Identification of Rgp1p, a novel Golgi recycling factor, as a protein required for efficient localization of yeast casein kinase 1 to the plasma membrane

Characterization of casein kinase I (CKI) in the budding yeast Saccharomyces cerevisiae has demonstrated that this activity is required for normal cell growth (Hoekstra et al., 1991; Robinson et al., 1992; Wang et al., 1992; Wang et al., 1996). YCK1 and YCK2 encode functionally redundant, plasma membrane-localized CKI proteins (Robinson et al., 1992; Wang et al., 1992; Vancura et al., 1994). Deletion of both YCK1 and YCK2 results in aberrant cellular morphology and loss of viability (Robinson et al., 1992; Wang et al., 1992). A yck^ts strain, which has a temperature-sensitive allele of YCK2 (yck2-2^ts) and a deletion of YCK1, retains 60% of wild-type kinase activity and displays relatively normal morphology at permissive temperature (Robinson et al., 1993). However, at the restrictive temperature, yck^ts cells show less than 25% of wild-type levels of kinase activity and undergo aberrant cell cycles with hyperpolarized growth and without cell division, resulting in dramatically elongated buds (Robinson et al., 1993). The elongated cell phenotype suggests that Yck could be required for regulating the switch from polarized to isotropic growth at mitosis. Yck2p is highly enriched in membranes of emerging and small buds, and at the mother-bud junction at the time of cytokinesis and cell separation (Robinson et al., 1999). Thus, Yck2p is correctly located to have a role in regulation of growth polarity. Moreover, the Yck protein kinases are required for internalization of plasma membrane proteins (Panek et al., 1997; Hicke et al., 1998). This effect is observed for a yck^ts strain even at permissive temperature, suggesting that wild-type kinase levels are necessary for efficient endocytosis. Together the phenotypic and endocytic defects point to a possible role for Yck activity in regulated cell surface remodeling.

Localization of Yck2p to the plasma membrane requires a functioning secretory pathway (our unpublished results). The secretory pathway is a highly regulated process involving proteins that are conserved among eukaryotes (Ferro-Novick and Jahn, 1994; Rothman, 1994). In yeast, vesicle-mediated protein transport to the plasma membrane has been extensively studied (reviewed in Kaiser et al., 1997). However, much of what is known about trafficking is based on proteins that enter the secretory pathway after translocation into the endoplasmic reticulum (ER). Yck2p has no signal sequence for ER translocation and is cytoplasmically oriented, arguing against an alternative method of entry into this pathway. Yck2p may interact with the cytoplasmic surface of secretory membranes by virtue of its presumed C-terminal isoprenyl modification, since removal of the consensus sequence for geranylgeranylation results in the inability both to associate with the plasma membrane (Vancura et al., 1994) and to interact with secretory membranes (our unpublished results). Yck2p interaction with membranes does not appear to be random, since association appears to proceed temporally from ER to Golgi to the plasma membrane. Therefore, specific trafficking factors linking Yck2p to internal membranes may direct transport of the kinase to the cell surface. Prenylated proteins are found in diverse membranes, but factors required for trafficking to those membranes and for specificity of localization have not been identified (Zhang and Casey, 1996; Sinensky, 2000). A previous genetic screen using the yck^ts mutant identified an AP-3 adaptor complex that acts in vesicular trafficking (Panek et al., 1997). However, loss of this complex has no effect on Yck2p localization (our unpublished results). To identify proteins acting in pathways that affect Yck activity or localization, we have isolated mutations that exacerbate the growth defect of a yck^ts strain. Using this synthetic enhancement screen, we identified two proteins that function in protein trafficking and indirectly in Yck2p trafficking. One of these, Tlg2p, is a member of the t-SNARE family of vesicle trafficking proteins (Holthuis et al., 1998a); the other is a novel protein encoded by the RGP1 gene (Aguilera et al., 1990).

Yeast strains (Table 1) were grown on standard rich (YEPD) or synthetic media (Sherman et al., 1986) at temperatures as indicated. Formamide, which destabilizes noncovalent bonds, was added to 3% in YEPD (Aguilera, 1994). Standard techniques were used for phenotypic and genetic analyses (Sherman et al., 1986).

Yeast transformation was by the lithium acetate procedure (Ito et al., 1983), modified by buffering all solutions with Tris-EDTA at Ph 8.0. All gene disruptions were constructed in vitro and introduced into diploid strains by one-step gene replacement (Rothstein, 1983). Strains confirmed by genomic Southern analysis to carry a disrupted allele were sporulated and haploid progeny were analyzed for cosegregation of the disruption marker and associated growth traits.

Standard techniques were used for DNA manipulation (Maniatis et al., 1982; Sambrook et al., 1989). Restriction and modification enzymes were used as recommended by the manufacturer (Promega, Madison, WI, USA). The Sequenase enzyme (US Biochemical, Cleveland, OH, USA) and dideoxy chain termination method (Sanger et al., 1977) were used with synthetic oligonucleotide primers (Integrated DNA Technologies, Coralville, IA, USA) to obtain DNA sequence from double-stranded templates. For cloning, the low-copy YCp50-based libraries of D. S. Conklin (unpublished) were used. DNA sequences were analyzed using the Genetics Computer Group (GCG; Madison, WI, USA) software package. Database searches with predicted protein sequences were carried out from the Saccharomyces Genome Database, using the Blast algorithm (Altschul et al., 1990).

