|Home | About | Journals | Submit | Contact Us | Français|
We have characterized LUV1/RKI1/TCS3/VPS54, a novel yeast gene required to maintain normal vacuolar morphology. The luv1 mutant was identified in a genetic screen for mutants requiring the phosphatase calcineurin for vegetative growth. luv1 mutants lack a morphologically intact vacuole and instead accumulate small vesicles that are acidified and contain the vacuolar proteins alkaline phosphatase and carboxypeptidase Y and the vacuolar membrane H+-ATPase. Endocytosis appears qualitatively normal in luv1 mutants, but some portion (28%) of carboxypeptidase Y is secreted. luv1 mutants are sensitive to several ions (Zn2+, Mn2+, and Cd2+) and to pH extremes. These mutants are also sensitive to hygromycin B, caffeine, and FK506, a specific inhibitor of calcineurin. Some vacuolar protein-sorting mutants display similar drug and ion sensitivities, including sensitivity to FK506. Luv1p sediments at 100,000 × g and can be solubilized by salt or carbonate, indicating that it is a peripheral membrane protein. A Green Fluorescent Protein–Luv1 fusion protein colocalizes with the dye FM 4-64 at the endosome, and hemagglutinin-tagged Luv1p colocalizes with the trans-Golgi network/endosomal protease Kex2p. Computer analysis predicts a short coiled-coil domain in Luv1p. We propose that this protein maintains traffic through or the integrity of the early endosome and that this function is required for proper vacuolar morphology.
The yeast vacuole or lysosome is an acidic, hydrolytic organelle that has many functions: osmoregulation, storage of nutrients, sequestration of toxins, and degradation and recycling of proteins, membrane, and carbohydrate. It is a dynamic and variable structure with several smaller clustered vacuoles found in growing cells; these coalesce into one large vacuole during cell stasis or low osmolarity (reviewed by Klionski et al., 1990 ). Genetic screens for loss of vacuole structure or function have established that the vacuole is formed from several distinct paths of SNAP/SNARE-mediated transport. Screens for missorting of vacuolar proteases have identified peptidase-deficient mutants (pep) (Jones, 1977 ) and vacuolar protein-sorting mutants (vps) (Bankaitis et al., 1986 ; Rothman and Stevens, 1986 ; Robinson et al., 1988 ; Rothman et al., 1989 ). Together, these genes define a clathrin-mediated pathway of vesicle transport from the Golgi, through the endosome or prevacuolar compartment (PVC), to the vacuole. An alternative, Adaptor Protein-3–mediated pathway is thought to bypass the endosome, leading from the late Golgi (trans-Golgi network [TGN]) directly to the vacuole (reviewed by Conibear and Stevens, 1998 ). Studies of endocytosis-defective mutants (end) have detailed the contribution to vacuole formation of actin-dependent endocytosis from the plasma membrane (reviewed by Wendland et al., 1998 ). Two other modes of transport to the vacuole, cytoplasm-to-vacuole targeting and autophagocytosis, deliver cytoplasmic material either directly to the vacuole or through the endosome (reviewed by Klionski, 1998 ). Many of the same genes have been isolated in these different genetic screens, indicating that there is substantial overlap in these transport pathways. These studies, and others of vacuole morphology mutants (vam) (Wada et al., 1992 ) or inheritance (vac) (Wiesman et al., 1990 ; Gomes de Mesquita et al., 1996 ), as well as in vitro assays of homotypic vacuole fusion (Nichols et al., 1987 ), have helped define both the common biochemical mechanisms shared among the pathways and specific players for each trafficking step.
Mutations that disrupt traffic between the TGN and the vacuole, anterograde and retrograde, often result in vacuole fragmentation. Raymond and coworkers (1992) characterized the vacuole phenotypes of the various vps mutants and grouped them into six classes, expanding the three classes of Banta and coworkers (1988) . Class D defines transport between the Golgi and the endosome/PVC; mutants in these genes have vacuoles that are thought to form from endocytic and alkaline phosphatase (ALP) pathway traffic. Class E proteins mediate vesicle transport out of the PVC to the vacuole and retrograde back to the Golgi. Class E mutants thus accumulate PVC material, but transport to the vacuole is not completely blocked and vacuoles still form in these mutants (reviewed by Conibear and Stevens, 1998 ). Class B mutants show moderately fragmented vacuoles; many of the proteins defined by these mutations have been found as two complexes, called the retromer, which is thought to mediate retrograde transport from the PVC to the Golgi (Horazdovsky et al., 1997 ; Seaman et al., 1998 ). Class C mutants show severe vacuole fragmentation, and these proteins, Vps18p, Pep5p/Vps11p, Vps16p, and Vps33p, form the RING complex, which is thought to link vesicle targeting components together at the vacuole and may supply target specificity (Rieder and Emr, 1997 ). Class C mutants disrupt traffic from the PVC to the vacuole and thus also affect the late stages of endocytosis (Dulic and Reizman, 1990 ).
The membrane fusion events that occur during transport to the vacuole are mediated by SNARE proteins, together with additional regulatory and fusion factors (SNAPs, NSF, and RAB GTPases). For example, vesicle traffic from the Golgi to the PVC in yeast requires Pep12p (Becherer et al., 1996 ) and Vti1p (Fisher von Mollard et al., 1997 ) as the SNAREs, Vps21p as the RAB (Horazdovsky et al., 1994 ; Singer-Kruger et al., 1994 ), Vps45p as the Sec1p homologue (Cowles et al., 1994 ; Piper et al., 1994 ), and Sec17p and Sec18p as the ubiquitous SNAP and NSF. SNAREs and RABs are usually unique to a particular transport step, and the study of all the yeast SNAREs has shown in finer detail the vesicular sorting steps in yeast (reviewed by Pelham, 1999 ). For instance, Vam3p defines the vacuole, Pep12p defines the late endosome or PVC, Tlg1p and Tlg2p define the early endosome and/or the TGN, and Sed5p defines the early Golgi. However, it has also become apparent from such studies that neither SNAREs nor RABs alone supply the specificity in vesicle and target fusion (Grote and Novick, 1999 ). Rather, it is currently thought that the SNARE model requires additional, accessory proteins to recruit or tether the appropriate SNAREs and/or RABs, and thus the transport vesicle, to the appropriate target (reviewed by Waters and Pfeffer, 1999 ). The retromer and RING complexes mentioned above are currently thought to fill this role of targeting specificity.
