Phosphatidylinositol 5-phosphate biosynthesis is linked to PIKfyve and is involved in osmotic response pathway in mammalian cells.

The cellular functions, regulation and enzymology of phosphatidylinositol (PtdIns) 5-P, the newest addition to the family of phosphoinositides (PI), are still elusive. Whereas a kinase that uses PtdIns-5-P as an intracellular substrate has been assigned, a kinase that produces it remained to be identified. Here we report that PIKfyve, the enzyme found to synthesize PtdIns-5-P in vitro and PtdIns-3,5-P(2) in vitro and in vivo, is responsible for PtdIns-5-P production in a cellular context. Evidence is based on examination of two groups of cell types by two independent approaches. First, [(32)P]orthophosphate-labeled cells (Sf9, 3T3-L1 fibroblasts, and 3T3-L1 adipocytes) that show a high pressure liquid chromatography (HPLC)-detectable peak of the PtdIns-5-P head group at basal conditions demonstrated a 20-50% increase in radioactive PtdIns-5-P amounts upon expression of PIKfyve(WT). Second, cell types (HEK293), in which the basal levels of radioactive PtdIns-5-P were undetectable by HPLC head group analysis, demonstrated higher in vitro type II PIP kinase-directed conversion of the endogenous PtdIns-5-P pool into PtdIns-4,5-P(2), when induced to express PIKfyve(WT). Conversely, a decrease by 60% in the conversion of PtdIns-5-P to PtdIns-4,5-P(2) was associated with induced expression of the dominant-negative kinase-deficient PIKfyve(K1831E) mutant in HEK293 cells. When 3T3-L1 fibroblasts and 3T3-L1 adipocytes were subjected to osmotic shock, levels of PtdIns-5-P measured by both approaches were found to decrease profoundly upon a hypo-osmotic stimulus. Together, these results identify PIKfyve as an enzyme responsible for PtdIns-5-P biosynthesis and indicate a role for PtdIns-5-P in osmotic response pathways in mammalian cells.

Given the pleiotropic functions of PIs, it is apparent that the kinases and phosphatases responsible for their synthesis and turnover play an essential role in cell regulation. Whereas the phosphatases and their contribution to the interconversion of the PIs are less well characterized (6 -8), phosphoinositide kinases have been identified for most PIs (1)(2)(3)(4). Studies with purified mammalian enzymes divide PIKs into a number of classes differing by the phosphorylation of specific hydroxyl groups in the inositol ring: PI 3-Ks, PI 4-Ks, and PI 5-Ks (1)(2)(3)(4). PI 3-kinases, which catalyze the phosphorylation at position D-3 in PtdIns, PtdIns-4-P, PtdIns-4,5-P 2 , and likely PtdIns-5-P, are further subdivided into three classes, i.e. I A and I B , II, and III, based on their structure, in vitro substrate specificity, and mode of activation. PI 4-Ks are also represented by several types. Some of them display restricted substrate specificity strictly directed toward position D-4 of PtdIns, but not of PIs, and are called PtdIns 4-Ks. Position 4 in D- 5-phosphorylated PtdIns, and to a lesser extent, in D-3-phosphorylated PtdIns, can be attacked by the enzymatic activity of another subclass of PI 4-kinases, known as type II PIPKs or PIP 4-Ks. Finally, position 5 can be phosphorylated by two subclasses of enzymes: PI 5-Ks (or type I PIPKs; Ref. 13) and PIKfyve (Refs. 14 and 15 and reviewed in Ref. 16), which display preferences for D-4-and D-3-phosphorylated PtdIns, respectively. Both enzymes are capable of converting PtdIns to PtdIns-5-P in vitro (13)(14)(15). It should be emphasized that often the in vitro determined substrate specificity does not reflect that in living cells, where factors such as substrate availability, accessibility and presentation, or enzyme regulation may alter the enzyme specificity. This, combined with the fact that each PI (with the exception of PtdIns-3,4,5-P 3 ) could be produced by dephosphorylation of higher phosphorylated PIs, indicates that PI biosynthesis and turnover are exceedingly complex.
Although studies with purified enzymes revealed PtdIns-5-P synthesis is supported by two enzymes, the kinase(s) responsible for PtdIns-5-P production in a cellular context remained unknown. The identification of this 5-phosphorylatd PtdIns metabolite, the latest addition to the PI family, was somewhat delayed because of its poor chromatographic separation from PtdIns-4-P on HPLC columns (17). Moreover, when detected in mammalian cells, PtdIns-5-P was found to represent only a minor fraction of PIs (17), thus making cellular studies even more problematic. PtdIns-5-P is a substrate for type II PIPKs (17) and, thus far, has been reported to occur naturally in resting mouse NIH3T3 fibroblasts in culture (17), human platelets (18), and Chlamydomonas cells (19). Negative results were reported in yeast (20,21), several mammalian cell types in culture (22), and Arabidopsis (23). Little is known about the intracellular regulation of PtdIns-5-P. In platelets, levels of PtdIns-5-P have been shown to increase acutely upon thrombin stimulation, whereas in Chlamydomonas cells PtdIns-5-P was up-regulated by hyperosmotic stress (18,19). Clearly, although the natural occurrence of PtdIns-5-P in the higher eukaryotic cells may play an important role in different cellular processes, the intracellular production, regulation of metabolism, and physiological functions of PtdIns-5-P are largely unknown. Therefore, to define the enzyme involved in PtdIns-5-P biosynthesis and a possible regulation of PtdIns-5-P levels, we have examined various cell types, in which accumulated radioactive PtdIns-5-P during [ 32 P]orthophosphate cell labeling is readily detected by HPLC head group analysis. In addition, mammalian cells, previously shown negative for radioactive PtdIns-5-P by HPLC (22), were examined in vitro for type II PIPK-directed conversion of PtdIns-5-P to PtdIns-4,5-P 2 . We report here that PIKfyve is responsible for PtdIns-5-P biosynthesis and that the osmoregulatory response in mammalian cells involves a robust change in PtdIns-5-P levels.

