml ␣-SNAP, the same amount of Rab9⅐GDI was significantly more effective at enhancing transport, suggesting that ␣-SNAP is a limiting component required for the detection of Rab9 function.
The ability of Rab9 to stimulate in vitro transport provides the most stringent confirmation that the transport event we have reconstituted here reflects the physiological process by which MPRs travel between endosomes and the TGN (19). Moreover, both the original and new transport assays show parallel requirements for each of the requisite protein factors identified to date. These findings strongly support the notion that we have reconstituted, endosome-to-TGN transport.
␣-SNAP is thought to attach NSF to membrane-bound SNAP receptors (reviewed in Ref. 3). We therefore tested whether ␣-SNAP and NSF showed synergy in their abilities to stimulate the transport reaction. Limiting cytosol yielded 35% maximal transport. NSF or ␣-SNAP added alone to limiting cytosol raised transport to 52 or 65%, respectively, of that measured with full cytosol. However, when both of these proteins were added in concert, transport was stimulated to ϳ110% of that seen with saturating amounts of cytosol (Fig. 7B).
In summary, these experiments demonstrate that ␣-SNAP can stimulate endosome-to-TGN transport. The stimulation by ␣-SNAP was specific, because it occurred in the presence of primaquine and also increased the stimulation observed in reactions containing Rab9⅐GDI complexes. Furthermore, ␣-SNAP increased stimulation of transport by NSF, supporting a role for NSF in endosome-to-TGN transport. 2

DISCUSSION
We have shown here that ␣-SNAP stimulates endosome-to-TGN transport in vitro. This is a strong indication that the SNARE machinery is also utilized for transport of MPRs from endosomes to the TGN. To reveal this requirement, we used a new in vitro assay that shows significantly greater cytosol dependence than our previous transport assay (20).
The new assay was designed to make use of the TGN localization of tyrosine sulfotransferase and required a cell line expressing MPR46HMY, a cation-dependent MPR modified at its N terminus to contain His-and myc-tags, as well as a site for 2 Y. Nakajima, C. Itin, and S. R. Pfeffer, submitted for publication. FIG. 6. A, Coomassie Blue-stained SDS-polyacrylamide gel electrophoresis of ␣-SNAP, NSF, and Rab9⅐GDI complexes. The bracket indicates the GDI and Rab9; the band above GDI is bovine serum albumin added to stabilize the complexes. Mobilities of size markers are indicated. B, ␣-SNAP stimulates endosome-to-TGN transport. ␣-SNAP was added to reactions containing 0.25 mg/ml cytosol (filled squares). Reactions were carried out at 37°C for 2 h as described in Fig. 4A. Cytosoldependent transport was obtained by subtracting transport measured in the absence of cytosol. 100% transport was defined as the transport obtained with 1 mg/ml cytosol. Shown is the average of two experiments carried out in duplicate. C, reactions were in the presence of 200 M primaquine (1470 cpm maximum).
FIG. 7. ␣-SNAP increases stimulation by Rab9⅐GDI complexes and NSF. Reactions (2 h) contained 0.25 mg/ml cytosol. Cytosoldependent transport was obtained by subtracting transport measured in the absence of cytosol. 100% transport was defined as the transport obtained with 1 mg/ml cytosol. Maximal transport was 2000 or 1150 cpm in A and B, respectively. tyrosine sulfation. We showed that MPR46HMY rapidly acquires endo consistent with its failure to be mislocalized to lysosomes for degradation. Moreover, the protein traversed from the cell surface to the TGN with the same kinetics as the cation-independent MPR in CHO cells. These features strongly support the conclusion that we have not altered the trafficking signals within MPR46. In addition, the fact that MPR46HMY was a substrate for tyrosine sulfotransferase showed that the consensus polypeptide derived from the cholecystokinin precursor can function when introduced near the N terminus of MPR46.
