Type I Phosphatidylinositol-4-phosphate 5-Kinases Synthesize the Novel Lipids Phosphatidylinositol 3,5-Bisphosphate and Phosphatidylinositol 5-Phosphate*

Inositol phospholipids regulate a variety of cellular processes including proliferation, survival, vesicular trafficking, and cytoskeletal organization. Recently, two novel phosphoinositides, phosphatidylinositol-3,5-bisphosphate (PtdIns-3,5-P2) and phosphatidylinositol- 5-phosphate (PtdIns-5-P), have been shown to exist in cells. PtdIns-3,5-P2, which is regulated by osmotic stress, appears to be synthesized by phosphorylation of PtdIns-3-P at the D-5 position. No evidence yet exists for how PtdIns-5-P is produced in cells. Understanding the regulation of synthesis of these molecules will be important for identifying their function in cellular signaling. To determine the pathway by which PtdIns-3,5-P2 and Ptd-Ins-5-P might be synthesized, we tested the ability of the recently cloned type I PtdIns-4-P 5-kinases (PIP5Ks) α and β to phosphorylate PtdIns-3-P and PtdIns at the D-5 position of the inositol ring. We found that the type I PIP5Ks phosphorylate PtdIns-3-P to form PtdIns-3,5-P2. The identity of the PtdIns-3,5-P2 product was determined by anion exchange high performance liquid chromatography analysis and periodate treatment. PtdIns-3,4-P2 and PtdIns-3,4,5-P3 were also produced from PtdIns-3-P phosphorylation by both isoforms. When expressed in mammalian cells, PIP5K Iα and PIP5K Iβ differed in their ability to synthesize PtdIns-3,5-P2 relative to PtdIns-3,4-P2. We also found that the type I PIP5Ks phosphorylate PtdIns to produce PtdIns-5-P and phosphorylate PtdIns-3,4-P2 to produce PtdIns-3,4,5-P3. Our findings suggest that type I PIP5Ks synthesize the novel phospholipids PtdIns-3,5-P2 and PtdIns-5-P. The ability of PIP5Ks to produce multiple signaling molecules indicates that they may participate in a variety of cellular processes.

of PI3K family members have now been identified which can catalyze the in vitro phosphorylation of PtdIns, PtdIns-4-P, and/or PtdIns-4,5-P 2 on the D-3 position of the inositol ring (25). Kinetic studies in 32 PO 4 -labeled cells suggest that PtdIns-3,4-P 2 and PtdInsP 3 are synthesized by the phosphorylation of PtdIns-4,5-P 2 on the D-3 position by PI3K to produce PtdIns-3,4,5-P 3 , followed by a dephosphorylation of PtdIns-3,4,5-P 3 on the D-5 position by a phosphatase to form PtdIns-3,4-P 2 (26). Studies in platelets provide evidence for an alternative synthetic pathway for D-3 phosphoinositide in which PtdIns is first phosphorylated on the D-3 position to produce PtdIns-3-P, followed by the phosphorylation of the D-4 position to form Ptd-Ins-3,4-P 2 (27)(28)(29).
A newly discovered phosphoinositide in the PI3K pathway has recently been described. Analysis of phosphoinositides present in smooth muscle cells revealed an inositol containing lipid with chromatographic properties similar to PtdIns-3,4-P 2 , which we proposed to be PtdIns-3,5-P 2 (30). Whiteford et al. (31) recently verified that PtdIns-3,5-P 2 is present in fibroblasts. Dove et al. (32) also found PtdIns-3,5-P 2 in Saccharomyces cerevisiae and showed that the level of this lipid is regulated by osmotic strength changes (32). The synthesis of PtdIns-3,5-P 2 is blocked by the PI3K inhibitor wortmannin (31). Analysis of the specific activity of the D-3 and D-5 phosphates indicates that PtdIns-3,5-P 2 is synthesized primarily by phosphorylation of PtdIns-3-P at the D-5 position (31,32). The specific activity data and the fact that PtdInsP 3 is absent in cells under conditions where PtdIns-3,5-P 2 is found argue that a phosphatase is unlikely to be involved in PtdIns-3,5-P 2 synthesis.