EXPERIMENTAL PROCEDURES
Cell Cultures-Conditions for maintaining mouse 3T3-L1 fibroblasts and their differentiation into insulin-sensitive adipocytes were previously described (14). Human embryonic kidney (HEK) 293 cells were maintained in DMEM, containing 10% fetal bovine serum, and the above antibiotics. Sf9 insect cells were maintained in a complete Grace's medium supplemented with 10% fetal bovine serum and 100 g/ml gentamycin.
Generation of Stable Cell Lines-Stably transfected doxycycline-inducible (Tet-On) cell lines expressing PIKfyve S WT (clone 9) or PIKfyve S K1831E (clone 5) were generated following the Tet-Off/Tet-On gene Expression System manual (Clontech). Briefly, PIKfyve S WT or PIKfyve S K1831E cDNA, released by XbaI-SalI from pBluescript IISKϩ (14), together with an HA-encoding adapter (flanked with BamHI and XbaI restriction sites) were cloned into the BamHI-SalI sites of the pTRE2hyg vector. The expected organization of the constructs was confirmed by restriction mapping. The new pTRE2hyg-based vectors linearized by SalI were used to transfect a HEK293 Tet-On cell line (Clontech) by LipofectAMINE as a transfection reagent. Transfected cells were selected by hygromycin treatment at 125 g/ml. Individual cell clonal lines were isolated by the help of cloning cylinders, propagated and then probed for a doxycycline-inducible expression of recombinant PIKfyve proteins by Western blot analysis with anti-HA polyclonal antibodies (a kind gift by Mike Czech) as described elsewhere (14,15). Experiments were carried out with cells seeded on collagen IV pre-coated plates to promote attachment.
Baculoviral and Adenoviral Infections-Baculovirus expressing PIKfyve S WT was generated and characterized previously (24). Sf9 cells seeded on 75-cm 2 flasks were infected with recombinant PIKfyve S WT baculovirus or control baculovirus and processed for [ 32 P]orthophosphate labeling 5 days post-infection. Recombinant adenoviruses, expressing HA-tagged PIKfyve s WT and GFP (AdPIKfyve), or GFP alone (AdEmpty) were generated by the AdEasy system (25) as described elsewhere (22). Viral stocks were purified by ultracentrifugation in two discontinuous CsCl 2 gradients and subsequent passage through a Nap 10 column (Sephadex G-25, Amersham Biosciences). Purified viral stocks were titrated, and the lowest dilution resulting in 100% infection of HEK293 cells 18 h postinfection (monitored by the GFP signals) was defined as a multiplicity of infection of 1. 3T3-L1 fibroblasts, 3T3-L1 adipocytes, or HEK293 cells were infected with AdPIKfyve or AdEmpty at a multiplicity of infection equal to 40, 40, and 1, respectively. These conditions yielded ϳ25-30% (3T3-L1 fibroblasts or adipocytes) and 100% infection efficiency (HEK293 cells) at day 3 and day 1 postinfection, respectively, when cells were processed for 32 P labeling.
Labeling of Cellular Phospholipids with [ 32 P]Orthophosphate and Lipid Extraction-Infected Sf9 cells were transferred from 75-cm 2 flasks into 50-ml Falcon tubes and then washed twice with a phosphatefree Grace's medium. Following 30 min of incubation, the medium was replaced with a phosphate-free Grace's medium supplemented with 0.5% bovine serum albumin and 1 mCi/ml [ 32 P]orthophosphate (PerkinElmer Life Sciences). After 2 h of incubation at 25°C, the cells were washed in PBS, and the lipids were extracted as described below. Basal or adenovirus-infected 3T3-L1 and HEK293 cells or stably transfected HEK293 cells were washed in phosphate-free DMEM and then labeled for 2.5 h at 37°C in phosphate-free DMEM supplemented with 0.5% bovine serum albumin, 2 mM sodium pyruvate, and 0.8 mCi/ml [ 32 P]orthophosphate as described previously (22,24). Cells were washed with ice-cold PBS, containing protease (1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, 5 g/ml leupeptin, 5 g/ml aprotinin, and 1 g/ml pepstatin) and phosphatase inhibitors (50 mM NaF, 10 mM sodium pyrophosphate, 25 mM sodium ␤-glycerophosphate, and 2 mM sodium metavanadate), and scraped with CH 3 OH, 1 M HCl (1:1) in the presence of 5 mM EDTA and 5 mM tetrabutylammonium sulfate. Extracted radiolabeled lipids were deacylated as described previously (15,22) and analyzed by HPLC. In some experiments, prior to lipid extraction, cells were subjected to treatment with different osmolytes.
Cell Treatment-Two days after seeding of 3T3-L1 fibroblasts (60-mm plates) or 8 -12 days following initiation of the differentiation program in these cells to acquire the adipocyte phenotype, and in some cases following [ 32 P]orthophosphate labeling, cells were incubated at 37°C for 10 min in DMEM (control stimulus), in DMEM, diluted to one-quarter strength with water (hypo-osmotic stimulus), or in DMEM containing either 0.6 M sorbitol or 0.2 M NaCl (hyperosmotic stimulus). Cells were placed on ice and washed once with PBS supplemented with protease and phosphatase inhibitors listed above. Cells were then scraped with CH 3 OH, 1 M HCl (1:1) containing 5 mM tetrabutylammonium sulfate and 5 mM EDTA, and the lipids were extracted as described above. Lipids were then dried under N 2 and stored at Ϫ77°C for PtdIns-5-P conversion assay or deacylated for HPLC analysis.
