|Home | About | Journals | Submit | Contact Us | Français|
Phosphatidylinositol-3,5-bisphosphate [PtdIns(3,5)P2] regulates several vacuolar functions, including acidification, morphology, and membrane traffic. The lipid kinase Fab1 converts phosphatidylinositol-3-phosphate [PtdIns(3)P] to PtdIns(3,5)P2. PtdIns(3,5)P2 levels are controlled by the adaptor-like protein Vac14 and the Fig4 PtdIns(3,5)P2-specific 5-phosphatase. Interestingly, Vac14 and Fig4 serve a dual function: they are both implicated in the synthesis and turnover of PtdIns(3,5)P2 by an unknown mechanism. We now show that Fab1, through its chaperonin-like domain, binds to Vac14 and Fig4 and forms a vacuole-associated signaling complex. The Fab1 complex is tethered to the vacuole via an interaction between the FYVE domain in Fab1 and PtdIns(3)P on the vacuole. Moreover, Vac14 and Fig4 bind to each other directly and are mutually dependent for interaction with the Fab1 kinase. Our observations identify a protein complex that incorporates the antagonizing Fab1 lipid kinase and Fig4 lipid phosphatase into a common functional unit. We propose a model explaining the dual roles of Vac14 and Fig4 in the synthesis and turnover of PtdIns(3,5)P2.
Phosphoinositides (PtdInsPs) are key determinants of the organelle identity code, which establishes and maintains organelle-specific properties such as size, shape, and function. They have important and diverse functions in membrane traffic, including regulation of membrane fission and fusion (Behnia and Munro, 2005 ; Di Paolo and De Camilli, 2006 ; Takenawa and Itoh, 2006 ). There are seven species of PtdInsPs (four in yeast) that are rapidly interconverted through the action of specific PtdIns kinases and phosphatases. Importantly, each PtdInsP has a stereotypical organelle distribution that regulates the localization of cognate effector proteins. PtdInsP recognition is predominantly achieved by short protein modules such as those of the FYVE and PH domain families (Lemmon, 2003 ; Balla, 2005 ; Behnia and Munro, 2005 ; Di Paolo and De Camilli, 2006 ; Takenawa and Itoh, 2006 ). Notably, these modules differ significantly in their affinity and specificity for each PtdInsP (Lawe et al., 2000 ; Balla, 2005 ). Consequently, PtdInsPs can direct an extensive network of membrane–protein interactions.
Genetic ablation of phosphatidylinositol-3,5-bisphosphate [PtdIns(3,5)P2] synthesis in Saccharomyces cerevisiae causes numerous vacuolar defects, including nonacidification, impaired vacuole-to-endosome retrograde transport, abnormal inheritance, and a prominently enlarged single-lobed vacuole (Efe et al., 2005 ; Michell et al., 2006 ). Similarly, PtdIns(3,5)P2-deficient cells from higher organisms also exhibit enlarged endosomes and impaired membrane traffic (Sbrissa et al., 2000 ; Nicot et al., 2006 ; Rusten et al., 2006 ). Thus, PtdIns(3,5)P2 governs events associated with endosomal compartments.
In yeast, PtdIns(3,5)P2 is generated by Fab1, which phosphorylates phosphatidylinositol-3-phosphate [PtdIns(3)P] at the 5-position. Fab1 is a conserved lipid kinase known as PIKfyve in mammals (Cooke et al., 1998 ; Gary et al., 1998 ; Sbrissa et al., 1999 ). Typically, Fab1 and its orthologues share a similar architecture: 1) an N-terminal FYVE domain, 2) a middle region related to the CCT/TCP-1/Cpn60 chaperonins that are involved in productive folding of actin and tubulin, 3) a second middle domain that contains a number of conserved cysteine residues (CCR) unique to the Fab1 family, and 4) a C-terminal lipid kinase domain related to PtdInsP kinases (Efe et al., 2005 ; Michell et al., 2006 ). Our understanding of how these domains modulate and coordinate membrane association and kinase activity of Fab1 remains poor. The mammalian PIKfyve depends on its FYVE domain and PtdIns(3)P for association with endosomes (Shisheva et al., 1999 ). However, the functional relevance for the FYVE domain of Fab1 remains unexplored. Similarly, despite being highly conserved, the roles of the CCT and CCR domains are practically uncharacterized (Efe et al., 2005 ; Michell et al., 2006 ).
In addition to Fab1, several vacuolar proteins that modulate cellular levels of PtdIns(3,5)P2 have been identified. These include Vac7, a transmembrane protein with no apparent orthologues outside of Fungi; and the conserved adaptor-like protein Vac14 (Bonangelino et al., 1997 ; Dove et al., 2002 ; Gary et al., 2002 ). In vac14Δ and vac7Δ cells, PtdIns(3,5)P2 levels are severely reduced and exhibit a phenotype similar to fab1Δ cells (Bonangelino et al., 1997 ; Dove et al., 2002 ; Gary et al., 2002 ). In contrast, deletion of ATG18 causes a drastic increase in PtdIns(3,5)P2 levels (Dove et al., 2004 ). Atg18 binds PtdIns(3,5)P2 and is thought to serve dual roles: as an effector of the lipid and as a negative regulator of Fab1 kinase activity (Dove et al., 2004 ). In addition, Fig4 is a PtdIns(3,5)P2-specific 5-phosphatase that was identified in a vac7Δ suppressor screen that partially rescued the levels of PtdIns(3,5)P2 (Gary et al., 2002 ). Therefore, Fig4 was originally proposed to directly counteract Fab1 activity (Gary et al., 2002 ; Rudge et al., 2004 ). Interestingly, recent work suggests that Fig4 is also necessary for the synthesis of PtdIns(3,5)P2; unlike wild-type yeast cells, fig4Δ cells do not exhibit a rapid and sharp increase in PtdIns(3,5)P2 levels upon exposure to high salt (Dove et al., 1997 ; Duex et al., 2006a ,b ). Additionally, deletion of FIG4 prevents the steady-state increase in PtdIns(3,5)P2 levels in atg18Δ cells (Efe et al., 2007 ). Curiously, Vac14 is also implicated in turnover of PtdIns(3,5)P2 (Duex et al., 2006b ). A key question is how Vac14 and Fig4 carry out their dual roles in PtdIns(3,5)P2 synthesis and turnover?
The molecular mechanisms that permit the Fab1 kinase and its regulatory proteins to govern PtdIns(3,5)P2 levels are poorly understood. Conceivably, these proteins form an intricate network of interactions that regulate PtdIns(3,5)P2. In yeast, the only established interaction as of yet is between Fig4 and Vac14 (Rudge et al., 2004 ; Duex et al., 2006b ), whereas assays testing for Fab1 interaction with its regulators have thus far been unsuccessful (Bonangelino et al., 2002 ; Rudge et al., 2004 ). In mammals, it is proposed that PIKfyve forms a ternary complex with ArPIKfyve and hSac3, the Vac14 and Fig4 orthologues, respectively (Sbrissa et al., 2007 ). However, the functional consequence for this complex remains to be examined. This study provides evidence that Fab1, Vac14, and Fig4 exist in a common vacuole-associated protein complex. Our observations suggest that Vac14 and Fig4 are codependent for interaction with the chaperonin-related domain of Fab1 and suggest a mechanism for the dual role of Vac14 and Fig4 in the synthesis and turnover of PtdIns(3,5)P2.
