Sphingolipid requirement for generation of a functional v1 component of the vacuolar ATPase.

There has been no previous indication that vacuolar ATPases (V-ATPases) require sphingolipids for function. Here we show, by using Saccharomyces cerevisiae sur4Delta and fen1Delta cells, that sphingolipids with a C26 acyl group are required for generating V1 domains with ATPase activity. Sphingolipids in sur4Delta cells contain C22 and C24 acyl groups instead of C26 acyl groups whereas about 30% of the sphingolipids in fen1Delta cells have C26 acyl groups and the rest have C22 and C24 acyl groups. sur4Delta cells have several phenotypes (vacuolar membrane ATPase, Vma-) that indicate a defect in the V-ATPase, and vacuoles purified from sur4Delta cells have little to no ATPase activity. These phenotypes are less pronounced in fen1Delta cells, consistent with the idea that the C26 acyl group in sphingolipids is necessary for V-ATPase activity. Other results show that the two V-ATPase domains, V1 and V0, are assembled and delivered to the vacuolar membrane in sur4Delta cells similar to wild-type cells. In vitro assembly studies show that V1 from sur4Delta cells associates with wild-type V0 but the complex lacks V-ATPase activity, indicating that V1 is defective. Reciprocal experiments with V0 from sur4Delta cells show that it is normal. We conclude that sphingolipids with a C26 acyl group are required for generating fully functional V1 domains.

little is known is the influence of membrane lipids on V-ATPase activity. Here we show that sphingolipids are necessary for V-ATPase activity in S. cerevisiae.
S. cerevisiae V-ATPase contains two components or domains, V 1 and V 0 , which associate to form an active V 1 V 0 complex on the vacuolar membrane (reviewed in Refs. [3][4][5]. The V 1 domain has ATPase activity and is composed of 8 different proteins. It can exist free in the cytoplasm or complexed with V 0 on the vacuolar membrane. The V 0 domain contains 5 protein subunits and is imbedded in the vacuolar membrane where it serves as a proton pore. The V 1 and V 0 domains are assembled and associate in the ER to form functional V 1 V 0 complexes, which are then transported from the Golgi to the vacuole (reviewed in Refs. 3 and 5). Alternatively, the two domains assemble independently and then associate once V 0 reaches the vacuole.
The ER is also where sphingolipid synthesis begins (reviewed in Ref. 6) with generation of ceramide. Ceramide is transported to the Golgi where it is converted sequentially into the complex sphingolipids inositol-phosphoceramide, mannoseinositol-phosphoceramide, and finally to mannose-(inositolphospho) 2 -ceramide. Complex sphingolipids are delivered to cellular compartments, particularly the plasma membrane (7) and to a lesser extent the vacuole (8).
One of the distinguishing features of sphingolipids in S. cerevisiae is the C26 acyl group. Fatty acids with 20 or more carbons, very long chain fatty acids (VLCFAs), are ubiquitous in nature, but little is known about their functions. VLCFAs are mostly found in the ceramide portion of sphingolipids. The importance of the C26 acyl component of S. cerevisiae sphingolipids was demonstrated by the isolation of mutant strains that do not make sphingolipids (9), but instead make a set of novel sphingolipid mimics in which ceramide is replaced by diacylglycerol (10). The presence of a C26 acyl group is the unique feature of these novel glycerolipids, which enables them to mimic some sphingolipid functions.
Further evidence for the essentiality of the C26 acyl group in sphingolipids comes from studies of the FEN1 (ELO2) and SUR4 (ELO3) genes. In fen1⌬ cells only 29% of the sphingolipids contain a C26 acyl group, the rest contain C22 and C24 acyl groups. The sphingolipids in sur4⌬ cells contain only C22 and C24 acyl groups and no C26s (11,12). Fen1p and Sur4p are components of the enzyme system that elongates C16 and C18 fatty acids to form VLCFAs. The exact function of Fen1p and Sur4p are unclear because the elongation system has not been fully characterized in any organism. sur4⌬ and fen1⌬ cells are viable, although they have many mutant phenotypes (reviewed in Ref. 13) and deletion of both genes is lethal (14).
A fraction of sphingolipids in higher eukaryotes also contain VLCFAs (15) and the mouse genes Ssc1 and Cig30 complement a sur4⌬ and a fen1⌬ mutant, respectively (16). Interestingly, the mice mutants Quaking and Jimpy, which develop intense tremors at the age of about 2 weeks as a result of severe demyelination of the central nervous system, have reduced levels of Ssc1 mRNA and reduced fatty acid elongation activity (17). Cig30 mRNA has the interesting property of being induced in brown adipose tissue when animals are exposed to cold temperature. These results suggest that VLCFAs are performing important, but unknown functions in mammals, and that some of these functions may be evolutionarily conserved.
Recently Kohlwein et al. (12) reported that sur4⌬ and fen1⌬ cells contain small vacuoles called fragmented vacuoles that fail to properly fuse to form larger vacuoles. This observation suggested to us that sphingolipids with a C26 acyl group are needed for some vacuolar function(s). Here we show that V 1 domains in sur4⌬ cells lack ATPase activity even though they associate with V 0 domains on the vacuolar membrane. Our data are the first to implicate sphingolipids with a C26 acyl group in the generation of a fully functional V 1 domain.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and Culture Conditions-Strains used in these experiments are listed in Table I. The sur4-⌬1::KAN and the fen1-⌬1::KAN alleles have codons 1-560 replaced with a kanamycin resistance cassette, generated by using the PCR and pUG6 as the template (18). Diploid cells transformed with deletion alleles were selected for G418 resistance (19). The expected deletion event was verified by PCR analysis of chromosomal DNA. Haploid offspring were obtained by sporulation and tetrad dissection. VMA13 under the control of its own promoter and having a Myc epitope immediately following the methionine start codon was carried in pRS316 (20).
