Type I Phosphatidylinositol-4-phosphate 5-Kinases

Type I phosphatidylinositol 4-phosphate (PtdIns(4)P) 5-kinases (PIP5K) catalyze the synthesis of phosphatidylinositol 4,5-bisphosphate, an essential lipid molecule in various cellular processes. Here, we report the cloning of the third member (PIP5Kγ) and the characterization of members of the type I PIP5K family. Type I PIP5Kγ has two alternative splicing forms, migrating at 87 and 90 kDa on SDS-polyacrylamide gel electrophoresis. The amino acid sequence of the central portion of this isoform shows approximately 80% identity with those of the α and β isoforms. Northern blot analysis revealed that the γ isoform is highly expressed in the brain, lung, and kidneys. Among three isoforms, the β isoform has the greatest V max value for the PtdIns(4)P kinase activity and the γ isoform is most markedly stimulated by phosphatidic acid. By analyzing deletion mutants of the three isoforms, the minimal kinase core sequence of these isoforms were determined as an approximately 380-amino acid region. In addition, carboxyl-terminal regions of the β and γ isoforms were found to confer the greatest V max value and the highest phosphatidic acid sensitivity, respectively. It was also discovered that lysine 138 in the putative ATP binding motif of the α isoform is essential for the PtdIns(4)P kinase activity. As was the case with the α isoform reported previously (Shibasaki, Y., Ishihara, H., Kizuki, N., Asano, T., Oka, Y., Yazaki, Y. (1997) J. Biol. Chem.272, 7578–7581), overexpression of either the β or the γ isoform induced an increase in short actin fibers and a decrease in actin stress fibers in COS7 cells. Surprisingly, a kinase-deficient substitution mutant also induced an abnormal actin polymerization, suggesting a role of PIP5Ks via structural interactions with other molecules.

Type I phosphatidylinositol 4-phosphate (PtdIns(4)P) 5-kinases (PIP5K) catalyze the synthesis of phosphatidylinositol 4,5-bisphosphate, an essential lipid molecule in various cellular processes. Here, we report the cloning of the third member (PIP5K␥) and the characterization of members of the type I PIP5K family. Type I PIP5K␥ has two alternative splicing forms, migrating at 87 and 90 kDa on SDS-polyacrylamide gel electrophoresis. The amino acid sequence of the central portion of this isoform shows approximately 80% identity with those of the ␣ and ␤ isoforms. Northern blot analysis revealed that the ␥ isoform is highly expressed in the brain, lung, and kidneys. Among three isoforms, the ␤ isoform has the greatest V max value for the PtdIns(4)P kinase activity and the ␥ isoform is most markedly stimulated by phosphatidic acid. By analyzing deletion mutants of the three isoforms, the minimal kinase core sequence of these isoforms were determined as an approximately 380-amino acid region. In addition, carboxyl-terminal regions of the ␤ and ␥ isoforms were found to confer the greatest V max value and the highest phosphatidic acid sensitivity, respectively. It was also discovered that lysine 138 in the putative ATP binding motif of the ␣ isoform is essential for the PtdIns(4)P kinase activity. As was the case with the ␣ isoform reported previously (Shibasaki, Y., Ishihara, H., Kizuki, N., Asano, T., Oka, Y., Yazaki, Y. (1997) J. Biol. Chem. 272, 7578 -7581), overexpression of either the ␤ or the ␥ isoform induced an increase in short actin fibers and a decrease in actin stress fibers in COS7 cells. Surprisingly, a kinase-deficient substitution mutant also induced an abnormal actin polymerization, suggesting a role of PIP5Ks via structural interactions with other molecules.