To isolate yck^ts synthetic enhancement mutations, an ade2 ade3 colony color sectoring assay was used. HPY15 (ade2 ade3), derived from JM162, was crossed to a yck1::ura3− yck2-2^ts strain (LRB503) to generate the yck1::ura3− yck2-2^tsade2 ade3 strains HPY16 (MATa) and HPY17 (MATα). HPY16 and HPY17 cells, transformed with pHRP5 (pYCK2 ADE3 URA3), were mutagenized to approx. 10% viability with ethylmethane sulfonate (EMS; Sherman et al., 1986). Mutagenized colonies were grown at semipermissive temperature for yck^ts, 30°C, on rich medium and 45 remained Sect⁻ after several platings to nonselective medium. Each mutant was backcrossed to the parental yck^ts strain to determine dominant or recessive and single gene character of mutations by segregation in meiotic progeny. After three serial backcrosses, 12 strains displayed consistent 2:2 segregation of synthetic defects. These were designated dwy (defective with yck^ts) mutants. Mutants were then crossed by one another and the 12 strains were assigned to 10 complementation groups. Additional crosses were carried out to YCK2⁺ strains to assay for conditional growth phenotypes in haploid progeny.

The DWY1 and DWY2 genes were cloned by complementation of the dwy1 and dwy2 conditional growth phenotypes. To test for identity of isolated clones with mutant loci, genomic sequences from library plasmids were subcloned into a URA3 integrating vector, linearized within the genomic fragment, and introduced into wild-type strains by transformation. Genomic Southern hybridization analysis confirmed integration at the appropriate loci. Genetic linkage of the Ura⁺ phenotype with the original dwy mutation was assayed by tetrad analysis. The dwy1 complementing region was narrowed down to a 2.1 kb HindIII-BamHI (the BamHI site is from the library vector YCp50) fragment from one of the library clones (pHP101) and was subcloned into pRS316, creating pHP102. This construct contains the entire TLG2 coding sequence with approx. 500 base pairs (bp) flanking each side. The dwy2 complementing region is contained on a BglII-SphI fragment from the original library plasmid (pHP203) and was cloned into YCp50, creating pHP209. This plasmid contains the entire RGP1 open reading frame as well as approx. 500 bp flanking each side.

The DWY1/TLG2 gene (HindIII-BamHI fragment in pUC18) was disrupted by substitution of a 4.2 kb PstI fragment from YEp13 containing LEU2 for a 336 bp PstI fragment (codons 126-242 of 397 total) within the gene (pHP117). The DWY2/RGP1 gene (BglII-SphI fragment in pUC18) was disrupted by replacing a 1350 bp EcoRI fragment within the coding sequence (codons 23-475 of 663 total) with a 2.5 kb EcoRI fragment from pJA50 (Elledge and Davis, 1988) containing HIS3/Kan^r, creating pHP212.

The pep12::LEU2 disruption strain was made by digesting pBJ4351 (kindly provided by the E. Jones laboratory, Carnegie Mellon University, Pittsburgh, PA, USA) with BamHI and SphI and transformation of the fragment into LRB861. The resulting diploid strain was sporulated and HPY204 was among the pep12::LEU2 meiotic progeny.

To assay carboxypeptidase Y (CPY) secretion, strains were overlaid with prewetted nitrocellulose filters and plates were incubated at 30°C. Cells were rinsed off the filter and the filter was immunoblotted with monoclonal CPY antisera (Molecular Probes, Eugene, OR, USA) and goat anti-mouse IgG conjugated to horseradish peroxidase (HRP, CalBiochem, San Diego, CA, USA), with subsequent detection using the ECL system (Amersham, Piscataway, NJ, USA). To examine steady state forms of CPY or Kex2p by immunoblot analysis, strains were grown in YEPD to approx. 3.5×10⁷ cells/ml. Total protein was harvested as described previously (Davis et al., 1993) and electrophoresed in SDS/10% polyacrylamide gels. Immunoblotting was carried out either with anti-CPY as above or using polyclonal anti-Kex2p (a gift from R. Fuller, University of Michigan, Ann Arbor, MI, USA) and goat anti-rabbit IgG conjugated to HRP (CalBiochem), with detection using the ECL system.

Yeast vacuoles were observed using the vacuole lumen-specific fluorescent probe CMAC (7-amino-4-chloromethylcoumarin; Cell Tracker Blue; Molecular Probes). Cells were grown in rich medium at 24°C to 1×10⁷ cells/ml, pelleted gently, and resuspended in 10 mM Hepes buffer pH 7.5 containing 5% glucose. CMAC was added to a final concentration of 100 μM, cells were incubated at 24°C for 20 minutes and observed by fluorescence microscopy using DAPI (UV) optics.

