The Yeast VPSl7 Gene Encodes a Membrane-associated Protein Required for the Sorting of Soluble Vacuolar Hydrolases*

ups1 7 mutants missort and secrete several vacuolar hydrolases. To analyze the role of the VPSl7 gene in vacuolar protein delivery, we have cloned this gene by complementation of the vacuolar protein sorting defects of a vpsl7-5 mutant. Disruption of the VPSl7 gene had no effect on the viability of haploid yeast cells, although they show an obvious defect in vacuolar morphology. ups1 7-disrupted cells contain numerous small vacuole-like compartments and also exhibit a severe defect in the sorting of carboxypeptidase Y (CPY), a soluble vacuolar hydrolase. 95% of CPY is missorted and secreted from the mutant cells. Vacuolar sorting of two other soluble hydrolases, proteinase A and proteinase B, is also affected, but to a lesser extent. Delivery and maturation of the vacuolar membrane protein alkaline phosphatase does not appear to be affected in a Avpsl7 strain.

ups1 7 mutants missort and secrete several vacuolar hydrolases. To analyze the role of the VPSl7 gene in vacuolar protein delivery, we have cloned this gene by complementation of the vacuolar protein sorting defects of a vpsl7-5 mutant. Disruption of the V P S l 7 gene had no effect on the viability of haploid yeast cells, although they show an obvious defect in vacuolar morphology. ups1 7-disrupted cells contain numerous small vacuole-like compartments and also exhibit a severe defect in the sorting of carboxypeptidase Y (CPY), a soluble vacuolar hydrolase. 95% of CPY is missorted and secreted from the mutant cells. Vacuolar sorting of two other soluble hydrolases, proteinase A and proteinase B, is also affected, but to a lesser extent. Delivery and maturation of the vacuolar membrane protein alkaline phosphatase does not appear to be affected in a Avpsl7 strain.
The DNA sequence of the V P S l 7 clone indicates that the gene encodes a 561-amino-acid protein with a calculated molecular mass of 63.1 kDa. The protein sequence is hydrophilic and contains no obvious N-terminal signal sequence or hydrophobic membrane-spanning domains, indicating that the Vpsl7p does not enter the secretory pathway. Using a Vpsl7p-specific polyclonal antiserum, we have demonstrated that the Vpsl7 protein is not modified with N-linked carbohydrates at any of its four potential N-linked glycosylation sites. The Vpsl7 protein, however, fractionates to a particulate fraction after centrifugation at 100,000 X g. Vpsl7p can be released from this particulate fraction by treatment with either Triton X-100 or urea, indicating that the Vpsl7p is peripherally associated with a crude membrane fraction.
Based on these results, we propose that the Vpsl7p functions on the cytoplasmic surface of some intracellular organelle, possibly the Golgi complex or an intermediate in Golgi to vacuole transport, to facilitate the sorting and delivery of soluble vacuolar hydrolases. Vacuolar membrane protein traffic, however, appears * This work was supported by Public Health Service Grant GM-32703 from the National Institutes of Health (to S. D. E.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

The nucleotide sequencefs) reported in thispaper has been submitted to the GenBankTM/EMBL Data Bank with accession numberfs) L02869.
$Supported by a research fellowship from the Deutsche Forschungsgemeinschaft.

I Supported as an investigator of the Howard Hughes Medical
Institute. To whom correspondence should be addressed Div. of to occur by a mechanism that is independent of Vpsl7p function.
Eukaryotic cells are highly compartmentalized containing many different membrane-enclosed organelles that are specialized for different tasks. These specialized compartments are largely maintained by a constant flow of a unique set of proteins that must be accurately sorted and delivered to each compartment. The five major targets for noncytoplasmic proteins in most eukaryotic cells include the secretory system, mitochondria, peroxisomes, lysosomes, and the nucleus. Proteins destined for the cell surface and the lysosome transit together through the early stages of the secretory pathway.
After translocation into the ER' and addition of core oligosaccharides, proteins that are not destined to be retained in the ER are transported to and through the Golgi complex via transport vesicles (Novick et al., 1980(Novick et al., , 1981Kaiser and Schekman, 1990). In the Golgi, most proteins undergo further post-translational modifications (addition of outer chain carbohydrates, elongation of 0-linked oligosaccharides (Tanner and Lehle, 1987)) prior to transport out of the late Golgi. Protein transport to the cell surface appears to occur by a default or bulk flow mechanism, whereas the transport of lysosomal proteins requires specific sorting signals (Kornfeld, 1986;Kornfeld and Mellman, 1989). Soluble lysosomal proteins of mammalian cells are modified with mannose 6-phosphate residues on N-linked carbohydrate chains, which are recognized by receptors that mediate sorting and delivery to the lysosome (Kaplan et al., 1977;von Figura and Hasilik, 1986;Kornfeld, 1987). In contrast, the sorting information present in yeast lysosomal/vacuolar proteins is not associated with any specific carbohydrate modification and instead appears to reside within the polypeptide backbone of these proteins (Clark et al., 1982;Schwaiger et al., 1982;Valls et al., 1987;Johnson et al., 1987;Klionsky et al., , 1990Klionsky and Emr, 1990;Winther et aL, 1991).
Genetic selections in yeast have identified more than 40 ups (ups for vacuolar protein sorting defective) complementation groups that are required for the sorting and delivery of vacuolar hydrolases (Bankaitis et al., 1986;Rothman and Stevens, 1986;Robinson et al., 1988;Rothman et al., 1989aRothman et al., , 1989b. All ups mutants missort and secrete vacuolar enzyme precursors instead of delivering them to the vacuole. Since protein secretion and post-translational modification seems to be normal in most of the ups mutants, these mutants are presumed to be either defective in components of the sorting 'The abbreviations used are: ER, endoplasmic reticulum; CPY, carboxypeptidase Y; PrA, proteinase A PrB, proteinase B; CDCFDA, 5(6)-carboxy-:!',7'-&1chlorofhorescein diacetate; ALP, alkaline phosphatase; SM, synthetic media; kb, kilobase(s); ORF, open reading frame; WT, wild-type. 559 machinery or in the assembly and maintenance of the acceptor compartment, the vacuole. The ups mutants have been grouped into three major classes, according to their vacuolar morphology . Class A ups mutants appear to have normal vacuoles, whereas class B mutants contain several small fragmented vacuole-like compartments, as shown by vacuole-specific fluorescent staining procedures . Class C mutants lack vacuolar structures and instead accumulate other membrane compartments including, vesicles, multilamellar membrane structures, and Golgi-like structures . Thus far, eight of t h e VPS genes have been cloned and characterized (Dulic and Riezman, 1989;Banta et al., 1990;Herman and Emr, 1990;Raymond et al., 1990;Rothman et al., 1990; Wada et al., 1990;Woolford et al., 1990;Herman et aL, 1991a;Preston et aL, 1991;Robinson et al., 1991;Paravicini et al., 1992).