PtdIns-5-P Mass Assay-PtdIns-5-P conversion assay was performed as described by Morris et al. (18). Briefly, PIs were isolated from cell lipid extracts on neomycin-coated glass beads prepared according to procedures published previously (26) using glyceryl-coated controlled pore glass beads, 200 -400 mesh (Sigma), stored at 25°C in water/ methanol (1:1; v/v). Twenty five l of packed neomycin beads equilibrated at room temperature in 500 l of chloroform, methanol, 425 mM ammonium formate (5:10:2; v/v) in glass microtubes were incubated for 1 h with duplicate samples of extracted lipids redissolved in 500 l of chloroform, methanol, 425 mM ammonium formate (5:10:2; v/v). All samples contained 5 nmol of PtdIns as carrier. After two 500-l washes in chloroform, methanol, 425 mM ammonium formate (5:10:2; v/v), PIs were eluted twice for 20 min at 25°C with 250-l portions of fresh chloroform, methanol, 2 M aqueous triethylamine bicarbonate (2:6:3; v/v). PtdIns (20 nmol) was added as carrier to all samples before drying them in a SpeedVac at low/medium heat setting for 2-3 h and storing at Ϫ80°C or proceeding further for PtdInsP conversion assay using bacterially produced and purified recombinant His-tagged type II␤ PIPK (cDNA was a kind gift by Richard Anderson;Ref. 27). Samples were vortexed vigorously with 45 l of II␤ PIPK assay buffer consisting of 50 mM Tris-HCl, pH 7.4, 80 mM KCl, 10 mM magnesium acetate, 2 mM EGTA, and 0.01% sodium deoxycholate. After addition of 1 l of recombinant type II␤ PIPK (5 g) and preincubation at 30°C for 5 min, 5 l of 50 M [␥-32 P]ATP (5 Ci) was added, and incubation was continued for 1 h at 30°C. The reaction was stopped with 200 l of 1 N HCl and extracted with 160 l of chloroform/methanol (1:1; v/v). Lower layers were washed twice with 100 l of methanol, 1 N HCl (1:1; v/v). Fifty l of the lipid samples were spotted on an oxalate-treated and activated (30 min at 120°C) TLC plate (Whatman, PE SIL G, 250 m), which was then developed in 65:35 (v/v) n-propyl alcohol, 2 M acetic acid. Dried plates were exposed 8 -24 h with Kodak X-Omat autoradiography film to detect PtdIns-4,5-P 2 product derived from PtdIns-5-P. Standard lipids containing 1-5 pmol of PtdIns-5-P were processed similarly. Control samples with no enzyme and/or no lipids were run in each experiment. The PtdIns-4,5-P 2 radioactive spot was confirmed by HPLC following lipid extraction from the silica scrapings and deacylation.
HPLC Analysis and Data Quantitation-Deacylated 32 P-labeled lipids were redissolved in water and analyzed by HPLC on a Whatman 235 (length) ϫ 4.60-mm (inner diameter) column packed with 5-micron Partisphere SAX (H 2 PO 4 2Ϫ ) eluted with a shallow ammonium phosphate gradient at a flow rate of 1.0 ml/min as detailed elsewhere (15,17). P]PtdIns-3,5-P 2 (synthesized from PtdIns-5-P, Echelon Research Laboratories) and the enzymatic action of immunopurified p85 PI 3-kinase), respectively, were used as standards. Fractions were collected every 0.25 min, and their radioactivity was analyzed simultaneously for 3 H-and 32 P-labeled standard and products, respectively, with 2.0 ml of ScintiVerse liquid scintillation mixture on a liquid scintillation counter (Packard Instrument Co.). The radioactivity of the TLC-scraped [ 32 P]GroPIns-4,5-P 2 formed during PtdIns-5-P conversion assay was analyzed with an on-line flow scintillation analyzer (Packard, Radiomatic 525TR). Typically, the radioactivity of the peaks was quantified by area integration and is presented as percent of the radioactivity determined in corresponding control samples. Because the separation of PtdIns-5-P from the PtdIns-4-P tail was not complete, for [ 32 P]PtdIns-5-P quantitation, the counts under the PtdIns-5-P peak on top of PtdIns-4-P tail were skimmed and presented as percent of the radioactivity determined in corresponding control samples by the same approach.
PIKfyve Lipid Kinase Activity-PIKfyve lipid kinase activity in vitro was analyzed with PIKfyve immunoprecipitates, derived from lysates of osmotically treated cells, incubated with PtdIns and [␥-32 P]ATP for 15 min at 37°C followed by subsequent TLC analysis of the extracted radiolabeled products, as described previously (15,24).

Identification of Cell Types with HPLC-detectable Basal Accumulation of Radiolabeled PtdIns-5-P-To gain insight into
the PtdIns-5-P biosynthesis, we sought to identify cell types in which, similarly to NIH3T3 fibroblasts (17), this 5Ј-phosphorylated PtdIns metabolite is accumulated upon labeling of resting cells and is readily detected upon HPLC separation of cellular lipids. The cell types tested included CHO-T, HEK293, COS-7, 3T3-L1 fibroblasts, 3T3-L1 adipocytes, and insect Sf9 cells. HPLC analyses revealed that only the latter three but not the former three cell types (Ref. 22 and this study, see below) demonstrated detectable basal accumulation of radioactive Ptd-Ins-5-P. Thus, using a shallow ammonium phosphate gradient in the HPLC separation of deacylated lipids extracted from [ 32 P]orthophosphate-labeled 3T3-L1 adipocytes, we detected a clear-cut peak co-migrating with the [ 3 H]GroPIns-5-P internal standard and separated from the peak of PtdIns-4-P by 50 -70 s ( Fig. 1). Quantitation from eight independent 32 P labeling experiments indicated that accumulated basal levels of [ 32 P]Ptd-Ins-5-P in this cell type were quite substantial, corresponding to as much as 11.5 Ϯ 3% of the [ 32 P]PtdIns-4-P levels. When detected in other cell types by the same methodology, the amounts of PtdIns-5-P were found to be comparable with Ptd-Ins-3-P amounts (Ref. 17 and see below). Intriguingly, in 3T3-L1 adipocytes accumulated amounts of radiolabeled Ptd-Ins-5-P were found to profoundly exceed that of [ 32 P]PtdIns-3-P by 4 -6-fold (Fig. 1).
