Fab1p Is Essential for PtdIns(3)P 5-Kinase Activity and the Maintenance of Vacuolar Size and Membrane Homeostasis

The Escherichia coli strain XL1-Blue (supE44 thi-1 lac endA1 gyrA96 hsdR17 relA1 F′ proAB LaqIq ZΔM15) was grown with standard LB media. Unless otherwise specified, the Saccharomyces cerevisiae strains used in this study were grown in standard YPD and SD minimal media with the necessary auxotrophic supplements. S. cerevisiae strains used in this study were as follows: (a) SEY6210 (MATα leu2-3,112 ura3-52 his3-Δ200 trp1-Δ901 lys2-801 suc2-Δ9; Robinson et al., 1988). (b) fab1Δ1 (SEY6210 fab1Δ::LEU2, this study). (c) fab1Δ2 (SEY6210 fab1Δ::HIS3, this study). (d) LWY1481 (MATα leu2-3,112 ura3-52 his3-Δ200 trp1-Δ901 lys2-801 suc2-Δ9 pep4-Δ1137 vac7Δ::HIS3; Bonangelino et al., 1997). (e) LWY2365 (MATα leu2-3,112 ura3-52 his3-Δ200 trp1-Δ901 lys2-801 suc2-Δ9 pep4-Δ1137 vac14-1; Bonangelino et al., 1997). (f) EMY119 (MATa leu2-3,112 ura3-52 fab1-2; lab strain).

All restriction enzymes were purchased from Boehringer Mannheim Biochemicals (Indianapolis, IN). T4 DNA ligase was received from GIBCO BRL (Gaithersburg, MD). PCR was performed using synthetic oligonucleotide primers and dNTPs purchased from GIBCO BRL. All PCR reactions used a 2:1 mixture of KlenTaq polymerase (Ab Peptides, Inc., St. Louis, MO) and Pfu polymerase (Stratagene, La Jolla, CA) as described in (Barnes, 1994). Otherwise standard molecular biology techniques were used (Maniatis et al., 1982). Yeast transformations were performed using the lithium acetate method (Ito et al., 1983) and genomic DNA was isolated using a modified version of the procedure described by Hoffman and Winston (1987).

The entire open reading frame of FAB1 was deleted in SEY6210 through transformation with a 1.1-kbp PCR fragment containing the HIS3 marker flanked by 52 nucleotides upstream and downstream of the FAB1 sequence. Histidine prototrophs were selected on histidine-free minimal media plates and confirmation of the disruption was done by PCR using two different sets of primers.

A 6,350 bp AatII-NruI fragment of the wild-type FAB1 gene was excised from pAY60 (pRS416-FAB1; Yamamoto et al., 1995). The resulting linearized 9,100-bp pRS416 vector containing ∼2,000 bp of flanking genomic sequence 5′ of the AatII site and 3′ of the NruI site was then transformed into EMY119, a strain carrying the fab1-2 allele (Yamamoto et al., 1995) and selected on uracil-free plates. Through homologous recombination and the subsequent purification of the gap-repaired plasmid, the fab1-2 allele was rescued from the genome.

The G2042/2045V point mutant construct was made using the gene SOE (splicing by overlap extension) and plasmid gap repair techniques (Muhlrad et al., 1992; Horton et al., 1993). For the gene SOE, two successive rounds of PCR were used to generate the final 1.9-kbp fragment containing the desired mutations. In the first round, the 5′ and 3′-PCR fragments were synthesized in two separate PCR reactions. For the second round of PCR, the products from the 5′ and 3′ reactions were diluted 100-fold and amplified in the presence of both outside primers to create the full-length fragment containing the desired mutations. pAY60 (Yamamoto et al., 1995) was then digested with NheI and NruI and the gapped vector was purified and cotransformed into the fab1Δ2 strain with the full-length gene SOE fragment. Plasmid DNA was rescued from yeast cells that had become uracil prototrophs, indicating successful gap repair of the pRS416-based plasmid. The gene SOE fragment containing the D2134R point mutation was made in a similar fashion. However, after synthesis of the full-length gene SOE fragment, it was digested with NheI and NruI and introduced into the gapped vector by direct ligation. Once transformed into E. coli the plasmid constructs for both point mutations were sequenced to confirm the presence of only the desired mutation. DNA sequencing was done at the UCSD CFAR core facility using the ABI Prism BigDye chemistry with Amplitaq DNA Polymerase FS. Gels were analyzed with a 373XL DNA Sequencer (PE Applied Biosystems, Foster City, CA). Correct constructs were then retransformed into fab1Δ2 for phenotypic analysis.

Primers specific to FAB1 were used to separately amplify a 350-bp and a 1,300-bp region upstream and downstream of codon 8, while also engineering a BglII site at codon 8. The 350-bp fragment was cut with SnaBI/BglII while the 1,300-bp FAB1 fragment was cut with BglII/Bsu36I. These fragments were then simultaneously ligated into pAY60 (Yamamoto et al., 1995) and pEMY105 (pRS426-FAB1) digested with SnaBI/Bsu36I. The resulting pRS416-FAB1-BglII and pRS426-FAB1-BglII constructs contain the full-length FAB1 gene with a unique BglII site engineered at codon 8. A triple-HA epitope repeat was also amplified using PCR primers that introduced BglII sites into both ends of the PCR product to make the triple-HA repeat in frame with the BglII site engineered into FAB1. This HA insert was ligated into pRS416-FAB1-BglII and pRS426-FAB1-BglII via the BglII sites.

Approximately 6 OD₆₀₀ units of yeast were harvested by centrifugation and resuspended in 2 ml of minimal media and allowed to recover at 24°C for 10 min. Express [³⁵S]-protein labeling mix (DuPont NEN, Boston, MA) was then added to a final concentration of 5 μCi/OD₆₀₀ unit. After a 20-min incubation at 24°C, cells were chased for 45 min at either 24 or 38°C with 5 mM methionine, 1 mM cysteine, 0.4% yeast extract and 2% glucose. The chase was ended with the addition trichloroacetic acid was added to a final concentration of 10%. Trichloroacetic acid precipitates were subjected to glass bead lysis and Fab1p was then immunoprecipitated as previously described (Gaynor et al., 1994) with antisera raised against a TrpE fusion of the Fab1 protein.