Using a gene fusion-based selection scheme for ups mutants (Bankaitis et al., 1986;Robinson et al., 1988), we identified 11 alleles of the upsl 7 locus, and each was found to missort precursor forms of the soluble vacuolar hydrolases carboxypeptidase Y (CPY), proteinase A (PrA), and proteinase B (PrB) . Further complementation analysis revealed that upsl 7 and pep21 define the same complementation group (Rothman et al., 1989a). Light and electron microscopic analysis have demonstrated that representative alleles of the ups2 7 complementation group exhibit class B vacuolar morphology. Many small vacuole-like compartments (10-30) are present i n the mutant cells, rather than the one to three large vacuoles normally seen in wild-type cells .
In this study, we report on the cloning and sequencing of t h e VPSl7 gene, the identification and localization of the VPSl7 gene product, and the biochemical and morphological consequences of a ups17 null allele. Our results support the model that there are alternative delivery mechanisms to the yeast vacuole and that the Vpsl7p appears t o be required for the efficient sorting of only a subset of vacuolar proteins.

EXPERIMENTAL PROCEDURES
Materials-Boehringer Mannheim and New England BioLabs (Beverly, MA) were the suppliers for all DNA restriction and modifying enzymes used in this study. Deoxynucleotides, 5-bromo-4chloro-3-indolyl-~-~-galactoside and isopropyl-8-D-thiogalactopyranoside were also purchased from Boehringer Mannheim. The DNA sequencing kit was from U. S. Biochemical Corp. and the T r a n~'~S label from E N Radiochemicals (Irvine, CA). [a-36S]dATP, ["PI orthophosphate, [CI-~~PI~CTP, and the multiprime DNA labeling kit were from Amersham Corp. Pall Biodyne transfer membranes were purchased from ICN Chemicals & Radioisotope Division (Irvine, CA). The Elutrap Electro-Separation Chamber was obtained from Schleicher & Schuell. Zymolyase IOOT (Kirin Brewery Co.) is a product from Seikagako Kogyo Co. (Tokyo, Japan). 5(6)-Carboxy-2',7'-dichlorofluorescein diacetate (CDCFDA) was from Molecular Probes, Inc. (Beaverton, OR). Freund's complete and incomplete adjuvants were obtained from GIBCO. The cellulose DC plates were from Merck (Darmstadt, Germany). All other chemicals including the antiserum to glucose-6-phosphate dehydrogenase were purchased from Sigma. The antiserum to PrB was a gift from Elizabeth Jones (Carnegie-Mellon University, Pittsburgh, PA) and the Kex2p-specific antiserum was obtained from Bill Wickner (University of California, Los Angeles, Los Angeles, CA). Antisera to CPY, PrA, and ALP were described previously (Klionsky and Emr, 1989;. Strains and Media- Table I describes all the Saccharomyces cereuisiue and Escherichia coli strains used in this study. Yeast cells were grown at 30 "C in YPD-rich medium or standard minimal medium, supplemented as necessary (Sherman et al., 1986) or in Wickerham's minimal proline medium (WIMP) (Wickerham, 19491, supplemented with 0.2% yeast extract (WIMPYE) and the necessary amino acids. All bacterial cells were grown at 37 "C in standard LB medium (Miller, 1972), supplemented with ampicillin (100 pg/ml). Genetic Procedures-Crosses, sporulation of diploids, and tetrad analyses were performed by standard genetic techniques (Sherman et al., 1986). One-step gene disruptions and integration experiments were done according to  and Orr-Weaver et al. (1983), respectively. Yeast cells were transformed using the lithium acetate method of Ito et al. (1983). E. coli transformation was carried out as described by Hanahan (1983).
Cloning of VPSl7-The VPS17 gene was cloned by transforming the yeast strain SEY17-5 containing the CPY-invertase fusion plasmid pCYI-50 (CEN4, ARSI, URA3) (Johnson et al., 1987) with a genomic yeast library, a pYCP50 derivative (CEN4, ARSl, LEUZ) containing large Sau3A yeast genomic inserts. The library was a generous gift from Phil Hieter (Johns Hopkins University School of Medicine, Baltimore, MD). Leu+ transformants were selected on SM glucose plates, replica plated onto SM fructose plates, incubated overnight at 26 "C, and subsequently analyzed by an invertase plate assay (see "Results") Paravicini et al., 1992). Vps' transformants were restreaked; the complementing plasmids were isolated as described by Sherman et al. (1986), and amplified in E. coli.
Nucleic Acid Techniques-Genomic yeast DNA was isolated from spheroplasts essentially as described by Sherman et al. (1986) and used for Southern blot analysis (Southern, 1975). Yeast mRNA was isolated by a hot phenol extraction as described previously (Kohrer and Domdey, 1991). Radioactively labeled DNA probes were generated according to the method of Feinberg and Vogelstein (1984). Double-stranded plasmid DNA templates used for sequencing were isolated by a small scale boiling method (Wilimzig, 1985) and denatured prior to sequencing. All other molecular biology techniques were performed as described in Ausubel et al. (1990).
