SAC1-like Domains of Yeast SAC1,INP52, and INP53 and of Human Synaptojanin Encode Polyphosphoinositide Phosphatases*

The SAC1 gene product has been implicated in the regulation of actin cytoskeleton, secretion from the Golgi, and microsomal ATP transport; yet its function is unknown. Within SAC1 is an evolutionarily conserved 300-amino acid region, designated a SAC1-like domain, that is also present at the amino termini of the inositol polyphosphate 5-phosphatases, mammalian synaptojanin, and certain yeast INP5 gene products. Here we report that SAC1-like domains have intrinsic enzymatic activity that defines a new class of polyphosphoinositide phosphatase (PPIPase). Purified recombinantSAC1-like domains convert yeast lipids phosphatidylinositol (PI) 3-phosphate, PI 4-phosphate, and PI 3,5-bisphosphate to PI, whereas PI 4,5-bisphosphate is not a substrate. Yeast lacking Sac1p exhibit 10-, 2.5-, and 2-fold increases in the cellular levels of PI 4-phosphate, PI 3,5-bisphosphate, and PI 3-phosphate, respectively. The 5-phosphatase domains of synaptojanin, Inp52p, and Inp53p are also catalytic, thus representing the first examples of an inositol signaling protein with two distinct lipid phosphatase active sites within a single polypeptide chain. Together, our data provide a long sought mechanism as to how defects in Sac1p overcome certain actin mutants and bypass the requirement for yeast phosphatidylinositol/phosphatidylcholine transfer protein, Sec14p. We demonstrate that PPIPase activity is a key regulator of membrane trafficking and actin cytoskeleton organization and suggest signaling roles for phosphoinositides other than PI 4,5-bisphosphate in these processes. Additionally, the tethering of PPIPase and 5-phosphatase activities indicate a novel mechanism by which concerted phosphoinositide hydrolysis participates in membrane trafficking.

and as a signaling molecule itself by regulating protein activities and through interactions with protein modules (3)(4)(5). Additionally, roles for PI(3)P, PI(3,5)P 2 and PI(3,4,5)P 3 in membrane movement have been defined (1, 6 -10). Homeostasis of phosphoinositides occurs both spatially and temporally via a plethora of lipase, kinase, and phosphatase activities, thereby providing several unique points of regulation (for reviews see Refs. [11][12][13][14][15][16][17]. A role for inositol lipid phosphatases in membrane trafficking has come from the characterization and cloning of synaptojanin, a mammalian neuronal inositol polyphosphate 5-phosphatase (5-ptase) involved in synaptic vesicle recycling (18). Additionally, studies of three yeast INP5 gene products (also known as SJLs) demonstrate that although they are collectively essential, the individual proteins also have nonredundant roles in regulating membrane trafficking, cell wall synthesis, osmo-sensitivity, and actin cytoskeleton structure (19 -22). Synaptojanin and the Inp5ps are members of a large 5-ptase gene family, conserved from yeast to humans, that specifically remove the D-5 position phosphate from certain phosphoinositides and/or inositol polyphosphates (15)(16)(17)(18)(19)(20)(21). Of interest, the domain structures of Inp52p and Inp53p, and to a lesser extent Inp51p, are analogous to synaptojanin, having a 300-amino acid 5-ptase domain that is flanked by amino-and carboxyl-terminal SAC1-like and proline-rich domains. Although the function of SAC1-like domains in synaptojanin, the Inp5ps, and Sac1p is unknown, genetic studies have shown that mutations in Sac1p overcome certain actin defects and bypass the requirement for Sec14p, a phosphatidylinositol/ phosphatidylcholine transfer protein required for generation of secretory vesicles from Golgi (23)(24)(25)(26). In addition, Sac1p function is required for proper transport of ATP into the endoplasmic reticulum (27).