Spheroplasted wild-type cells were subjected to gentle osmotic lysis as previously described (Gaynor et al., 1994). The resulting cellular extract was then sequentially centrifuged at 325 g for 5 min to remove unbroken cells, 13,000 g for 10 min to generate pelletable, P13 and soluble S13 fractions and 100,000 g for 30 min to yield P100 and S100 fractions. Protein present in all fractions was then precipitated using TCA to a final concentration of 5%. In the case of Fab1p and glucose 6-phosphate dehydrogenase, the extracts were derived from wild-type cells pulse-labeled with Express [³⁵S]-protein labeling mix for 20 min and chased with 5 mM methionine, 1 mM cysteine, 0.4% yeast extract for 40 min. Fractions were then immunoprecipitated for Fab1p and glucose 6-phosphate dehydrogenase using antiserum raised against these proteins as described (Gaynor et al., 1994). The presence of Kex2p and the 100-kD subunit of the vacuolar ATPase in specific fractions were determined using antisera specific to these proteins for immunoblotting and ECL as described in (Babst et al., 1997).

To label strains with myo-[2-³H]inositol (Nycomed Amersham Inc., Princeton, NJ), cells were grown for 24 h in YPD or SD minimal media (if selecting for a plasmid). 0.2–0.5 OD₆₀₀ units of cells were harvested, washed with 1 ml of synthetic media lacking inositol and used to inoculate 5 ml of inositol-free media containing 75 μCi myo-[2-³H]inositol. Cells were labeled 12 h at 22°C, harvested by centrifugation, washed with fresh synthetic media and then resuspended in 100 to 200 μL of synthetic media. An equal volume of 1.2 M NaCl was then added to the cells which were then incubated for a 10-min period to osmotically shock cells. This was carried out at 22°C unless otherwise indicated. Following this treatment, labeled cells were lysed in 1.5 ml of 1 M HCl/chloroform/methanol (1:1:1) by vortexing 10 × 30 s periods in the presence of 0.5 g of 0.25-mm glass beads. As previously described, cellular lipids were extracted, deacylated and analyzed by HPLC (Schu et al., 1993; Stack et al., 1995). Fractions eluting from an HPLC column (catalog no. 4611-1505; Whatman Inc., Clifton, NJ) were collected every 0.66 min. [³H]-PtdIns(3,5)P₂ standards were used to determine the elution point of PtdIns(3,5)P₂ from our column were a generous gift of Robert H. Michell (University of Birmingham, AL). The data presented are representative of multiple experiments. A variation of 10% was observed between experiments.

Approximately 1 OD₆₀₀ of yeast were harvested at an OD₆₀₀ of 0.6–0.8, and then labeled with the vital vacuolar dye FM4-64 as previously described (Vida and Emr, 1995). FM4-64 [N-(3-triethylammoniumpropyl)- 4-(p-diethylaminophenylhexatrienyl) pyridinium dibromide] was obtained from Molecular Probes Inc. (Eugene, OR). For temperature-shift experiments, cells were labeled at the permissive temperature and then shifted to the elevated temperature for the remainder of the incubation in the absence of the fluorophore.

For analysis of CPS, CPY, ALP, and AP1 processing, whole cells were labeled as previously described (Gaynor et al., 1994). Express [³⁵S]-protein labeling mix was added to a final concentration of 10 μCi/μl and cells were labeled during a 10-min pulse at 24°C. For temperature-shift experiments, cells were preshifted to 38°C for 30 min before labeling. The cultures were kept at the identical temperatures and chased as described above. After the chase, cells were harvested and lysates generated as previously described (Gaynor et al., 1994). Extracts were immunoprecipitated with antisera against CPY, ALP, CPS, and AP1, that have been previously characterized (Klionsky and Emr, 1989; Cowles et al., 1997b; antibody to API was a generous gift from Dan Klionsky, University of California, Davis, CA; Klionsky et al., 1992). Radioactive immunoprecipitates of CPS were treated with endoglycosidase H (DuPont NEN) as previously described (Gaynor et al., 1994). SDS-PAGE was done as described above however, gels were run at 25 mA constant current.

For observing the indirect immunofluorescence of the 60-kD subunit of the vacuolar ATPase, the method of (Redding et al., 1997) was used. In brief, exponentially growing cells were fixed in 4% formaldehyde and 0.1 M potassium phosphate, pH 6.5 for 16 h. The cell wall was removed by incubating cells with 45 μg/ml Zymolyase 100T (Seikagaku Kogyo, Tokyo, Japan) and cells were then treated with 1% SDS for 10 min. The fixed spheroplasts were first incubated in a 1:40 dilution of monoclonal antibody specific for the 60-kD subunit of the vacuolar ATPase (Molecular Probes Inc.), followed by successive incubations in goat anti–mouse IgG (1:3,000), and donkey anti–goat (1:200; Jackson ImmunoResearch Laboratories, Inc., West Grove, PA). GFP-ALP was transformed into the fab1Δ2 strain and observed as previously described (Cowles et al., 1997a). 50 OD₆₀₀ units of cells grown to mid-log phase were harvested for each sample analyzed by electron microscopy. All cells were grown in YNB supplemented with the required amino acids and 2% dextrose. Cells were then fixed and processed for electron microscopy as described (Rieder et al., 1996).

Immunofluorescence localization of an HA-tagged version of Fab1p was performed as previously described (Bonangelino et. al., 1997). In brief, cells overexpressing Fab1p-HA were resuspended in 0.1 M sodium phosphate buffer (pH 6.5) and fixed in 4.4% formaldehyde for 40 min at 30°C. Fixed cells were converted to spheroplasts by incubation in 1.2 M sorbitol in 0.1 M phosphate buffer (pH 6.5) with 1% mercaptoethanol and 150 g of oxalyticase (Enzogenetics, Eugene, OR) per ml for 15 min at 30°C. After adhering the fixed spheroplasts to 1% polyethyleneimine-coated wells, cells were incubated with monoclonal anti-HA (MMSR101; Berkeley Antibody Co., Inc., Richmond, CA) at a 1:200 dilution overnight (16 h), followed by affinity-purified rabbit anti-Vac8p (Wang et al., 1998) at a 1:50 dilution for 1 h. The primary mouse antibodies were detected by using a 1:200 dilution of Oregon-Green-488 conjugated goat anti–mouse (Molecular Probes) and the rabbit antibodies were detected by 1:200 dilution of Rhodamine red–conjugated goat anti–rabbit (Jackson ImmunoResearch Laboratories Inc.). Incubations with the fluorophore-conjugated secondary antibodies were performed in a dark humidity chamber for 1 h. Images were collected using an MRC 1024 Scanning Confocal head mounted on a Nikon Optiphot equipped with a 100× oil immersion lens.

Many of the phosphoinositides and phosphoinositide kinases (Fig. 1 A) present in mammalian cells are also found in the yeast S. cerevisiae. Homologs of the PtdIns 4-kinase (Stt4p and Pik1p; Flanagan et al., 1993; Yoshida et al., 1994), PtdIns 3-kinase (Vps34p; Schu et al., 1993), and PtdIns(4)P 5-kinase (Mss4p; Desrivieres et al., 1998; Homma et al., 1998) have been identified in yeast. Two groups recently have identified a novel lipid, PtdIns(3,5)P₂, in yeast and mammalian cells, that is produced by the sequential phosphorylation of PtdIns at the 3′ and 5′-positions (Dove et al., 1997; Whiteford et al., 1997). The PtdIns(3)P 5-kinase required for the synthesis of this previously undocumented lipid has not yet been identified.