Sequence Analysis-The DNA sequence of VPSl7 was obtained by generating exonuclease 111-mung bean nuclease deletion constructs from both ends of plasmid pS25 (pBluescript KS+, containing the 2.6-kb VPS17 BglII-SmaI fragment) and pCla2l (pBluescript KS+, containing the 2.1-kb VPSl7 ChI fragment) as described in the Stratagene manual. Overlapping clones were sequenced with T3-(5'-ATTAACCCTCACTAAAG-3') and T7-(5"AATACGACTCACTA-TAG-5') specific primers according to the dideoxynucleotide chain termination method (Sanger et al., 1977) using the Sequenase sequencing kit (U. S. Biochemicals). The sequence of the VPSl7 gene, as well as its predicted protein sequence were compared to the entries of the NBRF, GenBank, and EMBL databases, using the FASTA and TFASTA programs of the University of Wisconsin Genetics Computer Group sequence analysis package (Devereux et al., 1984) and at the NCBI, using the BLAST network service (Altschul et al., 1990, Karlin andAltschul, 1990).
Integration of the TRPl Marker Next to the VPSl7 Locus-Yeast strain SEY17-5 (ups1 7, trpl -A901) was transformed with 5 pg of linearized plasmid pKK17-12. Tryptophan prototrophic transformanta were selected and subsequently crossed to the wild-type strain SEY6210. After sporulation of the diploid strain, random spore analysis and tetrad analysis were performed.
Disruption of the VPSl7 Gene-Disruption of the VPSl7 gene was accomplished in several strains. The diploid strain SEY6210.5 was transformed with 10 pg of ClaI-digested pKK17-13 plasmid DNA. His+ transformants were selected, sporulated, and tetrads dissected. The same strategy was used to disrupt the VPS17 gene in the two haploid strains BHYlO and BHY11, generating KKYlO and KKY11.
Antiserum Production-An in-frame trpE-Vpsl7 fusion protein was expressed in E. coli XL1-blue cells, using the fusion construct pKK17-14. After induction with indoleacrylic acid, the insoluble protein fraction was purified as described previously (Paravicini et al., 1992). The titration of the Vpsl7p-specific antiserum used in this study showed that 2 pl of antiserum quantitatively precipitate Vpsl7p from 1 unit of cells at an optical density (ODmnm) of 1.
Labeling Yeast Cells, Immunoprecipitation, Electrophoresis, and Fluorography-Whole yeast cells and spheroplasts were labeled with Tran3% label as described earlier (Paravicini et al., 1992) with the following modification. The labeling reactions were chased by the addition of 1 volume of 2 X WIMPYE containing 10 mM methionine.
1 M sodium salycylate, 1% glycerol. Dried gels were fluorographed at In Vivo Phosphorylation Analysis-Whole cells were labeled with 32P04 as described by Herman et al. (1991a) except that bovine serum albumin was excluded from the labeling reaction. The vpsl7p immunoprecipitations were performed as described above except for the protein A-Sepharose beads which were washed three times with buffer A (50 mM Tris-HC1, pH 7.5,150 mM NaCl, 0.1 mM EDTA) containing 0.5% Tween-20 and three times with 0.5 M LiC1,O.l M Tris-HCl, pH 7.5. After electrophoresis, the SDS-polyacrylamide gels were soaked in water for 30 min, fixed for 30 min in 10% acetic acid, 10% trichloroacetic acid, 30% methanol, soaked in 1% glycerol for 30 min, and dried. The gels were exposed to x-ray film at -80 "C with intensifier screens.
Phosphoamino Acid Analysis-10 ODm units of BHY10, harboring the VPS17 gene on a 2-pm plasmid (pKKY17-2) were labeled with 32P04 as described above. The immunoprecipitated Vpsl7p was run on a 9% SDS-polyacrylamide gel, dried onto Whatman 3 " paper, and detected by autoradiography. The Vpsl7p containing gel piece was excised and treated for phosphoamino acid analysis as described by Meisenhelder and Hunter (1991) with the following modifications. The first and second dimension of the thin layer electrophoresis of hydrolyzed protein were run at 7 V/cm. Subcellular Fractionation of Vpsl7p"Fractionation experiments were done either with the wild-type strain BHYlO or with the same strain transformed with the overexpressing Vpsl7p construct (pKKY17-2). 50 units of cells at a ODm = 1 were harvested, spheroplasted with Zymolyase 100T, labeled with Tran"S for 15 min, and chased for 45 min with WIMPYE containing 10 mM methionine, as described above. The spheroplasts were sedimented in a clinical centrifuge at 2,000 X g, resuspended in 1.5 ml of lysis buffer (10 mM triethanolamine, pH 7.2,0.8 M sorbitol, 1 mM EDTA, 1 mg/ml bovine serum albumin, 1 mM phenylmethylsulfonyl fluoride, 50 pg/ml pepstatin, 100 pg/ml ora-macroglobulin, 100 pg/ml aprotinin, 50 pg/ml antipain-&hydrochloride) and dounced 20 times using a glass tissue homogenizer. The lysate was spun for 3 min at 500 X g at 4 "C. Subsequently, the pellet was re-extracted in the same volume of lysis buffer, centrifuged, and the combined supernatants spun again. The volume of this S5 supernatant was adjusted to 3 ml with lysis buffer. 0.5 ml of S5 was precipitated with 5% trichloroacetic acid and the rest was spun at 13,000 X g for 15 min at 4 "C to generate a P13 pellet and a S13 supernatant fraction. 0.5 ml of the S13 supernatant were directly trichloroacetic acid-precipitated. The P13 pellet was resuspended in 2.5 ml of lysis buffer and 0.5 ml were trichloroacetic acidprecipitated. 0.5-ml aliquots from the S13 supernatant fraction were adjusted to 1 ml with lysis buffer or with one of the following substances to the indicated final concentrations: 1 M NaC1,2 M urea, 2% Triton X-100. The lysates were incubated on ice for 30 min and subsequently centrifuged at 100,000 X g for 45 min in a Ti-70.1 rotor (Beckman Instruments, Inc.) at 4 "C. The supernatants were trichloroacetic acid precipitated, and the pellets were directly dissolved in 100 pl of boiling buffer. Immunoprecipitation was carried out as described above.

