INP51, a yeast inositol polyphosphate 5-phosphatase required for phosphatidylinositol 4,5-bisphosphate homeostasis and whose absence confers a cold-resistant phenotype.

Sequence analysis of Saccharomyces cerevisiae chromosome IX identified a 946 amino acid open reading frame (YIL002C), designated here as INP51, that has carboxyl- and amino-terminal regions similar to mammalian inositol polyphosphate 5-phosphatases and to yeast SAC1. This two-domain primary structure resembles the mammalian 5-phosphatase, synaptojanin. We report that Inp51p is associated with a particulate fraction and that recombinant Inp51p exhibits intrinsic phosphatidylinositol 4,5-bisphosphate 5-phosphatase activity. Deletion of INP51 (inp51) results in a "cold-tolerant" phenotype, enabling significantly faster growth at temperatures below 15 degreesC as compared with a parental strain. Complementation analysis of an inp51 mutant strain demonstrates that the cold tolerance is strictly due to loss of 5-phosphatase catalytic activity. Furthermore, deletion of PLC1 in an inp51 mutant does not abrogate cold tolerance, indicating that Plc1p-mediated production of soluble inositol phosphates is not required. Cells lacking INP51 have a 2-4-fold increase in levels of phosphatidylinositol 4,5-bisphosphate and inositol 1,4, 5-trisphosphate, whereas cells overexpressing Inp51p exhibit a 35% decrease in levels of phosphatidylinositol 4,5-bisphosphate. We conclude that INP51 function is critical for proper phosphatidylinositol 4,5-bisphosphate homeostasis. In addition, we define a novel role for a 5-phosphatase loss of function mutant that improves the growth of cells at colder temperatures without alteration of growth at normal temperatures, which may have useful commercial applications.

Sequence analysis of Saccharomyces cerevisiae chromosome IX identified a 946 amino acid open reading frame (YIL002C), designated here as INP51, that has carboxyl-and amino-terminal regions similar to mammalian inositol polyphosphate 5-phosphatases and to yeast SAC1. This two-domain primary structure resembles the mammalian 5-phosphatase, synaptojanin. We report that Inp51p is associated with a particulate fraction and that recombinant Inp51p exhibits intrinsic phosphatidylinositol 4,5-bisphosphate 5-phosphatase activity. Deletion of INP51 (inp51) results in a "coldtolerant" phenotype, enabling significantly faster growth at temperatures below 15°C as compared with a parental strain. Complementation analysis of an inp51 mutant strain demonstrates that the cold tolerance is strictly due to loss of 5-phosphatase catalytic activity. Furthermore, deletion of PLC1 in an inp51 mutant does not abrogate cold tolerance, indicating that Plc1p-mediated production of soluble inositol phosphates is not required. Cells lacking INP51 have a 2-4-fold increase in levels of phosphatidylinositol 4,5-bisphosphate and inositol 1,4,5-trisphosphate, whereas cells overexpressing Inp51p exhibit a 35% decrease in levels of phosphatidylinositol 4,5-bisphosphate. We conclude that INP51 function is critical for proper phosphatidylinositol 4,5bisphosphate homeostasis. In addition, we define a novel role for a 5-phosphatase loss of function mutant that improves the growth of cells at colder temperatures without alteration of growth at normal temperatures, which may have useful commercial applications.
Of central importance to the inositol signaling pathway is phosphatidylinositol 4,5-bisphosphate (PI(4,5)P 2 ), 1 which serves as a precursor to second messengers and is a signaling molecule itself (1)(2)(3). PI(4,5)P 2 has a direct functional role in signaling via interactions with actin-binding proteins, phospholipase D, and pleckstrin homology domains (3)(4)(5)(6)(7). Levels of PI(4,5)P 2 are tightly maintained by a balance of kinase, lipase, transferase, and phosphatase activities. The importance of this regulation is manifested by disease states that arise as a result of mutation in these enzymes. For example, in Drosophila deletion of the norpA phospholipase C ␤ gene results in blindness (8); ablation of an eye-specific CDP-diacylglycerol synthetase results in retinal degeneration that can be pharmacologically rescued by re-addition of PI(4,5)P 2 (9); and recent studies by Milligan et al. (10) demonstrate that the phosphatidylinositol transfer protein domain of the rdgB gene is necessary for proper photoreceptor function in response to prolonged light stimuli and serves to protect photoreceptor degeneration. In addition, the human disease oculocerebrorenal syndrome (Lowe's syndrome) arises from mutations in the OCRL-1 gene (11), which encodes an inositol polyphosphate 5-phosphatase (5-ptase) (12). 5-Ptase enzymes comprise a family of critical proteins that remove the 5-position phosphate from either soluble or lipid inositol polyphosphates, or both (reviewed in Refs. 13 and 14). Initially, types I and II 5-ptases were postulated to terminate inositol 1,4,5-trisphosphate (IP 3 ) and inositol 1,3,4,5-tetrakisphosphate-mediated Ca 2ϩ mobilization (15,16). However, recent characterization of several new 5-ptases provides evidence for their expanded role in cell function, including vesicular trafficking and Ras and cytokine signaling (13,17). These members include OCRL-1, synaptojanin (18), GA-5pt (19), p150 SHIP (20,21), and PIP 3 5pt (22). The mammalian 5-ptase family consists of at least 10 distinct members and is identified by two short peptide motifs, "(i/l)W(l/f/m)GD(l/f)D(y/f)R" and "P(a/s)W(c/t/s)DRIL" (23). Mutational analyses of these residues demonstrate their importance for substrate recognition and catalysis (23)(24)(25). It is unclear why so many 5-ptases exist and why in the case of Lowe's syndrome other 5-ptases are unable to compensate for defects in OCRL-1.