Buffered medium was prepared by the addition of 50 mM MES and 50 mM MOPS to YPD, and the pH was adjusted to pH 5.5 with NaOH. YPD plates supplemented with 100 mM CaCl 2 , 4 mM CaCl 2 , or buffered to pH 7.5 with 100 mM Hepes were prepared as described previously (21). YPD medium contained 1% yeast extract, 2% Bacto-peptone, and 2% dextrose.
Quinacrine Staining and Semiquantitative Quinacrine Assay-Vacuolar accumulation of quinacrine was assessed by fluorescence microscopy as described by Roberts et al. (22) or by a semiquantitative quinacrine assay (21). Fluorescence measurements of cell suspensions were done in a Beckman spectrofluorometer (excitation ϭ 419 nm, emission ϭ 425 nm) and the OD at 600 nm was monitored in a spectrophotometer.
SDS-PAGE, Immunoblotting, and Antibodies-SDS-PAGE and immunoblotting were performed according to procedures recommended for the Bio-Rad Tray-Blot S.D. Semi-Dry transfer cell (Bio-Rad Inc.). Monoclonal antibodies against Vph1p, Vma1p, Vma2p, CPY, and ALP were from Molecular Probes. Anit-Myc antibodies were from Roach Applied Science. Dr. Patricia Kane provided anti-Vma5p. Nitrocellulose membranes (Bio-Rad) were washed four times after the first and second antibody reactions with 0.1% phosphate-buffered saline containing 0.1% Triton X-100. Secondary antibody was anti-mouse IgG conjugated to alkaline phosphatase (Sigma). Membranes were incubated for 5 min facedown in ECF substrate (Amersham Biosciences) and fluorescent signals were collected by using a Molecular Dynamics Storm Phosphorimager and quantified by using ImageQuant software (version 5.1).
HPLC Analysis of LCBs and LCBPs-Lipids were extracted from whole cells, converted to fluorescent derivatives and analyzed by HPLC as described previously (23). Miscellaneous Procedures-Vacuolar membrane vesicles were purified by centrifugation on Ficoll gradients and Mg-ATPase activity was measured at 23°C as described previously (22) except that membranes were not homogenized before centrifugation on the second Ficoll gradient.
Vacuolar membrane vesicles were also purified by sucrose density gradient centrifugation as previously described (25), except that the concentration of sucrose was increased to prevent membranes from pelleting on the bottom of the centrifuge tube. For these experiments, 4 ml of the membrane fraction was overlaid onto a 32-ml gradient composed of equal volumes of 10, 30, 50, and 60% (w/v) sucrose. The gradient was centrifuged for 35 min at 100,000 ϫ g in a Sorvall AH629 rotor at 4°C and then fractionated starting from the top into 9 fractions of 4, 7, 2, 6, 2, 6, 2, 7 ml and the resuspended pellet. Fractions were frozen in liquid nitrogen and stored at Ϫ80°C. V 1 V 0 complexes on vacuolar membranes prepared by sucrose gradient centrifugation were dissociated by treatment with KI as described previously (26,27).
Cell-free protein extracts used for analysis of Vph1p (Fig. 3), were prepared as described by Kunz et al. (28). Cell-free protein extracts used for the analysis Vma1p, Vma2p, and Vma5p were prepared as described (29).
To measure the calcium-dependent ATPase activity of cytosolic V 1 domains, cells were grown, lysed and a high speed supernatant fraction was prepared as previously described (5). Proteins were precipitated by treatment with 5% trichloroacetic acid as described above, resuspended in and dialyzed against buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.5 mM p-methylsulfonylfluoride, 10% glycerol), and centrifuged on a glycerol gradient (30). Fractions (750 l) were collected from the top of the gradient. Fractions (9 -12) containing V 1 domains were located by immunoblotting for Vma1p, Vma2p, Vma5p, and Vph1p. The pooled V 1containing fractions were assayed for calcium-dependent ATPase activity as described previously (5). Values are expressed as the difference between assays performed with and without 1.6 mM CaCl 2 .