Recently, cDNAs encoding two isoforms of type I PIP5K have been cloned (10,11). Herein, we report molecular cloning of a third isoform of type I PIP5K (PIP5K␥) from a cDNA library of the murine pancreatic ␤-cell line MIN6 (12). This novel isoform has two alternative splicing forms of 87 and 90 kDa and is the most markedly stimulated by phosphatidic acid of the three isoforms. These molecular identifications revealed that PIP5K isoforms constitute a novel lipid kinase family, distinct from phosphoinositide 3-kinases, phosphatidylinositol 4-kinases, and diacylglycerol kinases. Demonstration of structural characteristics is essential for understanding the intracellular roles of these isoforms and the mechanisms by which they are regulated. Therefore, in this report, several aspects of the structural characteristics of these isoforms were also studied in vitro and in vivo. We found that a central region, consisting of approximately 380 amino acids, is sufficient for PtdIns(4)P kinase activity and that carboxyl-terminal regions are important for modulation of the kinase activities of these isoforms. We also found that expression of either the ␤ or the ␥ isoform leads to actin rearrangement in COS7 cells, as was the case with the ␣ isoform (13), and that the central region is sufficient for this effect. Furthermore, surprisingly, the expression of a kinase-deficient substitution mutant generated a similar effect in COS7 cells.

EXPERIMENTAL PROCEDURES
Cloning of Murine Type I PIP5K␥-A polymerase chain reaction (PCR) using degenerate oligonucleotide primers and screening of a MIN6 cell cDNA library were as described previously (10). To obtain a 5Ј sequence of PIP5K␥, a Marathon TM cDNA amplification kit (CLONTECH) was used according to the instructions of the manufacturer. The reverse transcription was performed using MIN6 cell poly(A) ϩ RNA and an antisense primer 5Ј-GGTGACGTAGAAGACA-GAGCC-3Ј. The first PCR was performed using adapter primer 1 (CLONTECH) and an antisense primer, 5Ј-CTTCACTGGGGAAGA-AGA TGC-3Ј. The second PCR was performed using adapter primer 2 (CLONTECH) and an antisense primer, 5Ј-GTGGCCCAGCTTCTTC-CCATG-3Ј. The first and second PCR reactions were conducted with inclusion of dimethyl sulfoxide (5%), without which only shorter products were obtained. Individual clones were sequenced following subcloning into pGEM-T vector (Promega, WI) as described above. The consensus of three independent clones confirms the sequence of the 5Ј region of PIP5K␥.
Northern Blotting-A murine multiple tissue Northern blot (CLONTECH) was hybridized according to the instructions of the manufacturer with an [␣-32 P]dCTP-labeled 0.4-kb Aor51HI-PstI fragment from the 3Ј portion of the ␥ isoform cDNA.
Production of Antibody Specific to the ␥ Isoform and Western Blotting-An oligopeptide, CASDEEDAPSTDIYF, was custom synthesized and conjugated to keyhole limpet hemocyanin (Research Genetics, AL) and injected into female New Zealand rabbits employing standard protocols (14). The COS7 cell lysates (10 g/lane) and murine brain lysate (50 g/lane) were subjected to SDS-polyacrylamide gel electrophoresis (PAGE) (7.5%) and then probed with the antisera raised against the above peptide (1:100 dilution). Blots were developed using ECL reagents (Amersham, UK).
Constructions of Mutant cDNAs-Substitution mutant cDNAs were constructed by oligonucleotide-directed mutagenesis. For example, the PIP5K␣-K138A mutant cDNA was generated using primers 5Ј-AT-GAATTCATCATCGCAACCGTTCAG-3Ј (underlined nucleotides encode a mutated alanine) and 5Ј-CTCCTGACTGCATGCAATACAGC-3Ј. The amino-and carboxyl-terminal deletion mutants were generated by using either endogenous restriction enzyme sites or a PCR-based strategy. Endonuclease BamHI, EcoRI, or NcoI digestion, followed by subcloning into the SwaI site of a cosmid vector (pAdex1CA) (16) generated mutants containing amino acids 1-456 with extra asparagine, 1-392, or 1-308 with extra lysine-leucine-isoleucine-lysine-leucine-valine, due to a poly-linker sequence of the vector, respectively. Amino-terminal deletion mutants were generated using an inner antisense primer, 5Ј-CTCCTGACTGCATGCAATACAGC-3Ј, and an appropriate sense primer containing a SalI site for connection with a sequence for the HA epitope. Carboxyl-terminal mutants were generated using an inner sense primer, 5Ј-CTCTATTCCACAGCCATGGAATCC-3Ј, and an appropriate antisense primer containing a stop codon and a BamHI site. The mutant cDNAs, confirmed by DNA sequencing, were subcloned into pBluescript containing a sequence coding the HA-epitope.