RESULTS
Cloning and Genetic Analysis of the VPSl7 Gene-Previously, we have described a genetic selection procedure, which was based on the sorting behavior of a CPY-invertase hybrid protein, to isolate ups mutants in yeast (Bankaitis et al., 1986;Robinson et al., 1988). Complementation analysis among the ups mutants revealed that there were 11 alleles in the ups17 complementation group ) and one of them, ups17-4, showed a temperature-sensitive growth phenotype. Detailed genetic analyses, however, demonstrated that the temperature-sensitive growth phenotype was tightly linked to the upsl 7 locus, but actually caused by a mutation in a neighboring gene. Therefore, we chose the upsl7-5 allele to clone the wild-type VPSl7 gene. The CPY-invertase sorting defect exhibited by the vpsl7-5 mutant was used to identify prospective VPSl7 clones. In wild-type cells (deleted for all endogenous genes encoding invertase, Asuc), the targeting information contained in the CPY portion of a CPYinvertase fusion protein leads to efficient sorting of the fusion protein to the vacuole (Johnson et al., 1987). However, ups mutant cells missort and secrete the fusion protein. The secreted invertase activity leads to a selectable phenotype, the ability to grow on sucrose, an invertase substrate, as a sole carbon source (Bankaitis et al., 1986;Robinson et al., 1988).
In order to clone the VPSl7 gene, the strain SEY17-5 (upsl 7-5, leu2-3, 112, harboring the CPY-invertase hybrid construct pCYI50 (Johnson et al., 1987)) was transformed with a YCp50-based genomic yeast library (CEN4, ARSI, LEU2). 15,000 Leu+ transformants were replicaplated onto SM fructose plates and were assayed for secreted invertase by a simple plate overlay assay (Paravicini et al., 1992). Using this assay, one can easily distinguish between colonies that secrete the CPY-invertase fusion protein (Vps-colonies turn red after a few minutes) and colonies that sort the CPYinvertase hybrid protein to the vacuole (Vps' colonies remain white). Among the 15,000 screened colonies only one colony remained white in the plate overlay assay, indicating that this transformant contained a plasmid that complemented the upsl 7 sorting defect. Curing this transformant of the library plasmid uncovered the Vps-phenotype. Subsequently, the complementing plasmid was isolated and amplified in E. coli.
Reintroduction of the rescued plasmid into  and other alleles of the ups17 complementation group (upsl 7-2, ups1 7-4) proved that this genomic clone, designated pKKY17, was able to complement the protein sorting defect exhibited by all ups17 mutants tested. This confirmed that the plasmid complemented the Vps-phenotype.
DNA restriction analysis of pKKY17 identified a 9-kb genomic insert (Fig. lA), and further subcloning and complementation experiments allowed us to localize the complementing activity to a 2.6-kb Hind11 fragment (Fig. 2B).
Integrative mapping was used to determine if this genomic fragment corresponds to the authentic VPS17 locus. We subcloned the complementing 5-kb BarnHIISphI DNA fragment (see Fig. 1B) into the integrating plasmid pRS304 (Sikorski and Hieter, 1989), which carries the selectable TRPl marker. This integrating plasmid (pKK17-12) was linearized at a single SmaI site within the complementing fragment to facilitate homologous recombination (Orr-Weaver et al., 1983). Following transformation of the yeast strain SEY17-5 (harboring the CPY-invertase fusion plasmid pCYI50) with the linearized plasmid, Trp' transformants were selected and subsequently crossed to the parental wild-type strain SEY6210. Following sporulation of the diploid strain, multiple tetrads were dissected and the spores analyzed. All analyzed asci showed the expected 2:2 Trp+/Trp-segregation pattern, but none of the spores was Vps-. We also analyzed 90 random spores and found that 43 were Trp' and 47 were Trp-. None of the 90 spores exhibited a Vps-phenotype. These results demonstrated that the complementing clone was tightly linked to the VPSl7 locus. The integrative mapping analysis also confirmed the earlier finding that the VPS17 gene is linked to the ADE2 locus , which has been mapped to the right arm of chromosome XV (Mortimer et al., 1989).
VPSl7 Sequence Analysis-Both strands of the 2.8-kb A EH pKIc117 t t P H BglII-Hind11 complementing fragment were sequenced (see "Experimental Procedures") and one long open reading frame (ORF) of 1653 base pairs was identified. This ORF has the potential to code for a 551-amino-acid polypeptide (Fig. 2 A ) . Using a 1.6-kb ClaI fragment (see Fig. IC), which is specific for this ORF, as a probe, we detected a single 1.9-kb polyadenylated RNA species on a Northern blot (Fig. 3B). The size of this mRNA is consistent with the predicted size of the VPSl7 ORF.

BBDBBOHllCKEECH S HI1 C SP t t t t t t t t t t t t t t t t t
Analysis of the deduced protein sequence indicated that the molecular mass of the Vpsl7 protein is 63.1 kDa, and according to the Kyte-Doolittle algorithm (Kyte and Doolittle, 1982), it appeared to be a very hydrophilic polypeptide (Fig.  2B). 30% of the amino acids in Vpsl7p are charged, and the protein has a predicted isoelectric point of 7.7. It contains four potential N-linked glycosylation sites (Asn-X-Ser/Thr) (Marshall, 1972), but no apparent N-terminal signal sequence or obvious hydrophobic membrane-spanning domains are predicted. Homology searches using the FASTA and TFASTA algorithms (Pearson and Lipman, 1988) of the UWGCG sequence analysis package (Devereux et al., 1984), or the BLAST program (Altschul et al., 1990;Karlin and Altschul, 1990), did not reveal any significant sequence similarity between Vpsl7p and any sequences in the GenBank, EMBL, or NBRF databases.

upsl 7 Null Mutant Strains Exhibit a Severe Vacuolar Protein Sorting Defect and Contain Numerous Small Vacuoles-
The one-step gene disruption technique ) was used to construct a Aupsl7 allele in the diploid strain SEY6210.5. Using the construct pKK17-13, we were able to replace almost all of VPSl7 (except amino acids 1-10) by the HIS3 gene (Fig. 1C). His+ transformants were sporulated and subjected to tetrad analyses. A segregation pattern of 2 Vps-, His+/2 Vps+, His-spores among all tetrads dissected was  observed, indicating that the VPSl7 null allele is not lethal, but results in a vacuolar protein sorting defect. There was also no temperature-sensitive growth defect associated with the Avpsl7 null mutant. All the spores were able to grow at 37 "C. The gene disruption was confirmed by isolating genomic DNA from all four segregants of one tetrad and analyzing it by Southern blot hybridization (Fig. 3A). Backcrossing the Avpsl7 strain to SEY17-5 demonstrated that the disruption indeed belongs in the ups17 complementation group. VPS17 gene disruptions were also performed in the two haploid strains BHYlO and BHY11; yielding the vpsl7 null mutant strains KKYlO and KKY11.