Our laboratory studies the serine/threonine protein phosphatase type 2B, or calcineurin. Yeast strains lacking calcineurin are viable under standard growth conditions (Cyert et al., 1991 ; Liu et al., 1991 ; Cyert and Thorner, 1992 ). To identify genes that might compensate for the loss of calcineurin, a genetic screen was performed to identify mutations that are lethal in combination with a calcineurin mutation. Mutants defective in known genes were identified in this screen and subsequent analyses, including cell wall biosynthesis genes fks1, kre5, kre6, kre9, mpk1, and vma mutants (Garrett-Engele et al., 1995 ; P. Garrett-Engele and M.S. Cyert, unpublished result). VMA genes encode subunits of the vacuolar membrane H+-ATPase and are required for vacuole acidification. This acidification is required to activate vacuolar proteases and phosphatases and to drive ion sequestration. Calcineurin regulates cell wall biosynthesis and ion homeostasis at least in part through transcriptional activation of genes involved in these processes (Mendoza et al., 1994 ; Cunningham and Fink, 1996 ; Matheos et al., 1997 ; Stathopoulos and Cyert, 1997 ). Other mutants were also identified in the synthetic lethality screen described above; here we characterize the gene defined by one of those mutants, LUV1/RKI1/TCS3/VPS54, and present evidence that the protein it encodes mediates traffic through, or the stability of, the TGN or early endosome.
Yeast strains used in this study are listed in Table Table1.1. Standard media and culture conditions were used (Sherman, 1991 ), except that twice the level of amino acids and nucleotides were added to synthetic media. Molecular cloning methods and yeast transformation were as described (Ausubel et al., 1987 ). FK506 was from Fujisawa (Osaka, Japan). Sequencing reactions used Sequenase (United States Biochemical, Cleveland, OH) according to the manufacturer's protocol. PCR reactions used Taq polymerase from Life Technologies/BRL (Grand Island, NY) in a MJ thermal cycler (MJ Research, Watertown, MA). All restriction enzymes were from New England Biolabs (Beverly, MA).
A luv1Δ::hisGURA3hisG allele was constructed in pBS-hisG-URA3-hisG (Alani et al., 1987 ) by inserting a 1.8-kilobase (kb) SacI–BglII fragment corresponding to a region just 5′ of the LUV1 ORF into SacI and BamHI sites on one flank of the hisGURA3hisG cassette. On the other flank, a 1.5-kb PstI–SalI fragment corresponding to the C terminus and flank of LUV1 was inserted into EcoRI and PstI sites. The 7.2-kb SacI–SalI luv1Δ::hisGURA3hisG fragment was excised and transformed into diploid yeast MCY3, which is heterozygous for a calcineurin mutation. Diploids were sporulated, and the genotype of luv1Δ::hisGURA3hisG (strain YMC1) was confirmed by Southern analysis. To obtain strain YMC4, strain SF838-9DaR2L1 was transformed with the 7.2-kb SacI–SalI luv1Δ::hisGURA3hisG fragment and selected on uracil-free medium.
An N-terminal GFP-LUV1 fusion was constructed with the use of PCR to amplify the LUV1 coding region. N- and C-terminal specific primers contained BamHI restriction sites, and the N-terminal primer was engineered to introduce a five-glycine linker before the first Luv1p methionine. The PCR product was digested with BamHI and ligated to similarly digested pTS545 (Carminati and Stearns, 1997 ), which contains enhanced Green Fluorescent Protein (GFP) under the control of the GAL1,10 promoter, to give plasmid pTS544::LUV1.
LUV1 contains a BglII restriction site at nucleotide 190. A 3X hemagglutinin (HA) fragment was amplified by PCR with the use of pTS515 (Marschall et al., 1996 ) as a template and primers containing BglII restriction sites. The product was digested with BglII and ligated to similarly digested LUV1 plasmid (in pRS315) (Sikorski and Hieter, 1989 ) to give pRS315::LUV1::3XHA.
For FM4-64 vacuolar staining, cells were grown in YPD to late log growth (OD600 = 5) to enhance vacuole structures. One OD600 unit of cells (i.e., the number of cells in 1 ml culture at OD600 = 1) was incubated on ice for 15 min with 32 μM FM4-64 dye (Molecular Probes, Eugene, OR) in 100 μl of YPD and washed in YPD at room temperature for 30 min as described (Vida and Emr, 1995 ). Similarly grown cells were incubated for 10 min with 2 μM quinacrine in pH 7.5 YPD as described (Roberts et al., 1991 ). Cells were washed with pH 7.5 YPD and viewed under FITC fluorescence. For Vma, carboxypeptidase Y (CPY), and ALP immunofluorescence, similarly grown cells were fixed, permeabilized, and incubated as described (Piper et al., 1997 ) with mouse mAbs to Vma2p, CPY, or ALP (Pho8p) (all from Molecular Probes). Secondary antibody was FITC donkey anti-mouse (Jackson Immunoresearch, West Grove, PA). GFP-Luv1p–expressing cells were grown in selective medium containing galactose and resuspended in YPD for one doubling before visualization or fixation. pBMKX22, a URA marked plasmid that contains KEX2 under the GAL promoter, and anti-Kex2p antibody were generous gifts of R. Fuller (University of Michigan, Ann Arbor, MI) and were used as described (Redding et al., 1991 ). Anti-Pep12 mAb was a generous gift from T. Stevens (University of Oregon, Eugene, OR) and was used at a 1:1000 dilution. Anti-HA antibody was acites fluid from Roche Molecular (Basel, Switzerland). FITC and Texas red–conjugated secondary antibodies were from Calbiochem (La Jolla, CA). Immunofluorescence, GFP, and FM4-64 were viewed on a Nikon (Garden City, NY) E600 microscope with the use of a Hammamatsu ORCA digital camera (Hamamatsu Photonics, Kyoto, Japan) and QED Imaging software (QED Imaging, Pittsburgh, PA) on an Apple (Cupertino, CA) Macintosh platform, and figures were prepared with the use of Adobe (Mountain View, CA) Photoshop.