Primers were designed to add ClaI sites (underlined; Cla-GFP-1: 5′ GCTGAATTCATCGATATGAGTAAAGGAGAAG 3′ and Cla-GFP-2: 5′ TTCGAATTCATCGATTTTGTATAGTTCATC 3′) by PCR to both the 5′ and 3′ ends of the coding sequence of the F64L, S65T bright GFP variant (Cormack et al., 1996; Robinson et al., 1999). The PCR product was cloned into pRSETB (creating pHRP29), transformed into BL21(DE3) cells for expression of GFP from the T7 promoter in pRSETB, and checked for function by monitoring colony fluorescence under UV light. Yellow fluorescent protein (YFP) is a variant of GFP (GFP-10c) with red-shifted excitation and emission wavelengths (Ormö et al., 1996), and is encoded on a plasmid kindly provided by R. Tsien (University of California, San Diego, USA). An NcoI-PvuII fragment that contained all of the YFP mutations was swapped with the corresponding GFP fragment using the construct described above, resulting in a ClaI-ended YFP fragment in pHRP30. For addition of YFP to TLG2 (in pUC18), a ClaI site was added immediately following the initiating ATG of TLG2 by PCR, creating pHP118. Primers were DWY1CLA-1: 5′ CGGATCGATTTTAGAG-ATAGAACT 3′ and DWY1CLA-2: 5′ CGGATCGATCATGTTTGTA-ACGAC 3′. YFP was then inserted into TLG2 at the ClaI site, creating pHP124. YFP:TLG2 was then cloned into pRS316 (HindIII-BamHI fragment), resulting in pHP125. The ClaI-ended YFP gene was added in-frame to the RGP1 coding sequence by insertion at a unique ClaI site corresponding to codons 6 and 7, creating pHP223.

A carboxy-terminal fusion of GFP coding sequence to SEC7 was kindly provided by O. Rossanese and B. Glick (University of Chicago, Chicago, IL, USA). The plasmid pUSE3+URA (Seron et al., 1998) was linearized by digesting at an unique Spe1 site within the SEC7 gene. This fragment was transformed into wild-type strain LRB759 for pop-in/pop-out gene replacement (Guthrie and Fink, 1991).

The GFP:yck2-2^ts fusion gene was created by swapping an XbaI-BclI fragment carried in pL2.992 (Robinson et al., 1999), which contains the GFP coding sequence and 280 bp of 5′ YCK2 sequence, for the corresponding fragment contained in a plasmid-borne yck2-2^ts gene (pL2.330; Robinson et al., 1993), creating the plasmid pRC2. The fusion gene was integrated at the wild-type YCK2 locus by excising an XbaI-SacI fragment that contained the entire coding sequence plus flanking regions and transforming into LRB856 (yck1Δ). Temperature-sensitive transformants were screened for GFP fluorescence and positive clones included HPY90. HPY90 was crossed to wild-type LRB759 and a Leu⁻ (YCK1⁺) fluorescent spore clone (HPY191) was isolated. HPY191 was crossed to HPY134 (tlg2Δ) or HPY185 (rgp1Δ) to create the strains HPY192 (GFP:yck2-2^tstlg2Δ) and HPY193 (GFP:yck2-2^tsrgp1Δ), respectively.

Expression of the GFP and YFP fusion proteins (except for GFP-Snc1p, see below) was monitored by fluorescence microscopy using an Olympus (Melville, NY, USA) AX-70 microscope equipped for DIC optics and epifluorescence. For observation of both GFP and YFP in the same cells, filter sets selective for S⁶⁵T GFP (JP1) and YFP (JP2) excitation and emission spectra (Chroma Tech. Corp., Brattleboro, VT, USA) were used. For observation of GFP only, we used the filter set 41001 (Chroma Tech.), which is selective for S⁶⁵T GFP, but has broader bandwidth than JP1. Images were captured with a Photometrics (Tucson, AZ, USA) cooled CCD (CE200A) camera using IPLab Spectrum software run by a Macintosh G3 computer. GFP-Snc1p was visualized by FITC optics and photographed with an Orca digital camera (Hamamatsu Photonics). All images for a given experiment were captured with the same exposure time and image composites were constructed using Adobe Photoshop.

To capture images of GFP and YFP fluorescence within the same cells, we used a custom IPLab Spectrum script written by K. Tatchell that automates switching between filter sets. This minimized the time between capture of the image pair, necessary due to rapid movement of the fluorescent structures within live cells. IPLab Spectrum software was also utilized to compare intracellular staining to membrane staining of GFP-Yck2p in mutants. Fluorescence levels were determined along linear regions across cells. The fluorescence level at membranes was set as the average of the peak levels generated at each end of the cell. The internal level was determined by calculating the average level within the cell section. Internal fluorescence level is expressed as a percentage of the membrane level.