To understand better the roles of 5-ptases in cell growth and in inositol metabolism in vivo, we initiated studies in a genetically tractable yeast Saccharomyces cerevisiae. Several genes involved in yeast inositol lipid biosynthesis and metabolism have been identified (see reviews in Refs. 26 and 27). In contrast, genes encoding inositol lipid phosphatases have not yet been defined. Importantly, the yeast genome sequencing project has enabled identification of open reading frames sharing similarity to mammalian inositol signaling genes. One such open reading frame, YIL002c, designated here as INP51, is highly similar to mammalian 5-ptases. We demonstrate that INP51 encodes a 108-kDa protein which has intrinsic PI(4,5)P 2 5-ptase activity. Characterization of both loss and gain of function mutants of inp51 demonstrates that 5-ptase activity is required for proper in vivo PI(4,5)P 2 maintenance. Our studies also suggest a role for Inp51p-regulated inositol metabolites in the growth of yeast at cold temperatures.

MATERIALS AND METHODS
Strains, Media, and Genetic Methods-Yeast strains used in this study are listed in Table I. The cells were propagated in standard rich (YPD) medium or in complete minimal medium (CM) containing either 2% glucose or galactose and, when appropriate, lacking the nutrient(s) to maintain selection for plasmids or markers. Standard procedures for yeast genetic manipulations were used (28,29). Strains were sporulated on 0.3% potassium acetate plates, and tetrads were dissected using a Zeiss Micromanipulator. The ability of a given yeast strain to propagate at temperatures below 15°C was assessed by dispersing cells onto an appropriate solid media plate and, after incubation for 2 weeks, observing the resulting growth either visually or under a dissecting microscope, as appropriate.
Targeted Gene Disruption and Complementation of INP51-The INP51 gene was obtained from cosmid 9687 (the gift of Mark Johnston, Washington University, St. Louis) which contained a 21,431-bp segment of chromosome IX. The cosmid DNA was digested with SacI and PstI endonucleases, and the 3.8-kb fragment containing the INP51 locus was purified and subcloned into pBluescript II SK ϩ (Stratagene, La Jolla, CA) to yield pBluINP51. The majority of the INP51 gene was removed and replaced with the selectable marker LEU2 (Fig. 1B). The LEU2 gene, obtained from an XbaI/PstI digestion of plasmid pJJ282 (obtained from Susan Wente, Washington University), was inserted at these sites into pBluINP51, to yield pinp51::LEU2. This plasmid DNA was linearized and transformed into W303 (MATa/MAT␣) cells (obtained from Susan Wente) by the standard Li ϩ acetate protocol (28). Stable Leu ϩ transformants were selected on Leu Ϫ plates and verified by Southern blot analysis (28) of genomic DNA digested with KpnI and PvuII endonucleases and probed with a radiolabeled 1.2-kb INP51 SacI/XbaI fragment (see Fig. 1B).
Complementation analysis was performed using centromeric plasmids harboring either INP51 or INP51D788A loci. Digestion of pBlu-INP51 with SacI and HindIII released a 3.8-kb INP51 locus, containing 775 bp of 5Ј-flanking sequence and the entire coding sequence, which was then inserted at these sites into pRS316 (Ref. 30; obtained from Susan Wente) to generate pRSINP51. The pRSINP51D788A plasmid was generated by a PCR-based mutagenesis approach in which the Asp 788 codon GAT was converted to Ala 788 codon GCT. Sense and antisense oligonucleotide 5Ј-AGATTTTGCGTACTAGTATCTCAT-3Ј and 5Ј-TAACTGTTCTAGTACTTCTCCCCTACTTAAAACCTAGCTGT-CCATGCTGGTAT-3Ј were used to PCR-amplify the 400-bp mutant fragment under reaction conditions recommended by the manufacturer of the Taq polymerase (Perkin-Elmer). The product was digested with SpeI and ScaI endonucleases, purified, and co-inserted along with a purified ScaI/HindIII INP51 726-bp fragment into the SpeI and HindIII sites of pRSINP51 to yield pRSINP51D788A. The region of the pRSINP51D788A generated by PCR was sequenced to confirm that only the appropriate mutation was made. The pRSINP51 or pRSINP51D788A plasmids were transformed into haploid inp51 mutant cells by the stand-ard Li ϩ acetate protocol (28), and Ura ϩ transformants were selected.
Inp51 Antibody Generation and Western Analysis-A carboxyl-terminal peptide of Inp51p having the sequence "CVENEDEPLFIER" was chemically synthesized (Protein and Nucleic Acid Chemistry Lab, Washington University, St. Louis) and injected into rabbits using standard protocols (Pocono Rabbit Farm, Canadensis, PA). Immunoreactive antisera was affinity purified using peptide sulfolink-agarose gel (prepared according to Pierce) Soluble and particulate protein fractions were made from appropriate haploid yeast strains as follows. Approximately 3 ϫ 10 8 cells were harvested at logarithmic phase by centrifugation (2,000 ϫ g), washed twice in 1 ml of ice-cold water, and resuspended in 250 l of 50 mM HEPES, pH 7.5, supplemented with 10 mM MgCl 2 , 50 mM KCl, and 1 mM phenylmethylsulfonyl fluoride (Lysis buffer). Cells were lysed by addition of 100 l of acid-washed glass beads (425-600 m diameter, Sigma) followed by vigorous reciprocal shaking (Mini Bead-Beater TM , BioSpec, Inc., Bartlesville, OK) for 2 min. Lysates were centrifuged at 20,800 ϫ g for 5 min, and the upper soluble fraction was transferred to a new tube. The particulate fraction was resuspended in 250 l Lysis buffer and homogenized by vigorous reciprocal shaking for 2 min. Proteins were separated by SDS-PAGE, immobilized onto nitrocellulose filters by electroblotting, and immunoblotted using affinity purified anti-Inp51p antibodies according to standard protocols (28).
Bacterial Expression of INP51 and Enzymatic Analysis-pBluINP51 was digested with XbaI and PstI, releasing a 2610-bp partial INP51 fragment encoding residues 163-946 that was gel-purified. INP51 residues 1-162 were made by PCR amplification of pBluINP51 template using sense and antisense primers, 5Ј-AGCTTGAGGATCCAAAAT-GAGACTCTTCATCGGTAGAAGA-3Ј and 5Ј-AGTGCGTATCTAGAAT-AGTAGAATGTTCCATCACT-3Ј, under conditions recommended by the manufacturer. The 516-bp product was digested with BamHI and XbaI, gel-purified, and co-inserted with the 2610-bp XbaI/PstI fragment into the BamHI and PstI sites of pCMV (31) to make pCMVINP51. The region generated by PCR was sequenced to confirm that no unwanted mutations were introduced. pCMVINP51 was digested with BamHI and PstI, and the 3.2-kb fragment containing the entire INP51 coding sequence was inserted into pTrcHisA (Invitrogen, Carlsbad, CA) at these sites to yield pTrcINP51.
