Purification and Characterization of SARlp, a Small GTP-binding Protein Required for Transport Vesicle Formation from the Endoplasmic Reticulum*

SEC12 encodes an integral membrane glycoprotein essential for vesicle formation from the endoplasmic reticulum (ER) in yeast. The SARl gene was discov- ered as a multicopy suppressor of a secl2“ strain and encodes a 21-kDa GTP-binding protein also required for protein transport from the ER to the Golgi apparatus (Nakano, A., and Muramatsu, M. (1989) J. Cell Biol. 109, 2677-2691). We have purified Sarlp to apparent homogeneity from cells harboring a galac-tose-regulated recombinant SARl. Purified Sarlp binds guanine nucleotides specifically and exhibits GTPase activity (0.001 min”). Nucleotide exchange and hydrolysis rates are greatly increased in the pres- ence of Mg2+ and nonionic detergents or phospholipids. An assay that the formation of a vesicle intermediate in ER to Golgi transport was devised that is dependent on the addition of purified Sarlp. assay employs from wild-type cells and cytosol of of Secl2p chroma-tography. The the of Sarlp and GTP to support vesicle budding. Sarlp prebound

have implicated SECl2,13,16,23, and SARl in vesicle formation from the ER (Novick et al., 1980;Kaiser and Schekman, 1990;Nakano and Muramatsu, 1989). Cell-free assays in yeast have been developed that reconstitute vesicle formation from the ER and protein transport to the Golgi apparatus in cytosol and ATP-dependent reactions (Baker et al., 1988;Ruohola et al., 1988;Groesch et al. 1990;Rexach and Schekman, 1991). The cell-free assay has corroborated genetic observations, implicating SeclBp, Sec23p, and Sarlp function in vesicle formation (Rexach and Schekman, 1991;d'Enfert et al., 1991a). An additional protein, p105 (Sec24p), that copurifies with functional Sec23p, is also essential for vesicle formation in vitro (Hicke et al., 1992). As a step toward our goal of reconstituting vesicle formation with ER membranes and purified cytosolic components, we report the purification and characterization of Sarlp, a small GTPase required for vesicle formation from the ER.
SARl was discovered as a multicopy suppressor of a temperature-sensitive secl2 strain and encodes a protein with sequence homology to the ras superfamily of small GTPases (Nakano and Muramatsu, 1989). Sarlp is a novel member of this family with highest amino acid identity to yeast ARFl (35% identity over 168 residues; Nakano and Muramatsu, 1989). The temperature-sensitive transport defect of secl2, which is reproduced in the in vitro transport assay (Rexach and Schekman, 1991), can be repaired at the restrictive temperature by adding excess Sarlp (Oka et al., 1991). Sarlp has been proposed to function in the vesicle formation step of ER to Golgi transport (d'Enfert et al., 1991a). Several other small GTPases have been identified in the yeast secretory pathway including Yptlp, Arflp, and Sec4p (Segev et al., 1988;Stearns et al., 1990;Salminen and Novick, 1987). Although the importance of small GTPases in the secretory pathway is well appreciated, a mechanism for GTPase function in vesicle formation or targeting is obscure. Thus, in vitro reconstitution of ER vesicle formation with purified and characterized components will allow us to examine one event that is likely to be a paradigm for organellar budding.