Metabolic labeling and immunoprecipitation of CPY immunoprecipitates were performed with the use of a method from the laboratory of T. Stevens (University of Oregon, Eugene, OR). One OD600 unit of early log-phase cells was incubated in 1 ml of Met-free and Cys-free synthetic medium, 50 mM potassium phosphate, pH 5.7, with 200 μCi of 35S Trans-label (New England Nuclear, Boston, MA) and 2 mg/ml BSA. Cells were pulsed for 10 min at 30°C and chased by addition of 10 μg/ml cold Met and Cys for 40 min, whereupon labeling was terminated on ice with 10 μl of sodium azide. Cells were sedimented from the medium, and 100 μl of 10× IP buffer was added to the medium fraction (final concentration: 10 mM Tris, pH 8, 0.1% SDS, 0.1% Triton X-100, 2 mM EDTA, 0.5 mM PMSF, 1 μg/ml leupeptin, 1 μg/ml pepstatin). The cell fractions were converted to spheroplasts by incubating for 30 min at 30°C in 1.4 M sorbitol, 50 mM Tris, pH 7.4, 2 mM magnesium chloride, 10 mM sodium azide, 0.3% 2-mercaptoethanol, and 30 μg/ml Zymolyase 100T. Spheroplasts were lysed by boiling in 0.5% SDS and diluted to 1 ml in IP buffer (1× final concentration). Samples were boiled for 5 min and precleared with 20 μl of protein A–Sepharose, and CPY was immunoprecipitated by sequentially adding 0.5 μl of purified anti-CPY antiserum (a generous gift from the laboratory of R. Scheckman, University of California, Berkeley, CA) and 10 μl of protein A–Sepharose per sample with 1-h incubations for each. Immunoprecipitates were washed once through a 30% sucrose cushion in IP buffer and twice with plain IP buffer and resolved by SDS-PAGE; protein bands were imaged with the use of Kodak (Rochester, NY) X-AR film and quantified with the use of a Bio-Rad (Richmond, CA) CS screen and a Molecular Dynamics (Sunnyvale, CA) phosphoimager with Molecular Analyst software.
Cellular extracts were prepared essentially as described by Holthuis et al. (1998) , who noted the addition of the energy poisons sodium fluoride and sodium azide to cells during spheroplasting. Early log-phase cells were sedimented, washed with 200 mM Tris, pH 8, 20 mM EDTA, 1% 2-mercaptoethanol, 5 mM sodium azide, and 5 mM sodium fluoride, converted to spheroplasts in the presence of 5 mM sodium azide and 5 mM sodium fluoride, and lysed by Dounce homogenization on ice in Tris/sorbitol buffer (50 mM Tris, pH 7.5, 200 mM sorbitol, 1 mM EDTA, with protease inhibitors added [1 μM each PMSF, pepstatin, benzamidine, 1 μg/ml each leupeptin, aprotinin, antipain, α-2-macroglobulin, and chymostatin]). Trypsin protease accessibility was determined by incubating cell extracts in 100 μl of 50 mM Tris, pH 7.5, 10 mM EDTA, 30 mM NaCl, 0.5% 2-mercaptoethanol, with or without l-1-tosylamide-2-phenylethylchloromethyl–treated trypsin, for 20 min at 25°C. Digestion was stopped by 10% trichloroacetic acid precipitation, and the proteins were identified by SDS-PAGE followed by Western blotting. For protein solubility, cell extracts were incubated on ice for 30 min in 150 μl of Tris/sorbitol buffer with the indicated compounds and then centrifuged for 30 min at 150,000 × g. For sucrose fractionation, 1 ml of extract from 50 ml of OD600 = 0.5 cells was layered onto 10 ml of a sucrose step gradient and buffered with 20 mM HEPES, pH 6.8, with steps of 1 ml at 60%, 2 ml each at 42, 36, 30, and 24%, and 1 ml at 18% sucrose (wt/wt). Gradients were centrifuged in an SW 41 rotor (Beckman Coulter, Fullerton, CA) at 150,000 × g for 18 h, and fractions were collected manually from the top. Sucrose concentration was determined on a Milton Roy (Ivyland, PA) refractometer (generously provided by R.T. Simoni, Stanford University, Stanford, CA). SDS-PAGE and Western blotting were performed with the use of Bio-Rad apparatus, Millipore (Bedford, MA) Immobilon membrane, antibodies at 1:1000 in TBS-Tween, 5% BSA, Amersham (Arlington Heights, IL) HRP-conjugated anti-mouse and anti-rabbit secondary antibodies (1:5000), and the Amersham ECL kit according to the directions of the manufacturer. Anti-HA mAb was from Roche Molecular and was used at a 1:1000 dilution. Anti-Kex2p was precleared against fixed kex2Δ yeast cells and used at a 1:500 dilution and detected robustly by sandwich amplification with the use of biotinylated anti-rabbit (1:1000) and HRP-streptavidin (1:1000) (Jackson Immunoresearch). Anti-Pep12p mAb was used at a dilution of 1:5000.
Previously, work from our laboratory identified mutants that depend on calcineurin for viability (Garrett-Engele et al., 1995 ). These mutants were generated in strain PGY2, which contains null mutations in both calcineurin catalytic subunits (cna1Δ1::hisG and cna2Δ1::HIS3) but contains a plasmid-borne copy of CNA2 that is galactose inducible. This study focuses on isolate #112 from that screen. Isolate #112 was transformed with a low-copy genomic LEU2 library (American Type Culture Collection [Rockville, MD] 77162; F. Spencer and P. Hieter, unpublished), and colonies were selected for their ability to grow on dextrose, i.e., without calcineurin. Six complementing plasmid clones were obtained; sequencing of the end of each insert indicated that they overlapped an 11-kb region of chromosome IV (Saccharomyces Genome Database). Subcloning and testing for complementation narrowed the complementing DNA to a 4.5-kb SalI–SacI genomic fragment containing YDR027c (this gene has since been reported as RKI1/TCS3/VPS54; in this report, we refer to this gene as LUV1). We constructed a null mutation in this gene, luv1Δ, that deleted the ORF (see MATERIALS AND METHODS). A strain containing this null mutation failed to complement isolate #112 in a diploid, and tetrad analysis indicated that the mutations contained in these two strains were allelic and that the sporulation efficiency for the luv1Δ/#112 diploid was unusually low. luv1 mutants (both the original isolate and luv1Δ) were extremely slow growing (but viable) in strains lacking calcineurin. luv1Δ alleles were used in the following analysis, but isolate #112 showed similar phenotypes as luv1Δ strains for all aspects tested.