The only known kinases that phosphorylate the D-5 position of the inositol ring are the type I PIP5Ks. As mentioned previously, the type II PIPKs have recently been demonstrated to phosphorylate the D-4 position of the inositol ring (16). While the function of PtdIns-3,5-P 2 in the cell is unknown, we were interested in the pathway by which it is synthesized and tested whether type I PIP5Ks can phosphorylate the D-5 position of PtdIns-3-P to form PtdIns-3,5-P 2 . We find that type I PIP5Ks do phosphorylate PtdIns-3-P at the D-5 position. We also find that these kinases can produce PtdIns-5-P from PtdIns and can synthesize PtdInsP 3 from either PtdIns-3-P or PtdIns-3,4-P 2 .

EXPERIMENTAL PROCEDURES
Materials-PtdIns and PtdIns-4-P were obtained from Avanti Polar Lipids Inc. and Sigma, respectively. Synthetic dipalmitoyl PtdIns-3-P, PtdIns-3,4-P 2 , and PtdIns-3,5-P 2 were synthesized as described (34 Plasmids-N-terminal HA-tagged murine cDNAs of type I PIP5K ␣ and ␤ in pBluescript were generated as described previously (5). The cDNAs of the PIP5Ks lacking the HA-tag were subcloned into the SalI and NotI sites of the bacterial expression vector pGEX4T. HA-tagged type I PIP5K cDNAs were also subcloned into the ClaI and NotI sites of the mammalian expression vector pEBB (provided by B. J. Mayer). The mutations PIPKI␣ D227A and PIPKI␤ D268A were generated using the CLONTECH Transformer Site-directed Mutagenesis Kit.
Expression and Purification of Recombinant PIP5Ks Isoforms-Glutathione S-transferase fusion proteins of the type I PIP5Ks were expressed in Escherichia coli and purified with glutathione-Sepharose beads as described previously (15). Proteins were eluted from glutathione-Sepharose beads with 50 mM Tris buffer, pH 8.0, containing 20 mM reduced glutathione and then concentrated in a Centricon filter after several washes with buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM MgCl 2 , 5 mM dithiothreitol). Glycerol was added to the proteins before storing at Ϫ80°C.
Cell Culture and Transfections-293 E1A cells were maintained in Dulbecco's modified medium containing 10% heat inactivated fetal calf serum. Cells were transfected with LipofectAMINE using 4 g of HAtagged PIP5Ks per 10-cm plate. Cells were harvested 18 h after transfection. Immunoprecipitation of PIP5Ks from 293 E1A Cells-Cells were rinsed with phosphate-buffered saline and then lysed in 600 l of lysis buffer (50 mM Tris, pH 7.5, 50 mM NaCl, 5 mM MgCl 2 , 1% Nonidet P-40, 10% glycerol, 1 mM dithiothreitol, 4 g/ml each leupeptin and pepstatin, and 200 M AEBSF). The clarified lysates were incubated with 1.5 g of an anti-HA antibody (12CA5 from Boehringer Mannheim) and protein A-Sepharose beads for 3 h at 4°C. The beads were then washed twice with lysis buffer and twice with TNM (50 mM Tris, pH 7.5, 50 mM NaCl, and 5 mM MgCl 2 ).
Periodate Oxidation-Periodate oxidation was performed as described with the following modifications (36). Deacylated lipids were incubated in the dark at 25°C with 10 or 100 mM periodic acid, pH 4.5, for 30 min or 36 h, respectively. The remaining oxidizing reagent was removed by adding 500 mM ethylene glycol and incubating the reaction in the dark for 30 min. 2% 1,1-dimethylhydrazine, pH 4.5, was then added to a final concentration of 1% and the reaction was allowed to proceed for 4 h at 25°C. The mixture was then purified by ion exchange using a Dowex 50W-X8 cation-exchange resin (20 -50 mesh, acidic form), dried, and applied to an HPLC column. PtdIns-3,4-P 2 and Pt-dIns-3,4,5-P 3 controls used in this experiment were prepared using recombinant Sf9 cell-expressed PI3K.