nents were obtained from Difco, and galactose was purchased from Sarlp Purification-Two liters of strain YPH5OO harboring plasmid pANY2-18 (GALI-SARI) was grown at 30 "C to late log phase in yeast nitrogen base supplemented with appropriate amino acids and 2% raffinose. Cells were harvested at room temperature and resuspended in 4 liters of 1% yeast extract, 2% peptone, and 2% galactose to induce expression from the GAL1 promoter. After growth at 30 "C for 14 h, cells (approximately 20,000 ODSOOnm) were harvested and washed once with ice-cold buffer 88 (0.25 M sorbitol, 20 mM K-HEPES, 1 mM Mg(OAc)z, and 150 mM KOAc, pH 6.8). The following steps of the purification were performed at 4 "C. The cell pellet was resuspended in 60 ml of buffer 88 supplemented with 1 mM phenylmethylsulfonyl fluoride and 0.5 mM dithiothreitol and lysed in a Bead-beater chamber (Biospec Products, Bartlesville, OK) containing volume of glass beads, by six 1-min periods of agitation. The lysate was decanted, and the glass beads were washed with 20 ml of buffer 88 to increase recovery. A combined crude extract was centrifuged at 20,000 X g for 20 min, and the resulting supernatant fluid was clarified by centrifugation at 100,000 X g for 60 min in a Beckman Ti45 rotor. Typically, 60 ml of this high speed supernatant material was recovered and divided into three 20-ml aliquots, quick frozen in liquid Nz, and stored at -70 "C. Each of the 20-ml aliquots was thawed on ice and loaded onto a 600-ml, 4.5 X 55-cm Sephacryl S-100 column (Pharmacia LKB Biotechnology Inc.) equilibrated with buffer 88. Protein was eluted with buffer 88 at a flow rate of 1 ml/min, and 8-ml fractions were collected. Sarlp, as determined by a Sarlp-dependent vesicle formation assay (described below) and specific immunoreactivity, eluted as a broad peak at a fraction corresponding to 7 kDa and was clearly resolved from the majority of protein (d'Enfert et at.,199fa). A pool of the peak Sarlp-con~ining fractions (45 ml) was adjusted to a final concentration of 0.02% Triton X-100 and frozen in liquid NP for storage at -70 "C. Three 5-100 pools were combined and diluted 4-fold with 0.02% Triton X-100 to reduce the KOAc concentration to 37.5 mM and loaded onto an 8-ml, 1 X 5-cm DEAE-Sepharose CL-4B (Pharmacia) column equilibrated with 20 mM K-HEPES (pH 6.8), 20 mM KOAc, I mM Mg(OAc)Z, and 0.02% Triton X-100. The DEAE column, operated at a flow rate of 0.5 ml/min, was washed with 50 ml of equilibration buffer and then 100 ml of equilibration buffer with 90 mM KOAc and 0.005% Triton X-100. Sarlp was eluted with 20 mM K-HEPES (pH 6.8), 200 mM KOAc, 1 mM Mg(OAc)*, and 0.001% Triton X-100. One-milliliter fractions were collected, and the peak six fractions, as judged by silver stain, were pooled, distributed into aliquots, and frozen in liquid nitrogen for storage at -70 "C. Sarlp was active for guanine nucleotide binding and vesicle formation after several months of storage at -70 "C.
Storage of protein at 4 "C resulted in a 50% loss of vesicle budding activity after 1 week. Protein could be thawed and refrozen once without detectable loss of nucleotide binding or vesicle budding activity.
Guanine Nucleotide-binding Assays-Nucleotide binding to Sarlp was quantified by a filter binding method (Northup et al., 1982). The binding reactions contained 0.1% Triton X-100, 0.5 mM dithiothreito], 0.25 mg/ml bovine serum albumin, 1.0 mM Mg(OAc)?, 25 mM K-HEPES (pH 6.8), 1-2 p~ guanine nucleotide (10,000 dpm/pmol), and 1-10 pmol of Sarlp in a total volume of 40 pl. Reactions were mixed on ice and then incubated at 29 "C for the times indicated in the figure legends. Reactions were stopped by dilution with 800 p l of icecold wash buffer (0.002% Triton X-100, 25 mM K-HEPES (pH 6.8), 100 mM NaCl, and 5 mM MgCIz), rapidly filtered onto nitrocellulose membranes (type HA 0.45 pm, Millipore), and washed four times with 2 ml of wash buffer; washes were completed within 20 S. The counts/min bound were determined after filters were dissolved in scintillation fluid (Universol, Heckman-Spinco). Counting efficiency was determined by adding known amounts of labeled nucleotide to mock treated filters. Background values were determined under various conditions by excluding Sarlp from the incubation mixture and subtracted from the values of mixtures containing Sarlp to calculate Sarlp-dependent binding. Background binding comprised less than 5% of Sarlp-dependent binding.