The subunit composition (Fig. 7) of V 0 and the V 1 V 0 complex were analyzed by using a published procedure to cross-link proteins before immunoprecipitation and SDS-PAGE analysis (31). The procedure was modified so that 1 OD of 600-nm units of spheroplasts were incubated with 50 Ci of Trans[ 35 S] label (ICN Inc., 1175 Ci/mmol, 5100607). After pretreatment of the sample with protein A-Sepharose beads (Sigma Inc.), 400 l a solution of 5% bovine serum albumin/phosphatebuffered saline containing 5 l of antibody solution was added, and the sample was incubated overnight on ice with mixing. Protein A-Sepharose (40 l of a 40% (v/v) suspension) was added to each sample and incubated for 2 h on ice with mixing. Immunoprecipitates were collected by centrifugation at 5,000 rpm for 5 min in a microcentrifuge. Pellets were washed four times in buffer (1% Triton X-100, 1% deoxycholic acid, 50 mM Tris-HCl, pH 7.5, 150 mM NaCl), and precipitated proteins were eluted from the beads by incubation for 10 min at 95°C in 50 l of 4ϫ SDS-loading buffer (50 mM Tris base, pH 6.8, 8% glycerol, 1.6% SDS, 4% ␤-mercaptoethanol, 0.04% bromphenol blue). Half of each precipitate was subjected to SDS-PAGE and phosphorimager analysis. Cells were prepared for indirect immunofluorescent microscopy by using a published procedure (22). The secondary antibody was goat anti-mouse IgG labeled with FluoroLink™ Cy3™ (Amersham Biosciences). Fluorescent images were obtained with a Nikon À clipse E800 fluorescence microscope equipped with a Nikon 100X/1.3 plan fluor oil-immersion objective and a Diagnostic instruments Spot camera controlled by Adobe Photoshop software. For Cy3 fluorescence, samples were excited at 510 -560 nm and viewed with a barrier filter of 570 -650 nm. Adobe PhotoShop software was used to process images. Protein concentrations were determined with the Bio-Rad DC protein assay kit with bovine serum albumin as a standard.

sur4⌬ Cells Have a Unique Set of Vma Ϫ Phenotypes That
Indicate Reduced V-ATPase Activity-The presence of fragmented vacuoles in sur4⌬ cells suggests that the C26 acyl group of sphingolipids is necessary for some vacuolar function. Functional vacuoles require a proton pump or V-ATPase to acidify their lumen. When the V-ATPase is defective a set of phenotypes, referred to as Vma Ϫ , are produced. For example, vma mutants do not grow on YPD buffered to pH 7.5 (32,33). We reasoned that sur4⌬ cells might have Vma Ϫ phenotypes if sphingolipids containing a C26 acyl group are needed for V-ATPase activity. Indeed, we found that sur4⌬ cells grow very poorly at pH 7.5 just like the authentic vma mutant vma2⌬ ( Fig. 1) (34). In contrast, fen1⌬ cells grow better at pH 7.5 ( Fig.  1), indicating that they are able to acidify their vacuoles more effectively than sur4⌬ cells.
Another Vma Ϫ phenotype is failure to grow on YPD plates containing 4 mM ZnCl 2 (34). We found that sur4⌬ cells have this phenotype whereas fen1⌬ cells do not since they grow almost as well as wild-type cells (Fig. 1). Other Vma Ϫ phenotypes include failure to grow in the presence of 100 mM CaCl 2 (35) or on medium containing a non-fermentable carbon source such as glycerol (36). We found that 100 mM CaCl 2 inhibits growth of sur4⌬ cells but does not inhibit growth of fen1⌬ cells ( Fig. 1), again supporting the idea that sur4⌬ cells are less able to acidify their vacuoles than fen1⌬ cells. In addition, sur4⌬ cells grow slowly on YP-glycerol while fen1⌬ cells grow slightly faster and both grow better than vma2⌬ control cells (Table II).
These data show that the Vma Ϫ phenotypes of sur4⌬ cells are nearly as pronounced as those in vma2⌬ cells and that the phenotypes are less severe in fen1⌬ cells. Based upon these phenotypes it appears that the V-ATPase is more impaired in sur4⌬ cells than in fen1⌬ cells.
To directly examine vacuolar acidification in vivo we used the lysosomotropic fluorescent dye quinacrine, which is taken up by cells and concentrated in acidified vacuoles; if vacuoles are not acidified the dye remains in the cytoplasm and gives a diffuse fluorescent signal (37). A strong fluorescent vacuolar signal was observed in wild-type cells ( Fig. 2A). No fluorescent signal was seen in sur4⌬ cells, which behaved like vma2⌬ control cells that do not acidify their vacuole. The fluorescent signal in fen1⌬ cells was lower than in wild-type cells but greater than in sur4⌬ cells.
The fluorescent microscopy results were verified by a semiquantitative spectrofluorometric assay of the relative fluorescence of populations of cells stained with quinacrine. By this technique, wild-type cells fluoresced strongly whereas sur4⌬ cells fluoresced near the background level measured in vma2⌬ cells ( Fig. 2 and Table II). The fluorescent signal in fen1⌬ cells was 60% of the wild-type level indicating that they are able to acidify their vacuoles better than sur4⌬ cells but not as well as wild-type cells. Taken together these data show that sur4⌬ cells have a unique set of Vma Ϫ phenotypes and that they fail to acidify their vacuoles. These phenotypes are less severe in fen1⌬ cells, which have some ability to acidify their vacuoles, although they do not acidify as well as wild-type cells.
V-ATPase Activity Is Reduced in sur4⌬ Cells and Ficoll Dissociates the V 1 V 0 Complex-The data presented in Figs. 1 and 2 and Table II suggest that vacuolar membranes isolated from sur4⌬ cells should have less V-ATPase activity than vacuolar membranes isolated from fen1⌬ cells and both should have less activity than those isolated from wild-type cells. In agreement with this prediction, vacuolar membranes isolated by Ficoll density gradient centrifugation from sur4⌬ cells had only 10% of the wild-type V-ATPase activity while those isolated from fen1⌬ cells had 25% of the wild-type activity (Table III).