PIP5Ks Phosphorylate PtdIns, PtdIns-3-P, and PtdIns-3,4-P 2 -
The recent description of the presence of PtdIns-3,5-P 2 in fibroblasts rekindled our interest in this phosphoinositide and how it is produced. Since it was recently shown that type II PIPK phosphorylates the D-4 position of PtdIns-5-P (16), the synthesis of PtdIns-3,5-P 2 must either be catalyzed by a type I PIP5K, a phosphatase, or a new enzyme. To further define the substrates and products of the type I PIP5Ks, we provided PtdIns, PtdIns-3-P, PtdIns-4-P, PtdIns-3,4-P 2 , and PtdIns-3,5-P 2 as substrates and analyzed the products. As our enzyme source, we used recombinant PIP5Ks I␣ and PIP5K I␤ (according to the nomenclature proposed by Ishihara et al. (5)) that were produced as glutathione S-transferase fusion proteins in bacteria. The products of the phosphorylation reactions were analyzed by TLC using a system that separates PtdInsP, Ptd-InsP 2 , and PtdInsP 3 , and visualized using a Molecular Analyst. As expected, the type I PIP5Ks converted PtdIns-4-P to Ptd-InsP 2 ( Fig. 1) (4, 5). Surprisingly, the enzymes were also capable of synthesizing other lipid products. Both PIP5K I␣ and PIP5K I␤ phosphorylated PtdIns-3,4-P 2 to produce Ptd-InsP 3 and phosphorylated PtdIns-3-P to produce PtdInsP 2 and Ptd-InsP 3 (Fig. 1). PtdIns was also converted to PtdInsP and Ptd-InsP 2 by the type I PIP5Ks (although bacterially expressed PIP5K I␤ was more active than PIP5K I␣ at phosphorylating PtdIns). In contrast, PtdIns-3,5-P 2 was a poor substrate for both enzymes (1/5 as good as PtdIns-3,4-P 2 ). The PtdInsP 2 produced in the reactions using PtdIns-3,4-P 2 and PtdIns-3,5-P 2 as substrate most likely resulted from a contaminant in either the synthetic lipid substrates (which were judged to be greater than 99% pure based on 1 H and 31 P NMR) or in the carrier lipid phosphoserine. The kinase-dead PIP5K mutants, PIP5K I␣ D227A and PIP5K I␤ D268A, did not catalyze the phosphorylation of any lipids, as expected (data not shown).
HPLC Analysis of the Products from the Phosphorylation of PtdIns-3-P and PtdIns-3,4-P 2 -To determine the identity of the products, they were deacylated and analyzed by anion ex- change chromatography using HPLC. As expected from previous studies (4,5), the PtdInsP 2 produced using PtdIns-4-P as a substrate for either enzyme was exclusively PtdIns-4,5-P 2 (not shown). The product of PtdIns-3,4-P 2 phosphorylation by either enzyme was confirmed to be PtdIns-3,4,5-P 3 , in agreement with a recent study by Zhang et al. (Fig. 2B) (33). However, in contrast to the findings of Zhang et al. (33), the deacylated PtdInsP 2 produced by phosphorylation of synthetic PtdIns-3-P with either type I isoenzyme was resolved into two peaks by HPLC analysis (75.5 and 77 min, Fig. 2A). The peak at 77 min comigrated with the glycerophosphorylinositol 3,4-bisphosphate (GroPIns-3,4-P 2 ) standard while the peak at 75.5 min FIG. 4. Resistance to periodate oxidation confirms that PtdIns-3,5-P 2 is the major product of the phosphorylation of PtdIns-3-P by type I PIP5Ks. The [ 32 P]PtdInsP 2 (7,572 disintegrations/min) and [ 32 P]PtdInsP 3 (2,121 disintegrations/min) products from the phosphorylation of PtdIns-3-P by bacterially expressed PIP5K I␤ were deacylated, and then exposed to periodate for a short time (30 min) to deglycerate (A), or a long time (36 h) to deglycerate and oxidize as described under "Experimental Procedures" (B). The resulting products were then analyzed by HPLC. migrated at the position expected for GroPIns-3,5-P 2 (31, 32). Zhang et al. (33) did not detect a second peak corresponding to GroPIns-3,5-P 2 when analyzing the reaction products of the phosphorylation of PtdIns-3-P by type I PIP5Ks. The best explanation for this discrepancy is that the anion exchange HPLC analysis done by Zhang et al. (33) did not separate GroPIns-3,4-P 2 from GroPIns-3,5-P 2 . We also detected two additional peaks in the reaction products of PtdIns-3-P phosphorylation by type I PIP5Ks (Fig. 2A). The peak at 98 min eluted at the known position for GroPIns-3,4,5-P 3 (Fig. 2A). The peak at 95 min (ϫ) was not identified.
