The yeast VPS5/GRD2 gene encodes a sorting nexin-1-like protein required for localizing membrane proteins to the late Golgi

The Golgi apparatus is a major intersection point for secretory and endocytic trafficking pathways and is involved in both post-transational modifications of proteins and protein sorting. The Golgi apparatus consists of a set of subcompartments (cisternae) each containing their own unique complement of resident proteins. The analysis of yeast sec mutant strains defective in discrete transport steps within the Golgi, as well as subcellular fractionation of Golgi enzymes, suggest that the yeast Golgi can be separated into at least three distinct compartments: the early, mid, and late Golgi (Cunningham and Wickner, 1989; Franzusoff and Schekman, 1989; Graham and Emr, 1991). Because of its roles in proteolytic processing of newly synthesized proteins and in protein sorting, the late Golgi of the yeast Saccharomyces cerevisiae appears to be the functional equivalent to the animal cell trans-Golgi network that is distal to the cis, medial, and trans-Golgi cisternae.

Multiple protein sorting events have been identified that involve the yeast late Golgi apparatus. For example, soluble vacuolar proteins such as carboxypeptidase Y (CPY) are sorted away from secretory proteins in the late Golgi apparatus because they contain vacuolar targeting signals and enter a pathway that leads to the vacuole. Over 45 VPS (vacuolar protein sorting) and PEP (peptidase deficient) genes are known to be required for delivery of CPY to the vacuole (Bankaitis et al., 1986; Jones, 1977; Rothman and Stevens, 1986). Mutations in the VPS genes cause CPY to be aberrantly secreted. Many of the VPS genes are thought to be directly involved in the initial sorting step itself. For example, the VPS10 gene has been shown to be an integral membrane protein that binds to CPY in the late Golgi compartment in a manner dependent on the vacuolar targeting signal of CPY (Cooper and Stevens, 1996; Marcusson et al., 1994). By analogy to the sorting of lysosomal hydrolases by the mannose 6-phosphate receptor (Kornfeld and Mellman, 1989), Vps10p-CPY complexes in the late Golgi are thought to be transported via vesicles to a prevacuolar compartment where CPY dissociates from the receptor. Vps10p would then cycle back to the late Golgi for several more rounds of sorting. Consistent with this idea, Vps10p is synthesized at a rate 20-fold lower than that of its ligand CPY and the ligand is bound to the receptor in a 1:1 stoichiometry (Cooper and Stevens, 1996). Mutation of a specific tyrosine-based targeting signal in its cytosolic domain causes Vps10p to be rapidly degraded in the vacuole (Cereghino et al., 1995; Cooper and Stevens, 1996) indicating that this signal is required for cycling.

Another protein sorting mechanism ensures that the late Golgi retains its resident membrane proteins while allowing transient membrane proteins to pass on to the plasma membrane or vacuole. While this process is poorly under-stood, proper localization of two of the resident membrane proteins, dipeptidyl aminopeptidase (DPAP) A and Kex2p, is known to be dependent on aromatic acid-based sorting signals in their cytosolic domains (Nothwehr et al., 1993; Wilcox et al., 1992). When these signals are removed by point mutations the resulting mutant late Golgi membrane proteins are mislocalized to the vacuolar membrane. In addition there is evidence indicating that Golgi localization may be achieved, at least in part, due to retrieval of DPAP A and Kex2p from a prevacuolar compartment (for reviews see Nothwehr and Stevens, 1994; Wilsbach and Payne, 1993a) that was identified by the analysis of cargo proteins during cell fractionation experiments (Schimmöller and Riezman, 1993; Singer and Riezman, 1990; Vida et al., 1993). While little is known of the machinery involved in the localization of these membrane proteins, the development of a fusion protein reporter construct (A-ALP) has enabled a genetic analysis of Golgi localization. This Golgi-localized fusion protein consists of the cytosolic domain of DPAP A fused to the transmembrane and lumenal domains of alkaline phosphatase (ALP). In Golgi retention defective mutant strains A-ALP is mislocalized to the vacuole where it becomes enzymatically active due to its proteolytic processing in the vacuole (Nothwehr et al., 1993). A genetic screen for such mutants (named grd for Golgi retention defective) recently identified 18 complementation groups (Nothwehr et al., 1996). A subset of the GRD genes are also required for CPY sorting and are allelic to previously identified VPS genes, while others appear to only be involved in localization of resident Golgi membrane proteins. Given that proper localization of Vps10p and resident late Golgi membrane proteins are dependent on similar types of aromatic amino acid-based signals and trans-acting factors, the two types of proteins may share a common membrane trafficking pathway between the late Golgi and prevacuolar compartment.