Reduced V-ATPase activity could be due to either a reduced specific activity or fewer molecules. To differentiate between these alternatives the concentration of Vph1p, a subunit of the V 0 domain, and Vma1p, Vma2p, and Vma5p, subunits of the V 1 domain, were measured by immunoblotting of whole cell protein extracts and of purified vacuolar membranes (Fig. 3). Whole cell protein extracts from wild-type, sur4⌬, and fen1⌬ mutants contained a similar level of each of the four proteins, indicating that the steady-state level of the proteins is similar in mutant and wild-type cells. Likewise, Vph1p was present in similar amounts in vacuolar membranes purified from the three strains. In contrast, the level of Vma1p Vma2p, and Vma5p in vacuolar membranes isolated from sur4⌬ and fen1⌬ cells was greatly reduced (Fig. 3). These results suggest that there is a defect in the V 1 domain in sur4⌬ and fen1⌬ cells or that the interaction between V 0 and V 1 is abnormal and V 1 or some of its subunits dissociate during vacuolar purification.
The interaction of V 1 with V 0 has been examined by treating vacuolar membranes with a low salt buffer containing ethylenediamine tetraacetate and then determining if particular protein subunits remained in the vacuolar fraction (high speed pellet) or became soluble (38). Because Ficoll purification removed V 1 subunits from vacuolar membranes, we used cell-free extracts in place of vacuolar membranes for these assays. Treatment with a low salt buffer did not reveal any difference in the level of Vma1p and Vma2p in the pellet and soluble fractions obtained with sur4⌬, fen1⌬, or wild-type cells (data not shown). We also determined if increasing concentrations of sodium carbonate (38) preferentially solubilized Vma1p and Vma2p in cell-free extracts made from sur4⌬ and fen1⌬ cells compared with wild-type cells. Vma2p but not Vma1p was more readily solubilized in the sur4⌬ and fen1⌬ samples (data not shown). Thus, whatever the nature of the abnormality in the interaction between Vma1p and Vma2p and V 0 in sur4⌬ and fen1⌬ cells, it must be fairly subtle and unique since, as far as we are aware, sensitivity of the V 1 V 0 complex to Ficoll has not been reported previously.
Vacuoles have also been partially purified by using sucrose density gradient centrifugation (25). We examined this procedure to see if V 1 domains remained attached to vacuolar membranes isolated from sur4⌬ cells. Sucrose gradient fractions were analyzed by immunoblotting for Vph1p to detect V 0 domains and for Vma1p and Vma2p to detect V 1 domains. The concentration of these three proteins peaked around fractions 5 and 6 in both the sur4⌬ and wild-type sample (Fig. 4). Thus, unlike the results obtained when extracts from sur4⌬ cells were centrifuged on Ficoll gradients, sucrose gradients yield a fraction in which the V 1 and V 0 domains remain associated.
Fractions 5 and 6 obtained from wild-type cells contain functional V 1 V 0 complexes because they have V-ATPase activity which has a slightly higher specific activity than that measured in vacuolar membranes isolated by Ficoll density gradient centrifugation (Table III). Even though the immunoblot of the sur4⌬ sample (Fig. 4) indicates that fractions 5 and 6 have both V 1 and V 0 domains, the fractions have only 20% as much V-ATPase activity as the wild type.
The data presented in this section show that V 1 domains do associate with V 0 on the vacuolar membrane in sur4⌬ cells but the association is abnormal because the Ficoll gradient procedure causes Vma1p, Vma2p, and Vma5p and possibly the entire V 1 domain to dissociate from V 0 domains. In addition, vacuolar membranes isolated from sur4⌬ cells by either Ficoll or sucrose density gradients have reduced V-ATPase activity.
V 1 Associates with the Vacuolar Membrane in sur4⌬ and fen1⌬ Cells-Both density gradient procedures for isolating vacuolar membranes subjects the sample to non-physiological conditions and could create artifacts such as by proteolysis. To try and avoid these potential complications, we examined the association of V 1 with V 0 on the vacuolar membrane of intact cells by using indirect immunofluorescence microscopy with anti-Vma1p or anti-Vma2p antibodies (33). In wild-type cells both antibodies localized to the vacuolar membrane as expected (Fig. 5). A similar localization is seen in sur4⌬ and fen1⌬ cells except that the vacuoles are fragmented and do not stain as uniformly as do vacuoles in wild-type cells (Fig. 5). Control cells lacking the vma2 gene show diffuse staining throughout the cytoplasm with anti-Vma1p antibody (data not shown) because V 1 subunits are not formed and no staining with anti-Vma2p is observed because the protein is absent (Fig. 5). These data verify those obtained by sucrose density gradient centrifugation and together the two sets of data establish two critical points about the V-ATPase in sur4⌬ and fen1⌬ cells. First, V 1 domains are assembled and, second, at least some of them do associate with V 0 on the vacuolar membrane.