Fifty-milliliter cultures of DH5␣ Escherichia coli strains (Life Technologies, Inc.) harboring either pTrc or pTrcINP51 plasmids were grown in LB broth to an A 600 of 0.6 and induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 14 h at 37°C. Cells were pelleted, and lysates were prepared by addition of 1 ml Lysis buffer, 300 l of glass beads, and vigorous reciprocal shaking as described above. Soluble and particulate fractions were prepared, separated by SDS-PAGE, and immunoblotted as described above.
5-Ptase activity was measured using either  H]myoinositol, grown to a density of 3 ϫ 10 7 cells/ml, harvested by centrifugation, washed twice with 1 ml of ice-cold water, and resuspended in 100 l of 0.5 N HCl. Extraction of soluble and lipid inositol phosphates was performed by adding 372 l of chloroform/methanol (1:2 v/v) and 100 l of glass beads, bead beating for 2 min, adding 125 l each of chloroform and 2 M KCl, bead beating again for 2 min, and centrifuging at 20,800 ϫ g for 5 min. The upper and lower phases containing water-soluble and lipid inositol phosphates, respectively, were separated and stored at Ϫ80°C no longer than 3 days prior to use.
Separation of Radiolabeled Inositol Phosphates-Silica-60 TLC plates (20 ϫ 20 cm, Merck) were prepared by submersion in 54 mM potassium oxalate containing 47.5% ethanol, 2 mM EDTA and baked a minimum of 60 min at 100°C prior to use. TLC plates were developed in a solvent mixture of 160 ml of chloroform, 60 ml of acetone, 52 ml of methanol, 48 ml of glacial acetic acid, and 28 ml of water. Radiolabeled lipids were detected by autoradiography after spraying plates with En[ 3 H]ance (DuPont) and quantified by scraping appropriate regions of silica into 10 ml of scintillation fluid and counting after equilibration for 10 h.
Alternatively, crude or TLC-separated radiolabeled lipids were subjected to deacylation (32), and resulting glycerophosphoinositols (groPI) were analyzed by HPLC. Crude radiolabeled lipids were dried by vacuum centrifugation and incubated with 300 l of methylamine reagent (prepared by bubbling gaseous methylamine (CH 3 NH 2 , Fluka, Buchs, Switzerland) into ice-cold 6.2 ml of methanol, 4.6 ml of H 2 O, and 1.5 ml of 1-butanol until the total volume reached 20 ml) at 53°C for 30 min and vortexing every 10 min. Cold 1-propanol (150 l) was added, and the mixture was vacuum centrifuged to dryness. The residue was resuspended in 200 l of H 2 O, extracted twice with 300 l of 1-butanol/petroleum ether/ethyl formate (20:4:1 v/v/v), each time discarding the upper layer, and stored at Ϫ80°C until use. Deacylation of TLC-separated lipids was performed by scraping appropriate regions of silica into a screw cap Eppendorf tube, pulverizing with a glass rod, and incubating with methylamine reagent as above. In some cases, [ 32 P]PI(3)P and [ 32 P]PI(3,4)P 2 (obtained from Andy Norris, Washington University) were included, thereby generating "internal" groPI(3)P and groPI(3,4)P 2 standards.
Alternatively, samples were equilibrated in 40 mM ammonium formate, pH 3.5 (AF), applied to a Partisil SAX (4.6 ϫ 250 mm, Whatman) column, and eluted using the following gradients: 40 to 400 mM AF over 30 min, 800 mM to 1.3 M AF over 33 min, 1.7 to 1.85 M AF over 10 min, 1.85 to 3 M AF over 47 min ("gradient 3" as described in Ref. 33). Fractions (1 ml) were collected and mixed with 10 ml of scintillation fluid (ScintiSafe TM , Fisher), and both 3 H and 32 P radioactivity was determined by counting. Samples of soluble inositol phosphates included internal standards of [ 32 P]Ins(1,4)P 2 and [ 32 P]IP 3 (provided by Cecil Buchanan, Washington University). It is noteworthy that using this system, slight decreases in resolution and elution times were noticed after repeated column use.
OCRL-1 Digestion of Radiolabeled PI(4,5)P 2 Obtained from inp51 Mutant Cells-Approximately 10,000 cpm of TLC-purified [ 3 H]PI(4,5)P 2 derived from radiolabeled inp51 mutant cells was combined with 100 g of PI, [ 32 P]PI(3)P, and [ 32 P]PI(3,4)P 2 standards and dried under a steady N 2 stream, and the residue was sonicated briefly in 100 l of 50 mM MES/Tris, pH 6.5, containing 5 mM MgCl 2 . Lipid substrate was incubated with 100 ng of purified recombinant OCRL-1 5-ptase (obtained from Xiaoling Zhang, Washington University) for 30 min at 37°C. Reactions were stopped by the addition of 300 l of methylamine reagent. The products were then deacylated and characterized by Partisil SAX HPLC as detailed above.
Overexpression of Inp51 in Yeast-pCMVINP51 was digested with BamHI and SacII, releasing a 3.2-kb fragment containing the complete coding sequence which was inserted into pBJ246 (ATCC 777452) at these sites to yield pBJ246INP51. pBJ246INP51 or pBJ246 plasmids were transformed into the haploid W303 by the standard Li ϩ acetate protocol (28). Ura ϩ strains were grown in 1 ml of CM lacking uracil (CM/ura Ϫ ) supplemented with 2% glucose to a density of 5 ϫ 10 7 cells/ml, washed 3 times in 1 ml of CM/ura Ϫ without sugar, resuspended in 1 ml of CM/ura Ϫ supplemented with either 2% glucose (uninduced) or 2% galactose (induced), and grown for 9 h at 30°C. Cellular extracts were separated by SDS-PAGE and analyzed by immunoblotting as described above. For inositol labeling studies the growth media were supplemented with 10 Ci/ml [ 3 H]myoinositol, and radiolabeled lipids were analyzed by TLC as described above.