LUV1 is predicted to encode an 889-amino acid protein (Figure (Figure1A)1A) with no predicted transmembrane domain, signal sequence, metal-binding site, or other identifiable motif (Saccharomyces Genome Database, Yeast Proteome Database, BLOCKS, MOTIF), except for a coiled-coil domain. Coils 2.1 (Lupas et al., 1991 ) gave a strong prediction of a coiled coil in the region between amino acids 280 and 320 (Figure (Figure1B),1B), regardless of search window size, and weaker predictions of a coiled coil that vary depending on window size around amino acids 180, 340, and 680. XREF searches with the use of the Luv1 protein sequence revealed three homologues of similar size, Schizosaccharomyces pombe SPAC2F3.10 (EMBL), Arabidopsis thaliana ATF24J7.50 (EMBL), and Caenorhabditis elegans CEY106G6H (EMBL), all proteins of unknown function. The strong prediction of a single coiled coil in the N-terminal third of the protein is shared with the homologues, which supports the prediction of this motif at this location in these proteins.
Calcineurin mutants are sensitive to Na+, Li,+ Mn2+, Cd2+, and high pH and are resistant to calcium ion (Nakamura et al., 1993 ; Breuder et al., 1994 ; Cunningham and Fink, 1994 ; Mendoza et al., 1994 ; Farcasanu et al., 1995 ; Tanida et al., 1995 ; Pozos et al., 1996 ; Withee et al., 1997 , 1998 ). vma mutants, which like luv1 were shown to require calcineurin function, share some of these calcineurin mutant phenotypes, including sensitivities to pH extremes, Zn2+, Mn2+, and Cd2+ (Anraku et al., 1992 ; Garrett-Engele et al., 1995 ). Therefore, we examined the growth properties of luv1 mutants and found that they were sensitive to Zn2+, Mn2+, Cd2+, and extremes in pH but not to Na+ or Li+ (Figure (Figure2;2; our unpublished results). luv1 was also sensitive to the drug FK506 (Figure (Figure2),2), consistent with its requirement for calcineurin. Calcineurin mutants are sensitive to hygromycin B (Withee et al., 1998 ), an aminoglycoside that inhibits protein translation. luv1 mutants similarly showed sensitivity to hygromycin B (Figure (Figure2).2). Additionally, luv1 was sensitive to caffeine, a cAMP diesterase inhibitor, and failed to grow at high (37°C) and low (11°C) temperatures (Figure (Figure2;2; our unpublished results). Calcium did not affect the growth of luv1; however, with an additional calcineurin mutation, the cells were very sensitive to this ion (our unpublished results). Unlike some of the genes shown to require calcineurin function, i.e., those involved in cell wall biosynthesis, luv1 mutants were not sensitive to hypoosmotic stress (our unpublished results).
The ion sensitivities of luv1 mutants are reminiscent of those of vma mutants and suggest that luv1 mutants are defective for vacuole function. We visualized vacuolar structures in luv1 cells with the use of the lipophilic dye FM4-64, which, at 25°C, is internalized in living cells through endocytosis, transported through the endosomal compartments, and accumulates at the vacuole (Vida and Emr, 1995 ). In wild-type cells at late log phase, numerous small vacuoles coalesce into one larger vacuole. When viewed under Nomarski differential interference contrast microscopy, luv1 cells did not have a visible vacuole. Instead, many (several tens) small vesicular bodies were seen dispersed throughout the cell; these vesicles also accumulated FM4-64 (Figure (Figure3,3, FM4-64). Vesicles were also seen in daughter buds. Given the absence of a morphologically identifiable vacuole, we named this gene LUV1 for Loss Upsets Vacuole.
We further examined luv1 cells for vacuolar acidification and the presence of two vacuolar proteins, CPY and ALP. The vacuole lumen is acidified by the vacuolar H+-ATPase, and the accumulation of a fluorescent dye, quinacrine, requires this acidification. luv1 cells were incubated with quinacrine and viewed under fluorescence microscopy. Wild-type cells showed quinacrine fluorescence in large vacuolar structures, whereas luv1 mutants displayed numerous punctate bodies of quinacrine staining (Figure (Figure3,3, quinacrine). A similar, irregular punctate pattern was seen for the 60-kDa subunit of the membrane-bound vacuolar H+-ATPase subunit, as detected by indirect immunofluorescence (Figure (Figure3,3, Vma). CPY is a lumenal protease that is transported to the vacuole by a receptor, Pep1/Vps10p (Marcusson et al., 1994 ; Horazdovsky et al., 1995 ). In wild-type cells, CPY localized to the vacuole lumen; in luv1 mutants, a punctate pattern was seen for CPY, as detected by indirect immunofluorescence (Figure (Figure3,3, CPY). ALP is a membrane-bound protein that reaches the vacuole by an alternative pathway to CPY (Cowles et al., 1997 ; Piper et al., 1997 ; Stepp et al., 1997 ). Wild-type cells contained ALP at the vacuolar membrane, whereas in luv1 mutants, punctate staining was observed (Figure (Figure3,3, ALP). These observations all indicate that the vesicles observed in luv1 mutants are fragmented vacuoles.