Similarly to 3T3-L1 adipocytes, HPLC analyses of deacylated radioactive lipids extracted from [ 32 P]orthophosphate-labeled insect Sf9 cells ( Fig. 2A) or 3T3-L1 fibroblasts (see Fig. 6A) revealed a 32 P-labeled peak whose elution time coincided with that of the [ 3 H]GroPIns-5-P internal standard. Quantitation from three independent 32 P labelings for either Sf9 cells or 3T3-L1 fibroblasts indicated that accumulated radiolabeled Ptd-Ins-5-P amounts represented 12.6 Ϯ 4.0 and 3.2 Ϯ 0.5% that of PtdIns-4-P, respectively. As opposed to 3T3-L1 adipocytes, the PtdIns-5-P levels in Sf9 cells and 3T3-L1 fibroblasts were comparable with the basal PtdIns-3-P amount in these cells (not shown). 32 P-PtdIns-5-P Amounts Are Increased upon Heterologous Expression of PIKfyve WT -By having identified cell types capable of accumulating HPLC-detectable basal PtdIns-5-P upon cell labeling with inorganic 32 P, we next examined the effect of expression of PIKfyve WT on levels of this lipid. Cells were transduced with PIKfyve WT or control viruses, and subsequent to labeling with [ 32 P]orthophosphate and extraction of the radioactive lipids, HPLC analyses were performed with deacylated products. As demonstrated in Fig. 2, Sf9 cell infection with baculovirus encoding PIKfyve WT or 3T3-L1 adipocyte transduction with adenovirus encoding PIKfyve WT resulted in accumulation of significantly higher amounts of [ 32 P]PtdIns-5-P peak that eluted identically with the [ 3 H]GroPIns-5-P internal standard. A similar increase of basal [ 32 P]GroPIns-5-P was also observed in 3T3-L1 fibroblasts, which expressed recombinant PIKfyve WT delivered by adenovirus-mediated gene transfer (not shown). Quantitation performed as detailed under "Experimental Procedures" and based on three independent experiments for each cell line revealed that this increase corresponded to 40 Ϯ 10, 23 Ϯ 8, and 20 Ϯ 5% over the basal radioactive PtdIns-5-P detected in control Sf9 cells, 3T3-L1 adipocytes, and 3T3-L1 fibroblasts, respectively, that were transduced with control viruses. Given the ϳ25-30% infection efficiency of the PIKfyve WT adenovirus in 3T3-L1 fibroblasts and 3T3-L1 adipocytes, as judged by the fluorescence detection of the GFP reporter, the calculated PtdIns-5-P increase due to expressed PIKfyve WT is apparently underestimated by at least 3-fold. In agreement with our previous in vivo and in vitro studies (15,22), expression of PIKfyve WT in the above cell types resulted in an increase of the PtdIns-3,5-P 2 radioactive levels as well, consistent with the ability of PIKfyve enzymatic activity to utilize PtdIns-3-P substrate. Together, these results demonstrate increased radioactive PtdIns-5-P amounts associated with PIKfyve WT expression and are consistent with the notion that intracellular PtdIns-5-P production is dependent on the PIKfyve enzyme.
Effect of PIKfyve WT Heterologous Expression in Cells with HPLC-undetectable Basal [ 32 P]PtdIns-5-P Levels-Whereas basal 3T3-L1 fibroblasts and adipocytes show well defined radioactive peaks migrating identically with the [ 3 H]GroPIns-5-P standard upon separation by HPLC, similar analysis in other mammalian cell types in culture such as COS-7, CHO-T and HEK293 cells failed to detect this lipid (22). Moreover, in contrast to 3T3-L1 fibroblasts and adipocytes, expression of high levels of PIKfyve WT in COS-7, CHO-T, and HEK293 cells (at ϳ100% cell infection efficiency) did not result in an accumulation of a quantifiable, clear-cut 32 P-radioactive peak corresponding to the elution time of GroPIns-5-P (22) (Fig. 3A). These results may indicate that PIKfyve lipid substrate preferences for PtdIns-3-P versus PtdIns vary among different mammalian cells. Alternatively, or additionally, technical limitations associated with a poor resolution of GroPIns-5-P from the descending edge of the bulky GroPIns-4-P peak upon HPLC and/or uneven specific activity of [ 32 P]ATP pools over the labeling period may have resulted in our inability to detect accumulationofbasal[ 32 P]PtdIns-5-PandtheexpectedPIKfyve WTdependent increase in the above cell lines. Consistent with these considerations are the data from the HPLC runs demonstrating a small shoulder with migration properties of [ 3 H]GroPIns-5-P in deacylated lipid samples extracted from 32 P-labeled HEK293 cells expressing high levels of PIKfyve versus control HEK293 cells (Fig. 3A). Therefore, to overcome the unfavorable HPLC detection of radiolabeled PtdIns-5-P in this cell type, we have used an alternative technique for detecting and quantifying intracellular PtdIns-5-P production (18). Resting upon the specificity of type II PIPK for position D-4 in PtdIns-5-P (17), this assay quantifies the TLC-resolved radioactive PtdIns-4,5-P 2 formed upon incubation of cell lipid extracts with [␥-32 P]ATP and recombinant type II PIPK (18). Because PtdIns-3-P but not PtdIns-4-P could be utilized to some extent as an additional substrate for type II PIPK, a subsequent HPLC head group analysis confirms the PtdIns-4,5-P 2 product formed. Recent application of this assay documented PtdIns-5-P production in human platelets, murine erythroleukemia, and Chlamydomonas cells (18,19,28). We have applied this assay to examine the basal and a plausible PIKfyve-dependent PtdIns-5-P synthesis in HEK293 cell lines that have been transfected to stably express PIKfyve WT (clone 9) upon doxycycline induction. Compared with endogenous PIKfyve, PIKfyve WT expression increased by ϳ7-fold 18 -48 h post-induction as documented by Western blot analysis with anti-PIKfyve antibodies (not shown and see Fig. 4A). PtdIns-5-P-conversion assay revealed substantial amounts of basal PtdIns-5-P in this cell type (Fig. 3B), which was calculated to be in the range of 10 -30 pmol/mg protein. More importantly, our data demonstrated that the induction of PIKfyve WT expression in HEK293 cells resulted in a significant rise of cellular PtdIns-5-P levels (Fig. 3B). This was evidenced by the observed 1.9 Ϯ 0.3-fold increase of the chromatographed PIP 2 radioactive spot (Fig. 3C), confirmed by HPLC head group analysis to be composed of PtdIns-4,5-P 2 but not PtdIns-3,4-P 2 (Fig. 3D). In contrast, doxycycline treatment of the parental HEK293 cells resulted in no change of the basal PtdIns-5-P to PtdIns-4,5-P 2 conversion, as judged by the detection of PtdIns-4,5-P 2 radioactive spots with similar intensity upon TLC (Fig. 3, B and C). Together, these results indicate, first, that whereas a [ 32 P]Ptd-Ins-5-P peak is undetectable by HPLC analysis in radiolabeled HEK293 cells, this 5-phosphorylated metabolite naturally occurs in this cell type, and, second, that similarly to 3T3-L1 fibroblasts and adipocytes, the route for PtdIns-5-P biosynthesis in HEK293 cells may entirely or partially rely on PIKfyve.
