Purification and Biochemical Characterization of a Mammalian Phosphatidylinositol 3,4,5-Trisphosphate 5-Phosphatase*

Characterization of the enzymes involved in metabo- lism of 3-phosphorylated inositol lipids and their subcellular localization revealed that in vitro a 5-phosphatase activity was responsible for the degradation of phosphatidylinositol 3,4,5-trisphosphate, whereas a 3-phos- phatase activity hydrolyzed phosphatidylinositol 3-phosphate and/or phosphatidylinositol 3,4-bisphos-phate. All these activities were localized in the cytosol. No phospholipase activities were detected. The cytosolic phosphatidylinositol 3,4,5-trisphosphate 5-phospha- tase activity was purified to near homogeneity using ion exchange, affinity, and size exclusion chromatography. Characterization of the purified phosphatase revealed that it is a magnesium-dependent 5-phosphatase that is able to hydrolyze phosphatidylinositol 4,5-bisphosphate and phosphatidylinositol 3,4,5-trisphosphate. The en- zyme is only partially inhibited by neomycin and vanadate but is strongly inhibited by phosphatidylinositol 4,5-bisphosphate and to a slightly lesser extent by phosphatidylinositol 4-phosphate.

Several investigations were recently carried out to identify specific targets for these 3-phosphorylated phosphoinositides in vitro, revealing that protein kinase Cs (22)(23)(24), the protein kinase C-related kinases (24) and probably phospholipase Ds (25,26) might be potential targets. However, none of these have been proven as yet to be specific PtdIns(3,4,5)P 3 binding proteins in vivo. In contrast, recent work by Hawkins and co-workers (11) and Hu and co-workers (27) identified the small G-proteins Rac-1 and Ras, respectively, as potential downstream targets in vivo, suggesting that 3-phosphorylated lipids might modulate these G-proteins. However, Ras has also been shown to be an upstream modulator of PI 3-kinase (28). Furthermore, no conclusive data have been obtained with respect to whether PtdIns(3,4,5)P 3 or PtdIns(3,4)P 2 is the actual second messenger for Rac-1-or Ras-mediated signal transduction. Thus, the definition of the second messengers remains unclear.
Lack of knowledge of metabolizing enzymes and availability of specific tools to modulate PtdIns(3,4,5)P 3 metabolism in vivo are slowing the pace of achieving much needed progress in identifying the important 3-phosphorylated inositol lipid for signal transduction. We describe here the purification and subsequent characterization of a PtdIns(3,4,5)P 3 5-phosphatase activity from rat brain tissue, one of the major PtdIns(3,4,5)P 3 metabolizing enzymes.

EXPERIMENTAL PROCEDURES
Materials-Inositol lipids, inositol phosphates, and inositol hexasulfate were obtained from Sigma. The inositol analogue InsS 3 , a sulfated deoxyinositol ((1R,2R,4R)-cyclohexane-1,2,4-trismethylenesulfonate), was purchased from Calbiochem. Amersham Corp. provided [␥-32 P]ATP, while hydroxyphenylglyoxal and N-sulfo-N-hydroxysuccinimide were from Pierce. The heterodimeric complex of the mammalian PI 3-kinase was expressed in Sf9 cells using the baculovirus system and subsequently purified as described earlier (29). Glutathione S-transferase-phospholipase C fusion protein was expressed in Escherichia coli (the bacteria were kindly provided by Dr. M. Katan, London) and affinity-purified. All chromatography columns were purchased from Pharmacia Biotech Inc.
In some instances the phosphoinositides were treated with PtdInsspecific phospholipase C in order to destroy the remaining PtdIns(4,5)P 2 , since the 3-phosphorylated inositol lipids are not a substrate for this enzyme (31,32). In this case the inositol lipids were dried under vacuum and subsequently dispersed in 2% cholate (w/v) in 0.2 M * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The 3-phosphatase assays were performed in principle as described for the 5-phosphatase assay (see above), except that after phase separation the removed water phase was treated with ammonium molybdate, and subsequently the phosphomolybdate complex was extracted using ethyl acetate. The extracted phosphomolybdate complex represented the orthophosphate in the reaction mix, whereas the remaining water phase contained the water-soluble inositol polyphosphates, the potential products of any phospholipase activity, if present. Activities were calculated as described above. HPLC analysis was performed as described earlier (33).