Disruption of PLC1-The complete coding region of PLC1 was replaced with the G418 gene using a PCR-based approach. G418 was amplified from plasmid pFA6-kanMX2 (34) by PCR using the sense primer 5Ј-AAACGTACAACGGTAAGGTCATTCACGCAGTGTATATG-CAGCTGAAGCTTCGTACGC-3Ј and the antisense primer 5Ј-CGTAT-TTATGAATATGTGTATTTGGCCGGAAAAAGATCGCCGCATAGGC-CACTAGTGGATCTG-3Ј (obtained from Joseph Heitman, Duke University) (where underlined bases correspond to PLC1 sequence and the remainder of the primer is from the pFA6-kanMX2 plasmid) according to conditions recommended by the manufacturer. The resulting plc1::G418 PCR product was introduced into diploid cells by the standard Li ϩ acetate protocol (28). Stable G418 R cells were selected for on YPD plates supplemented with 200 g/ml G418 (Calbiochem) and verified by PCR of genomic DNA template using the above sense primer and an antisense primer 5Ј-ATATACTCGAGAACTTTCTGCTGAAGA-GAAGGC-3Ј (obtained from Joseph Heitman) derived from the 3Ј-untranslated region of the PLC1 locus.

Identification of an Inositol Polyphosphate 5-Phosphatase in
Yeast-The TBLASTN search algorithm (35) was used to compare the amino acid sequences of the mammalian 5-ptases, type I, type II, and OCRL-1 (GenBank TM accession numbers Z31695, M74161, and M88162, respectively) to the predicted open reading frames in S. cerevisiae genomic sequence (available January, 1995). We identified a highly related open reading frame corresponding to locus YIL002c, located on chromosome IX, encoding a 946-residue gene product, designated here as INP51 for INositol polyphosphate 5-Phosphatase number 1. The carboxyl-terminal region of Inp51p is approximately 30% identical (Ͼ50% similar) over 300 amino acids to the catalytic domains of type II and OCRL-1 5-ptases. Importantly, Inp51p contains the two short sequence motifs common to all 5-ptases characterized to date (Fig. 1A). Two substitutions of the "PAWCDRILW" sequence of type II and OCRL-1 are found in Inp51p, Cys 787 3 Thr and Trp 792 3 Ser. In addition, the amino-terminal half of Inp51p has significant similarity to the S. cerevisiae 623-residue SAC1 and the predicted 879-residue YNL325c gene products (Fig. 1A). Although SAC1 is not an essential gene, sac1⌬ mutants are inositol auxotrophs and are cold-sensitive; moreover, particular sac1 alleles suppress certain act1 ts alleles (36) and a sac1⌬ mutation bypasses the effects of the sec14 -1 ts mutation (37,38). The segment of Inp51p with the greatest similarity to SacI (RTNCMD-CLDRTN motif in Fig. 1A) corresponds to the portion of SAC1 that when altered enables suppression of act1 and sec14 mutations (37,39). For example, sac1-8 and sac1-22 mutant alleles were found to have a Asp 337 to Asn mutation (39). The YNL325c gene product acts as a dosage suppressor of sac1 mutations, although a ynl325c null mutation does not bypass sec14 mutations (39). The two-domain structure of Inp51p most resembles the mammalian 5-ptase synaptojanin, a protein associated with synaptic vesicles (Fig. 1A).
As a means to address the role of INP51 in yeast cell func-tion, we constructed a mutant strain deficient in the protein and examined its phenotype. A null mutation, inp51⌬::LEU2, was constructed by replacing the majority of the INP51 coding sequence with a LEU2-selectable marker as described under "Materials and Methods" (also see Fig. 1B). Each mutant allele was introduced by DNA-mediated transformation into a diploid recipient (W303). Southern blot analysis was used to confirm the inp51⌬::LEU2/INP51 heterozygous diploid as evidenced by the presence of the mutant 5.3-kb and the wild-type 2.7-kb bands (Fig. 1C). Upon sporulation and tetrad dissection of the heterozygous diploid, four-spored tetrads were readily recovered in which the Leu ϩ marker segregated 2:2, confirming that inp51 mutants are viable (Fig. 1D).
In order to prove that INP51 encodes a protein expressed in yeast and to confirm that null mutations prevented production of Inp51p, we raised a specific rabbit polyclonal antibody directed against a 14-residue carboxyl-terminal peptide. Extracts prepared from a wild-type haploid derived from the parental strain and a null mutant were separated into soluble and particulate fractions. These proteins were resolved by SDS-PAGE and examined by immunoblotting ( Fig. 2A). In wild-type cells, a protein with an apparent molecular mass of 108 kDa, consistent with the predicted size of the INP51 gene product, was observed primarily in the particulate fraction (lane 1). It is noteworthy that greater than 80% of Inp51p was solubilized by preparing extracts in the presence of 1% Nonidet P-40 (data not shown). As expected, the inp51 mutant lacked the 108-kDa species, whereas many of the cross-reacting lower molecular weight proteins were still present (lane 2).
Inp51 Has Intrinsic 5-Phosphatase Activity-Based on the similarity of INP51 to synaptojanin, OCRL-1, and type II mammalian 5-ptases, we predicted that the Inp51p may function as a dual specificity phosphatase that hydrolyzes both lipid and soluble inositol polyphosphates. In the yeast S. cerevisiae, both putative substrates IP 3 and PI(4,5)P 2 have been identified (40 -42). Initially, we measured 5-ptase activity in extracts prepared from either haploid wild type derived from parental strain or inp51 mutants using IP 3 or PI(4,5)P 2 as substrates.
The only detectable 5-ptase activity we observed was against the lipid substrate (data not shown). The inability to find 5-ptase activity using IP 3 was puzzling in light of metabolism in mammalian cells but not unexpected as others have previously reported that yeast do not contain IP 3 5-phosphomonoesterase activity (43).