To identify proteins that participate in pathway(s) that impinge on, or are regulated by, Yck activity, we used a colony color sectoring assay (Bender and Pringle, 1991). yck^ts synthetic enhancement mutations were isolated at the semi-permissive temperature of 30°C (Robinson et al., 1993). Upon introduction of a YCK2 ADE3 plasmid, yck^tsade2 ade3 strains were Sect⁺ under nonselective conditions (retained viability upon plasmid loss, not shown). After mutagenesis, cells containing mutations that synthetically enhance the yck^ts growth defect should require wild-type Yckp activity for continued viability and thus display the Sect⁻ phenotype. Strains HPY16 and HPY17 were mutagenized with EMS (Materials and Methods) and 12 Sect⁻, dwy (defective with yck^ts) strains were isolated. All mutations were recessive, and were assigned to ten complementation groups. The dwy mutants were crossed to YCK⁺ and yck1⁻YCK2⁺ strains to determine if they displayed any growth defects in these genetic backgrounds. Mutations in DWY1 confer cold sensitivity (11°C) and mutations in DWY2 impart sensitivity to 3% formamide (not shown). Since mutations in DWY1 and DWY2 conferred both conditional growth phenotypes and the strongest synthetic growth defects, these were selected for further characterization.

The DWY1 and DWY2 genes were cloned from low copy genomic libraries by complementation of the dwy1 and dwy2 conditional growth defects in a YCK2⁺ background.Two overlapping library plasmids complemented the cold sensitivity of the dwy1 strain. After delimiting the complementing region and demonstrating linkage between the library sequences and dwy1-1 (not shown; Materials and Methods), we found that DWY1 is the previously described gene TLG2 (t-SNARE affecting a late Golgi compartment; Holthuis et al., 1998a). TLG2 encodes a member of the syntaxin (t-SNARE) family of vesicle trafficking proteins, which interact with cognate v-SNARES located on transport vesicles and mediate vesicle fusion (Sollner et al., 1993; Weber et al., 1998). Tlg2p function has been implicated in endocytic processes, in retention of late Golgi proteins (Abeliovich et al., 1998; Seron et al., 1998; Holthuis et al., 1998a), and in delivery of internalized Chs3p and Snc1p to the Golgi for redirection to the plasma membrane (Holthuis et al., 1998b; Lewis et al., 2000). Disruption of TLG2 in our genetic background confirmed that this gene is not essential for growth on rich medium at 30°C. However, it is essential for growth in the cold and in the yck^ts background, as double mutant spore clones were not recovered (Fig. 1A,B).

Eight overlapping library clones from chromosome IV complemented the dwy2 conditional phenotype and contain the RGP1 (reduced growth phenotype; Aguilera et al., 1990) gene. RGP1 is predicted to encode a 663-amino-acid protein that has no significant sequence similarity to any previously described protein and contains no distinctive signal sequences or any other described sequence motifs. Analysis of the rgp1::HIS3 disruption allele revealed that RGP1 is not essential for growth. However, it is required for growth in the presence of 3% formamide, for wild-type growth rates at temperature extremes in YCK⁺ strains (Fig. 1A), and for normal growth in the yck^ts genetic background (Fig. 1C). No function has been assigned to Rgp1p, but because it was isolated in the same genetic screen as Tlg2p, we explored whether it also has a role in protein trafficking. Hereafter, we will refer to DWY1 as TLG2 and DWY2 as RGP1, and unless stated otherwise, all further studies were done with strains disrupted for TLG2 or RGP1.

The simplest explanation for a mutation in a trafficking protein causing a synthetic growth defect with yck^ts is that this mutation affects targeting of Yckp to the plasma membrane. The yck2-2^ts gene product has reduced kinase activity even at permissive temperatures, so perturbing its localization to the plasma membrane would further lower Yckp activity at the cell surface. To test this hypothesis, the localization of a GFP-yck2^ts fusion protein was monitored in strains disrupted for rgp1 or tlg2 and carrying a chromosomal copy of GFP:yck2-2^ts. These strains contain the wild-type YCK1 gene to maintain viability. As shown in Fig. 2A, the GFP-yck2^ts fusion protein in otherwise wild-type cells is primarily plasma membrane-localized at 24°C. This localization is similar to the wild-type fusion protein (see below); however, internal fluorescence relative to plasma membrane fluorescence is increased by at least 10% over the wild-type ratio, suggesting that targeting of GFP-yck2^tsp to the cell surface is less efficient than that of wild type.

Consistent with the notion that these proteins affect Yck2p targeting, cells lacking Tlg2p or Rgp1p function show altered patterns of GFP:yck2^tsp fluorescence. Internal fluorescence relative to plasma membrane staining was quantitated using graphical representation of the fluorescence levels across single cells. A representative graph is shown below each set of cells and the region selected to generate the graph is depicted as a line through the cell being analyzed. For the wild-type strain expressing GFP-yck2^tsp, the internal fluorescence level averaged 84% (n=16) of that of the plasma membrane level. Internal fluorescence appears to be mostly excluded from vacuolar structures. tlg2 cells showed a higher level of internal fluorescence (Fig. 2A, middle panels), averaging 95% (n=25) that of the plasma membrane level. The small vacuoles in these cells (see below) are somewhat masked by internal fluorescence. The average internal level for rgp1 cells was similarly increased to 93% that of the plasma membrane (n=24, right panels). Prominent vacuoles are present by Nomarski optics in most of these cells (not shown) and fluorescence appears excluded from these structures as in wild-type cells.