Isolation of Subcellular Fraction from Rat Brain Tissue-Rat brains were removed from the animals and immediately frozen in liquid nitrogen. The frozen rat brains were stored at Ϫ70°C for several months. Frozen rat brains (ϳ200 g of wet tissue) were quickly washed with buffer A (1 mM EDTA; 10 mM benzamidine; 40 mM Tris, pH 7.4) before thawing. After thawing the rat brains were quickly homogenized in 3 ml of buffer A/g of wet tissue containing 1 mM phenylmethylsulfonyl fluoride and 0.1 mg/ml trypsin inhibitor (soy bean) using an Ultra-Turrax homogenizer. The crude homogenate was then further homogenized using a Dounce homogenizer and subsequently subjected to centrifuga-tion (8,000 ϫ g, 20 min). The pellet was rehomogenized using 100 ml of buffer A and centrifuged (8,000 ϫ g, 20 min). The combined supernatants were then subjected to an ultracentrifuge spin (100,000 ϫ g, 1 h). The supernatant ("cytosol") was either stored at 4°C (if used on the same day) or frozen at Ϫ20°C (storage for several months), whereas the corresponding pellet (particulate fraction) was resuspended in 50 ml of buffer A containing 0.5 M NaCl, homogenized (Dounce homogenizer), and again centrifuged (100,000 ϫ g, 1 h). The supernatant is referred to as the "NaCl extract," whereas the remaining pellet, resuspended in buffer A, is termed the "membrane" fraction. Both fractions were stored as described for the cytosol fraction. Protein was determined using a commercial dye-based assay (Bio-Rad). Sodium dodecyl sulfate-polyacrylamide gel electrophoresis was done as described (29).
Purification of the Cytosolic Phosphatase Activity-The cytosol (ϳ400 ml) of the rat brain tissue extract was loaded onto a 50-ml HiLoad Q column at a flow rate of 2 ml/min. After washing the column with buffer A, bound proteins were step-eluted using buffer A containing 0.1 M NaCl followed by 0.3, 0.5, and 1.0 M NaCl in buffer A. The eluates were fractionated and pooled according to their protein profile. The pooled fractions were subjected to phosphatase assays and stored at Ϫ20°C (in the presence of 50% (v/v) ethylene glycol) or at 4°C until they were employed for the next purification step.
The pooled fraction containing most of the phosphatase activity (0.3 M NaCl step) was diluted 3-fold and then loaded onto a 15-ml HiTrap S column at a flow rate of 2 ml/min. The bound material was step-eluted as described using 0.1 M, 0.2 M, 0.3 M, and 0.5 M NaCl in buffer A. Eluates were fractionated, pooled, assayed for phosphatase, and stored as described above. The fraction containing most of the phosphatase activity (0.2 M NaCl step) was then diluted 2.5-fold and loaded onto a 5-ml phosphocellulose column followed by a 5-ml HiTrap S column (run in series), thus enabling the binding of the breakthrough of the phosphocellulose column onto the S column. After loading and washing, the two columns were disconnected. The phosphocellulose column was stepeluted using 0.1, 0.2, 0.3, 0.5, 0.7, and 1.0 M NaCl in buffer A as described above. The HiTrap S column was eluted with 0.5 M NaCl and is referred to as the breakthrough of the phosphocellulose column.
The fractions containing most of the 5-phosphatase activity (0.1 M NaCl step from the phosphocellulose column) were then diluted 2.5-fold and subsequently loaded onto a 2-ml heparin-HiTrap column followed by a 1-ml Q-HiTrap column (run in series) at a flow rate of 1 ml/min. The columns were disconnected after loading and washing and subsequently step-eluted as described above. The heparin-bound protein was eluted using 0.1, 0.2, 0.3, 0.5, and 1.0 M NaCl in buffer A. The heparin pool (0.3 M NaCl step) was then loaded onto an S-300 size exclusion column (200 ml) equilibrated with 0.4 M NaCl in buffer A. Eluates were fractionated (0.3 ml/min; 15 min/fraction) and subsequently tested for protein and 5-phosphatase activity. The active fractions were pooled, diluted 5-fold, and finally loaded onto a 1-ml Q-HiTrap column. The bound protein was eluted using a gradient of 0.1-0.3 M NaCl in buffer A. The fractions (1 ml) were tested for protein and 5-phosphatase activity as described above. The active fractions were then concentrated by ultrafiltration using a 100-kDa cut-off membrane (Amicon).