Class B and C vps mutants show moderate and severe vacuole fragmentation, respectively. In these mutants, 31 to 97% of total CPY is missorted (Raymond et al., 1992 ), and as a consequence, it is secreted into the growth medium. Because luv1 mutants showed vacuole fragmentation, we examined CPY processing and sorting in luv1 cells. CPY is normally processed from a 67-kDa precursor (p1) in the endoplasmic reticulum to a fully glycosylated 69-kDa form (p2) in the Golgi. In the vacuole, precursors are cleaved to the 65-kDa mature (m) form by proteinase A (reviewed by Van Den Hazel et al., 1996 ). The amount of CPY secreted by luv1 was determined by metabolic labeling and immunoprecipitation. With the use of the method described (see MATERIALS AND METHODS), the luv1 mutant (YMC10; Table Table1)1) secreted 28% CPY into the medium (S) (Figure (Figure4).4). This was substantially less than we observed for other vps mutants; for example, vps17 (Table (Table1)1) secreted 89% CPY (our unpublished results). In luv1 mutants, the correctly sorted CPY was mature, whereas missorted CPY was the fully glycosylated, Golgi-modified p2 form (Figure (Figure4).4).
luv1 mutants have fragmented vacuoles and missort CPY. In this way, luv1 mutants resemble some vps mutants. We examined select vps mutants to determine if they exhibited drug and ion sensitivities similar to luv1. Representatives of each of the six phenotypic classes, A–F, were tested. LUV1 was disrupted in the parental strain for comparison (to give YMC4). Many of the vps mutants also displayed growth properties similar to those of luv1, such as sensitivity to high temperature, pH extremes, hygromycin, Zn2+ (except vps1), and caffeine (except vps1 and vps45) (Table (Table2).2). Unlike luv1, the vps mutants tested were not sensitive to Mn2+ or Cd2+. Interestingly, some of the vps mutants were sensitive to FK506, indicating that they also require calcineurin function for viability. Although vps35 (class A) and vps17 and vps41 (class B) showed the least severe vacuole defects and did not show FK506 sensitivity (our unpublished results), other vps mutants, such as pep5/vps11 (class C), vps15 and vps45 (class D), and vps1 (class F), did show FK506 sensitivity.
To further characterize Luv1 protein function, we examined its localization in wild-type cells. We first characterized Luv1p subcellular localization through biochemical fractionation. An HA-tagged LUV1 allele was constructed (see MATERIALS AND METHODS), and we determined that plasmid borne, low-copy HA-LUV1 complements both the temperature-sensitive growth and fragmented vacuole phenotypes of luv1 (our unpublished results). Western analysis of protein extracts from HA-LUV1 cells showed a single predominant protein band at ~105 kDa. Most HA-Luv1 protein sedimented from cell extract at 100,000 × g, indicating that it may associate with a transport vesicle, Golgi, or early endosomal membrane fraction (Figure (Figure5A).5A). We attempted to solubilize HA-Luv1p by pretreating cell lysates with various chemicals before centrifugation at 150,000 × g into pellet (P) and supernatant (S) fractions. HA-Luv1p was partially solubilized after treatment with 2% Triton X-100 and was fully solubilized by 2% 3-([3-chloramidopropyl]dimethylammonio)-2-hydroxy-1-propanesulfonate, 1 M NaCl, and 0.1 M Na2CO3, pH 11 (Figure (Figure5C).5C). The salt and carbonate solubility suggests that Luv1p associates with the P100 fraction as a peripheral membrane protein. In support of this, the HA epitope was undetectable after trypsinization of cell extracts, indicating that at least the N terminus of Luv1p is accessible to protease (Figure (Figure5B).5B).
To further characterize the complex with which Luv1p associates, the 13,000 × g supernatant (S13) cellular fraction from HA-Luv1p–expressing cells was subjected to equilibrium sedimentation through a sucrose gradient. Fractions were collected and analyzed by SDS-PAGE and Western blotting. Anti-HA antibody detected HA-Luv1p in two peaks, the major peak between 20 and 25% sucrose and the minor peak at 35% sucrose (Figure (Figure5D).5D). Pep12p is a PVC t-SNARE that mediates vesicle traffic between the Golgi and the PVC (Becherer et al., 1996 ). Anti-Pep12p antibody detected Pep12p in one peak that partially overlapped those fractions that contain the major peak of HA-Luv1p and did not detect Pep12p in the fractions that contain the minor peak of HA-Luv1p (Figure (Figure5D).5D). The protease Kex2p cycles between the TGN and an endosomal compartment (reviewed by Conibear and Stevens, 1998 ), and anti-Kex2p antibody detected Kex2p in the same fractions as HA-Luv1p, including the minor peak (Figure (Figure5D).5D). This indicates that Luv1p may be in the same subcellular structures as Kex2p (i.e., TGN and/or endosome) but is not in the same structures, or at least not all of the same structures, as Pep12p (i.e., PVC).
To determine the cellular localization of Luv1p, we fused GFP to Luv1p (see MATERIALS AND METHODS). Plasmid-borne, galactose-inducible GFP-LUV1 complemented the growth defect of luv1 mutants at 37°C and restored normal vacuolar morphology to this strain (our unpublished results). Cells expressing GFP-Luv1 protein showed the bulk of fluorescence as a few large dots visible inside the cell (Figure (Figure6A,6A, wt). Cells expressing GFP-Luv1p were fixed and immunostained to determine Pep12p localization. Pep12p was also detected as a few large dots inside the cell, although Pep12 dots did not colocalize with GFP-Luv1 dots (Figure (Figure6A,6A, wt, merge). In vps27, a class E vps mutant, prevacuolar material accumulates into a morphologically exaggerated PVC. In this mutant, Pep12p distribution is altered and accumulates in this exaggerated PVC (Piper et al., 1995 ; Voos and Stevens, 1998 ) (Figure (Figure6A).6A). The GFP-Luv1p fluorescence pattern in the vps27 mutant appeared the same as in wild-type cells (Figure (Figure6A,6A, vps27), indicating that Luv1p distribution is not altered in this mutant. Also, even in the vps27 mutant, GFP-Luv1p did not colocalize with Pep12p (Figure (Figure6A,6A, vps27). Together, these results indicate that Luv1p does not localize with Pep12p at the PVC.
We next compared Luv1p and Kex2p subcellular localization by means of immunofluorescence. Cells expressing low-copy HA-Luv1p and overexpressing Kex2p were fixed and double immunostained for HA and Kex2p. Both anti-HA and anti-Kex2p immunofluorescence patterns appeared as a few dots inside the cell, and an overlay of separate HA and Kex2p images indicated a substantial overlap of HA-Luv1p and Kex2p localization (Figure (Figure6B).6B). This confirms that Luv1p and Kex2p colocalize in the yeast cell.