PtdIns(4)P Kinase Assay-Immunoprecipitation was performed using a monoclonal antibody against the HA epitope (12CA5) and protein G-Sepharose 4 First Flow (Pharmacia Biotech Inc.). The immunoprecipitates were used for the PtdIns(4)P kinase assay. A standard assay for phosphorylation of PtdIns (4) For comparison of kinase activities among various constructs (see Table I and Figs. 4, 5, and 6), half an aliquot of the immunoprecipitate was used for triplicate kinase activity assay and the other half for Western blotting with the rabbit anti-HA-epitope IgG (MBL, Nagoya) and 125 I-labeled Protein A, by which the protein amount in the immunoprecipitate was estimated. Kinase activities were normalized with wild-type or mutant protein amounts. The signal intensities were measured with a BAS 2000 (Fuji Photo Film, Tokyo).
Immunofluorescence-COS7 cells were plated on coverlips in Dulbecco's modified Eagle's medium with 10% fetal calf serum and infected with recombinant adenoviruses the next day. After 18 h, cells were fixed with 3% paraformaldehyde and incubated with anti-HA monoclonal antibody (12CA5) in phosphate-buffered saline with 0.2% gelatin at room temperature for 45 min. After washing three times with phosphate-buffered saline-gelatin, cells were incubated with fluorescein isothiocyanate-conjugated anti-mouse IgG (DAKO) and rhodamine-conjugated phalloidin (Molecular Probes) for 30 min. Slides were observed under a Bio-Rad confocal microscope system (MRC 1024).
Materials-PtdIns(4)P from bovine brain and phosphatidic acid were purchased from Sigma. Oligonucleotides were custom synthesized and purchased from either Japan Bio-service Inc. (Saitama, Japan) or Becks Inc. (Itabashi, Tokyo).

Cloning of the Third Member of Type I PIP5K Family-A
polymerase chain reaction (PCR) using degenerated primers and subsequent screening of a MIN6 cell cDNA library, as described previously (10), identified a class of partial coding sequences with homology to type I PIP5K ␣ and ␤ isoforms. The third cDNA sequence, designated PIP5K␥, contains four overlapping clones, one of which has an additional 78-base pair sequence at the 3Ј terminus and possibly arises by alternative splicing of the PIP5K␥ gene. In-frame stop codons upstream from the first ATG codon of this cDNA could not be identified in the initial study. Therefore, to obtain an additional 5Ј sequence, an adapter ligation/PCR-based method (Marathon TM , CLONTECH) was employed. Although an additional 172 base pairs and another ATG codon were obtained, there were no in-frame stop codons in a 79-base pair sequence upstream from this ATG codon. Nonetheless, this ATG codon was concluded to be the initial translation codon for the following reasons. First, the ATG codon is in a favorable position for translation according to Kozak's rules (Fig. 1, Ref. 19). Second, as shown in Fig.  2A, recombinant proteins of the ␥ isoform with or without the 26 carboxyl-terminal amino acids expressed via adenoviral vectors migrated almost identically to either of the doublet bands (87 and 90 kDa) from brain tissue on SDS-PAGE. The ␥ isoform has two alternative splicing forms, consisting of 635 and 661 amino acids with calculated molecular masses of 69,563 and 72,469 Da, and isoelectric points of 5.40 and 5.27, respectively (Fig. 1). Because the 87-kDa protein was predominantly expressed in brain tissue ( Fig. 2A) and MIN6 cells (data not shown), the ␥ isoform without the 26 carboxyl-terminal amino acids was used in subsequent analyses. As shown in Fig. 3, the central portions of the three type I isoforms were found to be very similar (approximately 80% identity) in amino acid sequence. In addition, the amino-terminal sequence of the type ␥ isoform shows partial homology with that of the ␤ isoform (approximately 40% identity) whereas the carboxyl-terminal regions differed in length and amino acid sequence among the three isoforms. An entire coding sequence of murine cDNA homologous to human PIP5KII␣ (20, 21), which was recently revealed to be phosphatidylinositol 5-phosphate (PtdIns(5)P) 4-kinase (22), was also cloned from a MIN6 cell cDNA library (data not shown). The murine PIP5KII␣ consists of 405 amino acids, one residue less than its human counterpart, with only seven conserved amino acids differing between the two (Fig. 3).