All S. cerevisiae strains used in this study and their genotypes are listed in Table 1. Strains were grown in rich (YPD) or minimal (SD) media supplemented with the appropriate amino acids. Standard growth conditions and manipulations have been described previously (Rose et al., 1990 ).
Restriction enzymes were purchased from New England Biolabs (Ipswich, MA), and polymerase chain reactions (PCRs) were carried out with PfuUltra polymerase (Stratagene, La Jolla, CA) or ExTaq (Takara Mirus Bio, Madison, WI) for cloning and analytical reactions, respectively. Standard molecular biology techniques were used for all DNA manipulations (Maniatis et al., 1992 ). Yeast transformations were done as described previously (Ito et al., 1983 ). Yeast genomic preparations were performed with a modified protocol based on the QuickSpin mini-prep kit (QIAGEN, Valencia, CA). Briefly, yeast cell lysis and genomic DNA fragmentation was done by vortexing cells with glass beads in buffer P1 for 10 min. Subsequent steps were done as instructed by the manufacturer, except that DNA was washed once with PB buffer (QIAGEN) after binding to spin-columns.
PCR-amplified fragments were used to make genomic deletions and for epitope tagging as described previously (Longtine et al., 1998 ). All deletions and integrations were verified by PCR and/or Western blot analysis. Amino acid substitutions were introduced using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). Oligonucleotide sequences are available upon request.
pGEX4T3 (GE Healthcare, Little Chalfont, Buckinghamshire, United Kingdom) and pET23d(+) (Novagen, Madison, WI) were used to make glutathione transferase (GST) and T7-HIS6-tagged recombinant proteins, respectively. pJE181 (encoding Atg18), pBP74A, and pRB415A have been described previously (Efe et al., 2007 ). Constructs expressing FIG4 and fig4-1 were described in Gary et al. (2002) , and pRS416::FAB1 and pRS416::FAB1D2134R were characterized in Gary et al. (1998) . The methodology used to clone the various constructs used in this study can be found in Supplemental Material.
Cells were grown in the appropriate medium to OD600 ~1 and subsequently diluted to an initial OD600 of 0.6 in selective medium. Cultures were then serially diluted 10-fold, and 2 μl was spotted onto selective plates and grown at 26 or 38°C for 3 d.
Vacuoles were labeled with the fluorescent probes N-[3-triethylammoniumpropyl]-4-[p-diethylaminophenylhexatrienyl] pyridinium dibromide FM4-64 (Invitrogen, Carlsbad, CA) and/or 4-chloromethyl-7-aminocoumarin (Invitrogen) as described previously (Efe et al., 2007 ). Fluorescence and differential interference contrast (DIC) images of labeled cells and those expressing green fluorescent protein (GFP) fusion proteins were generated with a Delta Vision RT microscopy system (Applied Precision, Issaquah, WA). Specifically, data were acquired using an IX71 inverted microscope (Olympus, Tokyo, Japan) equipped with UV, fluorescein isothiocyanate, and rhodamine filters and coupled to a CoolSNAP HQ camera (Photometrics, Tucson, AZ). Background subtraction and intensity level were linearly adjusted using Adobe Photoshop CS (Adobe Systems, San Jose, CA).
Fluorescence intensity of FM4-64 and GFP channels of unprocessed TIFF images was quantified using ImageJ 1.36b (National Institutes of Health, Bethesda, MD). Typically, plot profiles were acquired by drawing a 5-pixel wide by 20- to 40-pixel-long line drawn from the midpoint of the cytosol to the vacuole lumen. Line positioning was randomly selected but avoided punctate structures on the vacuolar membrane. Fluorescence intensity values were exported to Excel 2004 (Microsoft, Redmond, WA). For graphical representation, intensity values were normalized against the highest intensity value to achieve a peak of 1. The vacuolar membrane was defined as the pixels with the highest FM4-64 intensity. A modified method was used to calculate the vacuole-to-cytosol (V/C) fluorescence ratio. A freehand line was drawn over >50% of the vacuolar membrane circumference, defined by the presence of FM4-64, and the average GFP fluorescence intensity was recorded. Mean cytosolic GFP fluorescence intensity was acquired by drawing a freehand area on >25% of the cytosolic area. These two values were then used to calculate the V/C fluorescence ratio. The two-tail, unpaired Student's t test was used to check for differences in V/C ratio between samples. Note that the contrast and the brightness of the images shown have been adjusted after acquiring numerical data.
Incorporation of 3H-labeled myo-inositol into phosphoinositides was analyzed as described previously by Rudge et al. (2004) . Briefly, 5 OD600 units of cells were labeled with 60 μCi of myo-[3H]inositol (GE Healthcare) in SD media lacking inositol for 1 h. Phospholipids were then precipitated in 9% perchloric acid for 5 min and washed in 1 ml of 0.1 mM EDTA. Deacylation was done in methylamine reagent (10.7% methylamine, 45.7% methanol, and 11.4% 1-butanol) for 50 min at 53°C, followed by evaporation in a vacuum chamber. Residual methylamine was eliminated by resuspending samples two times in 500 μl of sterile water and drying. After a third resuspension in water, an equal volume of extraction reagent (1-butanol/ethyl-ether/formic acid ethyl ester at a ratio of 20:4:1) was added, vortexed, and centrifuged. The aqueous phase containing the deacylated phosphoinositides was collected, and the extraction procedure was repeated twice more, followed by a drying step.
For quantitative analyses, dried pellets were resuspended in sterile water and 1 × 107 cpm were separated on a Partisphere SAX column (Whatman, Florham Park, NJ) attached to a high-performance liquid chromatography (HPLC) system (Shimadzu, Kyoto, Japan) and a 610TR on-line radiomatic detector (PerkinElmer, Whaltham, MA) by using Ultima Flo scintillation fluid (PerkinElmer). The HPLC and on-line detector were controlled with EZStart 7.2.1 and ProFSA 3.3 software, respectively, with final data analysis taking place in the latter.