Based on initial sequence analyses, Fab1p from S. cerevisiae, as well as its homologs from S. pombe and C. elegans, were proposed to be PtdIns(4)P 5-kinases (Yamamoto et al., 1995; Boronenkov and Anderson, 1995). A BLASTP search (NCBI; Altschul et al., 1997) with a COOH-terminal region of Fab1p (a.a. 2,000–2,278), containing the lipid kinase domain, reveals that the most closely related sequences with known enzymatic activities belong to a subset of PtdInsP kinases which include the Type I and Type II mammalian PtdIns(4)P 5-kinases. The Fab1p kinase domain therefore, is more closely related to the kinase domains of PtdInsP kinases (Fig. 1 B) than PtdIns kinases such as the 3-kinase Vps34p, the 4-kinases Stt4p and Pik1p, and the PtdIns kinase homologs Tor1p and Tor2p (Kunz et al., 1993).

Recent evidence supports the reclassification of the Type II PtdIns(4)P 5-kinases as PtdIns(5)P 4-kinases (Rameh et al., 1997) and this biochemical distinction of enzymatic activity is also borne out in sequence comparisons. Construction of a phylogenentic tree based on the pairwise alignment of the kinase domains of these groups, separates the Type I and Type II kinase domains into two classes (Fig. 1,B; Loijens and Anderson, 1996). Interestingly, Fab1p belongs to a third, distinct class of PtdInsP kinases when included in this analysis (Fig. 1,B; Loijens and Anderson, 1996). Consistent with this phylogenetic analysis, the yeast Mss4 protein has recently been identified as a PtdIns(4)P 5-kinase (Desrivieres et al., 1998; Homma et al., 1998; Fig. 1 B), indicating that this additional class does not appear to have been established simply on the basis of comparing lipid kinases from different species. These results suggest that the lipid kinase domain of Fab1p is distinct from analogous domains of known PtdIns(4)P 5-kinases and PtdIns(5)P 4-kinases, raising the possibility that Fab1p phosphorylates a PtdInsP substrate distinct from either PtsIns(4)P or PtdIns(5)P, perhaps PtdIns(3)P.

To determine if Fab1p is required for the production of specific phosphatidylinositol derivatives in vivo, we analyzed phosphoinositide levels in wild-type yeast cells and in a strain deleted for FAB1. SEY6210 (wild-type) and fab1Δ1 cells were labeled with myo-[2-³H]inositol for 12 h. At the end of this labeling period, cells were osmotically shocked with 0.9 M NaCl for 10 min. Hyperosmotic shock facilitates the analysis of PtdIns(3,5)P₂, which is difficult to detect in the absence of this treatment, without dramatically affecting the levels of other observable lipids (Dove et al., 1997). The labeled cells were then lysed in acidified chloroform/methanol and the cellular lipids were deacylated and identified by HPLC analysis (see Materials and Methods). This method enables the resolution and quantification of cellular PtdIns(3)P, PtdIns(4)P, PtdIns(3,5)P₂ and PtdIns(4,5)P₂ (Auger et al., 1989; Schu et al., 1993; Dove et al., 1997).

When this analysis is carried out using wild-type yeast cells, four distinct glycero-phosphoinositides corresponding to PtdIns(3)P (13,700 cpm, fraction 32-33), PtdIns(4)P (13,700 cpm, fraction 35-36), PtdIns(3,5)P₂ (2,600 cpm, fraction 66-67) and PtdIns(4,5)P₂ (4,100 cpm, fraction 73-74) are clearly separated (Fig. 2,A). The identities of the deacylated lipid products were confirmed using tritiated standards (data not shown). When an equal quantity of labeled cellular lipids isolated from fab1Δ1 cells are assayed in the same manner, the most dramatic effects are observed on PtdIns(3,5)P₂ levels, which are undetectable (<50 cpm; Fig. 2 A). Relative to wild-type cells, other phosphoinositide levels are also influenced in the fab1Δ1 mutant cells: PtdIns(3)P increases ∼100% (25,000 cpm), PtdIns(4)P decreases ∼50% (7,300 cpm) whereas PtdIns(4,5)P₂ decreases 60% (1,600 cpm).

To determine if the dramatic changes in PtdIns(3,5)P₂ levels seen in the fab1Δ1 strain resulted directly from a loss of Fab1p activity, we assayed PtdIns(3,5)P₂ in a fab1Δ2 + pfab1-2 (fab1^tsf) strain briefly shifted to nonpermissive temperature (Yamamoto et al., 1995; see Materials and Methods). At the permissive temperature, fab1^tsf cells have normal growth characteristics and wild-type vacuole morphology. However, shifting these cells to 37°C results in growth defects and aberrant vacuole morphologies (Bonangelino et al., 1997; Yamamoto et al., 1995). Wild-type and fab1^tsf cells were labeled for 12 h with myo-[2-³H]inositol at 22°C, as before. At the end of this labeling period, half of the wild-type and fab1^tsf cells were incubated with 0.9 M NaCl for 10 min at permissive temperature. The remaining cells were exposed to 0.9 M NaCl for 5 min at 22°C and shifted to the nonpermissive temperature of 38°C for the final 5 min of osmotic shock. Even at the permissive temperature (22°C) PtdIns(3,5)P₂ levels are reduced sevenfold (800 cpm) in the fab1^tsf cells compared to the wild-type control (Fig. 2), whereas the levels of other phosphoinositides remain only moderately affected (data not shown). In the fab1^tsf strain at the nonpermissive temperature, PtdIns(3,5)P₂ synthesis is dramatically affected, reaching undetectable levels (<50 cpm) after a 5-min shift. Compared to the wild-type control, the high temperature incubation results in no significant changes in the other lipids. Given the homology of Fab1p to the family of PtdInsP kinases and the fact that in yeast Vps34p is the only source of PtdIns(3)P, the rapid depletion of PtdIns(3,5)P₂ observed in the fab1 mutant cells suggests that Fab1p encodes a PtdIns(3)P 5-kinase.