Tissue Distribution of Type I PIP5K␥-Northern blotting analysis was performed using the ␥ isoform cDNA probe corresponding to the sequence close to the termination codon. A 4.8-kilobase mRNA was detected, as a major band, in murine poly(A) ϩ RNA from different tissues (Fig. 2B). The tissue distribution of the ␥ isoform differed from those of the ␣ and ␤ isoforms, being essentially restricted to the brain, lung, and kidneys.
The Novel cDNA Encodes Type I PIP5K Protein-To characterize the enzymatic activity of the third isoform, HA-tagged proteins of this isoform expressed in COS7 cells were immunoprecipitated using anti-HA-epitope monoclonal antibody 12CA5. The resulting immunocomplex exhibited PtdIns(4)P kinase activity (Fig. 2C). Although this thin layer chromatography separation did not provide information about whether the PtdInsP 2 produced was PtdIns(4,5)P 2 or PtdIns(3,4)P 2 , the close sequence similarity with the ␣ and ␤ isoforms (Fig. 3) strongly suggests that the ␥ isoform is also a 5-kinase. Furthermore, the PtdIns(4)P kinase activities of the third isoform increased by more than 10-fold when an equimolar amount of phosphatidic acid was added to the reaction solutions (Fig. 2D), demonstrating the novel murine cDNA to encode the type I PtdIns(4)P 5-kinase (23,24).
Comparison of Kinetic Activities of Type I PIP5K Isoforms-For initial characterization of members of the PIP5K family, kinetic parameters for the PtdIns(4)P kinase activity of these murine isoforms were studied. For this purpose, recombinant proteins of isoforms with the HA epitope were expressed and immunoprecipitated with the anti-HA-epitope monoclonal antibody. One-half of each immunoprecipitate was used for kinase assay and the other half for Western blotting with rabbit anti-HA-epitope polyclonal IgG. Lipid kinase activity was normalized with the protein amount estimated by Western blotting (for example, see Fig. 5B). Kinetic parameters for these isoforms are summarized in Table I. While affinities for PtdIns(4)P and ATP were similar among the three isoforms, the ␤ isoform had the greatest V max value, approximately 3.2fold and 1.7-fold higher than those of the ␣ and ␥ isoforms, respectively. Study of phosphatidic acid sensitivity revealed the ␥ isoform to be most sensitive to phosphatidic acid.
Deletion Analysis of the Type I PIP5K␣ Defines a Central Kinase Domain-Because type I PIP5K isoforms have no sequence homology with other lipid kinases including phosphoinositide 3-kinases, phosphatidylinositol 4-kinases and diacylglycerol kinases (25,26), it is of great importance to determine their structure and function relationships. To begin to address this issue, amino-or carboxyl-terminal deletion mutants of the type I PIP5K␣ isoform were constructed using endogenous restriction enzyme sites and PCR-based methods. As summarized in Fig. 4, while all four mutants with stepwise deletions from the carboxyl terminus to glutamine residue 400 have activity almost equal to that of the wild-type ␣ isoform, the PIP5K␣-(1-392)/EcoRI mutant has little or no PtdIns(4) 5-kinase activity. In contrast to the long dispensable region in the carboxyl terminus, amino-terminal deletions had a pronounced effect. Although the first 17-amino acid deletion did not alter kinase activity, deletion of only 31 amino acids from the aminoterminal region resulted in significantly reduced kinase activity (34 Ϯ 9% of the full-length ␣ isoform, mean Ϯ S.E., n ϭ 3), and a 46-amino acid deletion virtually abolished kinase activity.
These data suggest that an approximately 380 amino acid central portion of the PIP5K␣ isoform (amino acid residues 18 to 399) constitutes the kinase core domain. Indeed, as described below, this central portion alone retains kinase activity (see Fig. 6B). This region of the PIP5K␣ isoform has about 80% amino acid identity with the corresponding regions of both the ␤ and the ␥ isoform. The amino-terminal half of the kinase core domain is especially conserved among the three isoforms (more than 90%) (Fig. 3). In addition, sequence alignment between type I PIP5Ks and PIP5KII␣ suggests that the latter enzyme (PtdIns(5)P 4-kinase) consists essentially of the kinase region of type I PIP5Ks with approximately 40% identity (Fig. 3).