Yeast cultures were grown to OD600 ~0.6, and 20–60 OD600 units were used for coimmunoprecipitation (coIP) assays. Cells were spheroplasted and lysed in a 2-ml Dounce homogenizer in HEPES-lysis buffer (20 mM HEPES, pH 7.2, 50 mM potassium acetate, 200 mM sorbitol, and 2 mM EDTA), supplemented with 3× Complete protease inhibitor cocktail (Roche Diagnostics, Indianapolis, IN), 0.3 mM 4-(2-aminoethyl)benzenesulfonyl fluoride, 3 μg/ml pepstatin A, 3 μg/ml leupeptin, 3 mM benzamide, 3× chymostatin, and 3× aprotinin. Cleared cell lysates were solubilized in 0.25% Tween 20 and detergent-insoluble material spun down at 13,000 × g. IPs were done against the respective epitopes with 1 μg/ml monoclonal antibodies: anti-Myc (9E10; Sigma-Aldrich), anti-hemagglutinin (HA) (12CA5; Roche Diagnostics), or anti-FLAG (M2; Sigma-Aldrich) for 1 h at 4°C followed by the addition of GammaBind G-linked Sepharose beads (GE Healthcare) for 1–2 h. Beads were washed three times with detergent-containing HEPES-lysis buffer and two times without detergent. CoIPs were then analyzed by SDS-PAGE and Western blotting. Fig4 was detected with rabbit anti-Fig4 polyclonal antibodies used at a dilution of 1:10,000 (Rudge et al., 2004 ). Secondary antibodies were horseradish peroxidase linked and detected with ECL reagent (GE Healthcare). Antibody stripping was done with Blot Rejuvenation system (Millipore Bioscience Research Reagents, Temecula, CA).
For the salt-shock experiments, cells were grown as described above and exposed to 0.9 M NaCl for 0, 8, or 25 min. Cells were then quickly spun-down and lysed by vortexing in 300 μl of lysis buffer supplied with protease inhibitors and 300 μl of glass beads for 10 min at 4°C. The total volume was then increased to 1.0 ml with lysis buffer and processed as described above.
Cells were lysed as described above and unbroken cells were eliminated by two spins at 500 × g for 10 min. The supernatant was then centrifuged at 100,000 × g for 45 min to obtain the cytosolic (supernatant) and the membrane fractions (pellet). Samples were then precipitated with 10% trichloroacetic acid (TCA), washed twice with acetone, and dissolved in sample buffer. Subcellular fractionation was tested with monoclonal anti-Vph1 antibodies (10D7, 1:500; Invitrogen) and rabbit polyclonal anti-glucose-6-phosphate dehydrogenase antibodies (Sigma-Aldrich).
BL21 bacteria were used to express recombinant fusion proteins. Cells were grown to OD600 1.0 in SuperBroth media supplemented with ampicillin, and expression was induced with 0.3 mM isopropyl β-d-thiogalactoside overnight at 22°C. Cells were then collected and washed with lysis buffer. To purify GST-fusion proteins, cells were resuspend in 25 ml of GST lysis buffer (20 mM Tris, pH 7.5, 100 mM NaCl, and 2 mM β-mercaptoethanol, supplemented with protease inhibitors) and lysed with two passes through a French Press. Cell lysates were cleared with two high-speed spins and incubated with 1 ml of packed, washed glutathione-linked agarose beads for 2 h at 4°C. Beads were then washed twice with GST buffer A (50 mM Tris, pH 7.5, 1 mM EDTA, 100 mM NaCl, and 2 mM β-mercaptoethanol) and three times with GST buffer B (50 mM Tris, pH 7.5, 200 mM NaCl, and 2 mM β-mercaptoethanol). Proteins were eluted with 20 mM glutathione in GST buffer B, pH 8.0.
To purify HIS6-tagged proteins, cells were resuspend in 25 ml of HIS-lysis buffer (50 mM Tris, pH 8.0, 300 mM NaCl, 20 mM imidazole, and 2 mM β-mercaptoethanol, supplemented with protease inhibitors). Lysis and clearing were done as described above. Lysates were incubated with 1 ml of washed Ni2+-nitrilotriacetic acid agarose beads for 2 h at 4°C. Beads were then washed five times with His-lysis buffer, and proteins were eluted with His-elution buffer (50 mM Tris, pH 8.0, 300 mM NaCl, 250 mM imidazole, and 2 mM β-mercaptoethanol). All proteins were dialyzed overnight against GST buffer B with 2 mM dithiothreitol (DTT).
For in vitro binding reactions, GST fusion proteins were first immobilized onto 50 μl of glutathione-linked beads for 1 h in GST buffer B. HIS6-tagged proteins were then added to a total volume of 500 μl in GST buffer B + 2 mM DTT and incubated for 2 h. Beads were then washed five times with GST buffer B and eluted with 20 mM glutathione. Samples were concentrated by TCA precipitation and analyzed by Coomassie staining and/or blotting against the N-terminal T7-epitope present in the HIS6 fusion proteins.
The Fab1 lipid kinase is an ~250-kDa protein made up of several domains (Figure 1A; Efe et al., 2005 ; Michell et al., 2006 ). To further understand how these domains contribute to the function of Fab1, we adopted a structural-genetic approach based on rational mutant design and assessed PtdIns(3,5)P2 levels, growth, and vacuolar morphology.
To perturb the C-terminal lipid kinase domain (KIN), we truncated the entire kinase domain from residues 1652 to 2279 (Fab1ΔKIN encoded by pRB415A::Fab1ΔKIN) or used the Fab1D2134R allele [Fab1KIN(D/R)], a previously characterized mutation that ablates lipid kinase activity (Figure 1A; Gary et al., 1998 ). The FYVE domain was disrupted by eliminating the first 676 amino acids of Fab1 (Fab1ΔHindIII; encoded by pRB415A::Fab1ΔHINDIII) or more conservatively, by mutating cysteine262 to serine262 [Fab1FYVE(C/S)]. This cysteine is part of a basic cluster found in all FYVE domains that maintains structural integrity by coordinating a Zn2+ ion (Stenmark et al., 1996 ; Burd and Emr, 1998 ).
The remaining region between residues 800 and 1500 contains the CCT and CCR domains (Efe et al., 2005 ; Michell et al., 2006 ). Residues ~800 to ~1050 are related to the CCT chaperonin family, whose members catalyze folding of actin and tubulin (Michell et al., 2006 ). Residues ~1100–1500 contain multiple conserved cysteine and histidine residues and have been referred to as the CCR domain or as the PIPKIII-unique domain (Efe et al., 2005 ; Michell et al., 2006 ). Sequence alignment using the conserved domain database places the CCT and CCR domains within the GroL superfamily. Therefore, we refer to the combined CCT and CCR domains as the GroL-related region. To help delineate the function of these two domains, we mutated a conserved motif in the CCT region (T1017ILLR to I1017LLLA; Fab1CCT(T/I)) and cysteine1243 to alanine1243 in the CCR domain (Fab1CCR(C/A); Figure 1, A and B).