V 0 Domains Are Functional in sur4⌬ Cells but V 1 Domains Lack ATPase Activity-The nature of the V-ATPase defect in sur4⌬ cells was further elucidated by determining whether V 1 , V 0 or both domains are defective. For these experiments vacuolar membranes were isolated by sucrose density gradient centrifugation then treated with KI to dissociate V 1 (27). Samples were then centrifuged to give a vacuolar membrane pellet (P) containing V 0 and a supernatant fraction (S) containing V 1 . It has been shown previously that upon mixing the P and S fractions and dialyzing away the KI, V 1 and V 0 associate, as determined by immunoblotting, and V-ATPase activity is partially restored (35-40%) (27). We observed very similar results using the P and S fractions derived from wild-type cells. By immunoblotting, Vma1p and Vma2p (V 1 subunits) were more concentrated in the P fraction following dialysis compared with  the non-dialyzed sample (Fig. 6A, sample 2) showing that V 1 associates with V 0 . In addition, 35% of ATPase activity was restored in the dialyzed sample compared with only 9% in the non-dialyzed sample. The control experiment using wild-type V 1 V 0 complexes that were not treated with KI showed that all of the Vma1p and Vma2p were in the pellet along with Vph1p (Fig. 6A, sample 1). Dialysis reduced ATPase activity to 68% of the non-dialyzed sample. Similar mixing experiments with the P (V 0 ) and the S (V 1 ) fractions from sur4⌬ cells showed that dialysis does promote association of Vma1p and Vma2p with the P fraction, but it does not restore any ATPase activity (Fig.  6A, sample 4). The control experiment using sur4⌬ V 1 V 0 complexes that were not treated with KI showed that all of the Vma1p and Vma2p were in the pellet along with Vph1p and that dialysis reduced ATPase activity slightly from 18 to 13% (Fig. 6A, sample 3). Results from other control reactions are shown in samples 7 and 8 of Fig. 6.
Mixing the S (V 1 ) fraction from wild-type cells with the P (V 0 ) fraction from sur4⌬ cells followed by dialysis resulted in association of Vma1p and Vma2p with the P fraction and restoration of 36% of the ATPase activity (Fig. 6A, sample 5). A reciprocal mixing experiment with the P (V 0 ) fraction from wild-type cells the S (V 1 ) fraction from sur4⌬ cells showed that Vma1p and Vma2p associate with the P fraction but there was no restoration of ATPase activity (Fig. 6A, sample 6). These experiments demonstrate that V 0 domains in sur4⌬ cells are fully functional and can associate with wild-type V 1 domains to generate ATPase activity whereas the V 1 domains are capable of association with V 0 on the vacuolar membrane but they do not have ATPase activity. V 1 Domains from sur4⌬ Cells Lack Ca-ATPase Activity-Free V 1 domains do not show ATPase activity when assayed in the presence of Mg 2ϩ , but do show activity when assayed in the presence of Ca 2ϩ (20). Thus, to verify that V 1 domains in sur4⌬ cells lack ATPase activity we partially purified V 1 complexes by velocity centrifugation on glycerol gradients and assayed them for ATPase activity in the presence Ca 2ϩ . To remove background ATPase activity that was not stimulated by Ca 2ϩ , the assay was also done in the absence of Ca 2ϩ and this value was subtracted from the value obtained in the presence of Ca 2ϩ to give the Ca-stimulated activity. In preliminary experiments increasing concentrations of protein in the pooled V 1 -containing sucrose gradient fractions were assayed for Ca-stimulated ATPase activity (release of P i ). Activity was linear with increas-

FIG. 3. Vacuoles isolated from sur4⌬ cells by Ficoll density gradient centrifugation lack the V 1 subunits Vma1p, Vma2, and
Vma5p. Cell-free extracts and Ficoll-purified vacuolar membrane vesicles were examined by immunoblotting for the presence of the V 0 subunit Vph1p and the V 1 subunits Vma1p, Vma2p, and Vma5p. Samples (20 g of cell-free protein extracts and 10 g of purified vacuolar membrane vesicles) were incubated in cracking buffer for 10 min at 65°C for analysis of Vph1p or in SDS sample buffer for 5 min at 95°C for the analysis of other proteins.

FIG. 4. V 1 associates with V 0 on vacuolar membranes isolated from sur4⌬ cells by sucrose density gradient centrifugation.
Vacuolar membranes were isolated by centrifugation on sucrose gradients and fractionated into non-equal sized fractions as described in "Experimental Procedures." Fraction 1 is the top of the gradient and fraction 9 is the resuspended pellet. Part of each fraction (10 g of protein) was incubated in SDS-sample buffer for 5 min at 95°C prior to SDS-PAGE and then immunoblotted for the V 0 subunit Vph1p and the V 1 subunits Vma1p and Vma2p.

FIG. 5. V 1 domains associate in vivo with vacuoles in sur4⌬ cells. The location of the V 1 domain in cells was determined by indirect
immunofluorescence staining with anti-Vma2p primary antibody followed by Cy3-tagged secondary antibody (red images). Images taken by differential interference contrast optics are shown in black and white.
ing protein concentration (data not shown). A protein concentration that gave easily measured activity using V 1 domains from wild-type cells was chosen for kinetic analyzes. The kinetics of Ca-stimulated ATP hydrolysis for wild-type V 1 domains were linear up to about 5 min and then reached a plateau whereas the V 1 domains from sur4⌬ cells showed no ATP hydrolysis over the entire 30 min incubation period (Fig.  6B). These results verify those shown in Fig. 6A and based upon the combined data we conclude that V 1 domains in sur4⌬ cells lack ATPase activity.