In this study we have identified and characterized a gene required for both Golgi protein localization and CPY sorting, VPS5/GRD2. This gene encodes a hydrophilic protein similar to a human protein, sorting nexin-1 (SNX1), that interacts with the cytosolic domain of the epidermal growth factor (EGF) receptor and thereby regulates its intracellular trafficking. Mutations in VPS5 caused A-ALP, Kex2p, and Vps10p to be mislocalized to the vacuole. vps5 mutants also exhibited a strong CPY sorting defect presumably due to a failure to maintain Vps10p in the Golgi apparatus. A portion of Vps5 protein localizes to an endosome-like organelle that accumulates in class E vps mutants indicating that Vps5p may be required for recycling of Golgi membrane proteins.

Restriction enzymes and other enzymes used in subcloning procedures were obtained from New England Biolabs (Beverly, MA). [³⁵S]express label and [α-³⁵S]dATP were purchased from New England Nuclear (Boston, MA). Oxalyticase was obtained from Enzogenetics (Corvallis, OR). All secondary antibodies used from immunofluorescence experiments were from Jackson Immuno Research Labs. Inc (West Grove, PA). Other reagents were obtained from Sigma Chemical Co. (St Louis, MO) or as indicated.

Plasmids and yeast strains used in this study are indicated in Tables 1 and 2, respectively. The 2.75 kbp AatII-XhoI GRD2/VPS5 complementing fragment used to construct pSN289, pAH25, pAH26, pAH31, and pAH33 was initially isolated from pC-F1-1, a YCp50-based plasmid isolated from a yeast genomic library (Rose et al., 1987). The VPS5 gene replacement construct pAH25 was constructed by subcloning the 1.91 kbp PvuII-EcoRI HIS3 gene fragment from pJJ215 into the NcoI (blunted with Klenow enzyme) and EcoRI sites of pSN289. Disruption of VPS5 was accomplished by transforming XmnI/SnaBI digested pAH25 into wild-type yeast strains SNY17 and SNY36-9A, selecting for His⁺ prototrophs, and screening for the grd2Δ::HIS3 mutation by PCR Southern analysis.

In order to confirm that the cloned gene was in fact allelic to grd2-1, a yeast strain (AHY40) containing the gene disruption targeted by pAH25 was mated with strain SNY47-7D containing the grd2-1 allele (Table 1). The resulting diploid was found to exhibit a Grd^-phenotype. When the diploid was sporulated, and 19 tetrads dissected, it was observed that all four spores in every tetrad exhibited the Grd^-phenotype strongly arguing that the cloned gene is in fact GRD2. The same type of approach was used to determine whether GRD2 and VPS5 were the same gene. Strain AHY40 was mated with a vps5 mutant strain SEY6210-05149-05 (Robinson et al., 1988) and again a Grd^- phenotype was observed. Sporulation and dissection of 15 tetrads again showed that all four spores in every tetrad exhibited the Grd^- phenotype. Thus we conclude that GRD2 and VPS5 are allelic.

To epitope tag the VPS5 gene, a BglII site was introduced by site directed mutagenesis just 3’ of the initiator Met codon resulting in the Asn codon at position 6 being changed to a Leu codon. The resulting plasmid, was digested with BglII and an oligo duplex encoding the 9 amino acid influenza virus hemagglutinin (HA) epitope (YPYDVPDYA), and containing BglII-compatible cohesive ends, was inserted to construct pAH31. Immunoblots carried out on extracts from a strain carrying a low copy number plasmid containing the tagged allele (pAH31) using an anti-HA epitope antibody (see below) detected a single band at 95 kDa. When this construct was placed on a high-copy number plasmid (resulting in pAH33) the band increased in intensity while no corresponding band was detected in an extract made from a control strain lacking the VPS5-HA construct. The correlation of the abundance of the 95 kDa protein with dosage of the VPS5-HA allele demonstrated that the anti-HA epitope antibody does in fact detect epitope-tagged Vps5p. A VPS5 allele containing 3 copies of the HA epitope was constructed by subcloning a 126 bp BglII fragment from pSM492 (Tyers et al., 1993) into the BglII site described above resulting in plasmid pAH37.