To further analyze the vacuolar protein sorting defect of the vpsl7 null mutant and to compare it with the originally isolated vpsl7-5 allele, we analyzed the sorting of CPY in the * B. Horazdovsky, unpublished results. wild-type strain BHYlO and the two ups17 mutant strains SEY 17-5 (vpsl7-5) and KKY 10 (Avpsl7). After radioactively labeling yeast spheroplasts with Tran35S for 15 min and chasing them for 60 min, the spheroplasts were fractionated into a pellet (I, intracellular) and a media (E, extracellular/ secreted) fraction, CPY was immunoprecipitated from both fractions and analyzed on SDS-polyacrylamide gels (Fig. 4). Compared to the wild-type strain, where most of the newly synthesized CPY (>95%) is matured to the 61 kDa form after the 60-min chase, the ups17 null mutant and the vpsl7-5 strain exhibited a strong sorting defect (Fig. 4). The majority of CPY (95%) was present as the Golgi-modified 69-kDa p2 form in the extracellular media fraction (Fig. 4). Only a hint of mature CPY ((5%) could be detected inside the cell (Fig.  4). This vacuolar sorting defect of the null mutant strain was completely complemented when the wild-type VPS17 gene

gene.
A, Southern blot analysis of a diploid strain with one disrupted ups17 allele. Disruption of the VPS17 gene was performed in the diploid strain SEY6210.5, using the plasmid pKK17-13. Upon sporulation, tetrad dissection, and germination, genomic DNA was isolated from all four segregants and digested with ChI to completion. The cleaved DNA was separated on a 1.2% agarose gel, blotted, and probed with a radioactively labeled 2.6-kb BglII-SmaI fragment obtained from pKKY17. In wild-type segregants (I and 2), a 1.6-and a 2.1-kb fragment hybridized to the probe, whereas in the upsl7disrupted segregants (3 and 4 ) , only a 3-kb ChI fragment was detected. The numbers on the left side indicate the position of DNA markers in kb. B, identification of the VPS17 RNA transcript by Northern blot analysis. Poly(A)+ RNA was isolated from SEY6210, separated in a 1.2% agarose gel, blotted, and probed with a radioactively labeled 1.6-kb ChI-fragment obtained from pKKY17-8. Numbers on the right side indicate the position of RNA size markers in kb. The arrow indicates the position of the 1.9-kb VPSl7-specific mRNA.
We also analyzed vacuolar morphology in the vpsl 7 null mutant because originally isolated alleles of the vpsl 7 complementation group resulted in the accumulation of fragmented vacuolar structures . We used a vital fluorescent dye, CDCFDA, for staining vacuolar compartments of wild-type and ups17 null mutant cells. As shown in Fig. 5, disruption of the ups17 gene leads to a typical class B vacuolar morphology . It was possible to stain up to 10 small vacuolar organelles in one focal plane/ cell, whereas in a wild-type yeast cell, only one or two larger compartments were stained. Vacuole inheritance does not seem to be affected in this mutant as all buds contained vacuolar structures. Spheroplasts of BHYlO ( W T ) and KKYlO (Aupsl7) were labeled with T r a x~~~S label for 10 min and then chased for 0 or 60 min. The cultures were separated into a pellet ( I , intracellular) and a supernatant (E, extracellular) fraction by a l-min centrifugation at 13,000 X g. The amount of CPY in each fraction was determined by immunoprecipitation. The migration positions of the precursor and mature forms of CPY (plCPY, p2CPY, mCPY), PrA (proPrA, mPrA), and ALP (proALP, A L P ) are indicated. membrane protein, ALP. Spheroplasts of strains BHYlO (WT) and KKYlO (Avpsl7) were labeled with Tran35S label for 10 min and chased for 0 or 60 min. Each sample was fractionated into a pellet (I, intracellular) and a supernatant (E, extracellular) fraction, and then split into four equal aliquots. Each aliquot was immunoprecipitated with antisera specific either to CPY, PrA, PrB, or ALP. The three paneki in Fig. 6 show the results of the immunoprecipitates for CPY, PrA, and ALP in the wild-type and Aupsl7 strains. The processing and sorting of CPY in the wild-type cells ( W T ) was normal; after the 10-min pulse period, all three forms of CPY, the ER-specific p l form ( p l C P Y ) , the Golgi-specific p2 form (p2CPY), and the vacuole-specific mature form ( m C P Y ) were labeled, and after the 60-min chase, all CPY was matured and found inside the cell (Fig. 6). The Avpsl7 null mutant showed the expected severe sorting defect for CPY. After the 10-min pulse period, a significant portion of the Golgi-specific p2CPY was already secreted, and after the 60-min chase, about 90% of pZCPY was in the medium fraction (Fig. 6). Only a minor fraction of mCPY (<5%) could be detected inside the mutant cells (Fig. 6). Since the conversion of plCPY to p2CPY is normal in the ups17 mutant cells, there is no obvious defect in protein transport from the ER to the Golgi. However, Golgi to vacuole transport is almost completely defective.
The vacuolar protein sorting defect for another soluble vacuolar hydrolase, PrA, however, was different from that of CPY. In the wild-type strain, we observed normal processing and sorting of PrA; after the 60-min chase, proPrA was all matured and found inside the cell (Fig. 6). The Avpsl7 null mutant, however, showed an unexpected sorting defect for PrA. After the 60-min chase, there was a significant amount (60%) of mature proteinase A (rnPrA) inside the cells and only a minor fraction (10%) of proPrA was secreted (Fig. 6). 30% of PrA accumulated inside the cells and remained as the pro-form. This demonstrated that the sorting and processing defect for PrA in the Avpsl7 null mutant is not as severe as that seen for CPY. Essentially the same sorting/processing defect was observed for another soluble hydrolase PrB. Approximately 60% mPrB and 30% proPrB accumulated inside the mutant cells, whereas less than 10% of proPrB were secreted (data not shown). In a more detailed kinetic analysis of PrA and PrB processing in wild-type and Avpsl7 mutant cells, we also could show that the maturation of precursor forms of PrA and PrB to mature PrA and PrB occurs about 2 times slower than in wild-type cells (data not shown), which indicates that either the movement of newly synthesized PrA and PrB through the delivery pathway to the vacuole, or its processing, is delayed in the Avpsl7 mutant strain. This kinetic delay in PrA processing also explains the results reported previously for PrA processing in the original ups17 alleles . Since only a 30-min chase point was analyzed, the previous study overestimated the actual sorting defect.