Several lines of evidence confirm that we have reconstituted the transport of MPR46 from endosomes to the TGN. Most importantly, the reaction is stimulated by the Rab9 GTPase and is completely unaffected by primaquine. These characteristics are hallmarks of MPR recycling (20,23,34) and strongly rule out the possibility that we are measuring intra-Golgi transport or endoplasmic reticulum-to-Golgi transport. In further support of a vesicular transfer, the reaction is stimulated by ␣-SNAP and NSF and is inhibited by GDI, as would be expected of a Rab-dependent process.
At steady state in NRK cells and CHO cells, greater than 90% of 300-kDa MPRs reside in late endosomes (51). 3 Colocalization of MPR46 and the 300-kDa MPR has failed to reveal significant differences in the steady state distribution of these proteins between cellular compartments (cf. Ref. 52). Thus, it seems likely that the protein we follow begins transport in the late endosomal compartment. In addition, any protein that might have been present in the TGN at the beginning of the reaction will be masked by the cold pre-labeling protocol employed here.
Can we be certain that tyrosine sulfotransferase marks arrival in the TGN and not another cellular compartment? This enzyme is localized to the TGN in PC12 cells (36), MDCK cells (25), and IgM-secreting B cells (40). In one study by Spiess and co-workers (25), an overexpressed substrate appeared to acquire tyrosine sulfate before galactose and sialic acid when it was slowed in its export from the TGN by 20°C incubation of COS-7 cells but not MDCK cells. Although we have not directly immunolocalized tyrosine sulfotransferase in CHO cells, it seems most likely that the enzyme resides in the TGN because MPRs recycle to the TGN but not compartments containing galactosyltransferase in CHO cells (53). Moreover, Rab9 facilitates transport of cation-independent MPRs from late endosomes to the TGN in living cells and in vitro (19,34), the process reconstituted here.
Multiple features of the new in vitro transport assay can account for its increased cytosol dependence. In the new assay, MPR46HMY-expressing CHO wild type cells provide both donor and acceptor compartments. In contrast, the previous complementation scheme required addition of rat liver Golgi membranes (20) which do contain significant concentrations of peripheral membrane proteins (1,4). Furthermore, in the new assay, membranes are washed by centrifugation through a sucrose cushion, which removes cytosolic and, perhaps also, some peripherally associated membrane proteins. These differences readily explain the increased level of stimulation seen with added cytosolic proteins.
The new assay seemed especially limiting for ␣-SNAP. Wickner and colleagues (47) have shown that ␣-SNAP dissociates from yeast vacuolar membranes in the presence of ATP with a half-time of less than 10 min. This condition is very similar to the non-radioactive PAPS preincubation step used in our new assay, which involves a 15-min incubation with an ATP-regen-erating system. It is likely that the preincubation step has revealed a role for ␣-SNAP in endosome-to-TGN transport by depleting the protein from the membranes.
We have previously described an N-ethylmaleimide-sensitive protein required for MPR transport that differed from NSF (22). In contrast to NSF, this factor acted early in transport, presumably in the formation of transport vesicles. The fact that this protein was more sensitive to N-ethylmaleimide than NSF made it difficult for us to determine whether NSF was also required. Very recently, we have used mutant NSF proteins to demonstrate a role for NSF in endosome-to-TGN transport using our previous assay. 2 The enhanced cytosol dependence of the new assay also permitted us to confirm a role for NSF in endosome-to-TGN transport, without needing to inactivate endogenous NSF by treatment of membranes with N-ethylmaleimide.
The discovery of a role for ␣-SNAP in endosome-to-TGN transport will now allow us to investigate SNARE complex formation for this transport process. Our working hypothesis is that the Rab9 GTPase, present on transport vesicles in its active, GTP-bound conformation, recruits docking factors onto vesicles to direct their association with the TGN. We have proposed that this docking process leads to deprotection of the SNARE proteins, permitting SNARE pairing and subsequent fusion (48). The new in vitro transport assay will greatly facilitate the identification of novel docking factors, helping to expand our understanding of the link between Rab GTPases and the docking and fusion machinery.