Double-stranded sequencing was carried out employing the dideoxy chain termination method (Sanger et al., 1977). Database searches performed using the BLAST service of the National Center for Biotechnology Information showed that the VPS5/GRD2 gene region was sequenced by the Yeast Genome Sequencing Project (open reading frame YOR069W). A one nucleotide discrepancy was noted between our sequencing data for VPS5 (GenBank accession no. U73512) and the Genome Project data for YOR069W. As a result, the reported YOR069W open reading frame is much shorter than the VPS5 ORF. Sequence reported for the ALG8 gene (GenBank accession no. X75929), that flanks VPS5, agrees with our sequence data at the position in question. Multiple sequence alignments were performed using software from DNA Star Inc.

The procedure for immunoprecipitation of CPY was performed using a rabbit antibody against CPY as described previously (Vater et al., 1992). Likewise, immunoprecipitations of A-ALP, Vps10p, and Kex2p were performed using the procedure previously described for A-ALP (Nothwehr et al., 1993) and the following antibodies: rabbit anti-ALP serum (Nothwehr et al., 1996), rabbit-anti Vps10p serum (Cooper and Stevens, 1996), and rabbit anti-Kex2p serum (a generous gift from G. Payne). Gels were quantified using a Phosphorimager system (Fuji Photo Film Co.).

For subcellular fractionation of cells expressing epitope-tagged Vps5p, 50 absorbence units of strain AHY24 carrying plasmid pAH31 were harvested, washed with 10 ml of H₂O, and incubated at 30°C for 30 minutes in 15 ml of spheroplast buffer (50 mM Tris-HCl, pH 7.5, 1.4 M sorbitol, 1 mM MgCl₂, 10 mM NaN₃). The spheroplasts were then pelleted at 450 g and washed twice with 10 ml portions of 1.2 M sorbitol. The cell pellet was then resuspended in 10 ml of icecold lysis buffer (25 mM sodium phosphate, pH 7.4, 200 mM mannitol, 1 mM MgCl₂, 50 mM PMSF, 100 μg/ml leupeptin, and 100 μg/ml pepstatin A) and was incubated on ice for 20 minutes. The lysate was then centrifuged at 450 g and the supernatant (lysate fraction) was then centrifuged at 15,000 g for 10 minutes to generate supernatant (S15) and pellet (P15) fractions. The S15 fraction was then centrifuged at 200,000 g for 1 hour to generate supernatant (S200) and pellet (P200) fractions.

Equal absorbence equivalents of the lysate, P15, P200, and S200 were then loaded on SDS-PAGE gels and electroblotted. Blots were probed with the following antibodies: (1) the 12CA5 mouse monoclonal antibody against the HA epitope (Babco Inc., Berkeley, CA); (2) a rabbit antibody against DPAP A (Roberts et al., 1992); (3) a rabbit antibody against phosphoglycerol kinase (PGK); (4) a rabbit antibody against Vph1p (Hill and Stevens, 1994). Immunoblots were visualized using a chemiluminescent detection system from Amersham Life Science, Inc.

Colony blots of secreted CPY were performed as previously described (Roberts et al., 1991). Briefly, cells were applied to the surface of an agar plate, overlaid with nitrocellulose, and incubated for 14 hours at 30°C. The filters were then rinsed of cells using dH₂O and were processed as described above using an anti-CPY antibody.

The procedures for preparation of fixed spheroplasted yeast cells, attachment to microscope slides, and costaining of the A-ALP fusion protein and the 60 kDa subunit of the vacuolar proton-translocating ATPase (V-ATPase) using an anti-ALP polyclonal antibody and anti-60 kDa monoclonal antibody 13D11-B2 (Molecular Probes Inc., Eugene, OR) were previously described (Nothwehr et al., 1995; Roberts et al., 1991). Labeling of yeast vacuoles with 5(6)-carboxy-2’,7’-dichlorofluorescein diacetate (CDCFDA) was performed as described previously (Roberts et al., 1991). Yeast cells were photographed using an Olympus BX-60 fluorescence microscope (Olympus Co., Lake Success, NY).