As described above, the fluorescent lipophilic dye FM4-64 is endocytosed in living cells, travels through the endocytic pathway, and accumulates at the vacuole (Vida and Emr, 1995 ). Wild-type cells expressing GFP-Luv1p (YMC2) were loaded with FM4-64 dye and examined by fluorescence microscopy during the time course of FM4-64 endocytosis to the vacuole for GFP, FM4-64, or both. It was observed that GFP-Luv1p fluorescence colocalized with FM4-64 during that time course (our unpublished results). In cells held at 15°C, FM4-64 is endocytosed normally but accumulates at an endosomal compartment rather than at the vacuole (Vida and Emr, 1995 ). Wild-type cells expressing GFP-Luv1p were loaded with FM4-64 dye, incubated at 15°C, and again examined for GFP fluorescence, FM4-64, or both. At 15°C, most of the GFP dots also fluoresced with FM4-64, indicating colocalization (Figure (Figure7,7, merge). This indicates that GFP-Luv1p localizes to the endosomal compartment that accumulates FM4-64 under these conditions and that the compartment containing Luv1p is part of the endocytic pathway.
The vps class E mutations block transport out of the PVC both anterograde to the vacuole and retrograde to the Golgi; thus, these mutants accumulate an exaggerated PVC (reviewed by Conibear and Stevens, 1998 ). These mutants also have vacuoles that contain ALP, and the vacuoles are thought to form because the transport block to the vacuole is incomplete; in addition, the ALP pathway is still functional. We examined the effect of a luv1 mutation on vacuolar morphology in a vps27 (class E) mutant with the use of FM4-64 dye. Cells were loaded with dye at 25°C to promote full endocytosis of the dye and accumulation at vacuolar structures. Wild type (SF838-9DaR2L1) and vps27 mutants showed FM4-64 staining of large, lobed, normal vacuolar structures, whereas luv1 mutants (YMC4) showed vacuolar fragmentation (Figure (Figure8).8). In contrast to vps27 or the parental strain, the vps27 luv1 double mutant (YMC9) showed FM4-64 staining very similar to that of the luv1 single mutant, with numerous small vesicles and no visible vacuolar structures (Figure (Figure8).8). Also, as seen in luv1 mutants (Figure (Figure3),3), the vesicles seen in the vps27 luv1 double mutant stained with quinacrine, indicating that they are acidified (our unpublished results). This suggests that the same vacuolar fragmentation seen in the luv1 single mutant also occurs in the vps27 luv1 double mutant.
We have isolated a mutant of LUV1/RKI1/TCS3/VPS54/YDR027c that requires calcineurin for viability. To identify the function of the LUV1 gene product, we characterized the luv1 mutant, localized the Luv1 protein, and characterized biochemical properties of the protein. We propose that Luv1p acts at the late Golgi and early endosome to mediate the integrity of, or transport through, these compartments.
luv1 mutants display ion phenotypes (Figure (Figure2)2) that suggest some loss of vacuole function in these cells. The Mn2+, Zn2+, and Cd2+ sensitivities of luv1 mutants suggest that these ions are not effectively sequestered to the vacuole. These sensitivities may result from a loss of polyphosphate, which is thought to help sequester ions in the mature vacuole (reviewed by Klionski et al., 1990 ), or from a loss of vacuolar integrity. Many of these sensitivities are shared with vps mutants (Table (Table2),2), especially those that, like luv1, show severely disrupted vacuolar morphology. We show here that luv1, calcineurin, and some vps mutants are sensitive to the aminoglycoside hygromycin B. This sensitivity is likely not due to the effect of this drug on protein translation, because luv1 and calcineurin mutants show no sensitivity to cycloheximide, which inhibits translation. Dean (1995) has reported that glycosylation-defective mutants are sensitive to aminoglycoside drugs. However, luv1 shows correct CPY and ALP processing, indicating that the hygromycin sensitivity is not a result of glycosylation defects. In mammalian cells, aminoglycoside drugs, including hygromycin, have been shown to interfere with coatomer (COPI) coat formation and secretion (Hudson and Draper, 1997 ; Hu et al., 1999 ). Yeast COPI mediates retrograde vesicle transport from the Golgi to the endoplasmic reticulum (reviewed by Cosson and Letourneur, 1997 ). Perhaps hygromycin B inhibits the growth of some vps, luv1, and glycosylation-defective mutants by interfering with COPI-mediated transport.
luv1 mutations cause severe vacuolar fragmentation, yet the fragments seen in the luv1 mutant contain characteristic vacuolar proteins: the vacuolar H+-ATPase, CPY, and ALP (Figure (Figure3).3). The presence of the vacuolar H+-ATPase and vesicle acidification in luv1 cells indicates that the vacuolar H+-ATPase assembles and functions in luv1 mutants. Others have recently characterized luv1/tcs3/vps54 mutants and also noted vacuole defects (Bensen et al., 2000 ; Conibear and Stevens, 2000 ). Bensen et al. (2000) isolated luv1/tcs3 in a screen for synthetic lethality with a temperature-sensitive, partial-loss-of-function clathrin mutation. They found that luv1/tcs3 mutants show a defect in CPY sorting, in agreement with our findings, and also that ALP processing and sorting is normal in luv1/tcs3 mutants. Similarly, Conibear and Stevens (2000) reported no gross defects in ALP processing and sorting in luv1/vps54 and no defects in invertase secretion, although they noted significantly more CPY secretion. Either strain or protocol differences may account for these quantitative differences; however, qualitatively, both of these studies are consistent in finding that luv1 mutants show a CPY protein-sorting defect. Because the luv1 mutation appears to affect CPY sorting more than ALP and does not affect secretion at all, Luv1p may be required at a step after the CPY pathway diverges from the secretory and ALP pathways.