Dominant-negative Kinase-deficient PIKfyve K1831E Decreases PtdIns-5-P Production-If the contribution of PIKfyve enzymatic activity to the PtdIns-5-P cellular pool is physiologically significant, one would expect that expression of kinase-deficient mutants of PIKfyve with a dominant-negative effect (22) should arrest PtdIns-5-P production. To test this possibility we have generated a HEK293 stable cell line inducibly expressing dominant-negative kinase-deficient PIKfyve K1831E . Western blot analysis with anti-HA antibodies of the cell lysates derived from candidate PIKfyve K1831E -expressing clones selected one clone (clone 5) that demonstrated a significant induction of the mutant protein expression after 18 h of cell growth in the presence of doxycycline (Fig. 4A). The level of expression of PIKfyve K1831E in this clone was similar to that observed for PIKfyve WT -inducible expression of clone 9, exceeding by ϳ7fold the endogenous PIKfyve levels ( Fig. 4A and data not  shown). Importantly, PtdIns-5-P conversion assay revealed that the induction of PIKfyve K1831E expression was associated with a significant reduction of cellular PtdIns-5-P levels as manifested by the 60 Ϯ 4% decrease of the chromatographed PtdIns-4,5-P 2 radioactive spot versus non-induced cells (Fig. 4,  B and C). This result is consistent with the idea that in the cellular context a substantial portion of PtdIns-5-P is due to PIKfyve enzymatic activity.
PtdIns-5-P Dramatically Decreases upon Hypo-osmotic Shock-A condition shown to affect the turnover of PIs in yeast, plant, and mammalian cells is osmotic stress (19, 23, 29 -32). Related to PtdIns-5-P levels, a hyperosmotic increase has been reported in plants (19). To examine possible osmotically regulated changes in levels of PtdIns-5-P in mammalian cells, we have exposed 3T3-L1 fibroblasts or 3T3-L1 adipocytes for 10 min to DMEM containing non-permeant osmolytes in the form of non-ionic or ionic molecules (hyperosmotic treatment, 0.6 M sorbitol, or 0.2 M NaCl) or to DMEM made hypo-osmotic by 75% dilution with water. The lipids were then extracted and subjected to PtdIns-5-P conversion assay in the presence of [␥-32 P]ATP and recombinant type II PIPK. As illustrated in Fig. 5, treatment of both cell types with a hypo-osmotic solution induced a robust decrease of PtdIns-5-P intracellular production as evidenced by the 8.4 Ϯ 3.5-fold (3T3-L1 fibroblasts) and 5.5 Ϯ 1.5-fold reduction (3T3-L1 adipocytes) of the intensity of the PtdIns-4,5-P 2 spot (confirmed by HPLC head group analysis; data not shown). Treatment with osmotically active compounds, however, did not induce statistically significant changes in the cellular PtdIns-5-P levels in four experiments for each cell type (Fig. 5). Together these results demonstrate a robust reduction of cellular PtdIns-5-P levels upon 3T3-L1 cell exposure to a hypo-osmotic solution and suggest a role of Ptd-Ins-5-P as a regulatory intermediate in the osmotic response pathway in mammalian cells.
PtdIns-4-P Levels Remain Unchanged upon Hypo-osmotic Shock in 3T3-L1 Cells-Whereas the dramatic decrease of the type II PIP kinase-directed PtdIns-4,5-P 2 formation by hypoosmotic shock was clearly documented, it is still possible that this reduction affects the PtdIns-4-P rather than the PtdIns-5-P cellular levels. This is due to the fact that type II PIPK could still use PtdIns-4-P to make PtdIns-4,5-P 2 although at a rate 100 times less effective than PtdIns-5-P (18). Therefore, to confirm that PtdIns-4,5-P 2 was formed from PtdIns-5-P, but not from PtdIns-4-P, and to verify the conclusions for the robust decrease of the PtdIns-5-P cellular levels upon hypo-osmotic shock, we have examined the accumulation of radioactive PtdIns-4-P in these cells by HPLC. Following a 10-min hypoosmotic treatment of [ 32 P]orthophosphate-labeled 3T3-L1 fibroblasts, and a subsequent lipid extraction, deacylated samples were subjected to HPLC separation along with [ 3 H]GroPIns-5-P and [ 32 H]GroPIns-4-P internal standards. As demonstrated in Fig. 6, cell exposure to hypo-osmotic shock did not change the accumulated [ 32 P]GroPIns-4-P. Its amounts were similar to that detected in non-treated cells. Conversely, and consistent with the results obtained with PtdIns 5-P-conversion assay, hypo-osmotic shock decreased the accumulated [ 32 P]GroPIns-5-P amounts. In fact, hypo-osmotic shock induced a complete disappearance of the [ 32 P]PtdIns 5-peak seen in the control cells (Fig. 6A). These results demonstrate that whereas radiolabeled PtdIns-5-P amounts are decreased in 3T3-L1 cells, PtdIns-4-P remain unchanged upon hypo-osmotic shock, implying that the dramatic decrease of the in vitro PtdIns-4,5-P 2 formation by type II PIPK under this condition is due to a selective decrease in PtdIns-5-P levels.

FIG. 3. Basal PtdIns-5-P production and its increase upon PIKfyve WT expression in HEK293 cells determined by PtdInsP conversion assay.