The yeast grd mutants were identified by their failure to retain A-ALP (Nothwehr et al., 1996). One of the two grd2 alleles, grd2-1, was found to exhibit a moderate growth defect at 37°C that was tightly linked to the Golgi retention defect. The growth defect of grd2-1 was exacerbated when combined with a null allele of the PEP4 gene that encodes the vacuolar protease, protease A (Ammerer et al., 1986), whereas thepep4Δ mutation caused no apparent decrease in growth rate in an otherwise wildtype strain. The basis for the temperature sensitive growth phenotype of strains carrying the grd2-1 allele and the reason for the exaggeration of this phenotype when grd2-1 was combined with the pep4Δ mutation are not known, however, slow growth at 37°C is not a general consequence of a loss of GRD2 function since null mutants grew normally at 37°C (see below). To clone GRD2 we transformed a grd2-1 pep4Δ strain (SNY84-7D) with a yeast genomic library made in a centromere-based (CEN) plasmid (Rose et al., 1987) and screened for transformants that grew normally at 37°C and exhibited normal Golgi localization of A-ALP (Nothwehr et al., 1993). The library plasmid pC-F1-1 was rescued from one of the transformants that grew normally at 37°C. This plasmid was transformed back into strain SNY84-7D and was found to complement both the growth defect at 37°C and the A-ALP localization defect. The insert was mapped by subcloning portions of the ∼10 kbp genomic insert from pC-F1-1 into the low copy number plasmid, pRS316, and the resulting plasmids were assessed for complementation of the grd2-1 mutation. By this approach the complementing activity was refined to a 2.75 kbp AatII-XhoI fragment (Fig. 1A). Genetic linkage analysis was used to demonstrate that the cloned gene that complements the grd2-1 mutation corresponded to the GRD2 locus (see Materials and Methods).

Complementation analysis between the grd and vps mutants did not previously detect allelism between grd2-1 and any of the vps mutants (Nothwehr et al., 1996). However, a more recent complementation analysis carried out using a yeast strain carrying a null allele of GRD2 (see below) did detect a lack of complementation between the grd2Δ and vps5 mutant strains. Genetic linkage experiments demonstrated that GRD2 and VPS5 are the same gene (see Materials and Methods). We will hereafter refer to this gene as VPS5.

The 2.75 kbp AatII-Xho I complementing fragment (Fig. 1A) contained a single open reading frame that overlapped with the YOR069W open reading frame on chromosome XV identified by the yeast genome sequencing project (for details see Materials and Methods). This open reading frame identified for VPS5 (GenBank accession no. U73512) is predicted to encode a 675 amino acid hydrophilic protein of molecular mass 76.5 kDa that contains neither a signal sequence nor a potential transmembrane domain (Fig. 1B). GenBank database searches revealed homology between Vps5p and a family of proteins including human SNX1 as well as the yeast Ykr078W and Mvp1 proteins. SNX1 influences trafficking of the EGF receptor by directly interacting with its cytosolic domain and has been noted to share identity with Mvp1p (Kurten et al., 1996). Ykr078Wp is a putative yeast protein of unknown function. Mvp1p, required for sorting of CPY to the vacuole, was identified as a multi-copy suppressor of a dominant allele of Vps1p (Ekena and Stevens, 1995), a dynamin-like protein required for retention of Golgi membrane proteins and CPY sorting (Nothwehr et al., 1995; Vater et al., 1992; Wilsbach and Payne, 1993b). Identity between Vps5p, Ykr078Wp, and SNX1 was somewhat more extensive in the C-terminal regions of the proteins: 28% identity over 356 residues between Vps5p (residues 319 to 675) and SNX1 (Fig. 2B). While the level of similarity is low, blocks of identical amino acids are distributed throughout the entire length of the proteins and several blocks are shared by all three proteins.

To determine the phenotypic consequences of a complete loss of VPS5 gene function, VPS5 was disrupted in haploid cells by replacement of its open reading frame with a fragment containing the HIS3 gene (see Fig. 1A and Materials and Methods). The vps5Δ strains exhibited essentially normal growth at temperatures between 22°C and 37°C (data not shown).

Sorting of the soluble vacuolar hydrolase CPY was examined in strains carrying the vps5Δ allele. Strain AHY41 (vps5Δ) was transformed with either a CEN-based plasmid containing VPS5 or with vector alone. The vps5Δ strain carrying vector alone (pRS316) exhibited a dramatic CPY sorting defect (Fig. 2A) whereas a wild-type strain carrying pRS316 secretes little or no CPY as assessed using a colony immunoblotting procedure. The CPY sorting defect of the vps5Δ strain was complemented by the CEN-based plasmid containing VPS5.