The LUV1 gene product is predicted to contain a coiled-coil domain. Coiled coils are features shared by SNAREs and intermediate filament proteins. However, Luv1p does not otherwise resemble a SNARE, because it is much larger than SNARE proteins, lacks a transmembrane domain, and is solubilized by salt or high pH (Figure (Figure5C).5C). Luv1p does not resemble cytoskeletal proteins either, because the coiled-coil region is relatively small and is not predicted to form a structural rod. Nevertheless, there are several proteins that, like Luv1p, lack identifying functional motifs except for an N-terminal coiled coil (Figure (Figure1B).1B). Members of the exocyst complex, some vps class C proteins, and many vps class E proteins share this feature. The exocyst complex is a multimer of at least seven proteins (Bowser and Novick, 1991 ; Bowser et al., 1992 ; TerBush and Novick, 1995 ) that is thought to target transport vesicles to a specific intracellular location, the bud tip, by mediating vesicle–cytoskeleton interactions (TerBush et al., 1996 ). Although an exocyst-like complex has not been found at the PVC or vacuole, it has been proposed that the vps class C RING complex functions similarly to target vesicle fusion to a specific subregion of the vacuole (Rieder and Emr, 1997 ). Several of the components of the RING complex, Pep5/Vps11p, Vps16p, and Pep3/Vps18p, have predicted short coiled-coil motifs (Rieder and Emr, 1997 ). Perhaps Luv1p is part of an analogous complex at the TGN/early endosome (see below). The biochemical characteristics of Luv1p are consistent with this suggestion. Luv1p sediments with a high-speed pellet fraction (Figure (Figure5A,5A, P100), suggesting that it associates with transport vesicles, Golgi, and/or early endosome. Luv1p is solubilized from this fraction by treatment with high pH or salt (Figure (Figure5C),5C), indicating that it is a peripheral membrane protein. Recently, others have identified two proteins with which Luv1/Vps54p associates: Sac2/Vps52p and Vps53p (Conibear and Stevens, 2000 ). We also detected two proteins of the approximate molecular weight of Sac2/Vps52p and Vps53p in immunoprecipitations of HA-Luv1p (our unpublished results).
Like luv1, some of the class B vps mutants show vacuole fragmentation into numerous vesicles that are acidified and contain ALP (Raymond et al., 1992 ), and by these criteria, luv1 could be classified as a vps class B mutation. A complex of several class B Vps proteins, called the retromer, has been characterized to be required for endosome-to-Golgi recycling of membrane and cargo (Seaman et al., 1998 ; reviewed by Conibear and Stevens, 1998 ). Although Luv1p has not been found as part of the retromer complex, its localization and mutant phenotypes suggest that it may be involved at the same organelle or sorting step.
Most of the Luv1 protein shows colocalization with the TGN protease Kex2p (Figure (Figure6B)6B) and cofractionates on a sucrose gradient with Kex2p (Figure (Figure5D).5D). In agreement with this, Conibear and Stevens (2000) also report that Luv1/Vps54p colocalizes with the marker protein A-ALP at the TGN. Luv1p also appears to localize to an endosomal compartment, as demonstrated by substantial colocalization of a GFP-Luv1 fusion protein with FM4-64 dye in wild-type cells kept at 15°C (Figure (Figure7).7). However, this endosomal compartment is not the late endosome or PVC, because no colocalization was seen of GFP-Luv1p and the PVC t-SNARE Pep12p, even in a vps27 mutant (Figure (Figure6A)6A) that accumulates Pep12p in an exaggerated PVC (Piper et al., 1995 ; Voos and Stevens, 1998 ). Thus, we conclude that Luv1p localizes to both the TGN and an endosomal compartment. Others have shown that the endosomal/TGN syntaxin Tlg2p and FM4-64 colocalize very well at 15°C (Abeliovich et al., 1998 ) and that Tlg2p and Kex2p colocalize well (Holthuis et al., 1998 ; Lewis et al., 2000 ). The Golgi marker Sec7p was also shown to colocalize with FM4-64 dye at early stages of endocytosis, and Sec7p showed some colocalization with Tlg2p (Lewis et al., 2000 ). None of these proteins were shown to localize to the late endosome/PVC. These findings suggest that Luv1p, Kex2p, Tlg2p, Sec7p, and FM4-64 at 15°C may all colocalize in the cell at an early endosomal compartment and/or the TGN.
One current model of intracellular transport in yeast takes into account the existence of such an early endosome that is distinct from both the late endosome/PVC and the late Golgi/TGN (reviewed by Pelham, 1999 ). In this model, the TGN is where the secretory pathway and both the CPY and ALP vacuolar targeted pathways diverge, whereas the early endosome is where the CPY branch of vacuolar traffic meets endocytic traffic and where retrograde traffic flows back to the TGN. In this model, the Golgi does not directly receive endocytic traffic, and the late endosome/PVC does not directly receive TGN traffic, because both endocytic and TGN traffic are presorted through the early endosome. Within the framework of this model, we propose that Luv1 protein is required to stabilize and/or localize the early endosomal compartment and that luv1 phenotypes can be explained by mislocalized or abnormal early endosome, which subsequently matures to have vacuolar characteristics. Our reasoning is as follows. First, as described above, the luv1 mutation appears to preferentially affect the pathway defined by CPY transport, as opposed to the secretory or ALP pathway. This would suggest that Luv1p functions after the secretory and ALP transport pathways diverge from the CPY pathway, or after the TGN. Second, the vesicles visible in luv1 mutants display properties characteristic of vacuoles, yet they also rapidly accumulate FM4-64 dye at 15°C as at higher temperatures (our unpublished results, but compare with Figure Figure3).3). Thus, these vesicles also appear to have properties of early endosomes. Third, two additional defects have been noted in luv1/tcs3 mutants (E. Bensen and G. Payne, personal communication). One defect is a kinetic delay in Ste3p turnover. This indicates that there is an endocytosis defect in luv1/tcs3 mutants that compromises transport at some point between the plasma membrane and the vacuole. The other defect is that luv1/tcs3 mutants have reduced α-factor processing due to reduced stability of the Golgi-localized protease Kex2p (Bensen et al., 2000 ). This finding suggests that luv1/tcs3 cells have defects either in some step of retrograde transport between the early endosome and the Golgi or in sorting from the Golgi to the early endosome (reviewed by Wilsbach and Payne, 1993 ). Similarly, Conibear and Stevens (2000) noted proteolysis of Pep1p/Vps10p and the model protein A-ALP in luv1 mutants, indicating that retrieval from an endocytic compartment to the Golgi was impaired. Together, these defects suggest that in the absence of Luv1p, transport out of the early endosome is disrupted, in both directions, to the vacuole and back to the TGN. Fourth, the class E vps27 mutation incompletely blocks anterograde transport out of the PVC to the vacuole; thus, vps27 mutants still form vacuoles (reviewed by Conibear and Stevens, 1998 ). We find that a vps27 luv1 double mutant appears morphologically like a luv1 mutant, with an accumulation of many small vacuole-like vesicles but no large vacuole (Figure (Figure8).8). This observation suggests that luv1 disrupts an earlier compartment than the class E PVC or disrupts the remaining vacuolar transport in the vps27 mutant. Finally, we note that luv1 suffers growth defects at 37°C, at which temperature wild-type cells will accumulate endosomes (Mulholland et al., 1999 ). Together, these observations suggest that Luv1p acts to localize, stabilize, or facilitate traffic through the early endosome in wild-type cells, in vps27 mutants, and perhaps at high temperatures.
