Sequential Involvement of p115, SNAREs, and Rab Proteins in Intra-Golgi Protein Transport*

Delivery of transport vesicles to their receptor compartment involves tethering, priming, and fusion. Soluble NSF attachment protein-α (αSNAP) mediates the disruption of SNAREs by N-ethylmaleimide sensitive factor (NSF) and was employed to determine the hierarchy of proteins responsible for intra-Golgi protein transport. The N-terminal 23 amino acids of αSNAP are necessary for SNARE binding. The antibody 2F10 recognizes this SNARE interaction domain of αSNAP and inhibits intra-Golgi protein transport reversibly. This antibody was applied to modify the transport assay to determine the protein requirements relative to the action of αSNAP and NSF. We found that 1) p115 acts independently of αSNAP and NSF, 2) SNAREs are required after tethering and interact selectively after activation by αSNAP and NSF, and 3) Rab proteins act after SNARE activation and before fusion.

Vesicles mediate the transport between membrane-bound compartments of eukaryotic cells (1). To maintain the functional and morphological integrity of the cell, protein and lipid transport has to follow a selective and precise mechanism. After entry of cargo into the budding vesicle, it pinches off and is destined for the target membrane. Short range movement of vesicles is mediated by diffusion, whereas motor proteins move vesicles over a long distance on cytoskeletal tracks. Finally, vesicles dock to and fuse with the target membrane, resulting in the delivery of the cargo. Vesicular transport is a cyclic process accounting for the vectorial flux of cargo while maintaining the structural and functional integrity of the individual compartments (for reviews see Refs. [2][3][4][5]. Docking is the morphologically defined event of the association of a vesicle with a target membrane. Tethering is the biochemical equivalent and involves the formation of a stable interaction of a vesicle with a membrane. Priming refers to the activation of the membrane surfaces to obtain fusion competence (6). The SNARE 1 hypothesis (7) provides an explanation for specificity in vesicular transport; during targeting, the membrane-bound v-SNARE protein of a vesicle interacts with the corresponding t-SNARE of the target membrane forming a SNARE complex. SNAREs were initially identified in the neuronal system (7), and subsequently homologous proteins involved in other transport steps in the mammalian cell and in yeast were identified. SNAREs of the synapse and of the yeast vacuole are involved in docking and fusion of proteoliposomes and vacuolar membranes, respectively (8 -10). Several soluble proteins regulate the v-t-SNARE interaction. Members of the SNAP family of proteins, consisting of the ubiquitously expressed ␣and ␥SNAP and the neuronal ␤SNAP, bind to SNAREs (11,12). Subsequently, the ATPase NSF binds to SNAREs via the SNAPs (13), and ATP hydrolysis results in the disruption of SNARE complexes (7). Another family of proteins involved in vesicular transport is the Rab family of GTP-binding proteins (14). They cycle between a GTP-and a GDP-bound form, and both forms can be membrane-associated (15). The GDP dissociation inhibitor (GDI) forms a complex with Rabs in their GDP-bound form and removes them from the membrane (16 -18). GDP is exchanged for GTP by a guanine nucleotide exchange factor (19,20).
Although many proteins involved in the targeting of vesicles have been identified, no consensus exists about the order of involvement of these proteins (42,43). This study focused on the interaction of Golgi-derived transport vesicles with their target membrane in intra-Golgi transport. GOS28 was chosen as a representative of the SNARE family, and Rab6 was chosen as a representative of the Rab family. Both proteins and p115 are examined for their contribution in intra-Golgi transport relative to ␣SNAP and NSF.