The results demonstrate that Fab1p is required for normal PtdIns(3,5)P₂ levels. To address the role of the putative kinase domain of Fab1p in phosphatidylinositol metabolism, we used site-directed mutagenesis to make substitutions in highly conserved amino acid positions of the kinase domain. Specifically, Gly2042 and Gly2045 were simultaneously changed to valine residues (G2042/2045V) and Asp2134 was replaced with an arginine residue (D2134R; Fig. 3 A; see Materials and Methods). The GGxxG glycine triad motif containing Gly2042 and Gly2045 of Fab1p is reminiscent of the highly conserved GxGxxG motif present in the nucleotide binding site of numerous protein kinases (Bossemeyer et al., 1993; Hemmer et al., 1997; Sicheri et al., 1997; Xu et al., 1997). In the protein kinases, these glycine residues are included in the conserved glycine-rich loop and mutation of these residues to alanine or serine residues results in decreases in enzymatic activity (Hemmer et al., 1997; Grant et al., 1998). This comparison predicts that the substitution of these glycine residues to valine residues in Fab1p would result in commensurate impairment of lipid kinase activity. By similar sequence comparisons, Asp2134 of Fab1p appears to be analogous to the catalytically important Asp386 of the c-Src protein kinase (Xu et al., 1997) and mutation of this aspartate residue would be expected to severely affect phosphotransfer (Bossemeyer et al., 1993). An analogous mutation of this aspartate residue to an alanine residue in the Vps34p lipid kinase also lacked detectable amounts of PtdIns(3)P synthesis (Schu et al., 1993). Because of the apparent importance of this residue in catalysis, we decided to make the corresponding mutation in Fab1p.

Both the CEN-based G2042/2045V and the D2134R point mutant constructs were transformed into fab1Δ2 (strains fab1^G2042/2045V and fab1^D2134R, respectively). Immunoprecipitation of metabolically labeled Fab1p indicates that the proteins derived from the fab1^G2042/2045V and fab1^D2134R strains are stably expressed, as the levels of the Fab1p do not vary over the course of a 10- and 45-min chase at either 24°C (Fig. 3 B) or 38°C (data not shown). Interestingly, Fab1p from the fab1^tsf strain is also stable after a 45-min chase at the nonpermissive temperature (data not shown), indicating that despite its loss of kinase activity, the protein is not susceptible to rapid proteolytic degradation.

To determine the effect Fab1p kinase domain point mutations have on PtdIns(3,5)P₂ synthesis, the Fab1p G2042/ 2045V and Fab1p D2134R expressing strains were labeled for 12 h with myo-[2-³H]-inositol and hyperosmotically shocked with 0.9 M NaCl, as described above. Relative to the wild-type control (8,000 cpm of the PtdIns(3,5)P₂), the Fab1p G2042/2045V point mutant exhibits a 9-fold decrease in PtdIns(3,5)P₂ (900 cpm; Fig. 4). This reduced level of PtdIns(3,5)P₂ production in the fab1^G2042/2045V strain is consistent with the reduced levels of kinase activity observed when similar mutations are made in the analogous glycine residues of cAPK (Hemmer et al., 1997; Grant et al., 1998). PtdIns(3,5)P₂ is virtually undetectable (<50 cpm) in the fab1^D2134R strain containing the D2134R change in the putative catalytic loop of Fab1p (Fig. 4). In the same assay, levels of cellular PtdIns(4)P (data not shown) and PtdIns(4,5)P₂ (Fig. 4) did not change dramatically (<30%) as a result of these Fab1p kinase domain point mutations. However, increases of four- and sixfold were observed for the Fab1p substrate PtdIns(3)P in the fab1^G2042/2045V and fab1^D2134R strains, respectively.

PtdIns(3)P is synthesized from PtdIns in Golgi/endosomal membranes, and we have recently discovered that it is transported in a Vps-dependent manner from these compartments to the vacuole (Wurmser and Emr, 1998). Using GFP-RING-FYVE domain fusion constructs which, when expressed in vivo, specifically interact with PtdIns(3)P-containing membranes, the localization of PtdIns(3)P to vacuolar and endosomal membranes was confirmed (Burd and Emr, 1998). Thus, all organelles which contain PtdIns(3)P (Golgi, endosome, and vacuole) are potential sites of Fab1p activity. To localize Fab1p we examined wild-type cells by differential centrifugation.

Cells were converted to spheroplasts, osmotically lysed and cleared of unbroken cells by a low speed spin. The cell extract was then sequentially centrifuged at 13,000 g and 100,000 g yielding soluble fractions (S13 and S100) and pelletable fractions (P13 and P100; see Materials and Methods). All fractions were assayed for the presence of Fab1p, the 100-kD vacuolar ATPase subunit, the Golgi/endosome localized protease, Kex2p and the cytosolic marker glucose 6-phosphate dehydrogenase (G6PDH) using antibodies specific to these proteins. The majority of Fab1p is detected in the pellet fractions (30% P13, 40% P100) and the remainder (30%) is found in the S100 fraction (Table I). In this experiment, vacuolar membranes, identified by the presence of the vacuolar ATPase 100-kD subunit, are found predominantly in the P13 fraction (75%) and also in the P100 (25%). Golgi and endosomal compartments containing Kex2p are enriched in the P100 fraction whereas the cytosolic fraction (G6PDH) is predominantly S100. Fab1p, therefore, appears to be enriched in fractions that contain endosomal and vacuolar membranes, suggesting that Fab1p functions at one (or more) of these compartments.

To refine the localization of native Fab1p, CEN and 2μ versions of HA-tagged Fab1p (Fab1p-HA) have been constructed (see Materials and Methods). These constructs complement the growth and vacuolar morphology defects of the fab1 deletion strain (data not shown). Indirect immunofluorescence using the CEN construct did not produce a discernible signal above background (data not shown). However, at multicopy levels (8–10-fold overexpression; data not shown), we find that a pool of Fab1p-HA colocalizes with Vac8p (Fig. 5), a protein associated with the vacuolar membrane (Wang et al., 1998). Cytoplasmic accumulations of the Fab1p-HA protein are also evident. In some cells, no signal could be detected above background, presumably due to variable levels of Fab1p-HA overexpression within the cell population. Consistent with the subcellular fractionation results, which suggest that Fab1p is distributed among multiple membranes (Table I), this data indicates that Fab1p is associated with vacuolar and prevacuolar compartments.

Transformation of the fab1 null background with either of the Fab1p point mutant constructs results in a differential suppression of the fab1 phenotypes. As shown previously (Yamamoto et al., 1995), these defects include: (a) temperature-sensitive growth, (b) vacuole acidification defects and, (c) a 2.5-fold increase in vacuolar surface area, which leads to inappropriate nuclear segregation during mitosis. The temperature-sensitive growth of the fab1Δ2 at 30°C is complemented in cells expressing either of the Fab1p point mutant constructs. However, although the fab1^{G2042/ 2045V} strain grows even at 38°C, the fab1^D2134R strain does not. Similarly, rescue of the vacuole acidification defect as measured by quinacrine staining of the vacuoles in the fab1 mutants also varied. Vacuolar quinacrine fluorescence in both the fab1Δ2 and the fab1^D2134R strains is extremely weak, indicating a severe acidification defect. However, the fab1^G2042/2045V vacuole appears to have the same fluorescent intensity as the wild-type control (data not shown). By FM4-64 staining of the vacuolar membrane and Nomarski optics, the yeast vacuole appears as a multilobed structure. However, the vacuolar phenotype of the fab1^G2042/2045V and fab1^D2134R strains are similar to that of the fab1 null strain (Yamamoto et al., 1995) or the fab1^tsf after a 30-min shift to the nonpermissive temperature (Fig. 6). The vacuole in these mutants appears as a single grossly enlarged organelle that comprises the majority of the cell volume. Smaller compartments were also revealed immediately adjacent to the vacuole in the fab1^D2134R strain (Fig. 6) At the permissive temperature however, the vacuole morphology of the fab1^tsf strain is indistinguishable from the wild-type control (Fig. 6).