We first used HPLC to measure levels of [3H]inositol-labeled PtdIns(3,5)P2 [3H]PtdIns(3,5)P2] in fab1Δ cells expressing the above-mentioned alleles of FAB1 (Figure 1C). Cells expressing wild-type Fab1 from a CEN-based vector displayed [3H]PtdIns(3,5)P2 levels comparable with cells with an intact chromosomal copy of FAB1, whereas [3H]PtdIns(3,5)P2 was virtually undetectable in cells expressing Fab1KIN(D/R) (0.11 ± 0.01 vs. 0%; Figure 1C). In contrast, expression of Fab1FYVE(C/S), Fab1CCT(T/I), or Fab1CCR(C/A) produced intermediate levels of [3H]PtdIns(3,5)P2. Fab1CCR(C/A) showed the greatest decline in lipid levels [0.024 ± 0.01 or 22% of wild-type [3H]PtdIns(3,5)P2], followed by Fab1CCT(T/I) [0.046 ± 0.01 or 42% of wild-type [3H]PtdIns(3,5)P2] and then by the FYVE point mutant [0.07 ± 0.01 or 63% of wild-type [3H]PtdIns(3,5)P2, Figure 1C]. These data suggest that the Fab1 kinase is more sensitive to perturbations in the CCT and CCR regions compared with its FYVE domain, likely because the GroL-like region may regulate Fab1 kinase activity.
We next investigated whether the fab1 mutants in question could support growth under heat stress (Figure 1D). fab1Δ cells are incapable of growing at 38°C and exhibit a growth rate that is 2.5 times slower than wild-type cells at 26°C in rich medium (Yamamoto et al., 1995 ). Consistent with this, fab1Δ cells expressing Fab1KIN(D/R) grew slowly at 26°C and were growth deficient at 38°C (Figure 1D). This was in contrast to cells expressing wild-type Fab1 and Fab1FYVE(C/S), which grew comparably well at 26 and at 38°C (Figure 1D). fab1Δ cells expressing Fab1ΔHindIII also grew similarly well at 38°C, indicating that Fab1FYVE(C/S) is not a hypomorph of the FYVE domain (Supplemental Figure S1A). However, alteration of the GroL-related sequence affected growth under heat stress, whereas expression of both Fab1CCT(T/I) and Fab1CCR(C/A) ameliorated the growth of fab1Δ cells at 26°C, these mutant alleles were defective for growth at 38°C (Figure 1D).
Last, we explored the effect of the listed FAB1 alleles on vacuolar morphology (Figure 1E). Elimination of PtdIns(3,5)P2 by deletion of FAB1 or by inactivation of its kinase activity, leads to a conspicuously enlarged, single-lobed vacuole (Figure 1E; Yamamoto et al., 1995 ; Gary et al., 1998 ). Similarly, expression of Fab1CCR(C/A) did not rescue vacuolar morphology in a fab1Δ strain (Figure 1E). We also compared mutations in cysteine1247, cysteine1289, and cysteine1292 and found they all caused single-lobed, enlarged vacuoles in cells (data not shown). This implies that the cysteines in the CCR region participate in a crucial but uncharacterized function that modulates Fab1 kinase activity. By comparison, expression of Fab1CCT(T/I) and Fab1FYVE(C/S) mutant alleles resulted in vacuoles that were predominantly single-lobed but near wild-type size (Figure 1E), likely because sufficient PtdIns(3,5)P2 was maintained in these mutants.
These data indicate that the FYVE domain is largely dispensable for the synthesis of PtdIns(3,5)P2. Mutating the FYVE domain of Fab1 only abated the 3H-labeled lipid amount by ~40%, and the cells exhibited near-normal vacuolar morphology and survived during heat stress (Table 2). In comparison, perturbing the CCT and CCR regions produced more deleterious effects on the Fab1 lipid kinase activity; consequently, cells displayed larger vacuoles and/or were temperature sensitive for growth (Table 2). This indicates that the CCT and CCR domains are important regulators of the Fab1 kinase, despite the possibility that the applied mutations did not fully incapacitate these domains.
To determine the structural-genetic requirements for Fab1 localization to the vacuole, we visualized GFP-fusion proteins of Fab1 and of its mutant alleles (Figure 2). GFP fusion of wild-type Fab1 rescued vacuolar morphology (Supplemental Figure S1B) and PtdIns(3,5)P2 levels (data not shown). However, to avoid “dilution” of the fluorescence signal due to vacuole enlargement when expressing certain Fab1 mutants, we expressed GFP-fusion constructs at endogenous levels in wild-type yeast cells. Similar observations were obtained when these GFP-fusion proteins were expressed in a fab1Δ strain (Supplemental Figure S1B). The vacuolar limiting membrane was labeled with the FM4-64 dye, and quantification of the fluorescence intensity was done as described previously (Figure 2, A and B; Efe et al., 2007 ). We used two methods to analyze the behavior of the GFP-fusion proteins: 1) fluorescence line plotting to determine fluorescence correlation between GFP and FM4-64 signal (Figure 2A) and 2) quantification of the vacuole-to-cytosol (V/C) fluorescence ratio of the GFP signal (Figure 2B).
Consistent with Fab1 enrichment on the vacuolar membrane relative to the cytosol, the fluorescence intensity peak of Fab1-GFP correlated well with FM4-64 and exhibited a V/C = 1.78 ± 0.28 (n = 25; Figure 2, A and B). The V/C ratios for Fab1CCR(C/A)-GFP and Fab1KIN(D/R)-GFP were indistinguishable from the wild-type counterpart (1.88 ± 0.34, n = 27 and 1.71 ± 0.19, n = 28, respectively), suggesting that the kinase activity and the CCR region do not impact membrane association (Figure 2B). Mutation of other cysteines in the CCR region did not hinder localization of the lipid kinase as well (data not shown). However, although Fab1CCT(T/I)-GFP bound to the vacuolar membrane, the V/C ratio of this mutant was reduced compared with Fab1-GFP (1.47 ± 0.18; n = 23 vs. 1.78 ± 0.28; p <0.05), implying that the CCT region may have a role in stabilizing Fab1 on the vacuole. We also determined that more drastic alterations to Fab1, such as truncating the CCT domain (GFP-Fab1CCTΔ) or the kinase domain (GFP-Fab1KINΔ), still allowed for at least partial association with the vacuole (Supplemental Figure S2).
In striking contrast, Fab1FYVE(C/S)-GFP and GFP-Fab1ΔHindIII were both displaced to the cytosol (Figure 2A and Supplemental Figure S2). This is corroborated by a low V/C ratio of 1.07 ± 0.08 for Fab1FYVE(C/S)-GFP (Figure 2B; n = 24). Moreover, Fab1-GFP was completely cytosolic when expressed in vps34Δ cells (Figure 2A), suggesting that the FYVE domain and PtdIns(3)P are required for Fab1 association with the vacuole. Indeed, the FYVE domain of Fab1 (FYVEFab1) was sufficient for localization to the vacuole (Figure 3A). Its distribution was dissimilar from the endosome-specific FYVE domain of EEA1 (FYVEEEA1) coexpressed in the same cells (Figure 3A). Despite this difference in distribution, they were both cytosolic in vps34Δ cells (Figure 3A).