In order to verify that Inp51p harbored intrinsic 5-ptase activity, we heterologously expressed recombinant protein in the bacteria. Bacteria are especially useful because they lack measurable 5-ptase activity. The coding sequence of the INP51 gene was inserted into an isopropyl-1-thio-␤-D-galactopyranoside-inducible bacterial expression plasmid, and soluble extracts were prepared from induced cultures as described under "Materials and Methods." Immunoblot analysis of these proteins separated by SDS-PAGE (Fig. 2B) demonstrates that Inp51p is expressed as the expected 108-kDa product (although significant proteolysis occurs) in cells containing pTRCINP51 (lane 2) but not pTRC control cells (lane 1).
Soluble extracts were tested for 5-ptase activity using lipid and soluble inositol substrates. Micelles containing PI/ [ 3 H]PI(4,5)P 2 were treated with protein extracts, and the reactants were deacylated and separated by Partisphere Sax HPLC using "gradient 1" (Fig. 2C). The elution profile of radiolabeled groPI standards corresponding to groPI (8 min), groPI(3)P (20.5 min), groPI(4)P (22 min), and groPI(4,5)P 2 (43 min) was determined (top panel). The elution profile of control treated PI(4,5)P 2 shows a single peak at 43 min corresponding to groPI(4,5)P 2 indicating no hydrolysis occurred (middle panel). In contrast, recombinant Inp51p treatment results in the appearance of a groPI(4)P peak at 23 min and a diminution of the groPI(4,5)P 2 peak at 43 min, demonstrating that Inp51p hydrolyzes the 5-position phosphate (lower panel). This activity was determined to be time-and dose-dependent, having an estimated specific activity of 0.8 mol/min/mg (data not shown). Recombinant Inp51p exhibited no phosphomonoesterase activity using IP 3 as a substrate, consistent with the lack of this activity in yeast cellular extracts (data not shown).
Loss of INP51  Phenotype-The cold-sensitive phenotype of sac1 mutants and the possibility that INP51 may regulate levels of a lipid substrate prompted us to analyze growth of an inp51 mutant at temperatures ranging from 12 to 37°C. Haploid cells derived from the wild-type parental strain and inp51 mutant strains were dispersed onto solid media and incubated at temperatures of 12, 14, 23, 30, and 37°C for appropriate times. No differences in the size or numbers of colonies were observed between these strains at 23°C and above (Fig. 3A). In contrast, after 14 days at 12 or 14°C, we observed that the parental wild-type cells are unable to grow, whereas the inp51 mutants are able to grow as single colonies (Fig. 3A, W303a and JYY3).
We also tested the growth of these strains in liquid culture at cold temperature. Cultures of inp51 mutants and parental wild-type strains were inoculated at low densities (2 ϫ 10 5 cells/ml) and grown at 12°C for 200 h. The inp51 mutant cells reached saturation densities (2 ϫ 10 8 cells/ml) by 200 h, whereas the densities of wild-type cells remained low (less than 1 ϫ 10 7 cells/ml) and were only gradually rising (data not shown).
In order to ensure that the cold-tolerant phenotype was strictly due to disruption of the INP51 gene, we initiated complementation studies. A centromeric plasmid harboring the INP51 locus, including all of the coding sequence plus 775 bp of 5Ј-upstream sequence, was introduced into haploid inp51 mutant cells. Immunoblot analysis confirmed that Inp51p produced in inp51 pRSINP51 cells was of similar size, quantity, and distribution ( Fig. 2A, lane 3) to those found in haploid W303 cells (Fig. 2A, lane 1). When tested for growth on solid media at 12°C, we observed that the inp51 pRSINP51 cells are unable to grow as single colonies (Fig. 3B, JYY20).
Since INP51 contains a potentially functional SAC1 domain, we tested whether or not the cold-tolerant phenotype results from loss of 5-ptase activity. We generated a catalytically inactive INP51 by substitution of Asp 788 to Ala located in the PAWCD788RILW motif. Identical mutation of this conserved residue in mammalian 5-ptases results in loss of enzymatic activity (23,25). Immunoblot analysis of extracts prepared from inp51 pRSINP51D788A mutants demonstrates that level, size, and distribution of Inp51pD788A (Fig. 2A, lane 4) is similar to Inp51p present in haploid W303 cells (Fig. 2A, lane 1). These mutants are able to grow as single colonies on solid media at 12°C (Fig. 3B, JYY21), indicating that loss of 5-ptase activity is sufficient for the cold-tolerant phenotype.
It is noteworthy that our W303 parental strain grows relatively poorly at temperatures below 15°C in contrast to other S. cerevisiae strains which readily grow as single colonies at cooler temperatures. For example, a different W303 strain (obtained from Jeremy Thorner, designated W303 JT ) is able to grow as single colonies at 12°C. We therefore disrupted the INP51 gene in the W303 JT diploid background and obtained haploid progeny as described above. Haploid inp51 and wildtype cells derived from W303 JT parental strain were dispersed onto solid media, and after 10 days at 12°C the diameters of several single colonies were measured. The average colony diameter of inp51 (W303 JT derived) mutants was three times larger than the parental counterparts (W303 JT ) and two times larger than our original inp51 (W303) mutants (data not shown).
To characterize further the cold-tolerant phenotype, we de- were resolved by SDS-PAGE, and immunoblotted as described above. C, 5-ptase activity of Inp51p and product-proof. Partisphere SAX HPLC analysis using gradient 1 of groPIs that were prepared from standards (top panel), pTRC control treated PI(4,5)P 2 (middle panel) and pTRCINP51 treated PI(4,5)P 2 as described under "Materials and Methods." Peak elution times of groPI, groPI(3)P, groPI(4)P, and groPI(4,5)P 2 are 8-, 20.5-, 22-, and 43-min, respectively. termined if cold shock was sufficient to induce a growth difference between mutant and parental strains. Strains were streaked onto solid media, incubated at 12°C for 12, 24, or 36 h (cold-shock) and then transferred to 23 or 30°C for several days. No difference in growth between the wild-type and inp51 mutants was observed under any of these cold-shock conditions (data not shown).