Electron microscopic analysis not only highlights the enlarged vacuolar phenotype resulting from decreases in PtdIns(3,5)P₂ synthesis, but also shows the differential effects of Fab1p kinase domain point mutations on vacuole morphology. The electron-dense vacuole of the wild-type and fab1^G2042/2045V strains contrasts greatly with the electron-transparent nature of the vacuole in the fab1Δ1 and fab1^D2134R strains (Fig. 7). The opacity of the vacuole is likely to be a measure of vacuole acidification, as inactivation of the vacuolar ATPase results in electron-transparent vacuoles (Wurmser and Emr, 1998). Thus, the relatively severe acidification defect in the fab1^D2134R mutant compared to the fab1^G2042/2045V mutant strain correlates with the levels of PtdIns(3,5)P₂ synthesis. Fab1p kinase activity is therefore not only essential for maintenance of vacuolar size and acidification, but may also influence temperature-sensitive growth.

The Vps34p PtdIns 3-kinase is crucial in Golgi to vacuole transport of hydrolases, and recent evidence suggests that one function of PtdIns(3)P is to regulate RING-FYVE-containing proteins (i.e., Vac1p and Vps27p) that are essential for hydrolase transport from the Golgi to the vacuole (Weisman and Wickner, 1992; Piper et al., 1995; Burd et al., 1997). The finding that PtdIns(3)P can serve as a substrate for a PtdIns(3)P 5-kinase raises the possibility that the requirement for Vps34p in vacuolar protein transport arises through its role in the production of PtdIns(3,5)P₂ (Dove et al., 1997). Our findings that PtdIns(3,5)P₂ levels are specifically and dramatically decreased in fab1 mutant strains provides a direct means to assess this hypothesis.

Several trafficking routes from the Golgi to the yeast vacuole have been characterized. Soluble CPY and the type II integral membrane protein carboxypeptidase S, CPS, transit to the vacuole via an endosomal-dependent trafficking route, the CPY pathway (Banta et al., 1988). Alkaline phosphatase (ALP) however, arrives at the vacuole by a distinct pathway that bypasses the endosome, the ALP pathway (Cowles et al., 1997a,b). A third route for protein delivery to the vacuole delivers the protease aminopeptidase I (API) from the cytoplasm to the vacuole through autophagy (Scott et al., 1997). To address the effects that the loss of PtdIns(3,5)P₂ synthesis may have on each pathway, we assayed the proteolytic maturation of representative hydrolases from the three known pathways in the fab1 mutant strains. The conversion of the precursor hydrolase to the mature form occurs upon delivery of the hydrolase to a proteolytically competent vacuole.

Data from the original fab1-2 temperature sensitive allele supports the idea that PtdIns(3,5)P₂ is not required for anterograde flow of traffic to the vacuole via the CPY pathway. After a 60-min preshift at the nonpermissive temperature, CPY was properly sorted and processed (Yamamoto et al., 1995). Our data with the fab1^tsf strain also indicate that the other transport pathways to the vacuole do not depend on Fab1p function. As shown for CPY, ALP is also properly matured (Fig. 8,A) after a 30-min preshift to the nonpermissive temperature. The processing of two additional vacuolar hydrolases, CPS and API, was also found to proceed as in wild-type cells (data not shown). This sorting data from the fab1^tsf strain (Fig. 8,A) at the nonpermissive temperature combined with the PtdIns(3,5)P₂ quantitation (Fig. 2 B) from the same strain under similar temperature shift conditions indicates that Fab1p and PtdIns(3,5)P₂ are not directly involved in the CPY, ALP, or API sorting pathways. In contrast, after only a 5-min preshift to the nonpermissive temperature, the vps34^tsf missorts and secretes >95% of the pulse-labeled CPY (Stack et al., 1995).

As an additional test for protein trafficking to the vacuole, we used indirect immunofluorescence in fab1Δ2 cells with monoclonal antibody against the 60-kD vacuolar ATPase subunit (V₁ complex). This subunit is found distributed on the vacuolar membrane in wild-type cells and arrives at the vacuole via the CPY pathway (Rieder et al., 1996; Piper et al., 1997). In fab1Δ2 cells, the fluorescence for the 60-kD vacuolar ATPase subunit is coincident with the vacuolar membrane observed by Nomarski optics (Fig. 8,B). The ATPase subunit is similarly localized in the original fab1-2 temperature-sensitive strain after incubation at the nonpermissive temperature for 2 h (Bonangelino et al., 1997). Similarly, we used a GFP-ALP construct to localize ALP in the fab1Δ2 mutants. GFP-ALP also was properly localized to the vacuolar membrane in the fab1Δ2 cells (Fig. 8 B). Therefore, Fab1p function, and PtdIns(3,5)P₂ synthesis, is not essential for delivery of biosynthetic cargo to the vacuole via either the CPY or ALP transport pathways.

In the fab1^G2042/2045V and fab1^D2134R strains, with reduced and undetectable steady-state levels of PtdIns(3,5)P₂, respectively, we assessed the proteolytic processing of both CPS and ALP (Fig. 8,C). In both mutant strains, ALP maturation occurs at a rate indistinguishable from wild-type, even at the 15-min chase point (Fig. 8,C). Compared to the control, CPS processing also occurs with wild-type kinetics in fab1^G2042/2045V mutant cells (Fig. 8,C). However, there is a significant kinetic delay in CPS maturation in fab1^D2134R cells (Fig. 8 C). Based on chase times extending to 120 min (data not shown), the half-time for processing CPS to the mature form in wild-type and fab1^G2042/2045V cells is approximately 20 min at 24°C, whereas in the fab1^D2134R strain it is >120 min. Mature API is also detectable in both of these mutants, however with the fab1^G2042/2045V mutant, there is a slight kinetic delay in maturation that is more severe in the fab1^D2134R mutant (data not shown). These data indicate that all three vacuolar transport pathways are generally intact, as a significant pool of hydrolases delivered to the vacuole by these routes are matured independent of Fab1p kinase activity.