An alternative model that has been used to explain intracellular transport to the vacuole describes a single transport step from the TGN to the late endosome/PVC, another step from the plasma membrane to the late endosome/PVC (endocytosis), and retrograde transport from the late endosome back to the TGN (reviewed by Conibear and Stevens, 1998 ). This model does not specify the involvement of an early endosome in transport to the vacuole. Within the framework of this model, Conibear and Stevens (2000) have proposed that Luv1p/Vps54p acts to facilitate retrograde transport from the late endosome/PVC to the TGN. We have expanded this view of Luv1p function to incorporate the finding that the Luv1p-containing compartment also receives early endocytic traffic.
A recent report suggests that Luv1p may interact with microtubules. Smith and coworkers (1998) isolated luv1/rki1 in a synthetic lethal screen with an rbl2 null mutant. They reported that luv1/rki1 null mutants show phenotypes consistent with microtubule defects, namely spindle defects, incomplete tetrad formation from homozygous rki1Δ diploids, and sensitivity to cold (15°C) and the microtubule-depolymerizing drug benomyl. In our strain background, luv1 did not show benomyl sensitivity (our unpublished results). Smith et al. also demonstrated that Rbl2p binds in vitro to Luv1p/Rki1p and β-tubulin. However, neither Conibear and Stevens (2000) nor ourselves (our unpublished results) detected Rbl2p or β-tubulin in immunoprecipitates of Luv1/Vps54p. The role of microtubules in vesicle transport, organelle movement, and maintenance is well established in mammalian cell systems (reviewed by Bloom and Goldstein, 1998 ). In yeast, however, there are few studies implicating microtubules in vesicle transport: microtubules may be involved in autophagosome travel to the vacuole (Lang et al., 1998 ), and Golgi fragmentation has been reported in cells with disrupted microtubules (Rambourg et al., 1996 ). Any involvement with Luv1p/Rki1p with microtubules in transport is unclear and awaits further investigation.
We and others find that luv1, vma, and several vps mutants are FK506 sensitive and thus require calcineurin for viability (Table (Table2)2) (Garrett-Engele et al., 1995 ; Hemenway et al., 1995 ). Why do mutants that disrupt vacuolar function require calcineurin? One major function of calcineurin is activation of the Crz1p transcription factor, which results in increased expression of several genes, including PMC1, a vacuolar Ca2+-ATPase (Matheos et al., 1997 ; Stathopoulos and Cyert, 1997 ). The growth of vma2 crz1 double mutants is severely compromised (A.M. Stathopoulos and M.S. Cyert, unpublished results), and the FK506 sensitivity of vma2 mutants is suppressed by overexpression of PMC1 (L. Chen and M.S. Cyert, unpublished results). Thus, vma mutants seem to require calcineurin solely for its ability to promote PMC1 expression. In contrast, we found that the growth of a luv1 crz1 double mutant was equivalent to that of a luv1 mutant (16 tetrads dissected) and that the FK506 sensitivity of luv1 mutants was not suppressed by overexpression of either PMC1 or PMR1 (our unpublished results), the latter a Golgi-localized Ca2+-ATPase whose expression is also regulated by calcineurin via Crz1p (Matheos et al., 1997 ; Stathopoulos and Cyert, 1997 ). These observations suggest that calcineurin/Crz1p-regulated gene expression is not required for the viability of luv1 mutants. Thus, the role of calcineurin in maintaining luv1 mutant viability remains to be elucidated. Calcineurin has been shown to regulate vesicle transport in higher eukaryotic systems and to dephosphorylate several proteins required for clathrin-mediated vesicle recycling: dynamin, amphiphysin, and synaptojanin (reviewed by Marks and McMahon, 1998 ). The GTPase dynamin is generally thought to pinch or sever the neck of the developing vesicle, synaptojanin is a phosphatidylinositol phosphatase thought to modify lipid phosphate to recruit components to a site of vesicle budding, and amphiphysin is thought to recruit dynamin. Yeast cells do contain homologues of these proteins, and the dynamin Vps1p and the synaptojanin Inp53p are thought to act in trafficking from the Golgi to the vacuole (Conibear and Stevens, 1998 ; Bensen et al., 2000 ). However, in yeast, no studies have yet defined a role for calcineurin in the regulation of these proteins. In light of the work by Bensen and coworkers (2000) , showing that a luv1/tcs3 mutation is synthetically lethal with a clathrin mutation, and our results, a possible role of yeast calcineurin in a clathrin-mediated vesicle transport process is particularly intriguing.
We express our gratitude to T. Stearns for generously providing reagents, equipment, and advice, to E. Bensen, G. Payne, E. Conibear, and T. Stevens for selflessly sharing results and interpretations before publication, and to K. Bartz, T. Stevens, T. Sato, T. Darsow, S. Emr, K. Blumer, K. Redding, R. Fuller, E. Jones, and M. Rexach for reagents or reagent-grade discussion. We acknowledge the Saccharomyces Genome Database (Cherry et al., 1998 ) and the Yeast Protein Database (Costanzo et al., 2000 ) as being invaluable information resources for this work. M.J.C. was supported by National Institutes of Health grant 5 T32 GM07276. M.S.C. was supported by biomedical scholar award 92-42 from the Lucille P. Markey Charitable Trust, National Science Foundation Young Investigator award MCB-9357071, and funds from the Proctor and Gamble Company. This work was funded by National Institutes of Health research grant GM-48729 (to M.S.C.), which was also a source of support for M.J.C.