Mutations in the Putative Nucleotide Binding Region-In several protein and lipid kinases, a glycine-rich sequence followed by a lysine residue 10 -30 residues downstream constitutes an important region for phosphate-transfer reactions (25,27). Although there is no typical region for such a glycine-rich sequence, the region of amino acid residues 121 to 138 in the ␣ isoform is similar to the ATP binding domain of cyclic AMPdependent protein kinase (PKA, Fig. 5A). To examine the role of this region, two ␣ isoform mutants were constructed in which glycine 124 was substituted with valine (PIP5K␣-G124V mutant) or lysine 138 with alanine (PIP5K␣-K138A mutant). A lipid kinase assay revealed type I PIP5K␣-G124V to have 67 Ϯ 11% of the PtdIns(4)P kinase activity of the wild-type protein, while type I PIP5K␣-K138A had virtually no kinase activity (Fig. 5B).
Role of Amino-and Carboxyl-terminal Variable Regions of Type I PIP5Ks-As indicated above, these type I PIP5K isoforms consist of a similar central domain and variable aminoand carboxyl-terminal regions. To characterize these domains, amino-and/or carboxyl-terminal deletion mutants of the three isoforms, carrying the HA epitope, were constructed based on the results of deletion analysis of the ␣ isoform (Fig. 6A). As summarized in Fig. 6B, the central regions of the three isoforms showed essentially equivalent lipid kinase activities. There were no marked changes when amino-and/or carboxylterminal regions were deleted from the ␣ isoform. As shown in Table I, the ␤ isoform had approximately three-fold higher activity than the ␣ isoform. When the carboxyl-terminal region was deleted from the ␤ isoform, the activity was reduced to a level approaching that of the ␣ isoform. In addition, a deletion of the carboxyl-terminal region of the ␥ isoform also resulted in a reduction in its kinase activity. The phosphatidic acid sensitivities of these deletion mutants were also examined (Fig. 6C). The central regions alone of the three isoforms can be stimulated by phosphatidic acid. Although the ␥ isoform showed the highest sensitivity to phosphatidic acid, its carboxyl-terminal deletion mutant exhibited a magnitude of phosphatidic acid stimulation similar to that of the ␣ isoform.
Effects on Actin Polymerization of Overexpressing PIP5K Isoforms and Their Mutants in COS7 Cells-As reported previously, overexpression of type I PIP5K␣ via an adenoviral vector led to a decrease in typical stress fibers and an increase in disarrayed short actin fibers (13). In this study, we examined the effects of overexpressing three type I PIP5K isoforms, type II PIP5K␣, and their mutants on actin polymerization (Fig. 7). Abnormal reorganization of actin fibers was also observed in COS7 cells overexpressing either the ␤ or the ␥ isoform (Fig. 7,  B and C). These in vivo analyses using deletion mutants of the ␣ isoform are summarized in Fig. 4. All deletion mutants of the ␣ isoform with the complete kinase core domain induced abnormal actin reorganization in COS7 cells. In addition, COS7 cells expressing I␣-dNdC, I␤-dNdC, or I␥-dNdC, mutants in which both amino-and carboxyl-terminals are deleted from the three isoforms (Fig. 6A), had enormous amounts of short actin FIG. 2. Novel cDNA encodes 90/87-kDa type I PtdIns(4)P kinase with abundant expression in brain tissue. A, COS7 cells were infected with recombinant adenoviruses containing cDNAs for PIP5K-I␥(635) or PIP5K-I␥(661) without the HA-epitope sequence. The lysates from murine brain (50 g/lane) or infected COS7 cells (10 g/lane) were subjected to SDS-PAGE (7.5%). Expressed proteins were probed with anti-PIP5K␥ antisera. B, a mouse multiple tissue Northern blot (CLONTECH) was hybridized with specific probes for the ␥ isoform. Sk. mus., skeletal muscle. C, autoradiogram demonstrating the novel isoform to contain PtdIns(4)P kinase activity. Lysates of COS7 cells infected with a recombinant adenovirus encoding HA-tagged PIP5K␥ or a control virus (lacZ) were subjected to immunoprecipitation with anti-HA monoclonal antibody. The immunocomplex was assayed for PtdIns(4)P kinase activity as described under "Experimental Procedures." D, PtdIns(4)P kinase activity of the ␥ isoform is markedly stimulated by phosphatidic acid (PA). Lipid kinase reaction was performed using PtdIns(4)P as a substrate (100 M) in the presence and absence of PA (100 M).