The kinetic delay in hydrolase maturation observed in the fab1Δ (Yamamoto et al., 1995) or point mutant strains (Fig. 8 C) may be due to vacuolar acidification defects that result from the loss of Fab1p kinase activity. Loss of vacuole acidification is known to reduce the activity of proteinase A (Sorensen et al., 1994), a vacuolar protease that plays a key role in the maturation of many vacuolar hydrolases (Morano and Klionsky, 1994; Van Den Hazel et al., 1996). As a result, cells lacking vacuolar ATPase function exhibit kinetic delays in the conversion of many hydrolases to their mature forms (Morano and Klionsky, 1994). As assessed by quinacrine staining, the lipid kinase activity of Fab1p is also required for proper vacuole acidification. The normal quinacrine fluorescence in the fab1^G2042/2045Vstrain (data not shown), indicates that even low levels of PtdIns(3,5)P₂ are sufficient to maintain vacuolar pH. However, the fab1^D2134R mutant strain shows a severe vacuole acidification defect (data not shown). The processing delays observed in the fab1 point mutants are likely to be directly due to the slow maturation of certain hydrolases in a compromised vacuole. Consistent with this hypothesis, the degree of vacuole acidification in the two point mutants correlates well with the hydrolase maturation delays observed in these strains.

A number of observations suggest that an upstream regulatory factor(s) may be required for Fab1p activity. First, overexpression of FAB1 from a multicopy vector fails to cause elevations in endogenous PtdIns(3,5)P₂ (data not shown), suggesting that the concentration of Fab1-mediated synthesis of PtdIns(3,5)P₂ requires a limiting cofactor. Furthermore, immunoprecipitations of Fab1p from yeast protein extracts did not contain detectable PtdIns(3)P 5-kinase activity (data not shown), possibly due to the disassociation of Fab1p from an activator protein(s) during immunopurification. Finally, several of the recently described vac mutants exhibit phenotypes which are strikingly similar to the fab1 mutants, VAC7 is the best characterized. Like fab1, vac7 mutants exhibit acidification defects and an enlarged vacuolar phenotype (Bonangelino et al., 1997). This raises the possibility that Fab1p and Vac7p function together to regulate a common cell function (i.e., vacuole membrane homeostasis). To further assess whether Vac7p functions in the Fab1p pathway, we examined the morphology of vac7Δ cells by electron microscopy and assayed a vac7 mutant for changes in PtdIns(3,5)P₂ levels.

Relative to wild-type cells, which exhibited two to three electron-dense staining vacuoles when examined by electron microscopy, the vac7Δ (LWY1481) strain has a single enlarged vacuole that appears less electron dense (Fig. 7 and 9 A). Heterogeneous intravacuolar structures appear more prominently in vac7Δ cells than in fab1Δ1 cells (Fig. 7 B and 9 A). It is also important to note that aberrant membranes (vesicles and endosomal intermediates) do not appear to accumulate in the cytoplasm of the fab1Δ1 or vac7Δ cells, a hallmark of many mutants which have Golgi to vacuole transport defects (Cowles et al., 1994; Babst et al., 1997).

As the similar morphological defects of fab1Δ1 and vac7Δ cells suggest, if Fab1p and Vac7p function in the same pathway, then the mutation of VAC7 might also have effects on PtdIns(3,5)P₂ levels. Therefore, we analyzed PtdIns(3,5)P₂ in a vac7 mutant strain. Wild-type and vac7Δ cells were labeled 12 h with myo-[2-³H]inositol, osmotically shocked for 10 min with 0.9 M NaCl, lysed and analyzed by HPLC as described earlier. PtdIns(3,5)P₂ (5,300 cpm total) was observed in wild-type cells, however, PtdIns(3,5)P₂ was undetectable in vac7Δ cells (<50 cpm; Fig. 9,B). As seen in fab1Δ1 cells (Fig. 2,A), the vac7Δ mutant also exhibits a ∼50% decrease PtdIns(4,5)P₂ (Fig. 9 B), 100% increase in PtdIns(3)P and a 50% decrease in PtdIns(4)P (data not shown). Vac7p is not homologous to any other known protein and does not contain kinase motifs (Bonangelino et al., 1997). These data suggest that Vac7p, possibly in association with Fab1p, is required for PtdIns(3,5)P₂ synthesis.

Sequence comparisons of the putative lipid kinase domain of Fab1p (a.a. 2,022–2,263) with other kinases of known enzymatic activities categorized Fab1p into a unique subclass of general PtdInsP kinases (Fig. 1,B). Two distinct doubly phosphorylated derivatives of PtdIns have been identified in yeast, PtdIns(4,5)P₂ and PtdIns(3,5)P₂, and the enzyme activities involved in their stepwise synthesis already have been suggested (Fig. 1 A). Although its identity was not known, the existence of a specific PtdIns(3)P 5-kinase is supported by the sequential synthesis of PtdIns(3,5)P₂ in murine Ph cells from PtdIns(3)P, as well as the requirement for Vps34p activity in yeast for PtdIns(3,5)P₂ production (Dove et al., 1997; Whiteford et al., 1997). Given that the sequence of Fab1p kinase domain distinguishes it from the Type I and Type II PtdIns(4)P 5-kinases, Fab1p is the likely PtdIns(3)P 5-kinase in yeast.

Consistent with this hypothesis, our data show that fab1Δ1 cells have undetectable levels of PtdIns(3,5)P₂, whereas the amounts of other phosphoinositides are only moderately affected (Fig. 2,A). To determine if this decrease in PtdIns(3,5)P₂ is the direct result of Fab1p inactivation, we assessed the phosphoinositide levels in a strain harboring a temperature-sensitive allele of FAB1. Despite the fact that this mutant protein is stable during the shift to the nonpermissive temperature, cellular PtdIns(3,5)P₂ levels show a rapid decrease, reaching undetectable levels within 5 min at 38°C, whereas the amounts of other phosphoinositides do not change relative to wild-type cells treated in the same manner (Fig. 2,B). The differential loss of PtdIns(3,5)P₂ in cells expressing either of the two kinase domain point mutant constructs also provides evidence that Fab1p is the PtdIns(3)P 5-kinase (Fig. 4). By comparison with known structures of protein kinases, the glycine residues replaced in the Fab1p G2042/2045V mutant are located in a conserved glycine-rich loop region between two β-strands that form the ATP-positioning motif (Sicheri et al., 1997; Xu et al., 1997). Mutagenesis of these residues in cAPK and the insulin receptor results in reduced kinase activity (Hemmer et al., 1997; Odawara et al., 1989). Therefore, the residual PtdIns(3,5)P₂ synthesis observed in the fab1^G2042/2045V strain is consistent with Fab1p directly catalyzing the synthesis of PtdIns(3,5)P₂ from PtdIns(3)P. Additionally, mutation of the aspartate residue at position 2,134 eliminates the production of PtdIns(3,5)P₂. This aspartate is analogous to the catalytically critical Asp166 of cAPK (Bossemeyer et al., 1993). Mutation of the comparable residue in the v-fps and c-kit protein kinases as well as the yeast lipid kinase Vps34p results in the loss of kinase activity (Moran et al., 1988; Tan et al., 1990; Schu et al., 1993).