A, HEK293 cells were infected with adenovirus encoding empty virus or PIKfyve WT as indicated. One day post-infection, cells were labeled with [ 32 P]orthophosphate for 2 h, and deacylated lipids were analyzed by HPLC along with [ 3 H]GroPIns-5-P and [ 3 H]GroPIns-4-P internal standards as detailed in the legend to Fig. 2. Shown is an overlay of HPLC runs demonstrating that although a PtdIns-5-P-peak was not detected, more 32 P radioactivity within the fractions with the elution time of [ 3 H]GroPIns-5-P was found in the AdPIKfyve WT sample. B and C, parental HEK293 cells or HEK293 cells stably expressing HA-PIKfyve WT (clone 9) were seeded (60-mm dishes) in the presence or absence of doxycycline as described under "Experimental Procedures." Twenty hours following induction of the protein expression cells were washed and the lipids extracted. PIs were isolated on neomycin-coated glass beads and subjected to enzymatic conversion for 1 h at 30°C in the presence of [␥-32 P]ATP with (lanes 2-5) or without type II PIP kinase (lane 1). 32 P-Labeled lipids were separated by TLC and visualized by autoradiography. Shown is a representative autoradiogram (B) and quantitation from five independent experiments in duplicates presented as a percent of PtdIns-4,5-P 2 formed in PIKfyve WT -stable cells with no induction (C). D, [ 32 P]PtdIns-4,5-P 2 spot shown in B was scraped from the TLC plate and, following lipid extraction and deacylation, was subjected to HPLC analysis together with [ 3 H]GroPIns-4,5-P 2 standard (data not shown). Eluted is a single PIP 2 peak with migration properties of GroPIns-4,5-P 2 (arrowhead). tial decrease in accumulated amounts of the radioactive PIP 2 . Thus, PtdIns-3,5-P 2 and PtdIns-4,5-P 2 dropped to 36 and 62.5%, respectively, of the corresponding control values (Fig.  6B). Together, these results indicate that hypo-osmotic shock induces differential changes in the PI-radiolabeled amounts in fibroblasts.
Increase in Accumulated [ 32 P]PtdIns-4,5-P 2 Amounts upon Hyperosmotic Shock-Whereas a striking hypo-osmoticdependent reduction of PtdIns-5-P cellular levels was possible to be documented in Figs. 5 and 6, treatment of 3T3-L1 cells with osmolytes failed to document the expected increase in the PtdIns-5-P levels. This result suggests the possibility that Ptd-Ins-5-P may undergo a rapid turnover under hyperosmotic shock conditions. Because one pathway of PtdIns-5-P metabolism involves its intracellular conversion to PtdIns-4,5-P 2 by type II PIPK, we have examined the amounts of accumulated radiolabeled PtdIns-4,5-P 2 by HPLC, following a 10-min hyperosmotic treatment of 32 P-labeled 3T3-L1 fibroblasts. HPLC analysis of extracted and deacylated lipids demonstrated that both sorbitol (Fig. 7) and NaCl (not shown) resulted in accumulation of higher amounts into the 32 P-radiolabeled peak that eluted identically with [ 3 H]GroPIns-4,5-P 2 . This increase corresponded to 130 -170% over the control, as revealed by the combined results from three independent experiments for each condition. These results indicate a higher PtdIns-4,5-P 2 synthesis under hyperosmotic conditions that could have consumed PtdIns-5-P.
PIKfyve Activity in Vitro Remains Unchanged Upon Osmotic Shock-The results presented above demonstrating a PIKfyvedependent route in PtdIns-5-P biosynthesis and an osmoregulatory control of the intracellular PtdIns-5-P synthesis suggest the hypothesis that PIKfyve enzymatic activity may be affected by changes in osmolarity. To test this, we have examined PIKfyve lipid kinase activity in vitro by subjecting PIKfyve immunoprecipitates, derived from osmotically stimulated or control 3T3-L1 fibroblasts and adipocytes, to incubation in the presence of PtdIns and [␥-32 P]ATP and a subsequent TLC separation of the extracted lipids. The data demonstrated no changes in the in vitro formed PtdIns-5-P and PtdIns-3,5-P 2 in response to either hypo-osmotic or ionic/non-ionic hyperosmotic shock (data not shown). These results indicate that if the intracellular PIKfyve is subject to osmotic regulation, these changes do not result in a sustained alteration in the PIKfyve lipid kinase activity. DISCUSSION Since the discovery of PtdIns-5-P as a substrate for type II PIPK and its natural occurrence in NIH3T3 fibroblasts (17), studies related to this PtdIns metabolite have been surprisingly sparse. As a result, we know almost nothing about the intracellular synthesis, roles, and regulation of PtdIns-5-P. This lack of information is mainly due to the fact that PtdIns-5-P is only a minor fraction of PIs whose HPLC elution characteristics are very similar to that of the abundant PtdIns-4-P that comprises roughly 30 -40% of the PIs. Therefore, the 5-phosphorylated PtdIns metabolite was the last among the seven PIs to be identified and only after the optimization of elution conditions in the HPLC run to allow separation of PtdIns-5-P from PtdIns-4-P (17). When detected, PtdIns-5-P FIG. 4. Induced expression of dominant-negative PIKfyve K1831E decreases PtdIns-5-P production in HEK293 cells as determined by PtdInsP-conversion assay. HEK293 cells stably expressing HA-PIKfyve WT (clone 9) or HA-PIKfyve K1831E (clone 5) were seeded (60-mm dishes) in the presence or absence of doxycycline as described under "Experimental Procedures." Twenty hours following induction of the protein expression cells were washed, and lysates were collected and subjected to Western blotting (WB) with anti-HA antibodies (A). Alternatively, PIs were isolated from extracted lipids on neomycin-coated glass beads and subjected to enzymatic conversion for 1 h at 30°C in the presence of [␥-32 P]ATP and type II PIP kinase. 32 P-Labeled lipids were separated by TLC and visualized by autoradiography (B). Shown are a chemiluminescence detection of a representative immunoblot (A), an autoradiogram of a representative TLC plate (B), and quantitation of the TLC autoradiograms from three independent experiments presented as a percent of PtdIns-4,5-P 2 formed in PIKfyve K1831E stable cells with no induction (C).