Characterization of Enzymes Involved in Metabolism of 3-Phosphorylated Inositol
Lipids-In order to investigate the metabolism of the 3-phosphorylated inositol lipids, [ 32 P]PtdIns(3)P, [ 32 P]PtdIns(3,4)P 2 , and [ 32 P]PtdIns(3,4,5)P 3 were incubated in the presence of the corresponding inositol lipid precursor (PtdIns, PtdIns(4)P, and PtdIns(4,5)P 2 , respectively) using cytosolic and particulate fractions of rat brain tissue. The major degradation product of PtdIns(3,4,5)P 3 was identified as PtdIns(3,4)P 2 using HPLC head group analysis (data not shown; see also Fig. 5). Under these conditions no phospholipase activity nor 4-phosphatase activity could be identified (data not shown; see also Fig. 5). However, based upon [ 32 P]orthophosphate release there was detectable 3-phosphatase activity, although this activity was at least 10-fold lower in its ability to metabolize PtdIns(3,4,5)P 3 . Using [ 32 P]PtdIns(3,4)P 2 as a substrate revealed that this lipid is preferentially hydrolyzed at the 3-position of the inositol ring (see Fig. 1); no other phospholipase or phosphatase activity could be detected (data not shown). This PtdIns(3,4)P 2 3-phosphatase activity appears to account for the 3-phosphatase ac-tivity observed with PtdIns(3,4,5)P 3 as a substrate via the sequential activity of the 5-phosphatase and 3-phosphatase.
All three phosphatase activities metabolizing 3-phosphorylated inositol lipids are predominantly cytosolic activities (Fig.  1). Phosphatase activities were detected in the particulate fraction and could be solubilized by high salt treatment of the particulate fraction. However, the 3-phosphatase activities (PtdIns(3)P and PtdIns(3,4)P 2 as a substrate) were not inhibited by chelators (data not shown), and the PtdIns(3,4,5)P 3 5-phosphatase activity was found to be dependent on the presence of magnesium regardless of the subcellular localization (data not shown; see also Fig. 4).
Purification of the PtdIns(3,4,5)P 3 5-Phosphatase -Since the cytosolic 5-phosphatase activity was the major metabolizing enzyme for PtdIns(3,4,5)P 3 , this activity was used to purify and characterize the enzyme. Purification was performed as outlined under "Experimental Procedures" Table I. Preliminary experiments revealed that the phosphatase activity eluted as a broad peak in early steps when using gradient elution from columns. Thus, the initial chromatography steps of the purification were performed by applying step gradients. A Q-HiTrap column was employed as a first step for the purification of the phosphatase. Almost all of the detectable 5-phosphatase activity eluted between 0.1 and 0.3 M NaCl (0.3 M NaCl step), whereas the 3-phosphatase activity (probably PtdIns(3,4)P 2 3-phosphatase; see above) eluted earlier (0.1 M NaCl step).
The Q pool was then further purified using S-HiTrap, phosphocellulose, and heparin column chromatography. Since the phosphatase activity eluted in two fractions from the S-HiTrap (0.2 M and 0.3 M NaCl steps) and phosphocellulose column chromatography (0.1 M and 0.2 M NaCl steps), the fractions containing the most enriched 5-phosphatase activity were used for the next purification step. However, the side fractions were stored and later further purified using the same protocol. These fractions were found to contain the same 5-phosphatase activity as judged by polyacrylamide gel electrophoresis and subsequent Coomassie staining (data not shown; see also Fig. 3). Thus, the appearance of heterogeneity of the 5-phosphatase activity is due in part to the step elution technique employed for purification. However, although it cannot be ruled out that some of these side fractions contain in addition a different PtdIns(3,4,5)P 3 5-phosphatase activity.