Here we report studies aimed at elucidating the unique and overlapping roles of the INP5 gene products. We demonstrate that Inp52p, Inp53p, and synaptojanin have two lipid phosphatase activities, one as a novel polyphosphoinositide phosphatase (PPIPase) and the other as a PI(4,5)P 2 5-ptase. Remarkably, we show that these activities are encoded by two autonomous active sites provided by the SAC1-like and 5-ptase domains, respectively. PPIPase activity is biologically relevant because loss of this activity in cells results in significant changes in the levels of certain phosphoinositides. The unique substrate selectivity and regulatory properties of each activity suggests novel hypotheses regarding how individual and/or combinatorial control of PI(4)P, PI(3)P, PI(3,5)P 2 , and PI(4,5)P 2 are intimately involved in the regulation of vesicle trafficking and actin cytoskeleton organization.

EXPERIMENTAL PROCEDURES
Strains, Media, and Genetic Methods-Yeast strains used in this study are listed in Table I. Standard procedures for growth and yeast genetic manipulations were used as described (28). SacI mutant strain was kindly provided by Dr. Vytas A. Bankaitis (University of Alabama at Birmingham). In all cases, DNA produced by PCR amplification was sequenced to confirm that unwanted mutations were not present.
Production of Inp5 Proteins-Overexpression of Inp51p, Inp52p, and Inp53p in yeast was achieved on a triple knockout background. Construction of triple mutant strains required deletion of INP54 in previously reported double mutants (20). The INP54 coding region was deleted and replaced in a W303 diploid strain with the LEU2 gene by a PCR-based method (20). Haploid inp54 and appropriate haploid double mutants were mated, sporulated, and dissected to yield inp51 inp52 inp54, inp51 inp53 inp54, and inp52 inp53 inp54 triple mutants.
Overexpression of Inp5ps was accomplished in triple mutant yeast harboring plasmids containing the coding sequence of each Inp5 under the control of a galactose-inducible promoter. Plasmids were constructed using a PCR-based strategy in which the coding sequences were inserted into pBJ246 (ATCC 777452) to yield pGAL51, pGAL52, and pGAL53. Cells harboring plasmids were grown in CM/ura Ϫ containing 2% raffinose to an A 600 of 0.2, induced by the addition of galactose (2% final) for 6 h and harvested. Soluble extracts were prepared from 1 ϫ 10 8 cells by bead beating in 200 l of 250 mM Tris-HCl, pH 7.5, supplemented with 25 mM EDTA, 5 mM sodium pyrophosphate, 750 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, 1 mM NaVO 4 .
Mixed vesicle substrates were prepared by drying total cell or purified lipids under a N 2 stream and briefly sonicating in buffer containing RB buffer (50 mM HEPES, pH 7.4, 50 mM KCl, 3 mM EGTA). Detergent micelle substrate was prepared using RBT buffer (RB buffer supplemented with 0.2% Triton X-100). Enzymatic assays were performed in either RB or RBT buffer supplemented with 0.2 mg/ml bovine serum albumin, 2 mM MgCl 2 , and 10 mM dithiothreitol for 1 h at 37°C. Where noted, MgCl 2 was omitted to optimize PPIPase activity. Reactions were stopped, extracted with chloroform/methanol, deacylated with methylamine (21), and analyzed by HPLC (see below).
Expression of Recombinant Protein in Bacteria-The coding region of INP53 was inserted in-frame into pGEX-3X vector (Amersham Pharmacia Biotech) to yield pGEX53. The 5-ptase domain of INP53 (residues 490 -1107) was amplified by PCR and inserted in-frame into pGEX-3X to yield pGEX53-5pt. The SAC1-like domain of INP53 (residues 1-548) was constructed by digestion of pGEX53 with SnaBI and SmaI and religated to produce pGEX53-SAC. The SAC1-like domain of synaptojanin (residues 1-523) was amplified by PCR from a human brain pCMV sport library DNA (Life Technologies, Inc.) and inserted into pGEX-3X yielding pGEXSyn-SAC.