Newly synthesized CPY is modified with core oligosaccharides in the endoplasmic reticulum that are extended by the addition of mannose residues in the Golgi apparatus (Stevens et al., 1982). The Golgi-modified form (p2CPY) is then transported to the vacuole where it is proteolytically modified to the mature form (mCPY). In order to analyze the posttranslational modifications of CPY in the vps5Δ strain, and to more quantitatively measure the extent of CPY secretion, CPY was immunoprecipitated from cultures labeled with a mixture of [³⁵S]methionine and [³⁵S]cysteine (Fig. 2B). After a 10 minute pulse and 45 minute chase essentially all of the CPY synthesized by the wildtype strain was intracellular and present as the mature form (lane 1). In contrast, quantification of the gel shown in Fig. 2B revealed that 80% of the CPY synthesized by the vps5Δ strain was secreted (Fig. 2B, lane 4). Most of the intracellular CPY was found as the Golgi-modified precursor form (p2CPY; Fig. 2B, lane 3). Therefore, >90% of the CPY synthesized in vps5Δ cells fails to reach the vacuole within the 45 minute chase period.

Cells containing a VPS5 null allele also exhibited a vacuolar morphology defect. The vps mutants have been divided into five distinct classes based on morphological studies of strains carrying mutant alleles of VPS genes (Raymond et al., 1992a); vps5 mutants had class B vacuolar morphology characterized by highly fragmented vacuoles. We assessed the vacuolar morphology of cells carrying a complete deletion of VPS5 by immunofluorescence microscopy (Fig. 3). Whereas the isogenic wild-type control cells exhibited two or three spherical vacuolar structures per cell (Fig. 3B) the vps5Δ exhibited few if any vacuolar structures of normal size (Fig. 3D). Consistent with the defect defined for the class B mutants (Raymond et al., 1992a), the vacuolar structures in vps5Δ cells appear to have fragmented into minivacuoles dispersed throughout the cytoplasm. In addition, analysis of unfixed cells stained with the vacuole-specific vital dye CDCFDA also demonstrated the fragmented nature of vacuoles in vps5Δ cells (data not shown).

The ability of cells lacking Vps5p function to retain the resident late Golgi membrane proteins A-ALP and Kex2p was investigated. A loss of retention of the model late Golgi membrane protein A-ALP results in its mislocalization to the vacuole (Nothwehr et al., 1993), where the A-ALP lumenal domain is removed by a vacuole-specific protease. Similarly, loss of retention of Kex2p can be measured by the rate of its vacuolar protease-dependent degradation (Wilcox et al., 1992).

We measured the kinetics of vacuolar processing/degradation of these two proteins in vps5Δ cells by pulse-chase and immunoprecipitation. As previously reported (Nothwehr et al., 1993) little or no processing of A-ALP was observed in wild-type cells after a 10 minute pulse and 180 minute chase (Fig. 4A). However, the vps5Δ cells exhibited processing at the 20 minute time point with a half-time of processing of 50 minutes. This rate of delivery of A-ALP to the vacuole in grd2Δ/vps5Δ cells is the fastest yet documented amongst the grd mutants (Nothwehr et al., 1996) and thus indicates a severe A-ALP localization defect.

Immunoprecipitation of Kex2p from wild-type cells showed a slow but steady decline in Kex2p levels over a 120 minute period (Fig. 4B), a result similar to the previously reported 80 minute half-time of Kex2p degradation in wild-type cells (Wilcox et al., 1992). In contrast, degradation of Kex2p in the vps5Δ strain occurred much more rapidly and was essentially complete by the 90 minute time point (Fig. 4B). Both the processing of A-ALP and degradation of Kex2p observed in vps5Δ cells was dependent upon the PEP4 gene product, protease A (Fig. 4A,B, right hand panels). These results strongly argue that in the absence of Vps5p function both A-ALP and Kex2p fail to be retained in the Golgi and reach vacuolar compartments where they are acted upon by vacuolar proteases.

As an alternative approach to determine the destination of A-ALP in vps5Δ cells, wild-type and mutant cells were analyzed by indirect immunofluorescence microscopy with antibodies to A-ALP and Vma2p, a marker for the vacuolar membrane (Fig. 5). In wild-type cells A-ALP exhibits a cytoplasmic punctate staining pattern characteristic of the late Golgi apparatus (Nothwehr et al., 1993; Redding et al., 1991) that is clearly non-vacuolar (compare Fig. 5B and C). Comparison of the staining pattern of the vacuolar marker between the wild-type and mutant cells (Fig. 5C and F) indicates severe fragmentation of vacuoles in vps5Δ cells as shown in Fig. 3. Many but not all of the structures containing A-ALP in vps5Δ cells (Fig. 5E) colocalize with the fragmented vacuoles (Fig. 5F) indicating that A-ALP has reached the vacuole in these cells. The identity of the structures containing A-ALP that do not co-localize with the vacuolar membrane marker are not known but could correspond to vesicles, endsomes, or Golgi.