To the best of our knowledge, no reports exist suggesting a lipid partner for the FYVE domain other than PtdIns(3)P. Nonetheless, we postulated that the vacuolar distribution of the FYVEFab1 might occur by binding to PtdIns(3,5)P2, which is also eliminated in vps34Δ cells, or by binding to a vacuolar specific protein partner such as Vac7. Notably, the FYVEFab1 remained bound to the vacuole when expressed in fab1Δ or vac7Δ cells, excluding PtdIns(3,5)P2 and Vac7 as determinants of vacuolar localization (Figure 3B). In comparison, GFP fusion of the GroL-like region did not seem to associate with the vacuole and was predominantly cytosolic (Supplemental Figure S2).
Together, these observations suggest that the FYVE domain is the primary membrane targeting mechanism for Fab1 but that the CCT domain may play a role in stabilizing this association. Thus, it is possible that Fab1FYVE(C/S) might retain weak and transient association with the vacuole through a CCT-mediated interaction, explaining its ability to complement fab1Δ cells.
Rudge et al. (2004) previously showed that Vac14 and Fig4 mislocalize to the cytosol in fab1Δ cells. The authors suggested that this was an indirect effect because they could not detect an interaction between Fab1 and Vac14/Fig4. Never-theless, it remained possible that Fab1 targets Vac14 and Fig4 to the vacuole through protein–protein interactions, or indirectly, through the synthesis of PtdIns(3,5)P2 or even through protein kinase activity. To distinguish between a role for the Fab1 protein versus its catalytic activity, we used the kinase-dead Fab1 mutant Fab1KIN(D/R) (Gary et al., 1998 ). We labeled the vacuolar membrane with FM4-64 and acquired paired images of Vac14-GFP and FM4-64 by epifluorescence microscopy and quantified the V/C fluorescence ratio for Vac14-GFP (Figure 4). Vac14 function is not significantly affected by GFP fusion (Rudge et al., 2004 ).
As expected, the fluorescence peaks of Vac14-GFP and FM4-64 coincided well in FAB1 cells, but it failed to do so in fab1Δ cells (Figure 4A). The V/C ratio for Vac14-GFP in FAB1 cells (V/CFAB1) was 2.16 ± 0.55 (n = 20) versus 0.96 ± 0.2 (n = 20) in fab1Δ cells (V/Cfab1Δ; Figure 4B). In comparison, cells expressing fab1KIN(D/R) not only exhibited enlarged vacuoles but also the fluorescence profiles for Vac14-GFP and FM4-64 correlated well. Furthermore, the V/C ratio for Vac14-GFP in these cells was statistically indistinguishable from V/CFAB1 (1.88 ± 0.43, n = 21; p > 0.05; Figure 4). This indicates that Fab1 acts structurally to localize Vac14 to the vacuole and independently of its kinase activity or PtdIns(3,5)P2. Corroborating this conclusion, Vac14-GFP was displaced from the vacuole in cells expressing the FYVE-point mutant of Fab1, which itself is mislocalized to the cytosol (Figure 2, A and B). The V/C ratio for Vac14-GFP in fab1Δ cells and those expressing Fab1FYVE(C/S) was statistically the same (1.07 ± 0.18; n = 20 vs. 0.96 ± 0.2; p > 0.05).
We also explored the behavior of Vac14-GFP in cells expressing the Fab1CCT(T/I) and Fab1CCR(C/A) point mutants. The V/C ratio of Vac14-GFP in cells expressing Fab1CCT(T/I) or Fab1CCR(C/A) was 1.21 ± 0.13 (n = 20) and 1.19 ± 0.2 (n = 20), respectively, which was significantly less than wild-type V/C. Thus, perturbing the CCT and CCR domains of Fab1 led to a measurable dissociation of Vac14-GFP from the vacuole (Figure 4). However, it is important to point out that Vac14-GFP maintained partial association with the vacuole in the GroL-like domain mutants because their V/C ratios were statistically higher than V/Cfab1Δ (Figure 4B). This is consistent with the possible hypomorphic nature of these mutants and/or the presence of an additional site on Fab1 may contribute to localization of Vac14 (Figure 1). The localization of Fig4-GFP was similar to that of Vac14-GFP; expression of Fab1FYVE(C/S), Fab1CCT(T/I) and Fab1CCR(C/A) in fab1Δ cells did not rescue Fig4-GFP localization to the vacuole, whereas expression of Fab1 or Fab1KIN(D/R) did (Supplemental Figure S3). Together, these data suggest that Fab1 has a direct role in targeting Vac14 and Fig4 to the vacuole, perhaps via interactions with the CCT and CCR domains.
Our observations indicated that Fab1, Vac14, and Fig4 might assemble into a single protein complex in yeast cells. It was shown previously that Vac14 and Fig4 coIP and are mutually dependent for association with the vacuolar membrane (Rudge et al., 2004 ; Duex et al., 2006b ). Indeed, we observed coIP of Vac14-HA and Fig4 (Figure 5A). Nevertheless, previous attempts to demonstrate an interaction between Fab1 and Vac14 (or Fig4) were unsuccessful (Bonangelino et al., 2002 ; Rudge et al., 2004 ). However, we now demonstrate that Myc-tagged Fab1 was recovered with Vac14-HA (Figure 5A). Most notably, IP of Fab1-Myc retrieved not only Vac14-HA but also the Fig4 phosphatase as well (Figure 5A). Although it is possible that all three proteins interact with each other in pairwise combinations, we interpret these results as support for a protein complex that contains all three proteins together.
We then tested whether this putative Fab1 complex could be detected in both membrane and cytosolic fractions. We first fractionated whole cell lysates into cytosolic and membrane fractions and observed a relative distribution for Fab1-Myc, Vac14-HA, and Fig4 of ~70% versus ~30% between the membrane and cytosolic fractions, respectively (Figure 5B). The absence of the Vph1 vacuole membrane marker from the cytosolic fraction indicates the absence of contaminating membrane (Figure 5B). Additionally, this distribution was comparable with the relative distribution of GFP-fusion proteins by microscopy. We observed that Fab1-Myc and Fig4 were both recovered with Vac14-HA from the membrane fraction (Figure 5C). In contrast, IP of Vac14-HA from the cytosolic fraction recovered Fig4, but not Fab1-Myc (Figure 5C). These data indicate that Vac14 and Fig4 can interact in the cytosol but that association of Fab1 with Vac14 and Fig4 is stimulated and/or stabilized on the vacuolar membrane.
We also tested whether Fab1, Fig4 and Vac14 existed as monomers or whether they possibly displayed higher order organization. To test this, we simultaneously expressed in the same yeast cells two alleles of FAB1, FIG4, and VAC14 encoding two distinct epitopes (Figure 5D). We did not observe coIP between Fig4-Myc and Fig4-HA not between Fab1-Myc and Fab1-HA (Figure 5D). In contrast, Vac14-FLAG was retrieved with HA-Vac14 by IP (Figure 5D), which is consistent with yeast two-hybrid assays showing Vac14–Vac14 interaction (not illustrated; Dove et al., 2002 ). Overall, these results indicate that Fab1, Vac14, and Fig4 can physically associate to form a membrane-bound protein assembly. Likely, there are multiple copies of Vac14 per complex, whereas the Fab1 kinase and the Fig4 phosphatase may be present at one subunit per complex.