The highly purified 5-phosphatase activity was then further fractionated by size exclusion chromatography. As shown in Fig. 2, the 5-phosphatase activity eluted in a broad peak, slightly behind the major protein peak. The elution of the protein of the phosphatase preparation was similar to the elution of an IgG marker protein, suggesting that the 5-phosphatase activity has a native size of approximately 160 kDa. The active fractions were pooled and further processed on a 1-ml Q-HiTrap column, and the enzyme eluted as shown in Fig. 3A. As noted earlier, the 5-phosphatase activity eluted quite broadly at about 0.2 M NaCl. The activity comigrated with a 145-kDa protein revealed by polyacrylamide gel electrophoresis (Fig. 3B). Based upon gel filtration behavior it would appear that the 5-phosphatase is monomeric.
The Purified PtdIns(3,4,5)P 3 Phosphatase Activity Is a Magnesium-dependent 5-Phosphatase-The purified 5-phosphatase activity was employed for the optimization of the phosphatase assay, the analysis of the substrate specificity, and an initial screen of potential inhibitors for this phosphatase activity. Preliminary experiments already suggested that the phosphatase activity was inhibited by chelating agents (see above). Thus, the metal ion dependence of the 5-phosphatase activity was investigated using a mixed micellar assay. As shown in Fig. 4, the presence of magnesium was necessary for the enzyme activity, while other divalent cations (e.g. zinc, calcium, and manganese) did not support the 5-phosphatase activity. The magnesium dependence was also observed when the lipids were employed without detergent (data not shown). Most detergents (e.g. CHAPS, cholate, Nonidet P-40, cetyltrimethylammonium bromide) had either negligible or slightly activating effects (up to 2-fold stimulation) on the 5-phosphatase activity when substrate was presented in mixed micelles. However, when substrate was presented at a concentration 5 times higher than the critical micellar concentration, Nonidet P-40 and cetyltrimethylammonium bromide strongly inhibited the 5-phosphatase activity, whereas CHAPS and cholate further potentiated the enzyme activity (data not shown). Thus, anionic or zwitterionic detergents seem to be optimal for substrate presentation to the PtdIns(3,4,5)P 3 5-phosphatase.

5-phosphatase
The purification of the cytosolic PtdIns(3,4,5)P 3 5-phosphatase and the determination of protein and 5-phosphatase activity were as described under "Experimental Procedures." Due to different amounts of PtdIns(4,5)P 2 (see "Experimental Procedures") present in the employed batches of [ 32 P]PtdIns(3,4,5)P 3 the determination of the 5-phosphatase activity is only semiquantitative. The substantial increase in 5-phosphatase activity after concentration on an ultrafiltration device (100-kDa cut-off) is probably partially caused by the removal of small and medium molecular weight inhibitory proteins, partially due to the semiquantitative character of the assay. However, two independent purifications led to a pure 5-phosphatase preparation that had specific activities of approximately 1-2 units/mg protein.
Using PtdIns(4)P, PtdIns(3)P, or PtdIns(3,4)P 2 as a substrate revealed that indeed the preparation contains only trace amounts (about 100-fold lower than the 5-phosphatase activity) of these activities.
Two of the most potent inhibitors for the PtdIns(3,4,5)P 3 5-phosphatase are the phosphoinositides PtdIns(4)P and PtdIns(4,5)P 2 , which were able to inhibit the phosphatase activity at low micromolar concentrations (data not shown; see below). Using these lipids at a comparable concentration (0.5 mM) with respect to the other tested compounds revealed that these phosphoinositides completely blocked the 5-phosphatase activity.
Substrate Specificity of the PtdIns(3,4,5)P 3 5-Phosphatase-The observation that PtdIns(4,5)P 2 is a strong inhibitory compound led to an investigation of the substrate specificity of the purified PtdIns(3,4,5)P 3 5-phosphatase. Since only negligible amounts of a PtdIns(4)P 4-phosphatase were detected in the enzyme preparation, inhibition by PtdIns(4)P seems not to be caused by an intrinsic 4-phosphatase activity, which is also supported by the lack of any 4-phosphatase activity product in the HPLC analysis shown in Fig. 6. However, the preparation did contain substantial amounts of a PtdInsP 2 5-phosphatase activity, which could not be separated from the PtdIns(3,4,5)P 3 5-phosphatase activity by chromatographic techniques (data not shown).