To demonstrate that the PIP5K I␤ products shown in Fig. 2A were synthesized from the phosphorylation of PtdIns-3-P and not a contaminating lipid, we analyzed the ability of the type I PIP5Ks to phosphorylate [ 32 P]PtdIns-3-P which was enzymaticaly produced by PI3K. Bacterially expressed PIP5K I␤ was incubated with [ 32 P]PtdIns-3-P and unlabeled ATP for 10 min at room temperature. The lipid products were then extracted and analyzed. We found that PIP5K I␤ converted [ 32 P]PtdIns-3-P to [ 32 P]PtdIns-3,5-P 2 , [ 32 P]PtdIns-3,4-P 2 , and [ 32 P]Ptd-InsP 3 (Fig. 3).
Type I PIP5Ks Phosphorylate PtdIns-3-P to Produce PtdIns-3,5-P 2 -Although the major PtdInsP 2 product from the phosphorylation of PtdIns-3-P by type I PIP5Ks appeared to be PtdIns-3,5-P 2 based on its migration position, we further investigated its identity using periodate. Since this lipid was not PtdIns-3,4-P 2 or PtdIns-4,5-P 2 , it could only be PtdIns-3,5-P 2 , PtdIns-2,3-P 2 , or PtdIns-3,6-P 2 . The latter two products have never been described as existing in vivo, but can be distinguished from PtdIns-3,5-P 2 by their sensitivity to periodate treatment. Of this group, only PtdIns-3,5-P 2 lacks a viscinal diol for periodate attack of the inositol ring and is therefore resistant to cleavage. We treated the products of PtdIns-3-P phosphorylation with periodate. GroPIns-3,4-P 2 (which is sensitive to periodate) and GroPIns-3,4,5-P 3 (which is resistant to periodate) were also treated as controls. The reaction products were analyzed by anion exchange chromatography using an HPLC.
Short periodate treatment (30 min) resulted in the conversion of the GroPIns-3,4-P 2 and GroPIns-3,5-P 2 products to InsP 3 and the GroPInsP 3 product to InsP 4 due to cleavage of the glycerol moiety (Fig. 4A). Since PtdIns-3,4,5-P 3 lacks a viscinal diol and is therefore resistant to periodate cleavage (Fig. 4, E and F), we used the InsP 4 derived from the PtdInsP 3 present in our sample as an internal control for yield after prolonged periodate treatment (36 h). Prolonged periodate treatment resulted in the loss of 50% of InsP 4 (Fig. 4B). Similar treatment of Ins-1,3,4-P 3 resulted in almost complete loss of radiolabeled material (3% remaining) at the position of Ins-1,3,4-P 3 (Fig. 4, C and D). However, under the same conditions, approximately 30% of the GroPInsP 2 product of the type I PIP5K ␤ was recovered at the position expected for Ins-1,3,5-P 3 (Fig. 4B). When corrected for the reaction yield based on the treatment of PtdInsP 3 , 60% of the InsP 3 product was resistant to periodate. Similar results were obtained with the PtdIns-3-P phosphorylation products of PIP5K I␣. These results, in agreement with those in Figs. 2A and 3, indicate that the majority of the PtdInsP 2 produced by the Type I PIP5Ks is PtdIns-3,5-P 2 .
Synthesis of either PtdIns-3,4-P 2 or PtdIns-3,5-P 2 in our in vitro assays could have resulted from the dephosphorylation of PtdIns-3,4,5-P 3 by a contaminating phosphatase. To investigate this possibility we provided [ 32 P]PtdIns-3,4,5-P 3 labeled at the D-3 position by PI3K to both the ␣and ␤-isoforms of type I PIP5K and analyzed the products. We found no [ 32 P]PtdInsP 2 produced when the type I PIP5Ks were incubated with [ 32 P]Ptd-Ins-3,4,5-P 3 (not shown). These results suggest that the type I PIP5Ks can phosphorylate the D-4 or the D-5 position of Ptd-Ins-3-P, and that with relatively high frequency both sites become phosphorylated in a concerted fashion.