The processing of ALP (Fig. 6) was surprisingly unaffected in ups1 7 null mutant cells. After the 60-min chase period, in both wild-type and Avpsl7 mutant cells, all the radioactively labeled ALP was completely matured. Therefore, disruption of the ups1 7 gene seems to have a differential effect on the sorting and delivery of different vacuolar proteins.
Identifieation and Characterization of the Vpsl7 Protein-T o characterize the VPSI 7 gene product, we raised polyclonal antiserum against a trpE-Vpsl7 fusion protein. This trpE fusion protein contained 183 C-terminal amino acids of the Vpsl7 protein (amino acids 369-551). The fusion protein was expressed in E. coli cells, purified, and used to raise a polyclonal antiserum in a rabbit (see "Experimental Procedures"). Using this antiserum in immunoprecipitations, it was possible to detect a single 70-kDa protein in radiolabeled yeast cell extracts (Fig. 7A). This protein was about 5-8-fold more abundant when the VPSl7 gene was present on a multicopy plasmid, and the protein could not be detected in Avpsl7 cell extracts, or when preimmune serum was used (Fig. 7A). Based on these results, we conclude that this polyclonal antiserum specifically recognizes the Vpsl7 protein. Densitometric analysis of the levels of Vpsl7p relative to CPY (approximately 0.1% of total cell protein) suggested that Vpsl7p comprises <0.01% of total cell protein in logarithmically growing yeast cultures. The Vpsl7p appears to be a relatively stable protein, since in pulse-chase labeling experiments, we did not detect significant turnover of the protein even after a 90-min chase period (Fig. 7B).
The predicted molecular mass of Vpsl7p is 63 kDa, however, the observed molecular mass on a 9% SDS-polyacrylamide gel was 70 kDa. Although it is possible that the hydrophilic structure of the protein itself could account for this mobility shift, we also tested if the Vpsl7 protein undergoes any post-translational modifications. Thus, wild-type yeast cells were labeled with Tran35S label after 15 min of pretreatment with tunicamycin, a drug that inhibits N-linked glycosylation (Elbein, 1987). The size of the Vpsl7 protein was unaffected by this treatment, indicating that the four potential N-linked glycosylation sites are not used (Fig. 7B).
Because the VPS15 gene has been shown to encode a novel Ser/Thr protein kinase (Herman et al., 1991a) which plays a crucial role in the regulation of vacuolar protein sorting, we also tested if the Vpsl7 protein is phosphorylated in vivo. Whole yeast cells were labeled with [32P]orthophosphate, the cells were lysed with glass beads, and the Vpsl7p was immunoprecipitated from the clarified extracts. The phosphorylation analysis was done with the wild-type strain BHY10, BHYlO harboring the VPSl7 gene on a multicopy plasmid, and the Avpsl7 null mutant strain. As shown in Fig. 8, phosphorylated Vpsl7 protein was precipitated from the wildtype extract and significantly more from the Vpsl7p overexpressing strain, whereas no 32P-labeled Vpsl7p could be detected in the Avpsl7 null mutant strain. However, when we tested for phosphorylation of Vpsl7p in a ups15 null mutant, we found that phosphorylation of the Vpsl7p was independent of the Vpsl5 protein (data not shown). Although this observation does not exclude the possibility that Vpsl7p is a substrate of the Vpsl5 kinase, a detailed biochemical and mutational analysis will be required to further test this.
To determine if the Vpsl7p is phosphorylated on serine, threonine, or tyrosine, we performed a phosphoamino acid analysis. In uiuo [32P]orthophosphate-labeled Vpsl7p was immunoprecipitated and extracted from an SDS-polyacrylamide gel. After acid hydrolysis, the amino acid mixture was separated by thin layer electrophoresis in two dimensions (Fig.  8 B ) . Only 32P-labeled phosphoserine could be detected, even after prolonged exposure times. There was no indication of any phosphorylated threonine or tyrosine residues. Therefore, we conclude that the Vpsl7p is phosphorylated on 1 or more serine residues.

Subcellular Fractionation Studies Suggest That the Vpsl7p
Is Associated with a Membrane Fraction-We performed protease protection experiments to test if the Vpsl7 protein enters a membranous compartment in the yeast cell. Wildtype yeast spheroplasts were radioactively labeled and gently lysed with DEAE-dextran under conditions which disrupt the plasma membrane, but not internal organelles (Klionsky and Emr, 1989). Adding proteinase K, we found that Vpsl7p was rapidly degraded, whereas plCPY and pPCPY, which reside in the ER and Golgi compartments, respectively (Franzusoff and Schekman, 1989;Stevens et al., 1982), were resistant. plCPY and p2CPY were degraded only in the presence of Triton X-100 and proteinase K (data not shown). This indicates that the Vpsl7 protein is exposed to the cytoplasm of the yeast cell which is consistent with the observation that the four possible N-linked glycosylation sites of Vpsl7p were not glycosylated. Based on these results, we conclude that the Vpsl7 protein does not enter the secretory pathway.