EXPERIMENTAL PROCEDURES
Recombinant Material-Bovine His 6 -␣SNAP was purified from Escherichia coli lysate according to Ref. 12, and GST-VAMP was purified as described in Ref. 44. The truncated forms of GST␣SNAP designated GST␣SNAP-(1-156), GST␣SNAP- (24 -295), and GST␣SNAP-  were obtained by amplification of the corresponding DNA fragment of ␣SNAP with oligonucleotides introducing a BamHI site at the 5Ј end and an EcoRI site at the 3Ј end followed by ligation into the E. coli expression plasmid pGEX4T-1 (Amersham Pharmacia Biotech). GST␣SNAP and its deletion mutants were purified as described in Ref. 21, and His 6 -NSFmyc was purified according to Ref. 7. To bacterially express the head domain of p115 from rat (45), a DNA fragment coding for amino acids 1-651 was amplified by polymerase chain reaction introducing a BamHI site at the 5Ј end and an EcoRI site at the 3Ј end. The EcoRI site in the open reading frame was removed without affecting the translated amino acids. This fragment was ligated into pGEX4T-1. The p115 head domain was expressed as a fusion protein with GST and was purified by thrombin cleavage according to the manufacturer's instruction. The cytoplasmic domain of His 6 -GOS28 was expressed in E. coli (21). Insoluble protein (inclusion bodies) was dissolved in buffer A (8 M urea, 100 mM sodium phosphate, pH 8.0) and cleared by centrifugation. Soluble protein was bound to a nickel-nitrilotriacetic acid column (Qiagen) equilibrated with buffer A. After washing with 50 column volumes of buffer A, a linear gradient against buffer B (100 mM KCl, 10% (w/v) glycerol, 1 mM 2-mercaptoethanol, 20 mM imidazole, 50 mM Tris-HCl, pH 8.0) was applied. Refolded protein was eluted with a linear gradient of buffer B containing 500 mM imidazole. Protein concentration was determined by the method of Bradford (46).
Subcellular Fractions and Intra-Golgi Transport Assay-Extract from bovine brain membranes was prepared as described in Ref. 7, and GDI was purified as described in Ref. 51. 1 M KCl-treated (K-Golgi) membranes were prepared according to Ref. 52. Intra-Golgi protein transport assays were carried out at 37°C for 60 min as described (53,54). Reactions containing K-Golgi membranes were carried out in a volume of 100 l. For the two-stage assay, transport reactions were incubated with 500 ng of 2F10 on ice for 20 min and transferred to 37°C for 20 min. The reactions were placed on ice, and the cytoplasmic domains of GOS28, GDI, and anti-Rab6 were added, respectively. After the addition of 500 ng of His 6 -␣SNAP, the samples were incubated at 37°C for 1 h. The two-stage assays containing K-Golgi membranes were inhibited with 500 ng of 2F10 as described above, and anti-p115 antibody and 400 ng of His 6 ␣SNAP were added. Error bars indicate standard deviations of experiments carried out as triplicates (see Fig. 4) or as duplicates (see Figs. 5 and 6).

RESULTS
Domain Analysis of ␣SNAP-␣SNAP is required for the disassembly of SNARE complexes by NSF. This dissociation is a key event in vesicular protein transport; therefore, we chose ␣SNAP as a target to examine its role and, in conclusion, that of NSF and SNAREs in protein targeting. To determine the domain structure of ␣SNAP we carried out limited proteolysis with subtilisin, a protease of low sequence specificity. Protein domains of compact and rigid conformation are more resistant to proteolytic degradation by subtilisin, whereas flexible parts of a protein become preferentially hydrolyzed (55). To identify core domains of ␣SNAP we incubated His 6 -␣SNAP with increasing amounts of subtilisin. The proteolytic fragments were separated by SDS-PAGE. Fig. 1A shows a typical result of limited proteolysis of ␣SNAP. After transfer to a polyvinylidene fluoride membrane, partial N-terminal amino acid sequencing was performed on the four proteolytic intermediates indicated in Fig. 1A. The largest of the four intermediates starts at position 32 with the smaller fragments beginning at positions 93, 140, and 157 of ␣SNAP, respectively (Fig. 1B). A comparison of the apparent molecular weight of the proteolytic fragments with the calculated molecular weight suggests that the C terminus was not subject to proteolytic degradation (data not shown). The hydrophilic amino acid serine found in positions 24 and 35 flanks the site of the first proteolytic cut at position 32. Truncated GST␣SNAP fusion proteins were expressed in E. coli that lacked either the first 23 or 31 amino acids of ␣SNAP. In addition, amino acids 1-156 of ␣SNAP were fused to GST and expressed in E. coli.