Inp51 Regulates Levels of PIP 2 and IP 3 in Vivo-We next determined the effects of INP51 loss of function on metabolism of lipid and soluble inositol phosphates in vivo. Appropriate yeast strains were labeled to steady state with [ 3 H]inositol such that radioactivity reflected the relative mass levels of various inositol phosphates (labeling throughout eight cell doublings was sufficient to reach equilibrium). Radiolabeled lipids were isolated from haploid inp51 mutants, parental-derived wild-type, inp51 pRSINP51, and inp51 pRSINP51D788A strains, separated by oxalate TLC, and visualized by autoradiography. Five major and several minor lipid species are present in all four strains (Fig. 4A). Studies of inositol incorporation into phospholipids (PL) by Smith and Lester (44) (Fig. 4B). Amounts of PI recovered were not significantly different in any of the strains. In contrast, levels of PIP 2 increased 1.9-fold in an inp51 mutant (sample 1) as compared with parental wild type (sample 2); moreover, this increase was reverted by introduction of INP51 (sample 3) but not INP51D788A (sample 4). These data confirm that accumulation of PIP 2 requires loss of Inp51p 5-ptase activity. Additionally, similar patterns of PIP 2 lipid accumulation were observed in these strains grown at 12°C.
The accumulation of PIP 2 in inp51 null strains is consistent with Inp51p functioning as a PI(4,5)P 2 5-ptase; however, the TLC system used is not capable of separating individual isomers of PIP 2 nor is it capable of resolving PI(3)P, PI(4)P, and M(IP) 2 C. PI(3,4)P 2 or PI(3,5)P 2 has not yet been conclusively reported in S. cerevisiae; however, if these species are present only transiently or in very low levels they may have been easily missed. Thus, to unequivocally determine identity of the accumulating PIP 2 isomer and to determine the effects of Inp51p loss of function on PIP isomers, we utilized a strategy in which the radiolabeled lipids were deacylated to form water-soluble groPIs that are easily separated by strong anion HPLC. Crude radiolabeled phospholipids were deacylated and resolved by Partisil SAX HPLC using "gradient 3" as described under "Materials and Methods." We note that radiolabeled IPC, MIPC and M[IP] 2 C are "non-deacylatable" phospholipids as previously reported (44) and are therefore not detected in the soluble samples. The profile of inositol, groPI, groPI(3)P, groPI(4)P, groPI(3,4)P 2 , and groPI(4,5)P 2 standards showed elution in fractions 4 -8, 9 -13, 33-35, 35-38, 51-53, and 53-57, respectively (data not shown). Slight variability in elution times were observed depending on column age. Samples derived from parental wild-type cells show five peaks corresponding to above standards (except groPI [3,4]P 2 ) having radioactivities of 79,625, 499,010, 5900, 3433, and 1133 cpm, respectively (Fig.  4C, top panel). Radioactivities measured from the inp51 mutant sample are 72,295, 515,360, 5515, 3540, and 2370 cpm, respectively (Fig. 4C, middle panel). These data show that only groPI(4,5)P 2 levels change significantly (2-fold increase), that the PIP 2 lipid is not PI(3,4)P 2 , and that PIP levels are unaffected by INP51 loss of function.
Since we did not have a groPI(3,5)P 2 standard, these experiments were not sufficient to disprove that the PIP 2 was PI(3,5)P 2 . To distinguish between these species we treated the PIP 2 lipid with purified recombinant OCRL-1 5-ptase, reasoning that the possible products PI(3)P and PI(4)P could be identified based on groPI(3)P and groPI (4) HPLC. The elution profile demonstrates that the PIP 2 is converted to a single product peak (fractions 35 and 36) corresponding to groPI(4)P (Fig. 4C, lower panel), confirming that the PIP 2 is not PI(3,5)P 2 .
The consequence of INP51 deletion on in vivo metabolism of soluble inositol phosphates in yeast was also determined. Steady state [ 3 H]inositol-labeled inositol phosphates were prepared from parental wild-type and inp51 mutant cells harvested at log phase (3 ϫ 10 7 cells/ml) and separated along with internal [ 32 -P]Ins(1,4)P 2 and -Ins(1,4,5)P 3 standards by strong anion HPLC using gradient 3 as described under "Materials and Methods." Elution profiles of 3 H radioactivity shows 5 major peaks in fractions 2-8, 9 -15, 23-30, 44 -50, and 86 -90 corresponding to inositol, groPI, inositol monophosphates (InsP 1 ), inositol bisphosphate (IP 2 ), and IP 3 , respectively (Fig.  5A). Inspection of both chromatograms demonstrates a marked accumulation of the IP 3 peak in the inp51 mutant (lower panel) as compared with wild-type samples (upper panel). Quantitation of individual peaks from several experiments as well as a single experiment of samples prepared from an inp51 pRS-INP51 strain demonstrate that no significant changes occur in either Ins, IP 1 , or IP 2 isomers, whereas a 2.2-fold increase in the IP 3 peak is observed in inp51 mutants and that this increase is reversed by complementation with Inp51p (Fig. 5B). We note that the major IP 2 peak (fractions 45-47) does not co-elute with the internal [ 32 P]Ins(1,4)P 2 standard, and at this time its identity remains unknown. The IP 3 peak exactly coelutes with an internal [ 32 P]Ins(1,4,5)P 3 standard.
In addition to analysis of INP51 loss of function, we also determined the effect of INP51 gain of function. We utilized a galactose-inducible multicopy plasmid harboring the coding sequence of the INP51 gene. Overexpression of Inp51p protein was confirmed by Western blot analysis (Fig. 6A), and levels were estimated to be increased approximately 50-fold over en-dogenous and were primarily found in the particulate fraction. Overexpression of Inp51p did not result in an observable growth phenotype on solid media at temperatures ranging from 12 to 37°C (data not shown). We then analyzed levels of radiolabeled PI and PIP 2 obtained from steady state [ 3 H]inositollabeled cultures (Fig. 6B). Levels of PIP 2 decreased 35% in galactose-induced cells harvested from Inp51p overexpressing (sample 4) versus induced control cells (sample 3), whereas no changes in PI were observed. Additionally, we compared levels of PI and PIP 2 isolated from both strains grown in glucose (uninduced) (samples 1 and 2), and no significant changes were observed.