GTPase Assay-Conditions for GTP hydrolysis are identical to binding conditions except that the specific activity of [L~-~*P]GTP was 20,000 dpm/pmol and hydrolysis assays were performed in 10-pl volumes. Hydrolysis reactions were stopped by the addition of EDTA to a final concentration of 20 mM and placed on ice. One-microliter aliquots were spotted on polyethyleneimine cellulose plates (Sigma) and developed in 1 M LiCl, 1 M HCOOH (Wagner et al., 1987). The plate was dried and conversion to GDP over time was quantified after vale, CAI. exposure to a Phosphorimager plate (Molecular Dynamics, Sunny-Sarlp Assay in Microsome-based Vesicle Formation Reaction-Wild-type microsomes and cytosol were prepared form strain RSY607 (Wuestehube and Schekman, 1992) while a Sarlp-depleted cytosol was prepared from the Secl2p overproducer strain RSY658 (d'Enfert et at., 1991bl. The two-stage vesicle formation assay was performed as previously described (d'Enfert et a [., 1991a). Briefly, in the stage I reaction at 10 "C, 35S-prepro-a-factor is translocated into microsomes where three N-linked core-carbohydrate chains are attached yielding core-glycosylated pro-a-factor. Microsomes are then washed to remove untranslocated label and resuspended under various reaction conditions described in the figure legends and incubated at 29 "C in a total volume of 50 pl. Reactions are stopped by placing tubes on ice for 5 min then centrifuged at 12,000 X g at 4 "C for 5 min. Vesicles formed from microsomes remain in the supernatant fraction after centrifugation. The percent vesicle formation is quantified by Con A precipitation of protease-protected, core-glycosylated, pro-a-factor contained in the supernatant divided by the total protease-protected core-glycosylated pro-a-factor contained in a total reaction. A unit of Sarlp-dependent vesicle budding activity is defined as restoration of one-half maximal vesicle format.ion to a Sarlp-depleted cytosol (from RSY658) in a standard 50-pl reaction after 30 min. Material from each step of the purification was titrated to determine a linear response range for calculation of the amount required for one-half maximal budding.
Anti-Sarlp Antiserum-Antibodies were raised against a recombinant glutathione S-transferase-Sarlp fusion protein. The 600-base pair Rsaf-EcoRI fragment of pANY2-7 (Nakano and Muramatsu, 1989) was inserted into the S m d and EcoRl sites of plasmid pCEX3X (Pharmacia). This construct places the second exon of Sarlp at the C terminus of glutathione 5'-transferase. After isopropyl-l-thio-@-Dgalactopyranoside induction, this fusion protein was purified on a glutathione column according to the manufacturer's specifications (Pharmacia). Glutathione S-transferase-Sarlp fusion protein was used as antigen to raise polyclonal antisera in rabbits. Rabbits were immunized by subcutaneous injection with 150 pg of protein in Freund's complete adjuvant and boosted every 4 weeks with 100 pg of protein in incomplete adjuvant. Sarlp antiserum was used at a 1/ 2000 dilution for development of immunoblots by the enhanced chemiluminescence method (Amersham Corp.). Glutathione S-transferase-Sarlp (2 mg) was covalently linked to 1 ml of Affi-Gel-10agarose (Bio-Rad) as described by the manufacturer. This column was used to purify anti-Sarlp antibodies from serum by binding in Tris-buffered saline and elution with 0.2 M glycine-HCl, pH 2.2 (Harlow and Lane, 1988). Fab fragments from affinity-purified Sarlp antibodies were isolated after treatment with immobilized papain (Pierce Chemical Co.) as described by the manufacturer.
Other Methods-Sarlp depleted and desalted cytosol was prepared by gel filtration of a 100,000 X g supernatant (6 ml) prepared from strain RSY607 on a 200-ml(2.5 cm X 45 cm) Sephacryl S-100 column. The column was eluted with buffer 88, and 4-ml fractions were collected. The first six protein-containing fractions were pooled, concentrated 2-fold with a centriprep 10 concentrator (Amicon, Beverly, MA), and stored at -70 "C. SDS-PACE (Laemmli, 1970). Protein silver staining was performed as described by Morrisey (1981). For immunoblo~s, protein was transferred to nitrocell~ose (Towbin et al., 1979) and antigen detected by the enhanced chemiluminescence method (Amersham Corp.). Protein concentrations were determined (Bradford, 1976) with immunoglobulin as the standard.