The deletion mutants of ␣SNAP were examined for their ability to bind to neuronal SNARE complexes. SNARE complexes were prepared from bovine brain membranes by lysis with Triton X-100. ␣SNAP and its truncation mutants were incubated with the membrane extract and purified with glutathione-agarose beads. Neuronal SNAREs were detected by SDS-PAGE and transfer to a nitrocellulose membrane followed by Western analysis. Compared with full-length ␣SNAP, removal of 23 amino acids from the N terminus results in a decreased affinity for SNAREs, and removal of 34 amino acids renders SNARE binding undetectable. However, amino acids 1-23 of ␣SNAP are not sufficient for SNARE binding because a mutant ␣SNAP consisting of amino acids 1-156 does not bind SNAREs ( Fig. 2A).
The anti-␣SNAP antibody 2F10 interferes with the binding of SNAREs to ␣SNAP. 2 Considering the critical role of the N terminus of ␣SNAP in SNARE binding, we applied the deletion mutants of ␣SNAP in a Western analysis to determine the epitope of 2F10. As depicted in Fig. 2B, the antibody 2F10 recognized full-length ␣SNAP but none of the N-terminal truncation mutants. This coincidence suggests that the N terminus of ␣SNAP is directly involved in binding of neuronal SNAREs. The critical role of the N terminus of ␣SNAP in binding SNAREs is also supported by studies of the interaction of recombinant SNAREs with ␣SNAP (56) and electrophysiological experiments. In this study, a peptide derived from the N terminus of ␣SNAP inhibited calcium-induced exocytosis after microinjection into the giant axon of squid (57).
Development of a Two-stage Intra-Golgi Transport Assay-An in vitro transport assay that reconstituted vesicular intra-Golgi transport was developed by Rothman and co-workers (54,58). Subsequent characterization of the transport reaction resulted in a consensus that this assay allows measurement of the fusion of Golgi-derived transport vesicles with early Golgi elements (59,60), establishing this assay as a tool to identify and characterize components of the transport machinery (61). The antibody 2F10 inhibits the intra-Golgi transport reaction. 2 This inhibition is caused by a depletion of ␣SNAP from both cytosol and membranes. The ␣SNAP-2F10 complex is predicted to be soluble and to be minimally invasive during the transport assay. In the first step we determined the time span for which the transport assay is sensitive for 2F10. Standard transport reactions were started, and 2F10 was added after discrete time intervals. As shown in Fig. 3A, the transport signal becomes independent of ␣SNAP within 20 min. The remaining incubation time of 40 min at 37°C is required for the glycosylation of the vesicular stomatitis virusencoded glycoprotein. In the second step we tested whether the transport inhibition caused by 2F10 is reversible. Transport was first inhibited by 2F10 for 20 min, and then ␣SNAP was added. As depicted in Fig. 3B, the inhibition by 2F10 is reversible. The reversibility of the inhibition by 2F10 made it possible to divide the intra-Golgi transport reaction into two steps, enabling us to resolve requirements before and after ␣SNAP. In the first step, the reaction is inhibited by 2F10, allowing transport until the first ␣SNAP-dependent step is reached. Cells treated with N-ethylmaleimide accumulate docked vesicles at the Golgi apparatus (62). ␣SNAP and NSF were identi-fied as protein targets of N-ethylmaleimide (63), implying that the transport reaction is inhibited after the accumulation of docked transport intermediates. In the second step, this inhibition is reversed by the addition of recombinant ␣SNAP. At this point, substrates can be tested for their potential to inhibit the two-stage transport reaction, and the quality and quantity of inhibition can be compared with a standard one-stage transport reaction (Fig. 3C).