Expression of GST fusion proteins was performed in Escherichia coli strain BL21(DE3) (Novagen) as described previously (21) with the exception that 1 liter of LB medium was used for a 24-h induction at 23°C, and cells were disrupted by French press. Fusion proteins were purified by glutathione-Sepharose 4B chromatography according to manufacturer's directions (Amersham Pharmacia Biotech). Yeast Mutant Radiolabeling Studies and HPLC Analysis-Yeast strains were labeled over 10 cell doublings to steady-state in appropriate minimal medium containing [ 3 H]myo-inositol. Lipids were deacylated with methylamine reagent essentially as described (21) except that 1 mg of carrier lipid (brain extract, Sigma) was added to improve yield. Glycerophosphoinositols (groPI) were resolved by Partisphere SAX HPLC (Whatman 4.6 ϫ 125 mm). Samples were equilibrated to 10 mM ammonium phosphate, pH 3.5 (AP), applied to column, and eluted using the following gradient: 10 mM AP to 340 mM AP over 15 min, 340 mM AP to 1.02 M AP over 7.5 min, and isocratic 1.02 M AP for additional 5 min. Radioactivity was measured using the BetaRAM TM in-line detector (INUS, Tampa, FL).

RESULTS AND DISCUSSION
Sequencing of the Saccharomyces cerevisiae genome reveals four gene products, designated INP5s, that have the canonical catalytic motifs present in all inositol polyphosphate 5-phosphatases. Previously, we reported that recombinant Inp51p functions as a lipid-selective 5-ptase that converts PI(4,5)P 2 to PI(4)P but does not hydrolyze Ins(1,4,5)P 3 (21). We have recently found that Inp54p is also a PI(4,5)P 2 lipid selective 5-ptase. 2 These data coupled with published reports that yeast extracts do not have detectable Ins(1,4,5)P 3 5-ptase activity (21,30) suggest that the Inp5ps utilize lipid substrates. Candidate D-5 phosphorylated lipid substrates present in yeast include PI(4,5)P 2 and the newly discovered osmotic stress-induced PI(3,5)P 2 (31,32). To help explain the individual and overlapping cellular functions of the Inp5ps, we characterized their substrate selectivities.
To demonstrate that the PPIPase activity of Inp52p and Inp53p was intrinsic and not associated, recombinant Inp53p was produced in bacteria as a GST fusion protein (GST-53) and analyzed ( Fig. 2A). Treatment of mixed vesicle substrate with either control (not shown) or GST showed no detectable phosphatase activity. Purified recombinant GST-53 harbored activity that resulted in Ͼ60% decreases in all polyphosphoinositides. Treatment of detergent micelle substrate with GST-53 resulted in a 100% decrease in PI(4,5)P 2 and a quantitative increase in PI(4)P (not shown). These data are similar to those obtained with native Inp53p and demonstrate that both the PPIPase and 5-ptase activities are intrinsic properties of Inp53p and presumably Inp52p.
The PPIPase Activity Is Attributable to SAC1-like Domains-The unique biochemical properties and substrate selectivity of PPIPase and 5-phosphatase activities suggest that they are separable. Therefore, we separated partially proteolyzed recombinant Inp53p by size exclusion chromatography and assayed individual fractions for PPIPase and 5-ptase activities. This analysis confirmed that these activities are separable (not shown). Furthermore, the PPIPase and 5-ptase activities were detected in fractions eluting as low as 30 and 25 kDa, respectively (not shown). This was intriguing because multiple sequence alignments predict that the SAC1-like and 5-ptase domains are approximately this size. With this in mind, the 5-ptase and SAC1-like domains of INP53 were independently expressed as GST fusion proteins and analyzed ( Fig. 2A). Treatment of mixed vesicle substrate with the 5-ptase domain resulted in a 40% decrease in PI(4,5)P 2 and a quantitative increase in PI(4)P (GST53-5pt), demonstrating that it encodes a 5-ptase but not PPIPase activity. Remarkably, treatment of mixed vesicles with the SAC1-like domain resulted in Ͼ70% reduction in the levels of PI(3,5)P 2 , PI(4)P, and PI(3)P, whereas PI(4,5)P 2 levels did not change (GST53-Sac). Using detergent micellar substrate we found that the PPIPase activity of GST53-Sac was ablated and that the 5-ptase activity of GST53-5pt was stimulated (not shown), recapitulating data observed for the full-length GST53 protein. The PPIPase activity was found to be Mg 2ϩ -independent and inhibited by 2 mM Cu 2ϩ and Zn 2ϩ (80%) and by Mg 2ϩ (20%) (not shown). Addition of DTT markedly increased activity. Together these data demonstrate that both the SAC1-like and 5-ptase domains of Inp53p autonomously fold and are sufficient to encode the intrinsic PPIPase and 5-ptase activities found in Inp53p.