The Subunit Composition of the V 1 and V 0 Domains Appears Normal in sur4⌬ Cells-V 1 domains in sur4⌬ cells may have reduced ATPase activity and dissociate from the vacuolar membrane in the presence of Ficoll because a subunit is missing. To examine the subunit composition of V 1 and V 0 domains, cells were converted to spheroplasts, metabolically labeled with [ 35 S]amino acids, gently lysed, and then proteins were crosslinked with a reversible cross-linker. Samples were immunoprecipitated with monoclonal antibodies specific for Vph1p, Vma1p, or Vma2p and radioactive proteins in the immunoprecipitate were analyzed by SDS-PAGE and phosphorimaging.
The anti-Vph1p antibody used in these experiments only recognizes V 0 domains that are not complexed with V 1 (24) and is, therefore, useful for examining the subunit composition of V 0 domains. Radiolabeled proteins of 100 (Vph1p), 36 (Vma6p), 19 (identity unknown, Refs. 30 and 31) and 17 (Vma11p) kDa were immunoprecipitated by the Vph1p antibody in wild-type sur4⌬, fen1⌬, and vma2⌬ cells (Fig. 7A). We observed small variations in the relative intensity of bands from experiment to experiment, and the samples shown in Fig. 7 were chosen to represent the average of the results of three independent experiments. The data shown in Fig. 7A are in agreement with published results (e.g. Refs. 30 and 31) and indicate that the subunit composition of the V 0 domain is normal in sur4⌬ and fen1⌬ cells.
The anti-Vma1p and anti-Vma2p antibody used in these experiments recognize free V 1 and V 1 V 0 complexes (24,30). Radioactive proteins immunoprecipitated by anti-Vma1p and anit-Vma2p are shown in Fig. 7, B and C, respectively. Again, there were small variations is the intensity of some radioactive bands from experiment to experiment, but overall our results indicate that the subunit composition of free V 1 and the V 1 V 0 complex in sur4⌬ and fen1⌬ cells is similar if not identical to that in wild-type cells and to published data. The vma2⌬ cells served as a control for cells lacking V 1 and V 1 V 0 complexes (30).
A limitation of these data is that not all V 1 subunits are readily detected including Vma7p, Vma10p, and Vma13p. Vma7p and Vma10p are probably present in the V 1 domains we examined because if either protein were absence then V 1 and V 0 would not associate (39 -42). Vma13p is not necessary for V 1 -V 0 association and could be missing. To determine if it was present in vacuolar membranes isolated from sur4⌬ cells, sur4⌬ vma13⌬, and vma13⌬ cells were transformed with a vector carrying a VMA13 allele having a Myc epitope inserted immediately downstream of the methionine start codon (20).  1 and 3) with 300 mM KI to dissociate V 1 from V 0 . Reactions treated with KI were centrifuged to generate a pellet fraction containing V 0 domains and a supernatant fraction containing V 1 domains that were mixed (reactions 2, 4 -6) and dialyzed to promote domain association on not dialyzed to prevent association. To determine if the V 1 and V 0 associated, the reactions were centrifuged to yield a pellet (P) fraction containing V 0 domains and V 1 V 0 complexes (detected by immunoblotting for Vph1p) and a supernatant (S) fraction contain-ing V 1 domains (detected by immunoblotting for Vma1p and Vma2p). V-ATPase values represent the average of three determinations Ϯ S.D. Reactions 1, 3, 7, and 8 are controls. B, cytosolic V 1 domains were isolated from RCD390 (wild-type, F) and RCD410 (sur4⌬, E) cells on sucrose density gradients as described in "Experimental Procedures" and 200 g of protein was assayed for Ca-stimulated release of P i from ATP (i.e. ATPase activity). Values represent the average of three determinations Ϯ S.D.
Vma13p in the peak fraction containing vacuolar membranes (Fig. 7D, fraction 5) was similar in the sur4⌬ vma13⌬ and vma13⌬ samples, and the overall distribution of Myc-Vma13p was very similar in the two gradients. Immunoblots of the total cell-free extracts showed that the concentration of Myc-Vma13p was the same in the two strains as were the other Vma subunits examined (data not shown). We conclude that Vma13p associates with V 1 V 0 complexes in sur4⌬ cells.
Elevated Levels of LCBs and LCBPs Do Not Correlate with Reduced V-ATPase Activity-It has been noted previously that sur4⌬ and fen1⌬ cells accumulate LCBs but the levels have not been quantified nor have the species that accumulate been determined (11,12,43). Likewise, the concentration of long chain base phosphates (LCBPs) has not been measured.
We quantified LCBs and LCBPs by tagging them after extraction with a fluorescent reagent followed by HPLC (23). The analysis was done in two different strain backgrounds to see how similar or different they might be and if any differences correlated with mutant phenotypes. The five species of LCBs are at nearly identical levels in the two wild-type strains (Table  IV). In the two sur4⌬ strains all five species are elevated and their levels are similar in the two strain backgrounds except for C18-DHS and C20-DHS, which are less elevated in strain RCD410 (the W303 background). All five species are also elevated in the two fen1⌬ strains but there is more variability between strains and only the C16-DHS species have similar values.
LCBP values are also very similar in the two wild-type strains (Table V) and are quite low as we have reported for a wild-type strain related to RCD390 (44). All LCBPs show similar elevations in the two sur4⌬ strains except for C18-DHSP and C20-DHSP. All LCBPs are also elevated in the two fen1⌬ strains but the values vary between the strains.