Cold Tolerance Does Not Require IP 3 Produced via a Plc1p Pathway-Since loss of INP51 function leads to increases in both PI(4,5)P 2 and IP 3 , we were interested in determining whether phospholipase C activity is necessary for cold resistance. In yeast PLC1 encodes a PI-phospholipase C that hydrolyzes PI(4,5)P 2 to form diacylglycerol and IP 3 (45)(46)(47). TBLASTN comparison of the PLC1 sequence with the complete S. cerevisiae genome fails to identify another significantly related gene product indicating that a single gene is present. Therefore, we constructed plc1⌬::G418 single and plc1⌬::G418 inp51⌬::LEU2 double mutants as described under "Materials and Methods." Cells from these strains as well as from parental derived wild-type and inp51 mutant strains were dispersed onto solid media and grown at 23 and 12°C for 3 and 12 days. Consistent with previously published data at 23°C (45), on rich medium plc1 deletion results in a partially growth-compromised phenotype; moreover, the plc1 inp51 double mutant did not exhibit a significantly improved growth rate (data not shown). To quantify better the cold-resistant phenotype, we photographed regions of single colonies of strains grown at 12°C using a dissecting microscope (Fig. 7). We observed very slow growth in both wild-type  1 and 2) and particulate (lanes 3 and 4) protein fractions prepared from galactose-induced cell cultures were separated by SDS-PAGE and analyzed by immunoblotting with anti-Inp51p antibody. Extracts were prepared from cells containing pBJ246 (lanes 1 and 3) and from cells containing pBJ246INP51 (lanes 2 and 4). B, determination of PI and PIP 2 levels in control and Inp51p overexpressing cells. Radiolabeled lipids were prepared from cells grown in glucose (uninduced) and galactose (induced), separated by TLC, and quantified as m, whereas that of inp51 plc1 double mutant cells is 102.5 Ϯ 13.9 m (p Ͻ 0.001 for each as compared with wild type). These data demonstrate that production of IP 3 via a Plc1p pathway is not required for cold resistance induced by Inp51 loss of function.
We also measured the levels of IP 3 in the plc1 null strains using isotopic labeling studies with [ 3 H]inositol. Since our plc1 null strain did not propagate in minimal media, it was necessary to label in rich media. By using up to 50 Ci/ml [ 3 H]inositol, we were unable to detect IP 3 in either the plc1 single or plc1 inp51 double mutant strains, whereas 150 and 300 cpm of IP 3 was detected in parental and the inp51 null strains, respectively (compared with 600 and 1100 cpm in 20 Ci/ml minimal media from these strains, see Fig. 5). The reduced levels are a result of the poor labeling efficiency in rich media. In contrast to the plc1 single mutant, the plc1 inp51 double mutant was able to grow in minimal media. By using 20 Ci/ml [ 3 H]inositol, we were able to detect approximately 100 cpm of IP 3 in the plc1 inp51 double mutant, as compared with 1100 cpm for the inp51 mutant. Thus, it appears that yeast contain a Plc1p-independent pathway for production of IP 3 . Whether or not this "pool" of IP 3 is regulated by Inp51p and is involved in cold resistance remains uncertain at this time. DISCUSSION Many PI biosynthetic and metabolic enzymes are conserved from yeast to man. In the yeast S. cerevisiae, inositol is utilized to form PI, PI(3)P, PI(4)P, and PI(4,5)P 2 (reviewed in Ref. 26) as well as IPC-I, IPC-II, IPC-III, MIPC, and M(IP) 2 C (see review in Ref. 48). Genes encoding synthase, kinase, and lipase enzymes have been cloned from S. cerevisiae (reviewed in Refs. 26 and 27) including inositol 1-phosphate synthase, INO1 (49); PI synthase, PIS1 (50, 51); PI 4-kinases, PIK1 (52) and STT4 (53); PI(4)P 5-kinases, FAB1 and MSS4 based on strong homology to mammalian proteins (54 -56); PI 3-kinase, VPS34 (57); and phospholipase C, PLC1 (45)(46)(47). Sequencing the genomes of a multitude of organisms has resulted in the powerful ability to identify gene families conserved throughout evolution. We used this information to identify a 5-ptase, INP51, in the yeast S. cerevisiae.
Our studies indicate that INP51 encodes a lipid-selective 5-ptase whose activity is critically required to maintain proper cellular levels of PI(4,5)P 2 . In mammalian cells, several distinct 5-ptases have been classified into four groups based mainly on substrate selectivity (58) as follows: group I 5-ptases utilize only soluble inositol polyphosphates; group II enzymes utilize both lipid and soluble substrates; group III enzymes associate with PI 3-kinase and only utilize PI(3,4,5)P 3 ; and group IV enzymes associate with tyrosine-phosphorylated growth factor receptors and appear to utilize lipid and soluble substrates containing a D3 position phosphate. The domain structure of Inp51p indicates it is most like synaptojanin, a group II 5-ptase that hydrolyzes both PI(4,5)P 2 and IP 3 in vitro (18). However, since Inp51p does not utilize IP 3 in vitro, we suggest it should not be strictly classified as a group II enzyme and may represent a novel PI(4,5)P 2 -selective class. In vivo studies of inositol polyphosphate levels in cells lacking Inp51p show measurable increases in the levels of PI(4,5)P 2 and IP 3 . The increase in IP 3 in inp51 mutant cells is most likely a mass action effect that results from an accumulation of PIP 2 . The lack of measurable IP 3 5-ptase activity from either recombinant Inp51p or parental yeast cell extracts supports this hypothesis. Furthermore, overexpression of Inp51p decreases levels of PI(4,5)P 2 . Together these data clearly show that changes in cellular 5-ptase activity directly contribute to PI(4,5)P 2 homeostasis.