Type I PIP5Ks Produce PtdIns-5-P-The product of the reaction using PtdIns as substrate was also identified by deacylation and HPLC chromatography. As shown in Fig. 5, deacylated PtdInsP produced using TLC purified PtdIns as the substrate for the E. coli-expressed PIP5K I␤ enzyme migrates distinctly from GroPIns-3-P or GroPIns-4-P at the position of GroPIns-5-P (16). Similar results were obtained using bacterially expressed PIP5K I␣ as the enzyme source (data not shown). PtdIns-5-P has been found in cells (16), and this result suggests that it could be synthesized by type I PIP5Ks. A small amount of PtdInsP 2 , which was identified as PtdIns-4,5-P 2 , was also produced using PtdIns as substrate (not shown). Since the PtdIns used in this experiment was TLC purified to eliminate the possibility that PtdIns-4-P was present as a contaminant, this result suggests that type I PIP5Ks can phosphorylate both the D-4 and D-5 positions of the inositol ring. However, given that PtdIns-5-P is a poor substrate for type I PIP5Ks (16), it is likely that the type I PIP5Ks only synthesize PtdIns-4,5-P 2 from PtdIns in a concerted reaction at a low rate.
Substrate Preference of the Type I PIP5Ks-To investigate the relative substrate preferences, we determined the apparent K m , V max and V max /K m ratio for the various substrates of the bacterially expressed type I PIP5Ks. As expected, PtdIns-4-P was the favored substrate of both the ␣and ␤-isoforms. Bacterially expressed PIP5K I␣, which was significantly less active than PIP5K I␤, did not efficiently phosphorylate the other phosphoinositides tested. In contrast, PIP5K I␤ phosphorylated PtdIns-3-P about one-sixth as well as PtdIns-4,5-P 2 (V max /K m ratio of 92 pmol/min/mg compared with 564 pmol/ min/mg). Given the high ratio of PtdIns-4-P to PtdIns-3-P in mammalian cells (approximately 20:1), it is likely that PtdIns-4-P is the major in vivo substrate. However, it is certainly possible that the type I PIP5Ks account for the small amount of PtdIns-3,5-P 2 observed in vivo (approximately 1% of PtdIns-4,5-P 2 ) via phosphorylation of PtdIns-3-P at the D-5 position.
Type I PIP5Ks Expressed in Mammalian Cells Phosphorylate PtdIns-3-P, PtdIns-3,4-P 2 , and PtdIns-The experiments presented above were performed using enzymes that were expressed in bacteria. To determine whether type I PIP5Ks expressed in mammalian cells have the same properties, we repeated experiments with HA-tagged PIP5K I␣ and PIP5K I␤ produced in 293 E1A cells. Proteins were immunoprecipitated and then assayed for their ability to phosphorylate PtdIns-3-P, PtdIns-3,4-P 2 , and PtdIns. The products of the reactions were deacylated and analyzed by HPLC. The PIP5Ks expressed in 293 cells were both quite active, in contrast to the bacterially produced enzymes where PIP5K I␤ was significantly more active than PIP5K I␣. Both mammalian expressed type I PIP5K isoforms were able to phosphorylate PtdIns-3-P to produce PtdIns-3,5-P 2 , PtdIns-3,4-P 2 , and PtdIns-3,4,5-P 3 (Fig. 6,  A and D). They were also able to phosphorylate PtdIns-3,4-P 2 to produce PtdIns-3,4,5-P 3 (Fig. 6, B and E), and PtdIns to produce PtdIns-5-P (Fig. 6, C and F). In contrast to the bacterially expressed enzymes, however, the mammalian expressed PIP5Ks differed in their ability to synthesize PtdIns-3,5-P 2 relative to PtdIns-3,4-P 2 . When PIP5K I␣ phosphorylated Ptd-Ins-3-P, it produced more PtdIns-3,5-P 2 than PtdIns-3,4-P 2 (Fig. 6B), whereas PIP5K I␤ produced equal or more PtdIns-3,4-P 2 than PtdIns-3,5-P 2 (Fig. 6E). It is unclear whether this difference between PIP5K I␣ and PIP5K I␤ is the result of PIP5K I␣ being more efficient at synthesizing PtdIns-3,5-P 2 or less efficient at producing PtdIns-3,4-P 2 . DISCUSSION We have found that the type I PIP5Ks can synthesize the novel phospholipids PtdIns-3,5-P 2 and PtdIns-5-P by phospho- rylating PtdIns-3-P and PtdIns, respectively. Our results suggest a potential pathway for PtdIns-3,5-P 2 and PtdIns-5-P synthesis in vivo. We also find, in agreement with a recent report from Zhang et al. (33), that the type I PIP5Ks can phosphorylate PtdIns-3-P and PtdIns-3,4-P 2 to produce the important signaling molecules PtdIns-3,4-P 2 and PtdIns-3,4,5-P 3 . Taken together, our results demonstrate that in addition to their well established role of synthesizing PtdIns-4,5-P 2 , the type I PIP5Ks have the in vivo potential to synthesize the following four phosphoinositides: PtdIns-3,5-P 2 , PtdIns-3,4-P 2 , PtdIns-3,4,5-P 3 , and PtdIns-5-P. Since each of these phospholipids is likely to have distinct roles in signaling, the type I PIP5Ks may participate in a variety of cellular processes.