To determine the intracellular location of the Vpsl7p more precisely, we used differential centrifugation which is outlined in Fig. 9. Yeast spheroplasts (BHY10) were radioactively labeled with Tran3% label and lysed under conditions that do not destroy the integrity of internal organelles (Walworth et al., 1989) (see "Experimental Procedures"). After removing unbroken cells by a brief low speed spin (500 X g ) , this S5 lysate was spun at 13,000 x g, yielding a P13 pellet and a S13 supernatant fraction. The latter was subsequently spun at 100,000 x g to give rise to a PI00 pellet and a SI00 supernatant. The relative levels of Vpsl7p in each fraction were then assayed by quantitative immunoprecipitations. As shown in Fig. lOA, about 95% of Vpsl7p was found in the S13 supernatant fraction, and about 75% of this material pellets after the 100,000 x g spin. The remaining 25% appeared to be soluble, present in the SlOO fraction, together with other soluble cytosolic proteins, e.g. glucose-6-phosphate dehydrogenase (GGPDH) (Fig. lOC). A protein unrelated to Vpsl7p with slower mobility (which is normally only weakly detected by the Vpsl7 antibody) is also seen in the PlOO fraction, presumably because it is highly enriched in this fraction. In previous experiments we have shown that 90% of the ERspecific marker plCPY is found in the P13 pellet fraction (Herman et al., 1991a), and this fraction is also known to be enriched in nuclei, mitochondria, and vacuoles (Goud et al., 1988;Hurt et al., 1988;Walworth et al., 1989;Herman et al., 1991a). The lack of Vpsl7 protein in the P13 pellet fraction indicates that the Vpsl7p was not associated with these yeast organelles. Most of the Vpsl7p (75%) is in the PlOO fraction where most of the Golgi marker enzyme Kex2p (95%) (Cunningham and Wickner, 1989;Redding et al., 1991) fractionated (Fig. 1OC). This raises the interesting possibility that the particulate Vpsl7p might be associated with a Golgi compartment or possibly with small transport vesicles that sediment at 100,000 X g (Walworth et al., 1989). If the Vpsl7 protein interacts with a membranous compartment, then it should be possible to disrupt this interaction with Triton X-100 and 6' LYd# or A A (mluble proteios) (Golgi. small vericla. Uc.) 9. Outline of the subcellular fractionation scheme. Radioactively labeled spheroplasts were osmotically lysed, and unbroken cells were removed by a 500 X g spin. The remaining supernatant ( S 5 ) was subsequently spun for 15 min at 13,000 X g, which resulted in a P13 pellet and a S13 supernatant. The latter was spun at 100,000 X g for 45 Fig. 9. A, spheroplasts of BHYlO were labeled with Tran% label for 15 min, chased for 45 min, and osmotically lysed. The crude extract was spun at 500 X g to clarify the lysate, yielding the S5 supernatant, which was subsequently spun a t 13,000 X g giving rise to the P13 pellet and the S13 supernatant fraction. The latter was then spun at 100,000 X g, which resulted in a SlOO supernatant and a PlOO pellet fraction. Equivalent amounts were taken from each fraction at each step and subjected to immunoprecipitation with antisera specific for Vpsl7p, and the marker proteins GGPDH, ALP, and Kex2p. The migration position of Vpsl7p is indicated. The amount of Vpsl7p in each fraction is stated in % below each lane and is based on three independent experiments. B, equal amounts of the S13 supernatant fraction were treated for 30 min with 2% Triton X-100,l M NaCI, or 2 M urea and subsequently spun at 100,000 X g. The pellet and supernatant fraction were subjected to immunoprecipitations with Vpsl7p-specific antiserum. The migration position of Vpsl7p is indicated. The amount of Vpsl7p in each fraction is stated in % below each lane and is based on three independent experiments. C, the relative levels of Vpsl7p, GGPDH, ALP, and KEX2p in the P13, P100, and SlOO fractions are summarized. The subcellular fractionation analysis was done'as described as in A. The numbers are based on three independent experiments.

FIG.
possibly by other reagents. To test this, S13 supernatant fractions were treated with various reagents (1 M NaCl, 2% Triton X-100, 2 M urea) for 30 min on ice. The lysates were then spun at 100,000 X g for 45 min, and the supernatant and pellet fractions were immunoprecipitated with Vpsl7p-specific antiserum. Triton X-100 and urea treatment extracted most of the Vpsl7 protein from the PlOO pellet, whereas high salt treatment (1 M NaC1) only partially affected the association of Vpsl7p with the particulate PlOO fraction (Fig. 10B).
Additional evidence for a membrane association came from sucrose flotation experiments (Walworth et al., 1989), where we could show that a significant fraction of Vpsl7p fractionated together with other cell membrane markers (e.g. Kex2p) in a sucrose step gradient (60,50,30,0% sucrose from bottom t o top). Together, these findings suggest that the Vpsl7p is associated with a membranous component of the PlOO fraction, since the interaction is sensitive to the detergent Triton X-100 and because the Vpsl7p cofractionated with membranes in the sucrose step gradient analysis. It seems likely that this association is stabilized by hydrophobic interactions as urea treatment efficiently solubilized Vpsl'lp, but high salt treatment released only a portion of Vpsl7p from the particulate PlOO fraction.

DISCUSSION
We have shown earlier that ups17 mutants exhibit severe vacuolar protein sorting defects (Bankaitis et al., 1986;Robinson et al., 1988). To better understand the molecular function of the Vpsl7 protein in vacuolar protein delivery, we have cloned the wild-type VPS17 gene and analyzed its gene product. The gene codes for a hydrophilic protein of 551 amino acids that lacks an N-terminal signal sequence and hydrophobic membrane spanning domains. Protease protection experiments have shown that it is exposed to the cytoplasm, which is also supported by the observation that none of the four possible N-linked glycosylation sites are modified. Some insight into the possible function of the Vpsl7 protein came from subcellular fractionation studies. When wild-type yeast cells were fractionated by differential centrifugation, we found two pools of Vpsl7p, one completely soluble pool (25% of Vpsl7p) and a particulate pool (75%) that sediments at 100,000 x g. Five to eight-fold overexpression of Vpsl7p leads to an increase in the soluble pool which suggests a saturable interaction of Vpsl7p with the 100,000 X g pellet fraction (data not shown). Several lines of evidence indicate that Vpsl7p associates with a membrane. First, 75% of the Vpsl7p is found in the particulate fraction after a 100,000 x g centrifugation. Second, the interaction of Vpsl7p with the particulate fraction is sensitive to Triton X-100. Third, in sucrose density gradients, a significant portion of Vpsl7p fractionates together with cell membranes. Despite the fact that the Vpsl7p is very hydrophilic and has many charged amino acids, we believe that this association is mediated by a hydrophobic interaction, since Vpsl7p can efficiently be extracted from the particulate fraction with urea but not with high salt.