RESULTS
Functional Assay for Sarlp in Vesicle Formation-Sarlp is proposed to function in vesicle formation from the ER (d' Enfert et al., 1991a). We sought to purify active S a r l p a n d demonstrate function in a vesicle formation reaction. An initial assay that requires Sarlp for vesicle formation was devised by depleting the cytosol of Sarlp by o v e~r o d u c i n g SeclZp. Overproduction of SeclZp had been shown to recruit the soluble pool of Sarlp to a sedimentable membrane fraction (d'Enfert et aL, 1991b). This overproduced cytosol (OPC) was unable to support vesicle formation in vitro unless supplemented with Sarlp, thus allowing an assay for the purification of a functional form of Sarlp (Fig. 1). SARl overexpression under the GAL1 promoter resulted in a -50-fold increase in cellular Sarlp. In a wild-type strain, 80% of the Sarlp is membrane-bound; however, overexpression increases the concentration of soluble Sarlp contained in a 100,000 x g supernatant (d'Enfert et al., 1991a, Nishikawa and. Fractionation of the 100,000 X g supernatant material on a Sephacryl S-100 gel filtration column enriched Sarlp budding activity as it eluted at the position of 7-kDa proteins and was resolved from a majority of other proteins contained in a 100,000 x g supernatant ( Fig. 2 and Table I). This unusual elution position suggests an interaction of Sarlp with the gel filtration matrix, and, indeed, this property results in difficulties manipulating Sarlp; we have found that this protein binds to plastic surfaces and column resins unless detergent is added. Therefore, further purification required the presence of Triton X-100. Contaminating proteins in the S-100 pool (Fig. 2, lane 4 ) are resolved by DEAE chromatography. Sarlp eluted from the DEAE column a t 175 mM KOAc. Sarlp immunoreactivity and rescue activity coincided with the single protein that eluted from the DEAE column (not shown). SDS-PAGE demonstrated the purity of Sarlp at each step (Fig. 2), and Sarlp-specific activity in the budding assay is shown in Table I. A low yield of Sarlp was obtained, probably due to losses through hydrophobic interactions.
Sarlp purified to homogeneity restored vesicle formation activity to a Secl2p OPC (Fig. 3A, column 6). The addition of purified Sarlp to a saturating amount of WTC did not affect the budding efficiency (data not shown). Secl2p OPC supplemented with Sarlp exhibited similar but not identical kinetics to a cytosol prepared form a wild-type strain (Fig.  3B). The apparent lag in vesicle formation may reflect a requirement for Sarlp assembly into a complex prior to promoting vesicle formation.
Anti-Sarlp Antibodies Inhibit Vesicle Formation-As an independent verification of Sarlp function in vesicle formation, we tested the effect of Sarlp antibodies on a wild-type vesicle formation reaction. Potent inhibition of the vesicle formation reaction was observed with 0.5 pg of affinity-puri-  fied anti-Sarlp antibodies. The divalent nature of antibodies could interfere with the vesicle formation reaction by crosslinking vesicles to the ER membranes. Thus, anti-Sarlp Fab fragments were generated and tested for inhibition. A titration curve is shown in Fig. 4 where complete inhibition was achieved by adding 0.1 pg of affinity-purified anti-Sarlp Fabs. Addition of 0.2 pg of purified yeast Sarlp relieved the antibody block at antibody concentrations below 0.2 pg. Binding Properties-Sarlp expressed in Escherichia coli binds GTP after SDS-PAGE and renaturation on nitrocellulose (Oka et al., 1991). We determined the guanine nucleotidebinding properties by an assay in which nucleotide bound to Sarlp in solution is retained on a nitrocellulose filter. As shown in Table 11, binding requires M e (maximal concentration is a t 0.1 mM Mg+) and detergent; however, the detergent requirement may be replaced by soybean phospholipids. Triton X-100 (CMC = 0.016%) and P-octylglucoside (CMC = 0.7%) were equivalent and were required at concentrations above their CMC. A somewhat related situation has been reported for mammalian ARF which requires both detergent and phospholipids to display nucleotide binding (Weiss et al., 1989). The rate of guanine nucleotide binding by Sarlp was rapid and essentially complete after 20 min at 29 "C while the binding rate at 4 "C was markedly slower (Fig. 5A). This binding rate probably represents nucleotide exchange of

ER.