The observation that Vps5p is required for localization of resident late Golgi membrane proteins suggested a possible explanation for the defect in CPY sorting observed in vps5Δ cells. The sorting receptor for CPY, Vps10p, is known to bind to CPY in a late Golgi compartment and facilitate the entry of CPY into a pathway that ultimately leads to the vacuole. Therefore, it was possible that vps5Δ cells secrete CPY because of a failure to maintain the Golgi localization of Vps10p.

A loss of Golgi localization of Vps10p causes it to be mislocalized to the vacuolar membrane where it is degraded to a discrete faster migrating form (Cereghino et al., 1995; Cooper and Stevens, 1996). Immunoprecipitation of Vps10p from wild-type cells indicated that Vps10p remained as the intact Golgi form for at least 180 minutes chase (Fig. 4C). In contrast, degradation of Vps10p in vps5Δ cells occurred very rapidly with a half-time of less than 40 minutes. No degradation was observed in the vps5Δ pep4Δ double mutant indicating that proteolysis observed in vps5Δ cells occurred in the vacuole (Fig. 4C, right hand panel). These results indicate that the CPY secretion phenotype observed in vps5Δ cells is likely to be a secondary consequence of a failure to maintain localization of the CPY receptor in the late Golgi.

Cell fractionation studies were used to characterize the sub-cellular localization of Vps5p. To enable detection of Vps5p the 9 amino acid hemagglutinin (HA) epitope tag was introduced 6 codons after the initiator methionine codon of the VPS5 open reading frame. The introduction of the epitope tag did not disrupt function of Vps5p since the tagged allele (VPS5-HA) was able to complement the CPY sorting defect of the vps5Δ mutant strain (Fig. 2A). The epitope-tagged Vps5p protein was detected in whole cell extracts by immunoblot analysis as a 95 kD protein (for details see Materials and Methods). The apparent size of Vps5-HA was larger than expected considering the predicted sizes of the VPS5 orf (76.5 kDa) and the HA epitope (∼1.3 kDa).

Cell lysates were prepared from wild-type strain AHY24 carrying a low copy number CEN plasmid containing the VPS5-HA allele. Gentle lysis was achieved by resuspending sphero-plasts in hypotonic buffer and lysates were then subjected to centrifugation at 15,000 g to generate pellet (P15) and supernatant (S15) fractions. The S15 fraction was then centrifuged at 200,000 g to generate a slowly sedimenting microsome fraction (P200) and soluble protein supernatant fraction (S200). Fractions were then subjected to immunoblot analysis as shown in Fig. 6. The vast majority of the vacuolar membranes, marked by Vph1p (Manolson et al., 1992), were recovered in the P15 fraction whereas most of the late Golgi membranes, marked by DPAP A (Nothwehr et al., 1993; Roberts et al., 1992), fractionated in the P200 fraction. The soluble cytosolic marker PGK remained in the 200,000 g supernatant. These results are consistent with those from similar fractionation schemes (Marcusson et al., 1994). The majority of the Vps5p fractionated with the cytosol as soluble protein. However, ∼16% of Vps5p fractionated in the P200 fraction which contains Golgi membranes, vesicles, and possibly endosomal compartments. These results indicate that Vps5p may cycle between a soluble form and a form associated with a slowly sedimenting organelle.

To learn more about the nature of the organelle that Vps5p associates with, Vps5p was localized by indirect immunoflu-orescence microscopy. Successful application of this technique required expression of a VPS5 allele containing three tandemly repeated copies of the HA tag from a multicopy vector containing the 2μ origin of replication. The triple-tagged VPS5 allele was found to rescue the vps5Δ phenotype (data not shown) demonstrating that the addition of the tag did not prevent normal function of the encoded protein. In wild-type cells HA-tagged Vps5p exhibited a predominantly cytoplasmic diffuse staining pattern with the occasional appearance of faint punctate structures (Fig. 7B) consistent with the observation that the majority of Vps5p is soluble (Fig. 6).

The localization of Vps5p was also analyzed in the class E vps mutant, vps27, cells since the class E mutants have been shown to accumulate a novel prevacuolar, late endosomal organelle (Davis et al., 1993; Piper et al., 1995; Rieder et al., 1996). Studies conducted on a temperature sensitive allele of vps27 have shown that accumulation of the endosome-like organelle is due to defects in the rate of membrane trafficking from the organelle both to the vacuole and to the late Golgi compartment (Piper et al., 1995; Raymond et al., 1992a). In vps27Δ cells HA-tagged Vps5p localized to 1-2 discrete structures per cell that were positioned adjacent to the vacuole (Fig. 7E). These structures corresponded to class E organelles since they also contain the V-ATPase (Fig. 7F; Raymond et al., 1992a). The association of Vps5p with the class E organelle, taken together with the role of Vps5p in localization of late Golgi membrane proteins, supports the notion that Vps5p may carry out a role in cycling of membrane proteins from a late endosomal compartment back to the late Golgi.