The Contribution of p115, GOS28, and Rab Proteins to Intra- H]UDP-GlcNAc, and an ATP-regenerating system in a total volume of 50 l (54) were incubated at 37°C for 1 h. 500 ng of 2F10 were added after the indicated time intervals. Then the membranes were lysed by detergent, and the vesicular stomatitis virus-encoded glycoprotein was immunoprecipitated. B, reversibility of the inhibition by the antibody 2F10. An intra-Golgi transport reaction was carried out for 1 h at 37°C (lane 1). The same reaction was incubated with 500 ng of 2F10 on ice for 20 min and then for 1 h at 37°C (lane 2). Another transport reaction was incubated with 500 ng of 2F10 for 20 min on ice and subsequently at 37°C for 20 min. The reaction was transferred back on ice, 500 ng of ␣SNAP were added, and the incubation at 37°C was continued for 1 h (lane 3). C, model for the two-stage intra-Golgi transport reaction. The antibody 2F10 depletes the transport assay of ␣SNAP, resulting in a block of the assay at a step preceding the fusion of transport vesicles with the acceptor membrane. The reaction proceeds after the addition of ␣SNAP, and protein requirements relative to ␣SNAP and its interacting proteins NSF and SNAREs can be distinguished.
Golgi Protein Transport-The proteins p115, GOS28, and Rab6 have been implicated in intra-Golgi protein transport (for references see the Introduction), and we studied their role in the one-and two-stage transport assays. A polyclonal antiserum against the globular head domain of p115 was generated. Affinity-purified antibodies recognized a single band of 115 kDa of Chinese hamster ovary whole cell lysate in Western analysis (data not shown). We added increasing amounts of these antibodies to the transport reactions between 1 M KCl-treated Golgi membranes. As shown in Fig. 4 (top panel), only the one-stage transport reaction but not the two-stage reaction was inhibited by the anti-p115 antibody. The inhibition reached a saturation at 400 ng of antibody/100-l reaction that could be reversed by the addition of antigen (data not shown). Initially, the antibodies were tested in the standard intra-Golgi transport assay and did not inhibit protein transport (data not shown). Subsequently, K-Golgi membranes were used in the assay. Salt treatment removes p115 quantitatively from Golgi membranes, and cytosol becomes the only source of p115 (38). 100 ng of NSF had to be added to the transport reaction in these experiments because the 1 M KCl wash removes NSF likewise. The degree of GTP␥S inhibition is an indicator of to what extent vesicular transport as compared with uncoupled fusion contributes to the transport reaction (64). In our assay design, GTP␥S inhibited the reaction by 60% (Fig. 4, bottom panel), suggesting that most of the transport signal is due to vesicular transport intermediates.
We then examined the inhibition pattern of the cytoplasmic domain of GOS28 in the one-and two-stage transport assays.
The cytoplasmic domain of GOS28 containing a hexahistidine motif at the N terminus was expressed in E. coli. The protein was purified from inclusion bodies after solubilization in urea. The protein was bound to a nickel-nitrilotriacetic acid matrix, renatured by applying a gradient of a decreasing concentration of urea, and eluted by a gradient of increasing imidazole concentration. In SDS-PAGE a single band of 26 kDa was observed (data not shown). The renatured cytoplasmic domain of GOS28 was added to the one-and two-stage transport assay using standard membranes. In both cases, a similar degree of 70% inhibition was observed. In the two-stage reaction, a 2.5-fold smaller amount of GOS28 was sufficient to achieve inhibition (Fig. 5). Two control reactions for the specificity of GOS28 were carried out. 1) The effect of GOS28 can be neutralized by incubation with antibodies directed against GOS28 (data not shown); 2) the cytoplasmic domain of the plasma membrane SNARE synaptobrevin/VAMP was applied. Soluble recombinant synaptobrevin/VAMP was confirmed to be active by virtue of the fact that it forms a complex with the cytoplasmic domain of syntaxin 1a and SNAP25 that can be disassembled by ␣SNAP and NSF as described in Ref. 65. No inhibition by synaptobrevin/VAMP was observed in the transport reaction (Fig. 5).
The contribution of Rab proteins to the intra-Golgi transport was determined by adding GDI or antibodies against Rab6 (3A6) to the one-and two-stage transport assays. GDI depletes the transport reaction of all Rab proteins. Both GDI and anti-Rab6 inhibit the one-and two-stage reactions in a saturable manner by 50% (Fig. 6A and data not shown). Because GDI inactivates Rab proteins, the specific effect of both reagents was determined by simultaneous addition. The combination of GDI and anti-Rab6 exhibits a comparable inhibition that does not differ in a statistically significant manner when applied separately (Fig. 6B), suggesting that both proteins act specifically.