Although the in vivo lipid quantitation data from the fab1 mutants provide strong evidence that Fab1p is the functionally relevant PtdIns(3)P 5-kinase in vivo, it has not yet been possible to demonstrate the activity in vitro. Immunopurification of Fab1p from native yeast extracts did not contain detectable PtdIns(3)P 5-kinase activity (data not shown), nor were we able to demonstrate the production of PtdIns(3,5)P₂ in crude yeast lysates (data not shown). We suggest that these negative in vitro results may be due to the instability or lack of necessary cofactors required for Fab1p 5-kinase activity (see below) or the relative inefficiency of the enzyme under our in vitro conditions.

The catalytic activity and localization of most if not all phosphoinositide kinases are controlled by interactions with regulatory proteins in vivo. The p85/p110 phosphoinositide 3-kinase translocates to the plasma membrane through its interaction with the cytosolic tail of ligand-bound tyrosine kinase receptors (Hu et al., 1992), whereas the p110 phosphoinositide 3-kinase γ-isoform is activated in response to heterotrimeric G-protein–coupled receptor stimulation (Stoyanov et al., 1995; Stephens et al., 1997). Similarly, the Vps34p PtdIns 3-kinase binds a membrane associated serine/threonine protein kinase, Vps15p (Stack et al., 1993, 1995). This interaction localizes Vps34p to an intracellular membrane compartment and activates its PtdIns 3-kinase activity (Schu et al., 1993; Stack et al., 1995). We have made several observations which point to the existence of a protein cofactor(s) that is required for Fab1p activity. First, cells overexpressing Fab1p failed to exhibit increases in PtdIns(3,5)P₂ levels, suggesting that a protein critical for Fab1p activation may be limiting. Similarly, overexpression of Vps34p does not elevate cellular PtdIns(3)P levels (data not shown), presumably due to the limiting pool of activated Vps15p available to properly recruit and regulate Vps34p activity. Second, the lack of detectable PtdIns(3)P 5-kinase activity in Fab1p in vitro experiments also suggests that stimulatory signals from an upstream regulator may be absent from immunoprecipitated Fab1p or unstable in crude lysates.

The identification of a possible upstream activator for Fab1p came from a genetic screen in yeast for mutants exhibiting aberrant vacuole segregation structures during cytokinesis (Wang et al., 1996). This screen revealed several complementation groups displaying the enlarged vacuolar phenotypes of fab1 cells. Indeed FAB1 was isolated numerous times as well as a previously uncharacterized gene, VAC7 (Bonangelino et al., 1997). To test the possibility that Vac7p and Fab1p function in the same pathway, we assayed phosphoinositide levels in vac7Δ cells and found that PtdIns(3,5)P₂ was undetectable in this mutant strain (Fig. 9 B). Thus, Vac7p activity, like Fab1p, is required for PtdIns(3,5)P₂ production, suggesting that Vac7p functions upstream of Fab1p. Vac7p encodes a 1,165 a.a. protein that is enriched in vacuole membranes and bears no homology to lipid kinases (Bonangelino et al., 1997). It contains a potential transmembrane domain (a.a. 919–943) and is not stripped from the membrane fraction with high NaCl, urea, or Na₂CO₃ treatment indicating that Vac7p is an intergral membrane protein. Vac7p however, is unlikely to be critical for localizing Fab1p to the membrane as its deletion failed to increase the soluble pool of Fab1p (unpublished observation). This suggests that Vac7p may function to stimulate Fab1p PtdIns(3)P 5-kinase activity rather than recruit Fab1p to the membrane.

Other members of the FAB1/VAC7 gene family are likely to exist. In fact, a complementation group distinct from fab1 and vac7, designated vac14-1, has been identified that shares the vacuole acidification and morphological defects observed in fab1 and vac7 mutants (Bonangelino et al., 1997). Furthermore, in the vac14-1 mutant, both these defects are suppressed by the overexpression of FAB1 (Bonangelino et al., 1997). Therefore, Vac14p represents another protein that is likely to function upstream of Fab1p. Consistent with this model, vac14-1 mutant cells are also severely compromised in their ability to synthesize PtdIns(3,5)P₂ (unpublished observations). Efforts to clone VAC14 are currently underway. Interestingly, overexpression of FAB1 does not suppress the mutant phenotypes of vac7Δ cells (Bonangelino et al., 1997). Collectively, these data suggest that Fab1p, Vac7p and Vac14p function in a common signal transduction pathway which, through the activation of Fab1p, results in the localized production of PtdIns(3,5)P₂.

It has been previously demonstrated that a hyperosmotic shock is required for maximal PtdIns(3,5)P₂ accumulation in yeast. This was suggested to be due to the activation of a PtdIns(3)P 5-kinase (Dove et al., 1997). Our data does not exclude the possibility that Fab1p, or its putative upstream activators Vac7p and Vac14p, are sensitive to hyperosmotic stress. However, enzymes dedicated to the turnover of PtdIns(3,5)P₂, possibly corresponding to one or more of the four PtdIns(5)P 5-phosphatase homologs in yeast (Srinivasan et al., 1997; Stolz et al., 1998), also represent candidate activities that may be sensitive to hyperosmotic stress. The further characterization of upstream activators, such as Vac7p and Vac14p, and the biologically relevant PtdIns(3,5)P₂ turnover activity should help in distinguishing between these possibilities.

With the identification of PtdIns(3,5)P₂ in yeast, it was suggested that this lipid, rather than PtdIns(3)P, plays a role in Golgi to vacuole trafficking (Dove et al., 1997). Although previous data indicates the essential nature of PtdIns(3)P in anterograde traffic to the vacuole (Schu et al., 1993), it was argued that the apparent requirement for Vps34p in this pathway reflects the need for PtdIns(3)P solely as a substrate for the 5-kinase activity. However, our data show that Fab1p, and therefore PtdIns(3,5)P₂ synthesis, appears not to play a role in anterograde protein traffic, as the delivery of proteins along all known vacuolar transport pathways are unaffected in fab1 mutants. fab1 mutant cells do give rise to several phenotypes including, temperature-sensitive growth, a vacuolar acidification defect, as well as an inability to properly segregate genomic and vacuolar material upon cytokinesis (Yamamoto et al., 1995). The later phenotypes appear to be secondary consequences resulting from the enlargement of the vacuole, which is the most immediate defect caused by the inactivation of the fab1^tsf. Indeed, the normal vacuole morphology observed in the fab1^tsf at the permissive temperature contrasts dramatically with the enlarged vacuole phenotype of the fab1^G2042/2045V strain, suggesting that a threshold level of PtdIns(3,5)P₂ production is required for proper vacuolar morphology.