Vesicle formation assays were carried out as in Fig. 1, with 200 pg of wild-type cytosol and increasing amounts of affinity-purified anti-Sarlp Fab in the presence (0) or absence (U) of 0.2 p g of purified yeast Sarlp. Control budding efficiency is 20% in this experiment.
bound GDP from purified Sarlp. Saturation binding indicated that 67% of the Sarlp was guanine nucleotide bound under our optimal conditions. The reason for incomplete binding could be that a percentage of protein was inactive or the use Sarlp guanine nucleotide binding Sarlp (5 pmol of binding sites) was incubated with 2 p~ [cY-~*P] GTP (specific activity 10,000 dpm/pmol) for 30 min at 29 "C and bound label was determined by a filter binding assay (see "Experimental Procedures"). Complete includes 25 mM K-HEPES (pH 6.8), 0.1% Triton X-100, 0.5 mM dithiothreitol, 1 mM Mg(OAc),, and 0.25 mg/ml bovine serum albumin. Phospholipids or P-octylglucoside was added in the absence of Triton X-100. Values represent the mean f S.D. of triplicate determinations. of y-globulin as a protein standard may not accurately reflect the true Sarlp protein concentration. The off rates of prebound GDP, GTP, and GTPyS from Sarlp were determined by the filter binding assay (Fig. 5B). The apparent first order dissociation rates are 0.071 min", 0.091 rnin", and 0.043" for GDP, GTP, and GTPyS, respectively.
GTPase Activity of Sarlp-The rate of GTP hydrolysis by Sarlp was determined under conditions of maximal binding. Sarlp added to reactions is expressed in moles of nucleotidebinding competent protein, and [W~*P]GDP was quantified after separation of GDP from GTP by thin layer chromatography (Fig. 6A). Production of [e-"P]GDP was linear with time up to 90 min (Fig. 6 B ) and inhibited by the addition of 10 mM EDTA or a 100-fold excess of unlabeled GTP. The addition of a 100-fold molar excess of ATP had a minimal effect on GTP hydrolysis consistent with specificity for e anine nucleotides by Sarlp. A hydrolysis rate of 0.0011 min" (mol GTP hydrolyzed/mol Sarlp/min) was calculated from a time course (Fig. 6B). Production of GDP under these conditions in the absence of Sarlp was undetectable after 90 min (data not shown).
Sarlp-GTPyS Is Less Active in Supporting Vesicle Formation-We found that Sarlp bound GTP or GTPyS in the presence of soybean phospholipids (see Table 11). Therefore, we addressed whether these prebound forms of Sarlp were equally capable of supporting vesicle formation. Fig. 7 shows that Sarlp-GTPyS was reduced in its ability to support vesicle formation compared to a cytosol supplemented with Sarlp-GTP. A control shown in Fig. 7A, column 5, demonstrated that phospholipids and free GTPyS (0.2 PM) did not affect the degree of vesicle formation with a wild-type cytosol. This control was necessary because we were unable to separate efficiently free guanine nucleotide from Sarlp-bound nucleotide used in these reactions. A time course of vesicle formation with Sarlp-GTP and Sarlp-GTPyS is shown in Fig. 7B. At early times, Sarlp-GTPyS was unable to support vesicle formation; however, a t later time points, vesicle formation was apparent but about one-half that of Sarlp-GTP. This lag was likely due to an inability of Sarlp-GTPyS to support vesicle formation. Nucleotide exchange of GTPyS for GTP present in the Secl2p OPC may permit Sarlp function at later times.