FIG. 5. Decrease of PtdIns-5-P synthesis upon hypo-osmotic shock of 3T3-L1 cells. 3T3-L1 fibroblasts (A) or differentiated 3T3-L1
adipocytes (B) were incubated in DMEM alone (control), DMEM made hypo-osmotic by 75% dilution with water (Hypo), DMEM supplemented with 200 mM NaCl, or DMEM supplemented with 600 mM sorbitol for 10 min at 37°C as indicated. Cells were then washed with PBS, and the lipids were extracted as described under "Experimental Procedures." PIs were isolated on neomycin-coated glass beads and subjected to conversion by type II PIP kinase for 1 h at 30°C in the presence of [␥-32 P]ATP. 32 P-Labeled products were separated by TLC and visualized by autoradiography. Shown is a representative autoradiogram (A and B) and quantitation from five (C) and three (D) independent experiments for each cell type presented as a percent of the PtdIns-4,5-P 2 formed from PIP of control cells.
was typically seen as a small peak within the trailing edge of PtdIns-4-P long before the latter reaches the base line (Refs. 13,15,17, and this study). Apparently, low amounts of PtdIns-5-P are likely to be hidden within the PtdIns-4-P tail. Recently, an alternative approach for PtdIns-5-P intracellular determination has been based on the substrate preferences of type II PIPK for PtdIns-5-P versus PtdIns-4-P and the effective detection of PtdIns-4,5-P 2 by HPLC (18). The aim of this study was first to analyze whether the PIKfyve enzyme was responsible for PtdIns-5-P intracellular synthesis and second to examine plausible osmotic shock-related dynamics in PtdIns-5-P intracellular production. To this end, we have applied both available techniques for detection of intracellular PtdIns-5-P, i.e. an HPLC head group separation of radioactive PtdIns-5-P accumulated during cell labeling with [ 32 P]orthophosphate and the type II PIPK-directed conversion of PtdIns-5-P to PtdIns-4,5-P 2 . Our studies unequivocally demonstrate that higher or lower PtdIns-5-P levels are associated with cell expression of recombinant PIKfyve WT or kinase-deficient PIKfyve K1831E , respectively, providing the first clue for the biosynthetic route of PtdIns-5-P. Furthermore, a robust decrease in PtdIns-5-P cellular production was documented upon hypo-osmotic shock in mouse 3T3-L1 cells implicating PtdIns-5-P as a regulatory intermediate in the osmotic response pathway in mammalian cells.
The role of PIKfyve in PtdIns-5-P cellular production has been first suggested by the data demonstrating a powerful ability of a purified PIKfyve protein to support the conversion of PtdIns to PtdIns-5-P (14,15). Subsequent in vivo studies, however, both in yeast and several mammalian cell types heterologously expressing high levels of PIKfyve WT , failed to detect intracellular PtdIns-5-P production (21,22). Because an increase of PtdIns-3,5-P 2 was possible to be documented under these conditions, PIKfyve has been implicated in PtdIns-3,5-P 2 but not in PtdIns-5-P biosynthesis. However, the cell types tested so far all showed undetectable accumulation of radioactive PtdIns-5-P at resting conditions in in vivo labeling studies coupled to HPLC head group analysis (21,22). This implies that limitations of technical or other nature may have been preventing PtdIns-5-P production and/or detection. Among those, one could envision an upstream regulator(s) of PIKfyve PtdIns-5-P-synthesizing activity, which becomes a limiting factor particularly under conditions of PIKfyve overexpression in certain cells. Recent studies in yeast identifying Vac14 protein as an upstream activator of the yeast PIKfyve ortholog Fab1p lipid kinase (33,34) indicate that this assumption may be correct. Another group of limiting factors may include a rapid FIG. 6. Whereas the [ 32 P]PtdIns-5-P peak disappears, the [ 32 P]PtdIns-4-P amounts remain unchanged upon hypo-osmotic shock of radiolabeled 3T3-L1 fibroblasts. 3T3-L1 fibroblasts (60-mm dishes) were labeled with [ 32 P]orthophosphate for 2 h in phosphate/serum-free DMEM. The medium was then replaced with DMEM alone (control) or DMEM made hypo-osmotic by 75% dilution with water (Hypo) as indicated. Cells were then washed with PBS and lipids extracted as described under "Experimental Procedures." Deacylated lipids co-injected with the indicated [ 3 H]GroPIns-5-P, [ 3 H]GroPIns-4-P, and [ 3 H]GroPIns-4,5-P 2 standards were analyzed by HPLC. The elution times of GroPIns-3,4-P 2 and GroPIns-3,5-P 2 were determined from a parallel HPLC run with 32 P-labeled standards. Fractions, collected every 0.25 min, were monitored for 3 H and 32 P radioactivity by liquid scintillation counting. Shown is a representative HPLC elution profile of the region covering GroPIns-5-P and GroPIns-4-P elution times out of three independent experiments with similar results (A) and the quantitation of the indicated radioactive peaks presented as percent of control values (B).

FIG. 7. Increase in accumulated [ 32 P]PtdIns-4,5-P 2 amounts upon hyperosmotic shock of radiolabeled 3T3-L1 fibroblasts.
3T3-L1 fibroblasts (60-mm dishes) were labeled with [ 32 P]orthophosphate for 2 h in phosphate/serum-free DMEM. The medium was then replaced with DMEM (control) or DMEM containing 0.6 M sorbitol (Hyper). Cells were then washed with PBS and lipids extracted as described under "Experimental Procedures." Deacylated lipids co-injected with [ 3 H]GroPIns-4,5-P 2 internal standard were analyzed by HPLC. Fractions, collected every 0.25 min, were monitored for 3 H (peak indicated by arrow) and 32 P radioactivity by liquid scintillation counting. Elution times of [ 32 P]GroPIns-3,5-P 2 and [ 32 P]GroPIns-3,4-P 2 standards (arrows) were determined from a parallel HPLC run. Shown is a representative HPLC elution profile of PIP 2 s out of two independent experiments with similar results.