The existence of PtdIns-3,5-P 2 has only recently been verified in yeast, mammalian, and plant cells (31,32). Although present in resting cells, the level of this lipid appears to be regulated by osmotic strength changes (32). Since PtdIns-3,4,5-P 3 is absent in the cells in which PtdIns-3,5-P 2 is found, PtdIns-3,5-P 2 is unlikely to be produced by a phosphatase. Analysis of the specific activity of the D-3 and D-5 phosphates indicates that PtdIns-3,5-P 2 is synthesized primarily by phosphorylation of PtdIns-3-P at the D-5 position (31,32). We propose that the type I PIP5Ks are the kinases responsible for catalyzing this reaction in vivo. Given that the level of PtdIns-3,5-P 2 is regulated by osmotic strength, it will be interesting to determine whether the activity of the type I PIP5Ks is altered upon changes in osmotic pressure.
PtdIns-3,4,5-P 3 is primarily thought to be synthesized by the phosphorylation of PtdIns-4,5-P 2 on the D-3 position of the inositol ring by PI3K. Our results suggest an alternative synthetic pathway in which PI3K provides the initial substrate PtdIns-3-P by phosphorylating PtdIns on the D-3 position. Ptd-Ins-3-P is then phosphorylated by the type I PIP5Ks on the D-4 position to produce PtdIns-3,4-P 2 and then the D-5 position to produce PtdIns-3,4,5-P 3 . PtdIns-3,5-P 2 is less likely to be an intermediate in the synthesis of PtdIns-3,4,5-P 3 since it is a rather poor substrate for type I PIP5Ks. Evidence for this pathway is provided by studies in platelets which demonstrated that PtdIns is first phosphorylated on the D-3 position to produce PtdIns-3-P, followed by the phosphorylation of the D-4 position to form PtdIns-3,4-P 2 (27)(28)(29).
The type I PIP5Ks appear to be dual specificity kinases since they phosphorylate both the D-4 and D-5 positions of the inositol ring. However, while PtdIns-5-P is produced when PtdIns is used as a substrate for type I PIP5Ks, no PtdIns-4-P is detected. Similarly, PtdIns-3,5-P 2 is a poor substrate compared with PtdIns-3,4-P 2 for type I PIP5Ks. Therefore, the ability of type I PIP5Ks to phosphorylate the D-4 position of the inositol ring is primarily restricted to the substrate PtdIns-3-P. In general, the type I PIP5Ks favor phosphorylating the D-5 position of the inositol ring, whereas the type II PIPKs favor the D-4 position (16).
The proof of the existence of PtdIns-3,5-P 2 and PtdIns-5-P in cells and the finding that the type I PIP5Ks can synthesize these products in vitro increases the complexity of phosphoinositide signaling. The challenge remains to understand the function of the various phosphoinositides and the regulation of their synthesis and degradation. It will be interesting to determine whether extracellular signals that activate growth factor receptors and/or proteins known to interact with type I PIP5Ks such as Rho family small GTP-binding proteins (13)(14)(15) alter the specificity of the type I PIP5Ks for their various substrates.