Our results define the biochemical sequence of proteins involved in the targeting of vesicles during transport through the Golgi apparatus. The addition of recombinant ␣SNAP to an intra-Golgi transport reaction inhibited by the anti-␣SNAP antibody 2F10 results in a complete reversal of the inhibition. This observation enabled us to modify the intra-Golgi transport assay to determine the contribution and temporal action of p115, GOS28, and Rab6, three proteins known to be involved in intra-Golgi transport, relative to ␣SNAP and its interaction partner NSF. The data presented can be most easily explained by p115-mediated binding of vesicles to the target membrane followed by the activation of SNAREs on the vesicle and the target membrane, respectively. Rab proteins, in particular Rab6, are required for a step between SNARE activation and fusion.

DISCUSSION
The SNARE hypothesis predicted that SNAP and NSF bind to SNAREs paired in trans as the result of vesicle docking (7). This prediction placed SNAP and NSF at a defined step of the targeting mechanism. Meanwhile, it was shown for the fusion of yeast vacuoles that only t-SNAREs require priming (66) and that priming can take place without tethering (10). In this study, we examined protein transport through the Golgi apparatus, the central organelle of constitutive secretion. In interpreting our data and the data obtained in the yeast system, we propose that the function of ␣SNAP and NSF is to continuously activate the SNAREs on all membranes involved in constitutive transport. This activation would keep the machinery for the recognition and fusion of membranes continuously active for the consumption of newly delivered vesicles to ensure a high transport rate. As a consequence, no sequential order of priming relative to tethering or fusion can be assigned in constitutive secretion in vivo. In regulated secretion, the cycle of generation and consumption of vesicles becomes arrested and allows one to resolve mechanistic intermediates. The model systems studied suggest a role for ␣SNAP and NSF before and after vesicle fusion; in neuroendocrine cells, a readily releas-able pool of vesicles is docked to the plasma membrane and fuses after the influx of calcium ions from the extracellular medium (67,68). SNAP and NSF are required for a priming step that precedes the calcium-triggered fusion (69). This finding was confirmed in the temperature-sensitive Drosophila NSF mutant named comatose (6). Two lines of evidence suggest an additional role for SNAP and NSF after fusion; the temperature-sensitive phenotype of comatose is reversible, and the recovery kinetic corresponds to the kinetic of the Drosophila dynamin mutant shibire known to be required for the recycling of vesicles (70). In comatose mutants, v-t-SNARE complexes accumulate on synaptic vesicles after a temperature shift (71). This suggests that SNAP and NSF are required to separate vand t-SNAREs after the fusion to ensure that v-SNAREs enrich on recycling vesicles and t-SNAREs remain at the target membrane.
A promiscuous interaction of bacterially expressed SNAREs involved in unrelated transport steps has been observed (72,73). We compared the effect of the cytoplasmic domain of GOS28 and the neuronal SNARE synaptobrevin/VAMP in the standard intra-Golgi transport reaction. The neuronal SNAREs synaptobrevin/VAMP, syntaxin 1a, and SNAP25 are, to our knowledge, the only SNAREs that, after expression in E. coli, have been shown to form a complex that can be disassembled with ␣SNAP and NSF (65). Therefore, the v-SNARE synaptobrevin/VAMP was the protein of choice to examine the specificity of SNARE interactions in the intra-Golgi transport assay. The cytoplasmic domain of GOS28 inhibited the transport reaction in a dose-dependent manner, confirming previous studies that showed that an antibody against GOS28 inhibits intra-Golgi transport (21,22). No inhibition by the cytoplasmic domain of synaptobrevin/VAMP was observed. Our data suggest that SNAREs can interact in a specific manner, and we conclude that the structural organization of the cell is sufficiently preserved in the cell-free intra-Golgi transport assay to retain the key mechanisms in specific vesicular targeting. Compared with the one-stage reaction, the same inhibition results from a lower concentration of the cytoplasmic domain of GOS28 in the two-stage reaction. A kinetic and a steric effect can account for the observed difference. ␣SNAP binds to unpaired t-SNAREs (74), and a depletion by the antibody 2F10 gives the cytoplasmic domain of GOS28 a better access to syntaxin 5; alternatively, the simultaneous addition of ␣SNAP and the cytoplasmic domain of GOS28 in the two-stage reaction results in a higher association rate for the cytoplasmic domain of GOS28.