Depletion of Sarlp and GTP from a wild-type cytosol by gel filtration on a Sephacryl S-100 column revealed a complete dependence on GTP hydrolysis in vesicle formation (Fig. 8). Sarlp-GTPyS was devoid of budding activity with this cytosol GTP-t , : while Sarlp-GTP was active. GDP substituted for GTP; however, GDPPS, a nucleotide that is not efficiently phosphorylated, was unable to support vesicle formation. These results suggest that GTP hydrolysis by Sarlp is required for vesicle budding. GDP may substitute only by virtue of conversion to GTP by diphosphonucleotide kinase and the ATP and ATP regeneration system provided to stimulate vesicle budding. DISCUSSION We have purified Sarlp and characterized its nucleotide binding, exchange, and hydrolysis properties. Further, we demonstrated that the purified form functions in an in vitro vesicle formation assay. A comparison of Sarlp binding and hydrolysis properties with other purified GTPases reveals similarities and several significant differences. Common among GTPases is the ability to bind and hydrolyze GTP in a Mg'+-dependent manner. Unlike most small GTPaseu, Sarlp has a requirement for detergent or phospholipids to bind guanine nucleotide.
Interestingly, mammalian ARFs have a similar requirement for both detergent and phospholipids for maximal nucleotide binding (Kahn et al., 1988;Weiss et aZ., 1989). The reason(s) for this requirement are not known but may reflect the interaction of these proteins with intracellular membranes Serafini et al., 1991;Donaldson et al., 1991).
Although it is tempting to speculate how these altered rates of binding and hydrolysis translate into cellular function, increasing evidence suggests these parameters are tightly regulated in vivo by guanine nucleotide regulatory proteins (Bourne et al., 1991). In the case of yeast Raslp, a guanine nucleotide regulatory protein that activates GTPase activity is encoded by the IRA1/2 genes (Tanaka et al., 1990) while acceleration of GDP release is accomplished by the CDC25 gene product (Crechet et al., 1990;Jones et aZ., 1991). Another type of guanine nucleotide regulatory protein is reported to bind a small GTPase and slow the release of guanine nucleotide (Ueda et al., 1990). Thus, Sarlp hydrolysis and exchange rates are likely to be modulated in vivo dependent on context. A number of gene products involved in ER vesicle formation in yeast have been associated genetically with SARl and could modulate Sarlp guanine nucleotide interactions. Characterization of purified Sarlp binding and hydrolysis properties now allows us to investigate modulation by Sec proteins in vitro. We have recently found that Sec23p stimulates the G T P hydrolysis rate of Sarlp.' Sec23p is also required for * Yoshihisa, T., Barlowe, C., and Schekman, R., (1993) Science, in press. transport vesicle formation from the ER, suggesting GTP hydrolysis by Sarlp is required for this event.
Sarlp preloaded with GTPyS was less active in promoting vesicle formation than Sarlp-GTP. This observation is quite similar to that reported for GTPyS inhibition of a budding reaction with a wild-type cytosol (Rexach and Schekman, 1991). A cytosol that has been gel-filtered to remove both Sarlp and GTP does not support vesicle formation unless supplemented with GTP or GDP and Sarlp. GTPyS cannot replace GTP in this assay. We propose that GTP hydrolysis is required for vesicle budding; however, we cannot exclude the possibility that the Sarlp-GTPyS form malfunctions due to an inability to interact with the transport machinery.
As a majority of cellular Sarlp is membrane bound, it is interesting to note that soluble Sarlp is required for efficient vesicle formation in vitro. This suggests that Sarlp cycles between soluble or membrane-bound forms and vesicle budding is regulated by the attachment of soluble Sarlp to its target site on the ER membrane. Secl2p facilitates the attachment of Sarlp to membranes (d'Enfert et aZ., 1991b), perhaps by catalyzing the exchange of GDP to GTP on Sarlp. Membrane-bound Sarlp-GTP then could function in the assembly of a protein complex required for vesicle budding. After the complex has executed its function, Sarlp hydrolysis of the bound GTP may result in a disassembly of the complex and a recycling of budding components. The failure of endogenous membrane-bound Sarlp to function in the absence of the cytosolic Sarlp remains unexplained. Perhaps an additional rate-limiting recycling factor is necessary to mobilize membrane-bound Sarlp. The function of Sarlp in vesicle budding can be investigated more directly once the in vitro vesicle formation assay is reconstituted with purified components. I