The elevated levels of LCBs and LCBPs could be responsible for the lack of ATPase activity. However, if this were the case we would expect fen1⌬ cells (RCD393), not sur4⌬ cells (RCD389), to have a greater loss of V-ATPase activity because they have a higher total level of LCBs and LCBPs (Tables IV  and V).
To determine directly if LCBs dissociate Vma1p and Vma2p from V 0 , vacuolar membranes purified from wild-type cells were incubated with increasing concentrations of PHS. After incubation, samples were centrifuged to give a pellet and a supernatant fraction, which where analyzed by immunoblotting for Vma1p and Vma2p. Because PHS has detergent-like properties and could release proteases from vacuoles, we also immunoblotted for CPY to control for proteolysis. In the presence, but not in the absence of PHS, the level of Vma1p and Vma2p in the pellet fraction gradually decreased over the 120min incubation period (Fig. 8). However, the two proteins did not appear in the supernatant as would be expected if they were dissociating from V 0 . This result plus the fact that they disappeared (Fig. 8) at the same rate as CPY indicates that they are being digested by proteases. We conclude that LCBs do not dissociate Vma1p and Vma2p from purified vacuolar membranes.

DISCUSSION
Our analysis of sur4⌬ and fen1⌬ cells shows that sphingolipids containing a C26 acyl group are necessary for generation of a V 1 domain that is capable of hydrolyzing ATP. This conclusion is based upon an analysis of sur4⌬ cells, whose sphingolipids lack a C26 acyl group, and fen1⌬ cells, whose sphingolipids have about 30% of the wild-type level of C26 acyl groups (11,12). We predicted that cellular functions dependent upon sphingolipids with a C26 acyl group would be more disrupted in sur4⌬ than in fen1⌬ cells. Several results support this line of reasoning and our conclusion. First, we found that sur4⌬ cells have some but not all Vma Ϫ phenotypes, indicating a defective V-ATPase, and that some of these phenotypes are more pronounced in sur4⌬ than in fen1⌬ cells. For example, growth is more inhibited in sur4⌬ than in fen1⌬ cells on YPD plates buffered to pH 7.5 or when the plates contain 100 mM CaCl 2 , 4 mM ZnCl 2 , or 2% glycerol as the carbon source ( Fig. 1 and Table II). Second, V-ATPase activity in sur4⌬ cells is as defective as that in vma2⌬ cells, which lack V-ATPase activity, when measured by uptake of the vacuolar dye quinacrine (Fig.  2). In contrast, uptake in fen1⌬ cells is about 4-fold higher than in sur4⌬ cells or about 60% of the wild-type level. Finally, the ATPase activity of vacuoles purified from sur4⌬ cells by Ficoll gradient centrifugation is only 10% of the wild type while fen1⌬ vacuoles have 25% of the wild-type activity (Table III). These data show that V-ATPase activity is more impaired in sur4⌬ than in fen1⌬ cells.
To understand why sphingolipids with a C26 acyl group are needed for V-ATPase activity we determined if V 1 domains could associate with V 0 domains on the vacuolar membrane. Vacuolar membranes isolated on a Ficoll gradient were examined by immunoblotting to determine if V 0 and V 1 domains were present. The results showed that vacuolar membranes from sur4⌬ and fen1⌬ cells had very low, barely detectable levels of Vma1p, Vma2p, and Vma5p, indicating that the V 1 domain was not associated with the V 0 domain (Fig. 3). Since these three V 1 subunits are present in cell extracts (Fig. 3) their absence from purified vacuoles suggests that they mislocalize, do not assemble into a functional V 1 domain or V 1 associates abnormally with V 0 . The finding that V 1 and V 0 are found together in fractions of a sucrose density gradient where vacuolar membranes are located (Fig. 4, fractions 5 and 6), supports either of the latter two possibilities. In addition, indirect immunofluorescent microscopy on intact sur4⌬ and fen1⌬ cells showed that the V 1 subunit Vma2p was bound to vacuoles (Fig. 5). Together these data show that V 1 is able to associate with V 0 on the cytoplasmic face of the vacuolar membrane to form V 1 V 0 complexes in sur4⌬ cells, but the complexes are not stable in Ficoll (Fig. 3), have low ATPase activity (Table  III), and do not acidify vacuoles ( Figs. 1 and 2).
Reduced ATPase activity in sur4⌬ cells suggested that the V 1 domain was defective. To test this hypothesis, V 1 and V 0 domains were isolated and tested in vitro for association and for restoration of ATPase activity. Wild-type V 1 and sur4⌬ V 0 associated to produce V 1 V 0 complexes with ATPase activity. Wild-type V 0 and sur4⌬ V 1 also associated but the V 1 V 0 complexes had no ATPase activity (Fig. 6). We also partially purified cytosolic V 1 domains from wild-type and sur4⌬ cells and assayed them for calcium-dependent ATPase activity. V 1 domains from sur4⌬ cells completely lacked activity (Fig. 6B). These results establish that V 0 domains in sur4⌬ cells are normal, but the V 1 domains are defective and lack ATPase activity.
To understand why V 1 domains lack ATPase activity, we examined the subunit composition of V 1 and V 0 and found that they are the same in sur4⌬, fen1⌬, and wild-type cells (Fig. 7). Separate analysis of Myc-tagged Vma13p showed that it was present in V 1 domains in sur4⌬ cells (Fig. 7). The procedures used by us would not have detected the V 1 subunits Vma7p or Vma10p, although we infer that they are present because if either protein was missing from cells, V 1 domains would not associate with V 0 (39 -42). Thus, lack of a protein subunit is not likely to be the cause of the defect in ATPase activity in V 1 domains present in sur4⌬ cells.