Another vps gene product, Vps1p, has been previously shown to mislocalize A-ALP and Kex2p to the vacuolar membrane (Nothwehr et al., 1995; Wilsbach and Payne, 1993b). However, A-ALP reaches the vacuole in vps1Δ cells via initial delivery to the plasma membrane and subsequent uptake by the endocytic pathway, supporting the proposal that Vps1p performs a function at the late Golgi that is necessary for Golgi retention of A-ALP (Nothwehr et al., 1995).

If Vps5p is involved in retrieval of A-ALP from a prevacuolar compartment as the results above suggest, A-ALP should reach the vacuole in vps5Δ cells in a manner independent of the early endocytic pathway. To test this idea the vps5Δ mutation was combined with a rapid-onset temperature-sensitive mutant allele of END4, a gene required for the internalization step of endocy-tosis (Raths et al., 1993). A-ALP was immunoprecipitated from the resulting vps5Δ end4-ts double mutant strain that had been radioactively pulsed and chased at the non-permissive temperature of 36°C (Fig. 8). Substantial processing of A-ALP was seen after 60 minutes chase in the vps5Δ end4-ts double mutant (lane 8) the extent of which was similar to that observed for the vps5Δ single mutant (lane 4). The extent of A-ALP processing observed for both the vps5Δand the vps5Δend4-ts strains after 60 minutes chase was somewhat less than that shown in Fig. 4 for a vps5Δ strain, presumably due to a general negative effect of the elevated temperature on processing or transport kinetics. The conditions of this experiment clearly impose a block of the early endocytic pathway in the strains containing the end4-ts allele since an end4-ts vps1Δ strain exhibited a complete block in A-ALP vacuolar processing (data not shown) as previously reported (Nothwehr et al., 1995). Under these same conditions a vps1Δ strain with the wild-type END4 allele exhibits extensive processing of A-ALP after 60 minutes chase (data not shown; Nothwehr et al., 1995). These results indicate that Vps5p acts at a step distinct from Vps1p and are consistent with the model that Vps5p functions in retrieval from a prevacuolar compartment.

The process of membrane protein retention and protein sorting in the Golgi apparatus is a poorly understood and complex problem. Characterization of genes required for these processes is beginning to shed light on the mechanisms involved. Here we report the identification and characterization of the VPS5 gene and provide insight into the function of the Vps5 protein.

VPS5 is a nonessential gene that encodes a hydrophilic protein with a calculated molecular mass of 76.5 kDa. Epitope tagging of Vps5p was used to show that Vps5p has an apparent molecular mass of ∼94 kDa. Discrepancy between the predicted and apparent molecular masses of proteins is not uncommon (Novick et al., 1989; Sorger and Pelham, 1988). The lack of any recognizable signal sequence or transmembrane domains in the Vps5p sequence, coupled with the observation that most of Vps5p fractionates in the cytosolic fraction, strongly argues that Vps5p does not reach the lumen of secretory pathway organelles where it could be glycosylated. While it is not known if Vps5p acquires other posttranslational modifications that might decrease its SDS-PAGE mobility, it is of interest to note that several Vps proteins are phosphorylated including the product of another class B vps gene, VPS17 (Köhrer and Emr, 1993).

The resident Golgi membrane proteins Kex2p and A-ALP as well as the CPY receptor, Vps10p, were observed to be rapidly transported to the vacuole in vps5Δcells. The CPY missorting phenotype in vps5Δ cells is likely an indirect consequence of a lack of proper Vps10p trafficking and localization. Likewise, the rapid vacuolar degradation of Kex2p is consistent with the observation that vps5/grd2 mutant cells secreted unprocessed α-factor (Nothwehr et al., 1996), a mating pheromone normally processed by Kex2p in the late Golgi (Fuller et al., 1988). The most straightforward interpretation of these data is that Kex2p, A-ALP, and Vps10p normally cycle between the late Golgi and a late endosomal compartment and this cycling is disrupted in the vps5Δmutant.