To determine if loss of Tlg2p or Rgp1p activity also affects localization of wild-type Yck2p, mutant strains containing an integrated GFP:YCK2 fusion gene were generated. At 30°C wild-type cells carrying GFP:YCK2 show typical plasma membrane Yck2p association (Fig. 2B). However, in both mutant backgrounds the GFP-Yck2p fluorescence is more diffuse, with an increase in cytoplasmic staining. In wild-type cells, the average internal fluorescence level of GFP:Yck2p is 76% (n=21) of the fluorescence level at the plasma membrane. For both tlg2 (n=24) and rgp1 (n=20) cells, the internal fluorescence level is increased to approx. 90% of that at the cell periphery.

If Yck2p interacts with a sorting factor(s) during its transport, Yck2p overexpression could saturate its localization machinery. Consistent with this idea, when GFP-Yck2p is greatly overexpressed from a galactose-inducible promoter in wild-type cells, fluorescence is sometimes observed on vacuolar membranes, a localization rarely observed with lower kinase levels (Fig. 2C, left panels). In contrast to the faint vacuolar staining in wild-type cells, the fluorescence in rgp1 and tlg2 mutant cells appears equivalent on peripheral and vacuolar membranes (Fig. 2C, middle and right panels). We approximated vacuolar membrane versus plasma membrane fluorescence levels for cells overexpressing GFP-Yck2p from the GAL1 promoter (n=10 for each strain). In the 70% of wild-type cells that exhibit vacuolar fluorescence, the vacuolar membrane level is 72% that of the plasma membrane level. Strikingly, the vacuolar membrane level of fluorescence is as high or higher than that at the cell periphery in both tlg2 (101%) and rgp1 (114%) cells. These results are consistent with the missorting of a significantly larger fraction of the kinase in the mutant strains.

Since deletion of TLG2 or RGP1 results in similar effects on Yck2p localization, Rgp1p, like Tlg2p, could function in protein trafficking. Mutations in TLG2 confer a mild carboxypeptidase Y (CPY) sorting defect, causing missorting of approx. 20% of this vacuolar enzyme to the cell surface (Abeliovich et al., 1998), so we examined rgp1 mutant cells for this phenotype. Whereas wild-type and yck^ts strains do not secrete significant amounts of CPY (Fig. 3A), mutants defective in trafficking of this vacuolar protease, such as vps27 and pep12, secrete approx. 50% and greater than 75% of their CPY, respectively (Piper et al., 1995; Becherer et al., 1996) and were used for comparative purposes. We used a filter overlay assay to detect CPY secretion by the rgp1 strain and compared it to tlg2, vps27 and pep12 mutant strains. As shown in Fig. 3A, rgp1 cells secrete more of this vacuolar hydrolase than do tlg2 cells, and an amount intermediate between that of vps27 and pep12 mutant cells.

Immunoblot analysis was performed on total protein extracts to examine the intracellular steady state forms of CPY in these mutants. CPY is present in the ER as a 67 kDa precursor form (p1) and is modified in the Golgi to a larger p2 form of 69 kDa. Upon delivery to the vacuole, CPY is processed to its mature, active 61 kDa form, in part by Proteinase A (PrA), which is encoded by the PEP4 gene (reviewed in Jones et al., 1997). As illustrated in Fig. 3B, CPY in tlg2 and rgp1 strains is in the mature form, indicating that these mutations do not significantly affect processing of intracellular enzyme. Further, significant amounts of CPY are observed in rgp1 cells, suggesting that the missorting is not as severe as that observed in cells lacking Pep12p, in which internal, mature CPY is virtually undetectable in our hands.

Since vacuole morphology is often perturbed in mutants that missort vacuolar hydrolases (Raymond et al., 1992), and tlg2 mutant cells contain fragmented vacuoles (Holthuis et al., 1998a), vacuolar structure was examined in rgp1 cells. As Fig. 4 shows, at 24°C, most rgp1 mutant cells (>80%) contain highly fragmented vacuoles when visualized by staining with CMAC (staining was carried out in buffered 5% glucose to prevent coalescence of vacuole structures (Roberts et al., 1991; Materials and Methods). In these conditions, rgp1 vacuoles appear to be smaller and even more numerous than those in tlg2 mutant cells, whereas wild-type and yck^ts cells show one to several larger vacuoles. Staining of both tlg2 and rgp1 mutants with quinacrine, a weakly basic dye that accumulates in acidic compartments (Guthrie and Fink, 1991), demonstrated that these structures are acidified (not shown), indicating that the vacuolar H⁺-ATPase is targeted to and functional in these vacuoles.