As shown in Fig. 8, a 10-fold higher activity (V max ) for PtdIns(4,5)P 2 as compared with the PtdIns(3,4,5)P 3 5-phosphatase activity was detected for the enzyme. However, the apparent affinity of the enzyme for PtdIns(4,5)P 2 was surprisingly low (K m ϭ 100 M). The corresponding affinity for PtdIns(4,5)P 2 was almost 20 times higher (K m ϭ 5 M), even though the PtdIns(3,4,5)P 3 employed contained equimolar amounts of competing PtdIns(4,5)P 2 . Employing an almost pure PtdIns(3,4,5)P 3 preparation (10% PtdIns(4,5)P 2 present) revealed an even stronger affinity for PtdIns(3,4,5)P 3 (K m ϭ 1 M), suggesting that the K m for PtdIns(3,4,5)P 3 is indeed underestimated. DISCUSSION A PtdIns(3,4,5)P 3 5-phosphatase activity has been purified to near homogeneity and characterized. Based upon molecular weight and inhibitor sensitivity this activity is distinct from any of the inositol polyphosphate 5-phosphatase previously identified. Although the metabolism of the 3-phosphorylated inositol lipids has been previously investigated in vitro, the conclusion of these findings seem to be limited due to the use of nanomolar concentrations of these lipids (34,40,41). Here, we were able to produce sufficient amounts of PI 3-kinase to overcome limitations of lipid production. Thus micromolar concentrations of these inositol lipids, which are also close to the actual physiological levels of these lipids, were employed in the corresponding assays.
The cytosolic 5-phosphatase was subsequently purified to near homogeneity using ion exchange, size exclusion, and affinity chromatography. Since the PtdIns(3,4,5)P 3 5-phosphatase activity comigrated with a 145-kDa protein and showed a similar molecular mass on size exclusion chromatography, it is assumed that the enzyme exists in a monomeric form. The purified activity is dependent on magnesium, which cannot be substituted by other divalent cations. In fact, zinc, calcium, and manganese are able to inhibit substantially the magnesiumdependent 5-phosphatase activity at concentrations of 1 mM (data not shown).
Since the purified 5-phosphatase acts on phosphoinositides it was to be expected that neomycin, a phosphoinositide binding compound (39), would inhibit the PtdIns(3,4,5)P 3 5-phosphatase. However, neomycin showed only a partial inhibition at concentrations in excess of the substrate concentration. Thus, neomycin, which strongly inhibits PI 3-kinase (data not shown) and the PtdIns-dependent phospholipase C (43), is quite ineffective on the PtdIns(3,4,5)P 3 5-phosphatase. However, a recently discovered phosphoinositide phosphatase derived from the particulate fraction of rat brain tissue is also insensitive to neomycin (37), suggesting that inositol lipid phosphatases are better able to "compete" with neomycin for the phosphorylated substrate.