The product(s) formed by the PPIPase activity were characterized using purified PIP 2 and PIP mixed vesicle substrates (Fig. 2, B and C). Treatment of either lipid substrate with GST protein alone did not produce detectable hydrolytic activity. Full-length GST53 quantitatively hydrolyzed PIP 2 to form PI, and no PIP intermediates were detected. Treatment of PIP 2 with GST53-5pt protein converted PI(4,5)P 2 but not PI(3,5)P 2 to PI(4)P. In contrast, GST53-Sac protein hydrolyzed PI(3,5)P 2 to PI without producing detectable intermediates and did not utilize PI(4,5)P 2 . Treatment of PIP with GST alone results in no hydrolytic activity. Addition of GST53-Sac results in conversion of both PI(3)P and PI(4)P to PI. In contrast, PIP is not hydrolyzed by GST53-5pt protein (not shown).
To examine whether the SAC1-like domain of human synaptojanin harbors similar PPIPase activity, we expressed this domain as a GST fusion protein. Of interest, although the majority of reports of the biochemical activity of synaptojanin show that it is restricted to a type II 5-ptase, Chung et al. (33) reported that purified rat brain synaptojanin exhibits 4-/5phosphatase activity, which converts PI(4,5)P 2 to PI, although it was not determined whether this activity was intrinsic or a co-purifying contaminant such as the polyphosphoinositidase activity reported by Hope and Pike (34). Treatment of total cell substrate with purified GSTSyn-Sac resulted in Ͼ50% decreases in levels of PI(3)P, PI(4)P, and PI(3,5)P 2 (not shown). Again, PI(4,5)P 2 was not a substrate, consistent with the results obtained using GST53-Sac protein. Treatment of purified 3 H-PIP with GSTSyn-Sac results in 50% conversion of both isomers to PI (Fig. 2C). The activity is time-and dose-dependent and has similar biochemical properties to GST53-Sac PPIPase in terms of metal inhibition and DTT stimulation (not shown).
The inability of any SAC1-like domains tested to hydrolyze PI(4,5)P 2 is surprising because they are able to function as 3-, 4-, and 5-phosphatases. Attempts to detect PI(4,5)P 2 hydrolytic activity under a variety of conditions, including using commercially available PI(4,5)P 2 , have reproducibly failed (not shown). Thus, the mechanism by which SAC1-like domains discern their substrates remains uncertain and will be of significant future interest.
SAC1-like Domain PPIPase Activity Is Functional in the Cell-To ascertain if PPIPase activity of SAC1-like domains was functional in vivo, we examined the phosphoinositide levels in inp52 inp53 mutant strains. Previously, we reported that steady-state levels of PI(4,5)P 2 did not change in inp52 inp53 mutants (20), whereas two-fold increases were detected in inp51 null cells (21). However, these studies were performed using thin layer chromatography system not capable of resolving PI(4,5)P 2 and PI(3,5)P 2 . We therefore analyzed 3 H-inositol lipid levels in wild type and inp52 inp53 mutant strains using a deacylation and HPLC strategy. The levels of PI(4,5)P 2 , PI(3)P, and PI(4)P were similar in both strains (not shown). However, at early log phase (3 ϫ 10 6 ) the levels of PI(3,5)P 2 increased 5-7-fold in inp52 inp53 mutant cells (Fig. 3A). It is noteworthy that even at the increased levels, the amount of PI(3,5)P 2 was only 30% of the PI(4,5)P 2 , thereby explaining why this may have been missed in our previous study. Additionally, this difference was not observed at late log and stationary phase. These data are consistent with a role for Inp52p and Inp53p in regulating levels of PI(3,5)P 2 .