Based on the CPY sorting defect and abnormal vacuolar morphology seen in cells lacking Rgp1p, this protein could participate in protein trafficking. To ascertain where Rgp1p functions, a red-shifted green fluorescent protein variant, yellow fluorescent protein (YFP, Materials and Methods), was fused in-frame to the 5′ end of the RGP1 coding sequence. Introduction of this fusion on a low-copy vector complemented the formamide sensitivity of cells lacking RGP1 (Fig. 5A, left panel). By fluorescence microscopy, cells expressing the fusion protein showed staining of small punctate structures dispersed throughout the cell (Fig. 5C). The YFP variant was also fused in-frame to the TLG2 coding sequence following the initiating ATG of TLG2. Introduction of this fusion protein on a low-copy vector complemented the cold sensitivity of the tlg2 mutant (Fig. 5A, right panel) and the pattern of YFP-Tlg2p fluorescence (Fig. 5D) was similar to that of YFP-Rgp1p.

Golgi cisternae in yeast are observed in punctate staining patterns (Redding et al., 1991) and the structures labeled by the YFP-Rgp1 fusion protein are reminiscent of that organelle. Sec7p is required for secretory traffic through the Golgi (Franzusoff and Schekman, 1989) and the fluorescence pattern of a functional Sec7-GFP fusion protein reflects Golgi localization (Seron et al., 1998). To determine if YFP-Rgp1 is also Golgi-localized, we transformed low-copy vectors carrying the YFP fusion genes into cells that contain the SEC7:GFP fusion integrated at the wild-type SEC7 locus and imaged the fusion proteins with filter sets designed to selectively observe YFP or GFP. Fig. 5B demonstrates that the filter set selective for GFP does not pass significant amounts of yellow fluorescence, but the filter set selective for YFP passes a low level of green fluorescent signal. However, the YFP fluorescence is always significantly brighter than the GFP signal in the images using the yellow filter set. To assess colocalization of the fusion proteins, we compared spots in corresponding GFP and YFP images of individual cells. Percentage colocalization was determined by the ratio of overlapping spots to the total number of spots counted. Fig. 5C shows that the fluorescence of YFP-Rgp1p overlaps with that of Sec7-GFP. The calculated 85% colocalization (n=76 cells) suggests that Rgp1p functions at the Golgi. Only 60% colocalization (n=100 cells) was observed for the Tlg2 and Sec7 fusion proteins (Fig. 5D), which is in agreement with previous observations that Tlg2p distribution overlaps with both Golgi and endosomal markers (Abeliovich et al., 1998; Seron et al., 1998; Holthuis et al., 1998a; Conibear and Stevens, 2000).

Tlg2p has a role in retrograde protein trafficking at the Golgi (Holthuis et al., 1998a). The localization data and the common phenotypes of rgp1 and tlg2 cells together suggest that Rgp1p may also function in this process. One function of this retrograde pathway is to retrieve Golgi membrane proteins that have inadvertently been delivered to the late endosome (Roberts et al., 1992; Cooper and Stevens, 1996). Kex2p is one such Golgi enzyme that must be recycled from the endosome to avoid delivery to the vacuole, where it is degraded (Wilcox et al., 1992; Nothwehr et al., 1993). Thus, we assessed steady state levels of Kex2p in rgp1 mutant cells. If Kex2p is not recycled, the steady state Kex2p levels should be lower in rgp1 cells than in wild-type cells. We also assessed Kex2p levels in a vps27 strain, which, like the tlg2 strain, is defective in retrograde Golgi trafficking (Piper et al., 1995; Nothwehr et al., 1996), and in strains lacking pep4, since mislocalized Kex2p is not efficiently degraded in PrA-deficient vacuoles (Wilcox et al., 1992). As shown in Fig. 6, Kex2p levels are reduced by 60% in rgp1 strains as compared to wild type. This reduction is similar to that seen for tlg2 (50%) and vps27 (65%) cells. Kex2p protein levels are restored to wild type in the rgp1 pep4 double mutant strain, consistent with the idea that the reduced amount in the rgp1 strain reflects inaccurate Kex2p trafficking and subsequent vacuolar degradation rather than decreased Kex2p synthesis.

The SNARE protein Snc1p mediates fusion of exocytic vesicles with the plasma membrane (Protopopov et al., 1993). After exocytosis, Snc1p is endocytosed and delivered to the Golgi via an early endosomal compartment to participate in another round of transport to the cell surface (Lewis et al., 2000). This recycling is dependent on Tlg2p function, as cells lacking tlg2 accumulate Snc1p in internal punctate structures with very little cell surface localization (Lewis et al., 2000). To determine if Rgp1 is also required for recycling of Snc1p, we utilized a functional GFP-tagged Snc1p (Lewis et al., 2000). GFP-Snc1p has a polarized distribution at the cell surface of wild-type cells, where it is concentrated at the plasma membrane of small and emerging buds (Fig. 7, lower panels). In contrast, GFP-Snc1p is localized intracellularly in rgp1 mutant cells (Fig. 7, upper panels), which show a hazy cytosolic fluorescence that could correspond to transport vesicles as well as fluorescence at internal membranes. A defect in forward transport of Snc1p to the cell surface is most likely not the cause of its mislocalization in rgp1 mutant cells as secretion of invertase and transport of a GFP fusion to the hexose transporter Hxt2p to the cell surface (Kruckeberg et al., 1999) are indistinguishable from wild-type cells (not shown).