One important question is, what is the nature of this membranous material? Based on the differential centrifugation experiments, we can exclude the possibility that Vpsl7p is associated with mitochondria, nuclei, or vacuoles (Fig. lo), since it is found exclusively in the S13 supernatant fraction and not in the P13 pellet fraction, where most of the vacuolar, mitochondrial, and nuclear marker proteins accumulate. Our data indicate that consistent with a role in Golgi to vacuole protein sorting, Vpsl7p cofractionates with a Golgi marker. The majority of Vpsl7p (75%) pellets at 100,000 x g, together with the late Golgi compartment that contains Kex2 protein Wilcox and Fuller, 1991). This late Golgi compartment is predicted to be the site where vacuolar protein sorting takes place . Because this PlOO fraction also contains small transport vesicles (Walworth et al., 1989), we cannot exclude an interaction with an intermediate in the transport reaction. Based on these results, we assume that the Vpsl7 protein functions at an early stage of the sorting process. Whether it is involved in the actual sorting, packaging, or transport of vacuolar proteins is not yet clear.
In vivo phosphorylation and phosphoamino acid analyses have demonstrated that the Vpsl7p is phosphorylated on serine residue(s). This post-translational modification could account for the difference between the calculated molecular mass, based on the primary protein sequence (63 kDa) and the observed molecular mass on denaturing SDS-polyacrylamide gels (70 kDa). The finding that the Vpsl7 protein is phosphorylated on serine residues is of particular interest because another Vps protein, VpslBp, was shown to encode a novel Ser/Thr protein kinase that acts very early in the delivery pathway of vacuolar proteins (Herman et al., 1991a(Herman et al., , 1991b. Our finding that the Vpsl7p is phosphorylated in a Aupsl5 background, however, indicates that the Vpsl7p may not be a substrate of the Vpsl5 kinase. Further detailed biochemical and mutational studies of the phosphorylation site(s) of Vpsl7p will be required to rigorously rule out a role for Vpslfjp in the phosphorylation of the Vpsl7p. Mutational analyses of the Vpsl7p phosphorylation site(s) also should clarify the importance of this modification.
Gene disruption experiments have shown that a ups17 null mutant is viable, which is not too surprising, given that all ups genes analyzed to date have been found not to be essential for vegetative growth (Dulic and Riezman, 1989;Banta et al., 1990;Herman and Emr, 1990;Raymond et al., 1990;Rothman et al., 1990;Wada et al., 1990;Woolford et al., 1990;Herman et al., 1991a;Preston et al., 1991;Robinson et al., 1991;Paravicini et al., 1992). Unlike several other ups mutants (upsll, 15, 16, 18, 33, and 34) ) that show a temperature-sensitive growth defect, Aupsl7 null mutants are able to grow at elevated temperatures. However, strains deleted for theVPSl7gene do show a defect in vacuolar morphology . Comparing wild-type and Aupsl7 mutant strains, we could demonstrate that the null mutant accumulates many small vacuole-like structures. This vacuolar morphology defect is most likely due to the reduced flow of proteins to the vacuole, and therefore represents a secondary effect of the loss of Vpsl7p function. Although the Aupsl7 mutant does show a clear class B vacuolar morphology, these small vacuolar structures still fulfill many of the functions of an intact vacuole, as demonstrated by the fact that ups17 null mutants are resistant to osmotic stress and can sporulate . Other ups mutants (upsll, 15, 16, 18, 33, and 34) are sensitive to osmotic stress and sporulate poorly as homozygous diploids .
Our detailed analysis of the vacuolar protein sorting defect in the Aupsl7 null mutant strain indicates that the Vpsl7p is only required for the sorting of soluble vacuolar proteins. The sorting and processing of ALP, a vacuolar membrane protein, was not blocked in the mutant cells, whereas sorting and processing of the soluble vacuolar protein CPY was almost completely blocked. More than 95% of the Golgi-modified precursor form of CPY was missorted and secreted by the ups1 7 mutant. The processing and sorting defects for two other soluble vacuolar hydrolases, PrA and PrB, were partially affected (60% mature, 40% precursor). The observation that ALP processing is completely independent of Vpsl7p is consistent with previous observations with other ups mutants. Since ups35 and ups15 mutant cells also do not affect ALP sorting and/or maturation (Herman et al., 1991b;Paravicini et al., 1992), it raises the question of whether there are different delivery pathways to the yeast vacuole. In this respect, it is interesting to note that there is also more than one delivery pathway to the mammalian lysosome, a mannose 6phosphate-dependent pathway, used by many soluble lysosomal enzymes, and a mannose 6-phosphate-independent pathway, followed by several lysosomal membrane markers (von Figura and Hasilik, 1986;Kornfeld, 1987;Kornfeld and Mellman, 1989). In yeast, however, analyses of other vacuolar membrane prodeins as they are identified, will need to be examined to obtain further evidence that vacuolar membrane proteins travel to the vacuole by a Vpsl7p-independent path-way. Independent pathways do not necessarily mean that different vacuolar proteins travel to the vacuole by different transport vesicles, although we cannot exclude this possibility. However, because a number of ups mutants appear to mislocalize CPY, PrA, PrB, and ALP to the same extent Klionsky and Emr, 1989); completely independent delivery pathways seem to be unlikely. Based on the fact that the sorting signals in CPY (Johnson et al., 1987;Valls et al., 1987), PrA (Klionsky et al., 1988), and ALP (Klionsky and Emr, 1989) appear to be different, as they do not show any primary sequence similarity, these proteins may travel to the vacuole via the same vesicle but bind to different membrane receptor sites in the sorting compartment. Because the Vpsl7p does not resemble a membrane receptor, we propose that the Vpsl7 protein may associate directly or indirectly with a receptor complex(es) that participates in CPY, PrA, and PrB sorting to the vacuole. Receptor function is compromised in the absence of Vpsl7p such that CPY sorting is completely defective while PrA and PrB sorting is only partially blocked. Biochemical purification of the Vpsl7 protein together with any associated proteins or protein complexes should enable us to address this model.