p115 is involved in transcytosis in hepatocytes and is found on vesicles (45). Independently, it was isolated from cytosol in an assay measuring the transport of proteins between high salt-treated Golgi membranes (38,39). High salt treatment of membranes removes peripheral membrane proteins, including coatomer, resulting in an increased amount of uncoupled homotypic fusion (64). The antibody against p115 inhibited intra-Golgi transport only after removal of p115 by high salt treatment of the membranes. An inhibition was observed in the one-stage assay but not in the two-stage assay. This result suggests that p115 is recruited to the membrane in an ␣SNAPand NSF-independent manner and that the p115 N-terminal head domain does not become accessible to antibody during the transport reaction. Ultrastructural analysis of Golgi complexes reveals that vesicles are tethered to Golgi stacks by fibrous structures, unable to diffuse freely (75). p115 mediates binding of vesicles to Golgi membranes (41) and is a candidate for such a string protein. The partial inhibition of the intra-Golgi transport reaction caused by the anti-p115 antibody is consistent with the functional redundancy for interaction of vesicles and target membrane observed in a yeast strain harboring mutated p115. This conditional mutant yeast strain expresses a truncated form of Uso1p, the yeast homologue of p115 (39,45). The mutant Uso1p lacks the vesicle-binding domain (40,76), and the thermosensitive phenotype can be suppressed by the overexpression of SNAREs (77), suggesting that SNAREs can bypass the function of Uso1p and mediate the docking of vesicles.
GDI depletes membranes of Rab proteins resulting in a partial inhibition of the one-and two-stage transport assays. An involvement of Rab6 in intra-Golgi transport has been demonstrated (35,36), and the anti-Rab6 antibody showed an inhibition pattern comparable with GDI. This result suggests that Rab proteins, in particular Rab6, act after or parallel to ␣SNAP and NSF. The GTPase Rab5 is involved in the fusion of early endosomes (78), and several proteins interacting with its GTPbound form have been identified. One of them is the protein EEA1 (79). Early endosome fusion can be inhibited by GDI and can be stimulated by EEA1 (80). An increased concentration of EEA1 can suppress the inhibition by GDI. The partial inhibition of the intra-Golgi transport assay can be due to an EEA1like activity that bypasses the Rab6 requirement. A dual requirement for Rab proteins has been resolved for the fusion of yeast vacuoles. Antibodies against the ␣SNAP homologue Sec17p and the NSF homologue Sec18p have been compared with GDI and an antibody against the Rab homologue Ypt7p in a kinetic analysis. Under the assay conditions, GDI and Ypt7p act downstream of Sec17p and Sec18p (81). However, when the concentration of Sec18p present in the assay is increased 50fold over the amount necessary for saturation (82), the sensitivity of the fusion reaction toward GDI and anti-Ypt7p changes, and a contribution of Ypt7p at a step preceding Sec17p and Sec18p could be demonstrated (83,84). Biochemical and genetic studies of the endoplasmic reticulum-Golgi transport in yeast corroborate these observations. A genetic analysis of proteins involved in endoplasmic reticulum-Golgi transport revealed an interaction between the Rab homologue YPT1 and the p115 homolgue USO1 (77). In addition, high amounts of GDI remove Ypt1p and Uso1p from membranes (85). On the other hand, transport vesicles containing mutated Ypt1p remain fusion-competent, whereas acceptor membranes carrying the mutant of Ypt1p are fusion-incompetent (86). Two roles for Rab proteins and their associated proteins emerge in the yeast system. One precedes the requirement for Sec17p and Sec18p and catalyzes the tethering of vesicles to membranes.
Another not yet defined function of Rab proteins is accomplished after the ATP hydrolysis by Sec18p. In the case of the mammalian Golgi apparatus, vesicles are tethered to the cisternae by vesicle-bound giantin linked via p115 to cisternaebound GM130. Giantin and GM130 are members of the golgin protein family. Golgins share a Rab6-binding domain that may be required for targeting to the cisternae (87).