Elimination of TLG2 or RGP1 causes very similar trafficking phenotypes. If these gene products act only at the same trafficking step, combination of the two null mutations shouldhave no additive detrimental effect (Guarente, 1993). To test this, we generated double rgp1 tlg2 mutants by crossing single mutant strains. As shown in Fig. 8, meiotic progeny of this cross do not include viable double mutants, indicating that while Rgp1p and Tlg2p may both play a role in retrograde Golgi trafficking, they may also act in different pathways.

The t-SNARE Tlg1p has overlapping functions with Tlg2p (Holthuis et al., 1998a; Holthuis et al., 1998b; Lewis et al., 2000). We attempted to examine the requirement for Rgp1p in cells lacking tlg1, but deletion of this t-SNARE in our genetic background (derived from S288c) results in lethality, as it does in the W303 background (Coe et al., 1999).

The yeast CK1 protein kinases Yck1p and Yck2p are together essential for viability, and although they rely on the secretory pathway for localization, how they are targeted to the plasma membrane is unknown. To identify proteins that act in pathways that affect or are affected by Yck2p, a synthetic enhancement screen was carried out using a thermosensitive yck mutant. This screen identified mutations in TLG2 and RGP1, which cause a requirement for wild-type levels of Yck activity for normal cell growth.

Tlg2p is distributed on both Golgi and endosomal membranes, perhaps localized to structures at the junction between endocytic and exocytic pathways (Abeliovich et al., 1998; Holthuis et al., 1998a; Lewis et al., 2000). This t-SNARE is required for efficient endocytic turnover and for retrieval of Golgi resident proteins (Abeliovich et al., 1998; Seron et al., 1998; Holthuis et al., 1998a). Recently it has been demonstrated that Tlg2p also has a role in delivery of Chs3p (Holthuis et al., 1998b) and Snc1p (Lewis et al., 2000) to the Golgi from the cell surface.

RGP1 encodes a novel protein (Aguilera et al., 1990). Although Rgp1p shares no sequence similarity with any known trafficking protein, mutations in the RGP1 gene confer several traits in common with mutations in TLG2. Neither gene is essential for viability under optimal growth conditions, nor are they required directly for secretion as invertase secretion is unaffected in either mutant strain (not shown, see also Holthuis et al., 1998a). Further, cells with mutations in either display a fragmented vacuolar phenotype and affect the accurate trafficking of CPY, Kex2p and Snc1p.

Like Tlg2p, Rgp1p appears to be Golgi-localized. Although the overlap between Rgp1p and Sec7p localization is extensive, it is not complete. The lack of complete colocalization could reflect the fact that Sec7p is present on all Golgi compartments (Franzusoff et al., 1991) whereas Rgp1p functions at a subset of them. However, we observed Rgp1p not only in a subset of Sec7p-containing structures, but also in Sec7p-negative structures. Alternatively, as the localization study was done in live cells and the fluorescent structures were in motion, the lack of colocalization may be simply due to this movement. Indeed, when we compared the localization of YFP-Rgp1 to that of GFP-Rgp1, colocalization never exceeded 90% (not shown), suggesting that the colocalization observed between YFP-Rgp1 and Sec7-GFP is at the limit of overlap detectable in live cells using our microscope system. Thus, Rgp1 may act exclusively at the Golgi and not also at endosomal structures as has been proposed for Tlg2p.

Based upon its localization, the rgp1 phenotypes, and the similarities of these phenotypes to Golgi recycling mutants, we propose that Rgp1 participates in retrograde trafficking at the Golgi. A defect in this recycling pathway would explain some phenotypes exhibited by rgp1 cells. Failure to recycle the CPY sorting receptor Vps10p would result in default transit of some CPY to the cell surface (Cooper and Stevens, 1996). The decreased levels of Kex2p in rgp1 cells would stem from the inability to recycle this Golgi enzyme from an endosomal compartment, causing its degradation in the vacuole. Finally, the inability to deliver GFP-Snc1p to the Golgi for redirection into exocytic vesicles would result in its mislocalization.

The rgp1 and tlg2 mutants were isolated in the same genetic screen, and display phenotypes that suggest similar roles in recycling. Thus, Rgp1p could work in conjunction with Tlg2p in delivery of retrograde vesicles at the Golgi. However, several observations indicate that they do not function only at the same trafficking step. The mutant strains do not have identical phenotypes. Tlg2p is necessary for growth in the cold, as well as for efficient turnover of endocytosed proteins. In contrast, mutations in RGP1 confer sensitivity to formamide and a slow-growth phenotype especially at temperature extremes, but do not affect Ste3p endocytosis (not shown). If these proteins function only in Golgi recycling, no synthetic defects would be expected for cells containing both null mutations, yet this combination is lethal. Therefore, they may also act in different trafficking steps/pathways, one of which must be operational for viability. Tlg2p is present on endosomal membranes and a unique role for the t-SNARE at this compartment may be required when Rgp1p function is lacking.