The subcellular localization of Vps5p, as well as the analysis of the pathway by which A-ALP reaches the vacuole in vps5Δ cells, provide a framework for a model regarding the role of Vps5p in late Golgi membrane protein localization. About 16% of Vps5p was associated with a slowly sedimenting non-vacuolar structure. More importantly, Vps5p was found associated with the exaggerated endosome-like organelle present in the class E mutant vps27. Membrane proteins from both the late Golgi and the plasma membrane accumulate in class E compartments, and thus the organelle appears to be an intersection point for both the Golgi to vacuole and the endocytic pathways (Davis et al., 1993; Raymond et al., 1992b). While it is possible that Vps5p may have aberrantly associated with class E compartments due to its overproduction, this seems unlikely since Vps5p did not exhibit significant staining of other organelles and thus specifically interacted with the exaggerated endosomal structures. Delivery of A-ALP to the vacuole in cells lacking Vps5p was not affected by a mutation that blocked an early step in endocytosis indicating that A-ALP did not reach the vacuole via the plasma membrane. In contrast, A-ALP reaches the vacuole in cells lacking Vps1p via the plasma membrane, consistent with a proposed role for Vps1p in formation of vesicles from the late Golgi that enter the Golgi to vacuole pathway (Nothwehr et al., 1995). These data indicate that Vps1p and Vps5p act at different membrane trafficking steps. Taken together the data support a model in which Vps5p functions in retrieval of Golgi membrane proteins from a prevacuolar/late endosomal compartment back to the late Golgi. In the absence of Vps5p function Golgi membrane proteins are not able to cycle from the prevacuole back to the late Golgi and instead are directly transported to the vacuole via a default pathway.

The fragmented vacuoles in cells lacking Vps5p function fulfill many of the functions of vacuoles from wild-type cells. For example, both ALP and the V-ATPase localized to vacuoles in vps5Δ cells (Figs 3 and 5) and thus there do not appear to be any gross defects in trafficking of membrane proteins to the vacuole. Furthermore, the propeptide of newly synthesized ALP is processed normally in vps5 mutant cells indicating delivery to the vacuole (S. Nothwehr, unpublished data). Quinacrine staining in the class B vps mutants such as vps5 suggests that they are able to assemble active V-ATPase (Banta et al., 1988; Raymond et al., 1992a). Other processes requiring normal vacuole function can occur in vps5 mutants, such as sporulation and resistance to osmotic stress (Banta et al., 1988).

The underlying reason for the vacuolar morphology defect in vps5 mutant cells is not known but a role of Vps5p in vacuolar morphology is presumably indirect since Vps5p was not detected in association with vacuolar membranes. The morphology defect could arise from a reduced flow of soluble proteins to the vacuole, some of which could be important for maintenance of vacuole structure. This model would predict that an abrupt loss of Vps5p function would initially result in a breakdown in cycling of Golgi membrane proteins, such as Vps10p, and the vacuolar morphology defect would appear later. The test of such a model awaits the generation of a vps5 conditional allele.

Vps5p is homologous to human SNX1 and yeast Mvp1p. SNX1 interacts with the cytosolic domain of the EGF receptor in a functionally meaningful manner since overproduction of SNX1 dramatically increased the rate of internalization and lysosomal degradation of the receptor (Kurten et al., 1996). The function of Mvp1p may be related to that of Vps5p and SNX1 since it is also required for a protein sorting event: the sorting of CPY to the vacuole (Ekena and Stevens, 1995). However, its role is distinct from that of Vps5p since little if any destablization of the CPY receptor Vps10p has been observed in mvp1 mutants (L. Conibear and T. Stevens, personal communication) and mvp1 mutants have only a very modest effect on A-ALP localization (S. Nothwehr, unpublished). MVP1 genetically interacts with Vps1p, a yeast dynamin homolog (Ekena and Stevens, 1995). However, no such genetic interactions between VPS5 and VPS1 have been observed (S. Nothwehr, unpublished). In contrast, the relationship of Vps5p function to that of SNX1 may be more direct since both proteins are clearly involved in membrane protein sorting. The interaction of SNX1 with the EGF receptor cytosolic domain suggests the exciting possibility that Vps5p may mediate sorting of yeast Golgi membrane proteins via direct interaction with their cytosolic domains.

We acknowledge the work of Jeff Ahrens during the initial stages of this study. We thank Liz Conibear, Stephen Alexander, and John Rogers for their critical comments on the manuscript and Tom Stevens for antibodies. This work was supported by grants from the University of Missouri Research Board (RB 95-012) and the National Institutes of Health (GM 53449) that were awarded to S.F.N.