Previous work of Kearns et al. (25) suggested that sac1 mutant strains had large increases in mannosyl di-inositol diphosphorylceramide (M[IP] 2 C). Our biochemical studies and the observation that TLC systems, such as those used in the Kearns study, may not effectively resolve PIPs from M[IP] 2 C suggested that the M[IP] 2 C spot may have been misidentified. Therefore, we analyzed the levels of phosphoinositides in sac1 mutant strains using a deacylation/HPLC system. Analysis of groPIs from steady-state labeled CTY182 (wild type) and CTY244 (sac1⌬) strains lipids demonstrate that sac1 mutants exhibit ϳ10-, 2.5-, and 2-fold increases in PI(4)P, PI(3,5)P 2 , and PI(3)P, respectively (Fig. 3B). In contrast, levels of PI(4,5)P 2 did not significantly change. We then labeled cells with [ 3 H]serine, which labels ceramide but not PI lipids, and found no changes in any inositol ceramide lipids (not shown). Because inositol ceramides are resistant to cleavage by methylamine, we analyzed the nondeacylatable lipids from inositol labeled cells and again found no differences (not shown). Collectively, these data demonstrate that polyphosphoinositides and not M[IP] 2 C are regulated by Sac1p PPIPase activity. The dramatic accumulation of PI(4)P in sac1 mutants suggest that Sac1p regulates a major pool of PI(4)P, possibly at the Golgi. Remarkably, despite the 10-fold increases in PI(4)P, the levels of PI(4,5)P 2 did not increase, if anything they appear to decrease slightly (Fig. 3B). It is therefore possible that the pool of PI(4)P utilized by Sac1p is different or sequestered from a pool used to make PI(4,5)P 2 . This may occur through spatial sequestration or by a protein masking as was recently proposed by Stevenson et al. (35). Alternatively, the levels of PI(4,5)P 2 may be maintained by feedback mechanisms controlling phosphatase, lipase, and/or kinase activities. It is also noteworthy that upon deletion of Fig4p, another yeast SAC1-like domain-containing protein, only small changes were observed in steadystate phosphoinositide levels (not shown). One explanation may be that the pool of lipid that Fig4p accesses is relatively small or tightly regulated.
Conservation of a Mg 2ϩ -independent Catalytic Motif within Inositol and Protein Phosphatases-SAC1-like domains contain a highly conserved sequence motif, RTNCLDCLDRTN (Fig. 4A). Within this sequence is the CX 5 R(T/S) motif found in other metal-independent protein and inositol polyphosphate phosphatases including: the tyrosine and dual function protein phosphatases PTP1B and VHR; the inositol polyphosphate 4-phosphatase family; and the PI(3,4,5)P 3  effects of metal and DTT on PPIPase activity are consistent with the role of this motif in catalysis. These biochemical and sequence similarities indicate a possible evolutionary relationship among these proteins. Additionally, the lack of PPIPase activity in both native and recombinant Inp5p is likely explained by the presence of extensive mutations within this motif to RisafDsiekpN. Most notable are the mutations in the consensus residues C 3 A, R 3 K, and (T/S) 3 P.