Others have reported that sur4⌬ and fen1⌬ cells contain high levels of LCBs (11,12,43), but the levels have not been quantified nor have the species been identified. We were concerned that the high level of LCBs might impair ATPase activity. It  was recently shown that the reduced level of glucan synthase activity in sur4⌬ and fen1⌬ cells is due to elevated levels of PHS and DHS (43). Our analysis of LCBs shows that all five species are elevated in both sur4⌬ and fen1⌬ mutants in two different strain backgrounds, W303 and JK9-3d (Table IV). In the JK9-3d strain background, which we used for the majority of our experiments, there is a 44-fold increase in total LCBs in sur4⌬ cells and a 63-fold increase in fen1⌬ cells. The same trends hold for the mutants in the W303 background but the increases are smaller. If elevated LCBs were responsible for disruption of V-ATPase activity, then we would expect the activity to be reduced more in fen1⌬ than in sur4⌬ cells because fen1⌬ cells have a higher level of LCBs. Our results are just the opposite of this prediction and argue that elevated LCBs are not responsible for reduced ATPase activity. However, such arguments cannot eliminate the possibility that LCBs are interfering with V 1 function.
We also compared the total LCB content of fractions from a Ficoll gradient to see if there was any correlation between reduced V-ATPase activity in sur4⌬ and fen1⌬ cells and the level of these compounds. The vacuolar membrane fraction of wild-type cells had a very low, barely detectable level of LCBs, as did the two fractions below the membrane fraction (data not shown). The pellet at the bottom of the gradient had the highest level of LCBs, but the concentration was still very low. The level of LCBs was higher in the gradient fractions obtained from sur4⌬ and fen1⌬ cells. As with the total LCB values (Table  IV), the levels in the Ficoll gradient fractions are higher in fen1⌬ cells, suggesting that it is not the LCBs that are responsible for reduced V-ATPase activity. LCBPs were not detectable in any of the Ficoll gradient fractions nor in the cell-free extracts. Most likely they were degraded during the incubation period when cells were converted to spheroplasts.
We also determined if PHS added in vitro to purified vacuolar membranes could mimic what occurs in sur4⌬ and fen1⌬ cells and selectively release Vma1p and Vma2p from membranes. We found that neither protein was selectively released. In fact, the membrane V-ATPase was quite resistant to disruption by PHS and only at higher concentrations did the two proteins start to disappear from the membrane pellet (Fig. 8). Their disappearance, however, was not accompanied by their appearance in the soluble fraction. Rather their concentration decreased as a function of increasing PHS at the same rate as the luminal vacuolar protein CPY, indicating that protein loss was probably due to disruption of vacuolar membrane integrity and subsequent degradation by vacuolar proteases. Thus, wildtype V 1 V 0 complexes are very resistant to dissociation by treatment in vitro with PHS.
We also measured LCBPs, which have not been measured, and found that all five species were elevated in sur4⌬ and fen1⌬ cells. The total level is similar in both wild-type strains and in both sur4⌬ mutants, but is different in the fen1⌬ strains (Table V). The differences may be due to strain-specific variation in the activity of metabolic pathways that make and degrade LCBPs. Again, it seems unlikely that LCBPs are disrupting V-ATPase activity because in the W303 strain background their level is higher in fen1⌬ than in sur4⌬ cells, yet the Vma Ϫ phenotypes are less serve in fen1⌬ than in sur4⌬ cells (Table  II). However, further work will be necessary to eliminate the possibility that elevated LCBPs are responsible for the defective V 1 domain in sur4⌬ cells.
How might sphingolipids with a C26 acyl group affect the ATPase activity of V 1 ? One possibility is that ceramides, which are made in the ER (reviewed in Ref. 6), play a role in the assembly of V 1 in the ER (45,46). Another possibility is that complex sphingolipids in the Golgi influence maturation of V 1 as it transits to the vacuolar membrane. In sur4⌬ cells the ceramides and sphingolipids with C22 and C24 acyl groups would not substitute for the normal C26 groups and V 1 domains would assemble incorrectly. The ATPase defect would be less severe in fen1⌬ cells because about one-third of the ceramides and complex sphingolipids have C26 acyl groups. Alternatively, the RAVE protein complex (21) has recently been shown to be necessary for assembly of the V-ATPase (47) and some step in the action of RAVE may require sphingolipids with a C26 acyl group. Ficoll, a polymer of sucrose, may dissociate V 1 from V 0 by interacting with one or more V 1 subunits that are not correctly folded or assembled in sur4⌬ cells.
The results presented here are the first to indicate a role for C26 acyl groups and for sphingolipids in V-ATPase function. Their exact role will require further characterization of V 1 domains in sur4⌬ cells. Our results suggest that sphingolipids may be important for the activity of V-ATPases and related ATPases in other organisms. S. cerevisiae contains another type of V-ATPase located in the Golgi/endosomal compartments that is identical to the V-ATPase except that the V 0 domain contains Stv1p in place of Vph1p (48). Our results suggest that the functionality of this V-ATPase may also require sphingolipids with a C26 acyl group.