The importance of PI(4,5)P 2 in cell survival and as a regulator of several enzymes is well documented. Individual deletion of genes required for PI(4,5)P 2 synthesis including PI synthase PIS1, PI 4-kinases PIK1 or STT4, or putative PI(4)P 5-kinase MSS4 results in lethality. Moreover, electroporation of PI(4,5)P 2 -specific antibodies into yeast potently inhibits cell growth (59). Additionally, PI(4,5)P 2 modulates the activities of phospholipase D and certain actin-binding proteins such as profilin. Since only small changes in PI(4,5)P 2 are observed in loss or gain of function inp51 mutant cells, mechanisms may exist that are capable of sensing PI(4,5)P 2 levels enabling feedback compensation for changes in 5-ptase activity, for example by alterations in PI synthase, kinase, lipase, and/or other phosphatase activities. Alternatively, the small changes may reflect the possibility that Inp51p is restricted to hydrolyzing a distinct cellular pool of PI(4,5)P 2 via compartmentalization or sequestration. Indeed, upon completion of the S. cerevisiae genome two additional INP5-predicted gene products were identified, designated INP52 (YNL106c) and INP53 (YOR109w). All three genes share highly similar sequences as well as the synaptojanin-like two-domain structure. We report elsewhere a thorough genetic analysis of this gene family, and we show that simultaneous deletion of all three genes is lethal (75). During preparation of this manuscript, it came to our attention that an independent genetic study has also concluded that the INP51, INP52, and INP53 genes (which were designated SJL1, SJL2, and SJL3) constitute an essential gene family (76).
The significance of the SacI-like domain present in Inp51p is currently unclear. Mutant sac1 alleles suppress the secretory defects caused by certain actin (act1 ts ) alleles (36) and suppress the Golgi transport defects and inviability of a sec14 ts mutation (37). Sec14 is a PL transfer protein exhibiting specificity toward PI and phosphatidylcholine. SacI is an integral membrane protein, and null mutations result in inositol auxotrophy and cold sensitivity (36,38). Recently, Kearns et al. (39) demonstrated that defects in SacI bypass the requirement for Sec14 by increasing the pool of diacylglycerol in Golgi membranes. It was also reported that deletion of INP51 failed to "bypass" the requirement of Sec14 (39). However, it was initially reported that bypass suppression of act1 ts or sec14 ts was achieved by sac1 mutation and not sac1 deletion (36,37). Thus, perhaps similar mutations in the SacI domain of INP51 (see Fig. 1), for example Asp 3 Asn (residue 426 in INP51) substitution as found in sac1-8 or sac1-22 alleles (39), would be capable of bypass suppression. We also note that deletion of residues 2-490 of Inp51p results in a 2-fold increase in PI(4,5)P 2 accu- mulation in vivo indicating loss of 5-ptase function, whereas a point mutation of the Asp 426 3 Ala in the SacI domain of Inp51p does not (data not shown). It is uncertain at this time whether the loss of function of the SacI domain-less Inp51p is due to improper localization or loss of intrinsic activity.
The accumulation of PI(4,5)P 2 upon loss of INP51 5-ptase function correlates with the cold-tolerant phenotype. Whereas mutations in numerous genes have been shown to result in cold-sensitive phenotypes, we were unable to find reports in which loss of function causes cold tolerance. It is clear that this phenotype is due to loss of 5-ptase catalytic activity and does not require the production IP 3 via Plc1p. It is also of interest that deletion of INP51 in a cold-resistant strain (W303 JT ) further enhances growth at cold temperatures. Previous genetic analysis of mutations affecting growth of S. cerevisiae at low temperatures (60) demonstrates that mutations in the tryptophan biosynthesis machinery results in cold sensitivity even in rich media containing tryptophan. Presumably a tryptophan permease, such as Tat2 (61), is not functional at cold temperatures. Since our strain carries a trp1-1 allele we analyzed tryptophan permease function by measuring tryptophan uptake at 12°C. We observed that inp51 mutants had significantly increased rates of uptake as compared with the very slow uptake of our parental strain (data not shown).
How increases in PI(4,5)P 2 may function to regulate cold tolerance is unclear. PI(4,5)P 2 may cause this effect directly, although this seems unlikely given it represents only a minor fraction (0.01%) of the total yeast phospholipid. It may be that increases in PI(4,5)P 2 lead to increases in downstream metabolites which mediate the cold resistance; however, our studies demonstrate that PLC1 is not involved. Alternatively, it is attractive to suggest that PI(4,5)P 2 mediates its effects through a signaling pathway which may regulate lipid biosynthesis, cytoskeletal organization, and/or vesicular trafficking.
The ability to induce cold tolerance by loss of a single enzyme function may have useful applications in agriculture and fermentation industries. In plants, defining the molecular bases for chilling tolerance is of significant academic and commercial interest (reviewed in Ref. 62). The cold tolerance of inp51 mutants appears quite similar to chilling tolerance in Arabidopsis. One hypothesis for chilling tolerance in plants centers around membrane fluidity and/or permeability. Several studies in Arabidopsis provide a correlation between increases in fatty acid unsaturation and resistance to cold (63-65). For example, FAD2 is believed to encode a desaturase, and a fad2 mutant of Arabidopsis is cold-sensitive (65,66). Moreover, studies of cold adaptation in carp demonstrate an up-regulation of ⌬ 9 -desaturase activity upon cooling fish from 30 to 10°C (67). In S. cerevisiae the gene OLE1 encodes a ⌬ 9 -desaturase (68) which may be of particular interest to test with our inp51 mutants. In addition, it has been shown that cold shock induces an immediate rise in intracellular free Ca 2ϩ in Arabidopsis (69). In yeast, IP 3 does not appear to be involved in Ca 2ϩ signaling and regulation of intracellular Ca 2ϩ (70 -74). Thus, it would be worth testing whether increases in Ca 2ϩ concentration in our solid media result in changes in growth at 12°C. Alternatively, "cold-shock" exposure of organisms results in dramatic changes in protein synthesis. We therefore analyzed Inp51p levels in cells grown at 30 and 12°C; however, no changes were observed (data not shown). Regardless of the mechanism, cloning of INP51 homologs in Arabidopsis and studies of loss of 5-ptase function may provide important clues in defining a molecular basis for chilling tolerance.