These data appeared to indicate that kinase activity of PIP5K would be necessary for abnormal actin polymerization. However, surprisingly, expression of a kinase-deficient substitution mutant, PIP5K␣-K138A, also led to a decrease in typical stress fibers and an increase in short actin fibers (Fig. 7F). In addition, as was the case with the wild-type ␣ isoform (13), COS7 cells expressing the kinase defective substitution mutant also exhibited decreased adhesion activity. They became rounded and readily detached from the bottoms of culture dishes (data not shown).

TABLE I Kinetics of type I PIP5K isoenzymes
The relative V max for PtdIns(4)P kinase were obtained after normalization of the V max values with protein amounts estimated by Western blotting. The normalized V max value of the ␣ isoform was taken as 1.0.   Figs. 4 and 6. B, alignment of PIP5Ks. Identical residues among three or four enzymes are boxed in black. Gaps are represented by dashes.

DISCUSSION
A novel isoform of PIP5K was identified in this study. This novel isoform (PIP5K␥) migrated at 90/87 kDa on SDS-PAGE, suggesting this isoform to be identical or closely related to the type Ib isoform (90 kDa) previously purified (23). The co-existence of three isoforms of the type I PIP5K in insulin-secreting clonal cells suggests that these isoforms have specific functions in vivo. Since the 90-kDa isoform has been reported to have higher activity than the 68-kDa isoform in restoring Ca 2ϩregulated catecholamine release from cytosol-depleted neuroendocrine cells (7), it appears likely that the ␥ isoform plays an important role in regulated secretion. Type I PIP5K and phospholipase D are postulated to be involved in exocytotic processes (8). Stimulation of PtdIns(4,5)P 2 synthesis by phosphatidic acid, which may be generated by phospholipase D, was considered to be important in this process (8). The highest sensitivity to phosphatidic acid of the PIP5K␥ isoform might be a reason for the 90-kDa isoform playing a more active role in Ca 2ϩ -regulated secretion (7). More specific expression of the ␥ isoform in the brain may reflect an important role of this isoform in neurotransmitter release.
Molecular cloning of type I PIP5K isoforms allows study of the structure-function relationships of these important enzymes. Recent studies revealed that type I PIP5K isoforms can phosphorylate several lipid substrates other than PtdIns(4)P (28). In this study, we studied the structure-function relationships with regard to PtdIns(4)P 5-kinase activity since PtdIns(4)P is the preferred substrate of type I PIP5K isoforms (28). The minimal kinase core domain of the type I PIP5K isoforms was determined to be an approximately 380-amino acid region. The amino acid sequence in this region does not, however, contain typical sequences homologous to known protein or lipid kinase domains, the exception being one which exhibits weak homology with a phosphate binding loop of PKA. Almost complete loss of PtdIns(4)P kinase activity by substitution of lysine 138 with alanine suggests that this lysine residue plays a role similar to lysine 72 in PKA, which was proposed to interact with the ␣and ␤-phosphate groups of ATP (29). It was also found that the amino-terminal half of the kinase core domain is highly conserved among type I PIP5K isoforms. Especially, in the region spanning residue 80 to 161 of the ␣ isoform, 98% of 82 residues are identical or conserved among three isoforms. In addition to the putative nucleotide binding domain, there may be domains essential for lipid kinase activities in this region.
Among the three type I isoforms, the ␤ isoform has the greatest V max for PtdIns(4)P kinase activity while the ␥ isoform is most markedly stimulated by phosphatidic acid. Our results using deletion mutants indicate an important role of the carboxyl-terminal regions for these characteristics. Since the type I PIP5K isoforms have recently been reported to phosphorylate PtdIns(3)P and PtdIns(3,4)P 2 (28), it would be intriguing to examine whether amino-and/or carboxyl-terminal sequences are involved in recognition of these different substrates. In addition, these regions might be important for possible associations with other unknown molecules. Further studies are needed to elucidate roles of these amino-and carboxyl-terminal variable regions.