Concerted and Independent Action of PPIPase Activity-The assignment of PPIPase activity to SAC1-like domains represents a significant step forward in understanding how its substrates or products may regulate vesicle trafficking, actin organization, and other cellular events as summarized in Fig. 4B. Loss of PPIPase activity in sac1 mutants results in profound changes in the cellular levels of phosphoinositides, especially PI(4)P. In the case of all the "sec14 bypass" mutants, it is thought that the common downstream component is expansion of the diacylglycerol pool (25). The accumulation of phosphoinositides in sac1 mutants may expand the pool of diacylglycerol directly through Plc1p-mediated turnover of PI(4)P or indirectly through alterations in novel phosphoinositideregulated lipid metabolic pathways that lead to increased diacylglycerol production. Although the largest metabolic changes are observed in PI(4)P, significant changes are also found in PI(3)P and PI(3,5)P 2 . Thus, it is possible that these lipids regulate events related to diacylglycerol production. The ability of defects in Sac1p and presumably PPIPase activity to overcome defects in act1-1 mutant is intriguing. Much focus has been placed on PI(4,5)P 2 in regulating actin cytoskeleton; however, our data indicate levels of this lipid do not significantly change in sac1 mutants. Thus we suggest a novel role for PI(4)P, PI(3)P, and/or PI(3,5)P 2 in regulating actin organization. Furthermore, our work provides insight into possible roles for PPIPase activity and phosphoinositides in the regulation of ATP transport into the lumen of the endoplasmic reticulum (27).
The fact that Inp52p, Inp53p, and synaptojanin possess both PPIPase and 5-ptase activities is novel and necessitates reevaluation of the relative roles of PI(4,5)P 2 versus PI(4)P, PI(3)P, and PI(3,5)P 2 in synaptic and general vesicle trafficking. It is intriguing to speculate as to why the two catalytic domains are found in nature to be tethered or alone (Fig. 4B). One possible reason may be that tethering enables localized substrate hydrolysis. In a concerted model, the 5-ptase domain may hydrolyze PI(4,5)P 2 to produce PI(4)P, signal event one, which is then sequentially converted to PI by the SAC1-like domain, signal event two. These events may occur rapidly or slowly awaiting the appropriate regulatory queues. Alternatively, the lipid substrates for 5-ptase and PPIPase activities may be independent and/or localized at different membranes. To this end, each activity may be involved at different steps in vesicle trafficking, for example one during vesicle formation and the other during vesicle docking or fusion. Perhaps PPIPase activity functions to reduce polyanion charge facilitating changes in protein recruitment or by directly reducing anionic lipid-lipid repulsion forces encountered during membrane fusion. Additionally, PPIPase activity may be required to replenish PI stores that are critical for re-initiation of signaling events and general membrane homeostasis. Importantly, the distinct biochemical properties and substrate selectivity of each domain indicate multiple points of potential regulation, thereby enabling combinatorial signaling modes for proteins such as Inp52p, Inp53p, and synaptojanin. Clearly, both 5-ptase and PPIPase activities are important in regulating membrane trafficking. It is interesting that Inp51p and certain alternative splice variants of synaptojanin have lost the PPIPase functional domain.
Lastly, our report represents the first description at the molecular level of a PI(3)P 3-phosphatase, a PI(4)P 4-phosphatase, a PI(3,5)P 2 3-/5-phosphatase, and collectively a PPIPase. Of significant recent interest is the regulation of PI(3,5)P 2 . PI(3,5)P 2 is transiently present during osmotic stress (32), and its synthesis by Fab1p, a PI(3)P 5-kinase, may be involved in delivering vacuolar proteins via endosome pathway, especially the formation of multi-vesicular bodies (MVB) (7-10). The osmo-sensitive phenotype and abnormal vacuolar structure of certain inp5 mutants suggest their involvement in the regulation of PI(3,5)P 2 (19 -21). Our data demonstrate that the PPIPase activity of Inp52p, Inp53p, synaptojanin, and Sac1p are biologically relevant regulators of PI(3,5)P 2 . Whether or not any of these proteins and PPIPase activity plays a role in regulation of MVB formation or MVB recycling remains to be established.