Recently, a complex composed of Vps52p, Vps53p and Vps54p was isolated that localizes to the Golgi and is required for retrograde trafficking (Conibear and Stevens, 2000). This complex was identified in a screen for mutants that secrete CPY and are synthetically lethal with end4. Although Rgp1p appears to function at a similar step as the Vps52/53/54 complex, we have not observed synthetic effects upon combining rgp1 and end4. Thus, if Rgp1p works in conjunction with the Vps52/53/54 complex, it is not required for the Vps52/53/54p function which, when lacking, results in the requirement for End4p. One possibility is that Rgp1p function increases efficiency of Golgi recycling. In this case, loss of Rgp1p function would slow delivery, but not completely block it. Only upon loss of both efficiency and specificity (such as with tlg2 or vps52/53/54 mutations) does the essential nature of retrograde trafficking become apparent. A role for Rgp1p as a guanine nucleotide exchange factor for Ypt6, in conjunction with Ric1p, was recently determined (Siniossoglou et al., 2000). This role fits well with our genetic data.

The most probable reason for the isolation of tlg2 and rgp1 in the yck^ts synthetic lethal screen is their effects on localization of the Yck kinases. Yck2p requires the secretory pathway for plasma membrane delivery and most likely travels on the cytoplasmic surface of secretory vesicles (our unpublished results). As this delivery appears to be proceed in a regulated fashion (our unpublished results), we speculate that Yck interacts with a trafficking factor(s) for accurate delivery. Recently, it was demonstrated that targeting of farnesylated Ras to the plasma membrane also requires the secretory pathway; however, no specific factor(s) linking it to the cytoplasmic face of vesicles has been identified (Choy et al., 1999). Mutations in either tlg2 and rgp1 affect efficient delivery of GFP-Yck2 and GFP-yck2^ts proteins to the cell surface. Thus, Tlg2p and Rgp1p may be required for recycling of a membrane-bound factor that interacts with Yck2p during its transit to the plasma membrane. If the rgp1 and tlg2 mutations perturb Yck2p transport indirectly due to a defect in recycling of its sorting factor to the Golgi, overexpression of the kinase in these mutants should result in pronounced missorting, as efficient recycling would become crucial under these conditions. This is what was observed upon overexpression of Yck2p in tlg2 and rgp1 mutants. Based on these results, it is possible that factors directing transport of prenylated proteins on the cytoplasmic surface of vesicles utilize classical trafficking pathways for their recycling. De novo synthesis of the Yck2p trafficking factor could allow delivery of sufficient wild-type Yck kinase to the membrane for it to perform its essential function in rgp1 and tlg2 mutant cells. However, when the yck2^ts kinase is the sole source of Yck activity in the rgp1 or tlg2 mutant cell, inefficient delivery coupled with reduced activity would be lethal, even at permissive temperature. Alternatively, the Yck kinases, which regulate internalization of several plasma membrane proteins (Panek et al., 1997; Hicke et al., 1998; Marchal et al., 2000), could enter the cell on endocytic vesicles and be recycled back to the cell surface. An endosome-plasma membrane recycling pathway appears to deliver a mutant form of the major plasma membrane ATPase, Pma1-7p, to the plasma membrane, supporting the idea that recycling to this membrane can occur in yeast (Luo and Chang, 2000). Either explanation leads to the prediction that other mutations that affect recycling of proteins to the Golgi from the cell surface could result in mislocalization of Yck2p. This is in fact what we observed upon loss of the Vps52/53/54 complex (our unpublished results). We also predict that synthetic effects would be observed upon removal of this complex in a yck^ts background. Interestingly, mutants that affect trafficking at the late endosome do not result in synthetic lethality with the yck^ts mutation (our unpublished results). The proposed Yck2p targeting factor may therefore be sorted away from other endocytosed proteins in the early endosome and may not travel through the late endosome before reaching the Golgi. A similar scenario was proposed for the trafficking of Snc1p (Lewis et al., 2000). Experiments to identify the Yck trafficking factor(s) are currently underway.

We thank K. Tatchell for help with microscopy and N. Mathias and K. Tatchell for critical evaluation of this manuscript. We also thank S. Emr, R. Fuller, B. Glick, E. Jones, G. Payne, T. Stevens, K. Tatchell, R Tsien and the members of their laboratories for generously providing strains, plasmids and reagents. We are grateful to Kim Marks, Allison Jerome and Jennifer Morgan for superb technical assistance. This work was supported by NSF grant MCB-9601294 to L.C.R. H.R.P. was supported by Louisiana Educational Quality Support Fund doctoral student training award GF11 to D. J. O’Callaghan and R. E. Rhoads.