To better understand how the Fab1 complex forms, we performed coIP experiments in different mutant backgrounds. We first determined that Vac14 and Fig4 bound to each other in the absence of the Fab1 kinase (Figure 6A). This is consistent with a putative Vac14–Fig4 subcomplex observed in the cytosolic fraction (Figure 5C). We also established that the Fab1 kinase and the Fig4 phosphatase were not retrieved together from vac14Δ cell lysates, which supports a role for Vac14 as a linker between the kinase and the phosphatase (Figure 6A). Remarkably though, Fab1 and Vac14 interacted poorly in the absence of Fig4 (Figure 6A), which suggests that Fig4 may play a regulatory/structural role in the assembly of the Fab1 complex.
Deletion of VAC7 leads to undetectable levels of PtdIns(3,5)P2 (Bonangelino et al., 1997 ; Gary et al., 2002 ). Because the mechanism for this regulation remains unknown, we postulated that Vac7 might be important for the association of Fab1 with Vac14-Fig4. However, deleting VAC7 did not prevent Fab1 from interacting with Fig4, seemingly excluding this hypothesis (Figure 6A). Conversely, hyperosmotic shock of yeast cells causes a dramatic increase in PtdIns(3,5)P2 levels after 5–10 min, followed by an abatement after 20–30 min (Dove et al., 1997 ; Duex et al., 2006b ). Nonetheless, changes in PtdIns(3,5)P2 did not seem to correlate with changes in the amount of Fab1 complex; we did not observe a significant change in the levels of Fig4 coIPed with Fab1-Myc upon exposure to salt-shock for 0, 8, or 25 min (Supplemental Figure S4). Together, these results indicate that Vac14 and Fig4 are mutually dependent for interaction with Fab1 and may explain the concomitant role of Vac14 and Fig4 in the synthesis and turnover of PtdIns(3,5)P2 (Duex et al., 2006a ; Efe et al., 2007 ).
Atg18 was identified as an effector of PtdIns(3,5)P2 that regulates vacuolar morphology and membrane recycling. Atg18 also regulates PtdIns(3,5)P2 levels—deletion of Atg18 leads to an approximately ninefold increase in PtdIns(3,5)P2 levels (Dove et al., 2004 ). We showed previously that this increase is dependent on FIG4 (Efe et al., 2007 ). Additionally, Fig4 is required to increase PtdIns(3,5)P2 levels during hypertonic shock (Duex et al., 2006a ,b ). This unexpected role for Fig4 in the synthesis of PtdIns(3,5)P2 is likely explained by our observations—Fig4 is necessary for formation and/or stability of a productive Fab1-Vac14 interaction.
Conceivably, Fig4 may stimulate PtdIns(3,5)P2 synthesis by mediating direct protein–protein interactions or through its phosphatase activity, possibly by acting as a protein phosphatase (Duex et al., 2006b ). To differentiate between these possibilities, we used a catalytic-inactive FIG4 allele, fig4-1, which was initially isolated as a suppressor of the vac7Δ phenotype (Gary et al., 2002 ; Rudge et al., 2004 ).
The basal level of [3H]PtdIns(3,5)P2 in atg18Δ fig4Δ cells was 0.13 ± 0.03% of total [3H]PtdIns, which is similar to wild-type [3H]PtdIns(3,5)P2 levels (0.11%; Figure 6B). When Atg18 was expressed episomally—recreating the fig4Δ condition—[3H]PtdIns(3,5)P2 levels were slightly decreased (0.07 ± 0.01%). By contrast, expression of wild-type FIG4 in atg18Δ fig4Δ cells recapitulated [3H]PtdIns(3,5)P2 levels in atg18Δ cells (0.81 ± 0.05 vs. 0.9%; Figure 6B). Similarly, expression of fig4-1 in atg18Δfig4Δ cells produced even higher levels of [3H]PtdIns(3,5)P2 (1.37 ± 0.11% of total PtdIns; Figure 6B), likely because of abated turnover of the lipid. These results support the notion that Fig4 provides structural stability to the Fab1–Vac14 interaction independently of its phosphatase activity. To confirm this notion, we tested whether the Fab1 complex was reconstructed in atg18Δ fig4Δ cells expressing vector, FIG4 or fig4-1. As shown in Figure 6A, the absence of Fig4 prevented coIP between Fab1-Myc and Vac14-HA (Figure 6C). In contrast, expression of FIG4 or of the phosphatase-dead fig4-1 allele permitted corecovery of Fab1-Myc and Vac14-HA (Figure 6C). Therefore, these observations support a structural role for Fig4 in stabilizing a productive Fab1 complex.
Expression of Fab1CCT(T/I) or of Fab1CCR(C/A) crippled vacuole association of Vac14 and Fig4 (Figure 4 and Supplemental Figure S3). This suggests that the GroL-like region of Fab1 might be a binding site for Vac14 and/or Fig4. To test this hypothesis, we expressed a HA-tagged GroL-like fragment in fab1Δ cells and assessed the recovery of FLAG-tagged Vac14 and Fig4 by coIP. As shown, both Vac14-FLAG and Fig4 were recovered with the HA-GroL domain but not from control cells expressing the empty vector (Figure 7A). Importantly, mutating the TILLR motif in the CCT domain to ILLLA blocked interaction with Vac14 and Fig4, which is congruous with the release of Vac14 and Fig4 from the vacuole membrane in cells expressing Fab1CCT(T/I) (Figure 3 and Supplemental Figure S3). Additionally, deletion of FIG4 in fab1Δ VAC14-FLAG cells prevented coIP of Vac14-FLAG with the HA-tagged GroL-like fragment (Figure 7B), which recapitulated the requirement for Fig4 to stabilize the interaction between Vac14 and full-length Fab1 (Figure 6).
To further corroborate the interaction between Vac14 and the GroL-like region in vivo, we used fluorescence microscopy. To circumvent the cytosolic distribution of the GroL-like region of Fab1 (Supplemental Figure S2B), we engineered a chimeric protein composed of the GroL-like domain and of the first 134 amino acids of ALP (ALP134). ALP134 contains the transmembrane domain and the AP3-pathway sorting motifs that target ALP to the vacuole (Klionsky and Emr, 1989 ; Bryant et al., 1998 ). Indeed, GFP-GroL-ALP was clearly targeted to the limiting membrane of the vacuole (Supplemental Figure S2B).