PtdIns-5-P turnover coupled with uneven specific activity of the [ 32 P]ATP pool during cell labeling and/or unfavorable Ptd-Ins-5-P versus PtdIns-4-P detection under HPLC analysis, discussed above. Overcoming these restrictions was made possible in the present study, at least in part, by applying two strategies. First, usage of cell types in which the accumulation of radioactive PtdIns-5-P peak upon HPLC head group analysis was evident at resting conditions, and second, usage of the PtdIns-5-P enzymatic conversion in cell types with undetectable levels of PtdIns-5-P under HPLC analysis. Both strategies unequivocally demonstrated an increase of PtdIns-5-P production associated with expression of high levels of PIKfyve WT in both mammalian and insect cells, thus establishing the biosynthetic pathway in PtdIns-5-P metabolism. That this pathway is likely physiologically relevant is further substantiated in the present study by the data demonstrating a marked decrease of PtdIns-5-P levels by inducing the expression of the dominantnegative kinase-deficient PIKfyve K1831E mutant. Because PIKfyve is an evolutionarily conserved protein found also in plants (16) where a substantial basal PtdIns-5-P production has been recently documented (up to 18% of the PtdIns-4-P amounts; Ref. 19), we suggest a biosynthetic mechanism that operates due to plant PIKfyve in these cells. It remains to be re-examined by PtdIns-5-P conversion assay whether the reported lack of PtdIns-5-P production in yeast cells (20,21) is associated with technicalities in detection or truly reflects a more restricted substrate specificity of Fab1p toward 3-phosphorylated PtdIns. In any case, Fab1p is able to convert PtdIns into Ptd-Ins-5-P in vitro (21).
One important aspect of the present studies is the observation that the intracellular PIKfyve lipid kinase activity synthesizes two lipid products. Thus, we reproducibly detected a simultaneous increase of both PtdIns-5-P and PtdIns-3,5-P 2 production upon transduction of 32 P-labeled 3T3-L1 adipocytes with adenovirus encoding PIKfyve WT (22) (Fig. 2B). This was also evident in the HEK293 stable cell line, where we documented higher PtdIns-5-P levels by mass assay (Fig. 3B) and an increase of the accumulated [ 32 P]PtdIns-3,5-P 2 by HPLC analysis of in vivo labeled cells induced to express PIKfyve WT (Ref. 22 and data not shown). This dual specificity implies that distinct intracellular signals may differentially regulate PIKfyve lipid substrate preferences. Although direct experimental evidence is currently unavailable, the assumption that such differential regulation may operate in the cellular context is supported by our recent observation documenting altered lipid substrate preferences for either PtdIns-3-P or PtdIns upon substitution of Lys-2000 or Lys-1999 in PIKfyve, respectively (35).
A key observation in the present study is the demonstration that PtdIns-5-P synthesis is profoundly decreased upon hypoosmotic shock in 3T3-L1 fibroblasts and adipocytes. Because these cells showed no hyperosmotically dependent increase in PtdIns-5-P, but a substantially augmented PtdIns-4,5-P 2 production, we suspect a concomitant activation of type II PIPK that should be tested in future studies. Whereas reports related to osmotically regulated PtdIns-5-P synthesis in mammalian cells are, to the best of our knowledge, unavailable, a recent study (19) in Chlamydomonas cells has demonstrated a rapid and sustained increase in PtdIns-5-P upon treatment with 300 mM NaCl for a period of 15 min. Intriguingly, in the same study, the PtdIns-4,5-P 2 levels were back to normal at the 15-min time point (19), suggesting a less efficient type II PIPKdirected conversion of PtdIns-5-P in this cell system. Together, these data are consistent with the concept that acute changes in PtdIns-5-P levels are involved in the cellular osmoregulatory pathway, and this mechanism may be evolutionarily conserved.
It is tempting to speculate that PtdIns-5-P, like other PIs, has a direct signaling potential and is able to selectively recruit specific yet-to-be identified protein intermediates with relevance to osmoprotective cell responses. Noteworthy, osmotic shock of 3T3-L1 adipocytes is shown to stimulate the glucose transport activity, translocation of GLUT4, and membrane ruffling. These cellular effects resemble those elicited by insulin, and studies indicate that these two stimuli share both common and unique signaling pathway mechanisms (36 -38). It would be interesting in future studies to examine a plausible regulatory role of PtdIns-5-P in the molecular mechanism of osmotic shock-and insulin-induced responses in this cell type. Moreover, dominant-negative kinase-deficient PIKfyve K1831E has been shown recently to inhibit insulin-regulated GLUT4 translocation in 3T3-L1 adipocytes (39).
Our data demonstrating PIKfyve-dependent intracellular PtdIns-5-P production would imply a role for PIKfyve in osmotically regulating PtdIns-5-P levels. Although this assumption may be correct, the complex enzymology of PIs, including that of PtdIns-5-P, precludes us from making such a conclusion. Moreover, there is the unfavorable fact that we were unable to detect sustained osmotically dependent regulation in the in vitro PIKfyve lipid kinase activity. Theoretically, levels of Ptd-Ins-5-P could decrease due to osmotically dependent inhibition of PIKfyve, PtdIns-4,5-P 2 4-phosphatase, and PtdIns-3,5-P 2 3-phosphatase activities or activation of type II PIPK and Pt-dIns-5-P phosphatase. However, no such phosphatases have been clearly described thus far (6 -8). Also, it should be emphasized that hypo-osmotic shock in 3T3-L1 cells caused a substantial drop in accumulated [ 32 P]PtdIns-3,5-P 2 and Ptd-Ins-4,5-P 2 amounts corresponding to 2.8-and 1.6-fold, respectively. These results and considerations eliminate the action of the first two hypothetical candidate phosphatases, making PIKfyve a more likely candidate for osmoregulation. Thus, the simultaneous inhibition of PtdIns-5-P and PtdIns-3,5-P 2 production both associated with the dual lipid kinase activity of PIKfyve is consistent with the idea for a hypo-osmotic downregulation of PIKfyve activity in vivo, although it remains to be experimentally proven in future studies.
In conclusion, we demonstrate here that PIKfyve lipid kinase activity is responsible for biosynthesis of PtdIns-5-P, a lipid whose natural occurrence was identified several years ago but whose kinase remained elusive. The turnover of PtdIns-5-P likely involves further conversion to PtdIns-4,5-P 2 by type II PIPK, supporting the previously suggested alternative pathway of PtdIns-4,5-P 2 biosynthesis. It is also possible that Ptd-Ins-5-P is a precursor for an alternative production of PtdIns-3,5-P 2 that would require the specificity of a separate PI 3-kinase. The fact that hypo-osmotic shock of 3T3-L1 cells induces a robust decrease of PtdIns-5-P levels indicates that PtdIns-5-P may be a regulatory intermediate in osmoprotective response pathways in mammalian cells.