Taking these observations concerning the size (molecular weight) and the inhibitor characteristics together, it can be surmised that the purified PtdIns(3,4,5)P 3 5-phosphatase is unique and not the same as one of the known phosphatases. For example, there is a 160-kDa inositol polyphosphate 5-phosphatase in rat brain cytosol (36), but this activity is strongly inhibited by 2,3-glycerolbisphosphate or ATP and only partially inhibited by 2 mM concentrations of calcium (36). In contrast, the PtdIns(3,4,5)P 3 5-phosphatase is not inhibited by ATP (data not shown) nor 2,3-glycerolbisphosphate at 1 mM concentrations but is nearly completely inhibited by 1 mM calcium. Recently a cytosolic 155-kDa phosphoinositide 5-phosphatase was characterized, which was inhibited by 2,3-glycerolbisphosphate (38). As mentioned above, this inhibitor does not block the PtdIns(3,4,5)P 3 5-phosphatase activity; furthermore the 155-kDa phosphatase could accept manganese for magnesium, which is in contrast to the behavior of the PtdIns(3,4,5)P 3 5-phosphatase. However, three monoclonal antibodies raised against the 155-kDa 5-phosphatase derived from bovine brain (38) cross-react with the purified 145-kDa 5-phosphatase from rat brain, but only if Coomassie-stainable amounts are employed. Two of these antibodies were much weaker in their cross-reactivity as compared with the third antibody, suggesting that at least one epitope might be shared between these two 5-phosphatases (data not shown). Since these two enzymes have different molecular weights as judged by SDS-polyacrylamide gel electrophoresis (data not shown) and do not display similar biochemical properties, the 155-kDa enzyme from bovine brain is at best no more than a relative of the 145-kDa 5-phosphatase activity from rat brain. Thus, although earlier investigations described 5-phosphatase activities with comparable size, these activities display distinctly different properties from the PtdIns(3,4,5)P 3 5-phosphatase described here.
Whereas the PtdIns(3,4,5)P 3 5-phosphatase activity is not strongly inhibited by neomycin or vanadate, relatively strong inhibition was observed with either PtdIns(4)P or PtdIns(4,5)P 2 . As there are only negligible amounts of PtdIns(4)P 4-phosphatase activity present in the purified PtdIns(3,4,5)P 3 5-phosphatase preparation, the inhibition by PtdIns(4)P is due either to a competition for the substrate binding site or an allosteric effect on the PtdIns(3,4,5)P 3 5-phosphatase. Inhibition by PtdIns(4)P is unlikely to be due to lipid presentation, since a mixed micellar assay was employed, and similar concentrations of the parent lipid PtdIns were actually activating for the 5-phosphatase activity. While the PtdIns(4)P inhibition could not be due to substrate specificity of the PtdIns(3,4,5)P 3 5-phosphatase (see above), PtdIns(4,5)P 2 inhibition might be caused by the competitive hydrolysis of this inositol lipid by the PtdIns(3,4,5)P 3 5-phosphatase. Indeed, the purified PtdIns(3,4,5)P 3 5-phosphatase contained a strong PtdIns(4,5)P 2 5-phosphatase activity, which could not be separated by the chromatographic procedures employed (data not shown). Furthermore, both 5-phosphatases showed comparable inhibitor characteristics (data not shown), suggesting that the purified PtdIns(3,4,5)P 3 5-phosphatase is as well a PtdIns(4,5)P 2 5-phosphatase. However, although the purified 5-phosphatase reveals a 10-fold higher activity for PtdIns(4,5)P 2 , the corresponding apparent affinity is much lower as compared with PtdIns(3,4,5)P 3 as substrate. This would suggest that although both substrates could be hydrolyzed by the purified 5-phosphatase activity, the actual preferred substrate is PtdIns(3,4,5)P 3 . Thus, it would be appropriate to refer to this enzyme as a "PtdIns(3,4,5)P 3 5-phosphatase." The substrate specificity of the purified enzyme would be responsible for the semiquantitative nature of the PtdIns(3,4,5)P 3 5-phosphatase assay employed here. Since the PtdIns(3,4,5)P 3 used in routine assays contains some PtdIns(4,5)P 2 , the actual specific activity could be much higher, and of course, the estimated V max for PtdIns(3,4,5)P 3 would then be an underestimate, supporting the suggestion that the purified activity is preferentially a PtdIns(3,4,5)P 3 5-phosphatase.
The relative abundance of this PtdIns(3,4,5)P 3 5-phosphatase and the fact that this activity represents the major PtdIns(3,4,5)P 3 5-phosphatase activity in cytosol implies that the purified activity is responsible for the degradation turnover of the supposed second messenger PtdIns(3,4,5)P 3 . Since the 5-phosphatase is localized in the soluble fraction, a mechanism to attract this enzyme to the membrane, where its substrate is present, is likely to exist. This would provide a potential means of regulation of the PtdIns(3,4,5)P 3 5-phosphatase and hence of the levels of the second messenger PtdIns(3,4,5)P 3 .