Expression of HA-ALP in fab1Δ cells failed to mobilize Vac14-GFP to the vacuole (Figure 7, C and D). In sharp contrast, HA-GroL-ALP efficiently recruited Vac14-GFP to the vacuole. Introducing the ILLLA mutation into HA-GroL-ALP (which generated HA-GroLT/I-ALP) resulted in loss of Vac14-GFP localization to the vacuole (Figure 7C). Comparing the V/C ratio for Vac14-GFP in cells expressing HA-ALP, HA-GroL-ALP, or HA-HA-GroLT/I-ALP substantiated our conclusions. The Vac14-GFP V/C ratio in cells expressing HA-ALP and the mutated HA-GroL-ALP were 1.04 ± 0.11 (n = 29) and 0.98 ± 0.12 (n = 35), respectively, and they were statistically indistinguishable (p > 0.05; Figure 7D). By comparison, the V/C ratio in HA-GroL-ALP–expressing cells was 1.72 ± 0.32 (n = 30) and was considerably greater than either HA-ALP or the mutated form (p < 0.05 for both comparisons; Figure 7D). In conclusion, we propose that the GroL-like region of Fab1 is an important docking site for Vac14 and Fig4, albeit it remains possible that Vac14/Fig4 may also interact with other regions on Fab1. The exact molecular events that translate into kinase activation upon Vac14-Fig4 binding will require further analysis.
To corroborate the above-mentioned observations and test for direct interaction, we used bacterially expressed recombinant proteins. GST-Vac14 and GST-GroL-like domain were purified from bacterial extracts by using glutathione-coupled agarose beads. We also expressed and purified Vac14 and Fig4 proteins tagged with an N-terminal T7 epitope and a C-terminal HIS6 epitope (Figure 7E). We first tested and confirmed Vac14 interaction with itself in vitro; T7-Vac14-HIS6 was recovered with GST-Vac14 but not with GST coupled to glutathione-agarose beads (Figure 7F). Similarly, T7-Fig4-HIS6 bound to GST-Vac14 but not to GST alone (Figure 7F), showing that Vac14 and Fig4 interact directly. This is consistent with yeast two-hybrid analysis showing Vac14–Vac14 and Vac14–Fig4 interactions (not illustrated; Dove et al., 2002 ).
We also tested whether recombinant GroL-like region of Fab1 was sufficient to interact directly with T7-Vac14-HIS6 and/or T7-Fig4-HIS6. We reproducibly observed enhanced recovery of T7-Vac14-HIS6 with a GST-fusion of the GroL-like region compared with GST alone (Figure 7F). We could also detect an interaction between GST-GroL and T7-FIG4-HIS6, although this interaction was weak and may reflect a nonspecific affinity by the GroL-related fragment. These results suggest that recombinant GroL-like region of Fab1 can directly bind to Vac14, and perhaps Fig4, in vitro.
In conclusion, these observations support a model where Vac14 and Fig4 bind to the GroL-like region of Fab1 to form a vacuole-bound complex in the cell to regulate PtdIns(3,5)P2 levels. Our data suggest that Vac14 and Fig4 are mutually dependent for binding to Fab1 and for association with the vacuole. This can explain how Vac14 and Fig4 are concomitantly involved in synthesis and turnover of PtdIns(3,5)P2.
Disruption of the FYVE domain of Fab1 weakly affected PtdIns(3,5)P2 levels and cellular function. This relatively weak phenotype is not likely to result from incomplete inactivation of the FYVE domain because cysteine262 is at a vital position for folding of FYVE domains. Moreover, there was no apparent detriment to fab1Δ cells expressing a Fab1 mutant with a full deletion of its FYVE domain (Fab1ΔHindIII) relative to cells expressing wild-type Fab1 or Fab1FYVE(C/S) when grown at high temperature. There is precedent for these results as well; the FAB1 allele that was originally cloned from a fab1Δ complementation screen lacked its FYVE domain (Yamamoto et al., 1995 ). Consistent with this, several Fab1 orthologues lack a FYVE domain (Michell et al., 2006 ).
In contrast, Fab1 was delocalized from the vacuole upon perturbation of its FYVE domain or ablation of PtdIns(3)P synthesis. Therefore, it is challenging to reconcile loss of membrane association while maintaining function (Figure 1). At least three plausible events could result in the appearance of a weak phenotype instead of extensive lipid loss that should accompany Fab1 membrane dissociation: 1) Fab1FYVE(C/S), possibly through its CCT domain, may transiently interact with the vacuolar membrane to synthesize a small amount of PtdIns(3,5)P2, but 2) the extent of lipid loss is mitigated by the concomitant dissociation of the phosphatase from the vacuole and 3) such a transient association may favor the kinase over phosphatase activity, possibly because Fab1 is more proximal to the membrane.
Notably, the FYVE domain of Fab1 was sufficient to bind to the vacuole, whereas the FYVE domain of EEA1 distributed strictly to endosomes. Nevertheless, both FYVE domains required PtdIns(3)P for membrane association. Because vacuolar PtdIns(3)P is not thought to be abundant and is converted to PtdIns(3,5)P2 by Fab1, this suggests an additional targeting mechanism for the FYVE domain of Fab1. Recently, we showed that Atg18 not only depends on PtdIns(3,5)P2 for association with the vacuole but also that it requires an interaction with Vac7 (Efe et al., 2007 ). Nevertheless, the FYVE domain of Fab1 maintained its vacuolar distribution in vac7Δ cells, indicating that Vac7 does not play a role in targeting the FYVE domain of Fab1 (Figure 3B). Further analysis is required to identify additional targeting mechanisms of Fab1 to the vacuole.
The middle region of Fab1 is highly conserved across all of its orthologues. It consists of two distinct areas, the CCT and CCR domains. We refer to this combined region as GroL-related because recent comparisons with the conserved domain database retrieved the CCT and CCR regions over the GroL chaperonin. The GroL-related region is not critically important for vacuolar targeting, albeit it may stabilize Fab1 on the vacuole as suggested by the mild delocalization of Fab1CCT(T/I) to the cytosol. Nevertheless, mutations in the CCT and CCR domains impeded Fab1 activity and function. Although it is likely that Fab1CCR(C/A), and especially Fab1CCT(T/I), are hypomorphs rather than representing full impairment of the respective domains, our data still support an important role for the GroL-related region. This is also consistent with inactivation of PIKfyve upon deletion of the CCT domain (Sbrissa et al., 1999 ; Ikonomov et al., 2001 ).
The exact function of the GroL-related region remains poorly understood. Our data suggest that at least one function of the GroL-related region is to bind Vac14 and Fig4. Conceivably, the GroL-like region of Fab1 translates Vac14-Fig4 binding into Fab1 activity, which is consistent with reduced lipid levels in vac14Δ and in the CCT/CCR mutants (see below for Fig4 details). Possibly, docking of Vac14-Fig4 onto the GroL-like region of Fab1 might induce a conformational change that exposes the kinase domain to its substrate. Hyperactive mutants such as fab1-5 and a number of vac14Δ-suppressing fab1 mutants may bypass the need for Vac14 by constitutively stabilizing Fab1 in the open/active conformation (Gary et al., 2002 ; Duex et al., 2006b ).