We envision three possible models for how Fab1p activity is responsible for maintenance of vacuolar size. (a) PtdIns(3,5)P₂ may be required for vacuole ion homeostasis. Aside from its function as a site for macromolecular degradation, the vacuole is an important ion storage site which plays a role in regulating intracellular ion concentrations (Klionsky et al., 1990). Therefore, PtdIns(3,5)P₂ could be an activator of a vacuolar ion transporter and loss of this lipid could affect vacuolar morphology through changes in its osmolarity with respect to the cytoplasm. In mammalian cells, PtdIns(4,5)P₂ has been shown to be a positive regulator of inward-rectifying potassium channels (Huang et al., 1998; Sui et al., 1998). (b) Alternatively, PtdIns(3,5)P₂ could regulate the turnover of vacuolar membranes by mediating its invagination at the vacuole (or endosome), allowing for degradation within the vacuole lumen. Invaginations at the vacuole or endosome have been suggested by the presence of lumenal vesicles within the vacuole (Wurmser and Emr, 1998) and endosome (multivesicular body; Gruenberg and Maxfield, 1995). Consistent with this idea, fab1Δ1 and fab1^D2134R kinase domain mutant strains have significantly fewer intravacuolar structures. (c) PtdIns(3,5)P₂ may direct the efflux of vacuolar membranes to earlier secretory compartments. According to this model vacuolar membrane homeostasis requires a balance between anterograde traffic (mediated by PtdIns(3)P) and retrograde traffic (perhaps through PtdIns(3,5)P₂; Fig. 10). Disruption of PtdIns(3,5)P₂ synthesis might prevent membrane recycling and result in the accumulation of membranes at the vacuole. This predicts that PtdIns(3,5)P₂ may be required for the formation or docking/fusion of retrograde carrier vesicles emanating from the vacuole. Indeed, the existence of a retrograde pathway from the vacuolar membrane to earlier secretory compartments has also been suggested based on the recycling of a recombinant protein construct from the vacuole to the Golgi (Bryant et al., 1998).

Therefore, this model suggests that PtdIns(3)P may itself act as an internal membrane marker which is synthesized early in the Golgi to vacuole transport pathway and imprints the carrier vesicles with a tag that is subsequently decoded/recognized at a distal site in the pathway. Fab1p may recognize and convert this anterograde-specific lipid tag into a second signaling lipid that then directs vacuolar membrane efflux or turnover. Recently, however, it has been shown that a significant fraction of PtdIns(3)P is delivered to the vacuole as cargo where it is degraded by lumenal hydrolases (Wurmser and Emr, 1998). This necessitates that while a pool of PtdIns(3)P is transferred to the lumen of the vacuole for degradation, a separate or earlier pool of PtdIns(3)P remains available as a substrate for Fab1p in the cytoplasmic leaflet (Wurmser and Emr, 1998; Burd and Emr, 1998). This sequential signaling system at distinct organelles in the Golgi to vacuolar signaling pathway would allow the cell to maintain vacuole size/function (homeostasis).

The involvement of phosphoinositides in the spatial and temporal regulation of membrane trafficking argues that the various phosphorylated derivatives must be synthesized and degraded at specific sites along the pathway. For instance, the lipid PtdIns(4,5)P₂ may not only direct vesicle fusion (Hay and Martin, 1993; Hay et al., 1995), but its conversion by a PtdIns(5)P 5-phosphatase may also play a role in synaptic vesicle endocytosis (McPherson et al., 1994, 1996). Similarly, mutations within the PDGF tyrosine kinase receptor that prevent the recruitment of the mammalian phosphoinositide 3-kinase cause defects in the sorting and downregulation of this receptor (Kapeller et al., 1994). Interestingly, the internalization of the receptor occurs normally, indicating that the subsequent endosome to lysosome trafficking step is affected. Thus, the regulated synthesis and turnover of specific lipids at distinct organelle sites is required for the regulation of membrane trafficking. The isolation of Vps34p as a vacuolar protein sorting mutant lead to the hypothesis that the regulated production PtdIns(3)P on a specific membrane directly effects or recruits necessary factors for proper trafficking (Stack et al., 1993). Candidate downstream targets of PtdIns(3)P in yeast have already been identified, in part, based on the observation that the mammalian EEA1 gene product specifically interacts with PtdIns(3)P-containing liposomes (Patki et al., 1997). This interaction has now been determined to be mediated by the RING-FYVE finger domain (Burd and Emr, 1998) that is also present in Vps27p and Vac1p, proteins previously shown to be required for Golgi to vacuole protein trafficking (Piper et al., 1995; Burd et al., 1997). The binding of PtdIns(3)P by this domain may serve to recruit these proteins to a specific membrane site and/or modulate the activity of these and other transport factors. Interestingly, this domain is also present at the NH₂ terminus of Fab1p and may function to localize (and/or activate) Fab1p to membranes containing its substrate, PtdIns(3)P.

Specific downstream effects of phosphoinositides are often mediated by proteins which bind directly to specific lipid ligands (Martin, 1997). Regulation of membrane homeostasis by Fab1p is likely to be mediated by a set of proteins that specifically interact with PtdIns(3,5)P₂. These PtdIns(3,5)P₂-specific binding proteins could correspond with proteins that commonly play roles in membrane trafficking events, such as coat proteins, SNAREs, or rab GTPases. As the ongoing VAC mutant screen has been successful in defining candidate upstream regulators of Fab1p activity, it also promises to be instrumental in identifying downstream effectors of Fab1p. Our results with Fab1p confirm the use of yeast in defining signal transduction pathways on the basis of an observable phenotype and ordering these factors within the pathway based on in vivo biochemical analyses.

We thank Michael McCaffrey and Chris Hofeditz for carrying out electron microscope analysis and figure preparation (Immunoelectron microscopy Core B of Program Project grant CA58689 headed by M. Farquar). We also appreciate the helpful scientific discussions with Chris Burd, Markus Babst, and Tamara Darsow.

This work was supported by NIH grants CA58689 (S.D. Emr and A.E. Wurmser) and GM50403 (L.S. Weisman). S.D. Emr is supported as an investigator of the Howard Hughes Medical Institute.

a.a.

amino acid

ALP

alkaline phosphatase

API

aminopeptidase I

CPS

carboxypeptidase S

CPY

carboxypeptidase Y

G6PHD

glucose 6-phosphate dehydrogenase

PITP

PtdIns transfer protein

PtdIns(4)P

phosphatidylinositol 4-phosphate