We also found that overexpression of any one of the three isoforms led to the production of massive amounts of short actin fibers while disrupting actin stress fibers in COS7 cells. A surprising result was that a kinase defective mutant, PIP5K␣-K138A, induces similar effects. The mechanism by which PIP5K isoforms and the kinase-deficient substitution mutant induce such effects remains to be determined. The causal relationship between short actin fiber formation and disruption of actin stress fibers is also unclear. It has been reported that expression of PtdIns(4,5)P 2 5-phosphatases in COS7 cells decreased the number of actin stress fibers via the hydrolysis of PtdIns(4,5)P 2 bound to actin regulatory proteins (13,30). An  (37); INS-R, human insulin receptor (38); amino acid numbers of (putative) ATP binding lysine residues (bold letters) are indicated. B, the effects of mutations in the putative ATP binding domain of PIP5K␣. Kinase activities were normalized with the amounts of immunoprecipitated enzymes estimated by Western blotting. PtdIns(4)P kinase activity of the wild-type ␣ isoform was taken as 100%. Data are presented as means Ϯ S.E. of three independent experiments, each performed in triplicate.

FIG. 4. Deletion mutants of the type I PIP5K␣ isoform.
A, HAtagged mutant proteins were immunoprecipitated from lysates of COS7 cells infected with recombinant adenoviruses, subjected to SDS-PAGE and blotted with anti-HA polyclonal antibody. B, deletion mutants are schematically represented with their relative kinase activities at 200 PtdIns(4)P and with their effects on actin reorganization. Effects on actin reorganization are represented as ϩ when more than 70% cells expressing one of deletion mutants exhibited an increase in short actin fibers and a decrease in stress fibers, as Ϯ when 10 -70%, and Ϫ when less than 10% (for typical cells, see Fig. 7).
opposite mechanism (i.e. via an increase in PtdIns(4,5)P 2 ) is unlikely to lead an increase in short actin fibers in cells overexpressing 5-kinases since the kinase-deficient mutant induced a similar effect, although the possibility of endogenous 5-kinase activity playing some part in the effect cannot be ruled out. Indeed, it was reported that overexpression of type I PIP5K isoforms in COS7 cells did not increase cellular levels of Pt-dInsP 2 (31), providing evidence that effects on actin reorgani-zation of overexpressing PIP5Ks were not mediated by the kinase activity of overexpressed proteins. The fact that the kinase inactive mutant induces actin reorganization similar to that seen with the wild-type enzyme suggests that structural interactions with other as yet unknown molecules mediate this effect. Small GTP binding proteins, Rac and Rho, are possible candidates (32)(33)(34). In this regard, it should be noted that the structure of the kinase core domain was found to be sufficient FIG. 6. Roles of amino-and/or carboxyl-terminal regions of the three type I PIP5Ks. A, schematic representation of amino-and/or carboxyl-terminal deletion mutants of the type I PIP5K isoforms. B, PtdIns(4)P kinase activities of deletion mutants. Assays were performed using immunoprecipitated wild-type and mutant enzymes expressed in COS7 cells at a PtdIns(4)P concentration of 200 M and ATP at 50 M. Kinase activities were normalized with the amounts of immunoprecipitated enzymes estimated by Western blotting. PtdIns(4)P kinase activity of the wild-type ␣ isoform was taken as 1.0. Data are presented as means Ϯ S.E. of at least five independent experiments, each performed in triplicate. *, Difference from the relevant wild-type isoform at p Ͻ 0.05. C, stimulation of the PtdIns(4)P kinase activity of deletion mutants by phosphatidic acid. Assays were performed using immunoprecipitated wild-type and mutant enzymes at a PtdIns (4) for inducing abnormal actin polymerization. There may be a binding site for such an interacting molecule within the kinase core domain. Future studies should be designed to identify the molecules interacting with PIP5K isoforms.
Recent findings suggest that PIP5Ks play various roles in signaling pathways, by participating in the synthesis of a number of phosphoinositides (28). The present results suggest that structural interactions are also important in PIP5Ks functions. Much research remains to be done in order to elucidate the complex signaling pathways in which these lipid kinases are involved.