Vac7, Vac14, Atg18, the Fab1 lipid kinase and the Fig4 lipid phosphatase are the known components of an elaborate interaction network that modulates PtdIns(3,5)P2 synthesis and turnover. Atg18, Fig4, and Vac14 buttress the complexity of this pathway by performing multiple tasks. In Atg18, it is a PtdIns(3,5)P2 effector responsible for vacuole-to-endosome traffic and vacuole morphology. Remarkably, Atg18 is also a negative regulator of Fab1 whose elimination leads to a dramatic increase in PtdIns(3,5)P2 levels (Dove et al., 2004 ). Likewise, Fig4 and Vac14 play a dual role: regulating the turnover and synthesis of PtdIns(3,5)P2 (Gary et al., 2002 ; Rudge et al., 2004 ; Duex et al., 2006a ,b ; Efe et al., 2007 ). The mechanism enabling this dual functionality was unknown and was particularly challenging to explain because previous attempts to show an interaction between Fab1 and Vac14/Fig4 were unsuccessful (Bonangelino et al., 2002 ; Rudge et al., 2004 ). Nonetheless, we identified a vacuolar signaling complex consisting of Vac14 and the antagonizing Fab1 lipid kinase and Fig4 lipid phosphatase, where Vac14 and Fig4 are mutually dependent for interaction with Fab1 in the cell (Figure 8). The discrepancy between our work and that of others probably lies within technical difference; for example, we used 13-tandem copies of Myc to tag Fab1 for increased sensitivity and efficiency, a particularly important point because Fab1 is a very large, low-abundance protein. Regardless, our observations are consistent and provide a mechanistic insight into the observed complex interplay between Fab1, Fig4 and Vac14 as discussed below.
First, our observations indicate that Vac14 and Fig4 are tethered to the vacuolar membrane by binding to Fab1, not by an indirect mechanism as suggested previously (Rudge et al., 2004 ). Additionally, the mutual dependency of Vac14 and Fig4 for binding to Fab1 explains their mutual dependency for vacuole association. Second, our data support a mechanism in which Vac14 binds to and activates Fab1 through its GroL-like domain, not through an intermediate as proposed previously (Rudge et al., 2004 ). The exact molecular events leading to Vac14-mediated activation of Fab1 will require additional studies. Third, although Vac14 is clearly required for Fab1 activity, vac14Δ mutants are also impaired for down-regulation of PtdIns(3,5)P2 (Duex et al., 2006b ). Likely, this is because in the absence of Vac14, Fig4 cannot bind Fab1, and therefore, it has reduced access to PtdIns(3,5)P2. Fourth, the mechanism by which Fig4 sustains synthesis of PtdIns(3,5)P2 in response to salt shock and in atg18Δ cells was not understood, although it was postulated to also function as a protein phosphatase that modulates Vac14 or Fab1 (Duex et al., 2006a ,b ; Efe et al., 2007 ). Instead, our data imply that Fig4 sustains PtdIns(3,5)P2 synthesis by stabilizing the Fab1–Vac14 interaction independent of its catalytic activity—this is especially true in conditions that hyperactivate Fab1 such as atg18Δ and hyperosmotic shock.
Fig4 seems to contribute two opposing inputs to modulate PtdIns(3,5)P2—the phosphatase activity comprises the negative input, whereas maintenance of the Fab1–Vac14 interaction constitutes the positive signal. This illuminates why elimination of the entire Fig4 protein has little effect on basal PtdIns(3,5)P2, whereas elimination of its phosphatase activity alone (fig4-1 allele) leads to an approximately threefold increase in this lipid (Gary et al., 2002 ). In fig4Δ cells, there is a concomitant loss of the positive and the negative input, resulting (perhaps serendipitously) in unaffected PtdIns(3,5)P2. In contrast, fig4-1 stabilizes the Fab1–Vac14 interaction (positive signal) in the absence of PtdIns(3,5)P2 turnover, resulting in a significant increase in PtdIns(3,5)P2 levels. In fact, the level of PtdIns(3,5)P2 in the absence of Vac14 and Fig4 is only ~0.1 times that of wild-type cells but in the presence of Vac14 and Fig4-1 it is 3 times that of wild-type (thus, ~30-fold difference; Bonangelino et al., 2002 ; Dove et al., 2002 ; Gary et al., 2002 ; Rudge et al., 2004 ). Because elimination of phosphatase activity cannot solely explain this increase, it suggests that binding of Vac14-Fig4 to Fab1 stimulates its intrinsic kinase activity.
The assembly of the Fab1 complex is likely to be a controlled process. However, Vac7, Atg18, and hyperosmotic shock are not likely to regulate PtdIns(3,5)P2 levels by modulating the assembly/disassembly of the Fab1 complex. The amount of the Fab1 complex did not seem to increase upon deletion of ATG18 nor upon hyperosmotic shock. Similarly, Fab1 and Fig4 still interacted in vac7Δ cells, suggesting that Vac7 is not required for Fab1 complex assembly. In fact, this is consistent with the unperturbed vacuolar distribution of Fab1, Vac14, and Fig4 in vac7Δ cells (Rudge et al., 2004 ).
It might be useful to compare the Fab1 and PIKfyve complexes (Sbrissa et al., 2007 ). Notably, the union of the kinase and phosphatase activities into one complex is conserved in yeast and mammalian cells. However, in previous studies of the PIKfyve complex, the authors did not elaborate on the role of hVac14 and hSac3 (Fig4 orthologue) in the formation and stability of the PIKfyve complex nor did they explore whether hSac3 positively regulates PtdIns(3,5)P2 synthesis, as is the case for Fig4. Strikingly, knockout mice for mFig4 (mouse orthologue of hSac3) exhibit reduced PtdIns(3,5)P2 levels (Chow et al., 2007 ), indicating that both mammalian and yeast Fig4 phosphatases may indeed have a role in PtdIns(3,5)P2 synthesis.
It will be interesting to unravel the details of the Fab1 complex and study how Fab1 and Fig4 coordinate their kinase and phosphatase activities within the context of a common protein complex. The conserved nature of this complex indicates a fundamental requirement to tightly regulate PtdIns(3,5)P2. Possibly, this may permit rapid and local adjustments in PtdIns(3,5)P2 levels in specific subdomains of the vacuole membrane, which may assist in the generation of transport vesicles responsible for retrograde transport from the vacuole (Bryant et al., 1998 ; Dove et al., 2002 ). It is tempting to suggest that PtdIns kinase-phosphatase coupling may be a common mechanism to modulate PtdInsPs. Recently, the mammalian VPS34 PtdIns 3-kinase and the MTM1 PtdIns(3)P 3-phosphatase were shown to interact (Cao et al., 2007 ).
We thank Drs. Bong-Kwan Han, Jason MacGurn, Dan Baird, and Chris Stefan for helpful suggestions and discussion. We acknowledge the initial work performed by Drs. Simon Rudge and John Gary. R.J.B. is supported by the Jean-François St.-Denis Cancer Fellowship from the Canadian Institutes for Health Research, and D. T. is supported by Human Frontiers Fellowship HSFP LT00634/2006-c. S.D.E. was formerly supported by the Howard Hughes Medical Institute and is presently supported by a Cornell University Research Award.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E08-04-0405) on July 23, 2008.