Identification of Multiple Phosphoinositide-specific Phospholipases D as New Regulatory Enzymes for Phosphatidylinositol 3,4,5-Trisphosphate*

In the course of delineating the regulatory mechanism underlying phosphatidylinositol 3,4,5-trisphosphate (PI(3,4,5)P3) metabolism, we have discovered three distinct phosphoinositide-specific phospholipase D (PI-PLD) isozymes from rat brain, tentatively designated as PI-PLDa, PI-PLDb, and PI-PLDc. These enzymes convert [3H]PI(3,4,5)P3 to generate a novel inositol phosphate, d-myo-[3H]inositol 3,4,5-trisphosphate ([3H]Ins(3,4,5)P3) and phosphatidic acid. These isozymes are predominantly associated with the cytosol, a notable difference from phosphatidylcholine PLDs. They are partially purified by a three-step procedure consisting of DEAE, heparin, and Sephacryl S-200 chromatography. PI-PLDa and PI-PLDb display a high degree of substrate specificity for PI(3,4,5)P3, with a relative potency of PI(3,4,5)P3 ≫ phosphatidylinositol 3-phosphate (PI(3)P) or phosphatidylinositol 4-phosphate (PI(4)P) > phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) > phosphatidylinositol 3,4-bisphosphate (PI(3,4)P2). In contrast, PI-PLDc preferentially utilizes PI(3)P as substrate, followed by, in sequence, PI(3,4,5)P3, PI(4)P, PI(3,4)P2, and PI(4,5)P2. Both PI(3,4)P2 and PI(4,5)P2 are poor substrates for all three isozymes, indicating that the regulatory mechanisms underlying these phosphoinositides are different from that of PI(3,4,5)P3. None of these enzymes reacts with phosphatidylcholine, phosphatidylserine, or phosphatidylethanolamine. All three PI-PLDs are Ca2+-dependent. Among them, PI-PLDb and PI-PLDc show maximum activities within a sub-μm range (0.3 and 0.9 μmCa2+, respectively), whereas PI-PLDa exhibits an optimal [Ca2+] at 20 μm. In contrast to PC-PLD, Mg2+ has no significant effect on the enzyme activity. All three enzymes require sodium deoxycholate for optimal activities; other detergents examined including Triton X-100 and Nonidet P-40 are, however, inhibitory. In addition, PI(4,5)P2 stimulates these isozymes in a dose-dependent manner. Enhancement in the enzyme activity is noted only when the molar ratio of PI(4,5)P2 to PI(3,4,5)P3 is between 1:1 and 2:1.

and PI(3,4)P 2 are produced by PI 3-kinase in response to a wide array of external stimuli (1,2). These two phosphoinositides and their downstream effector Akt constitute the key component of a major signaling pathway that acts both to stimulate cell growth and to prevent apoptosis (3)(4)(5). In view of their physiological importance, the metabolism of these lipid second messengers has been the focus of many recent investigations. Evidence indicates that they are not susceptible to hydrolysis by any known phospholipase C (6) and that different types of phosphatases mediate the major degradative pathway via dephosphorylation. For example, there exist multiple inositide polyphosphate 5-phosphatases that transform PI(3,4,5)P 3 to PI(3,4)P 2 , through which the ratio of these two lipid second messengers is controlled. These enzymes include PI(3,4,5)P 3 5-phosphatases (7,8), SHIP (SH2-containing inositol 5-phosphatase), (9 -11) or SIP (signaling inositol polyphosphate 5-phosphatase) (12), and synaptojanin (13). Especially noteworthy is the discovery that the PTEN tumor suppressor displays a PI(3,4,5)P 3 3-phosphatase activity (14), thus terminating the second messenger activities of PI(3,4,5)P 3 by converting it to PI(4,5)P 2 . This finding provides an intricate link between PI(3,4,5)P 3 regulation and tumorigenesis (15). Tumor cells with mutant forms of PTEN lack such an off-switch mechanism for PI 3-kinase, thereby containing high levels of PI(3,4,5)P 3 and PI(3,4)P 2 and high endogenous Akt activity (16,17). Loss of PTEN function has been found in a variety of common human cancers, including breast, prostate, and brain cancer (18), and may attribute to the inability of cancer cells to undergo apoptosis (17).
In an effort to gain insight into the complex machinery that regulates PI(3,4,5)P 3 , we have synthesized [1-3 H]PI(3,4,5)P 3 to examine its metabolite fate in rat brain extracts. Here we report the identification of three distinct cytosolic PI-PLD isozymes that convert [1-3 H]PI(3,4,5)P 3 to a novel inositol phosphate [1-3 H]Ins(3,4,5)P 3 and PA. The present data raise a possibility that these PI-PLDs act as a regulator of PI(3,4,5)P 3 in vivo. This premise connotes physiological implications conforming to that of the PTEN tumor suppressor. Furthermore, because PA is generated, these isozymes may provide a putative link between PI 3-kinase and other signaling pathways mediated by PA or its metabolites. * This work was supported by National Institutes of Health Grant R01 GM53448. 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  H]Ins(4,5)P 2 were synthesized according to a modification of the synthetic methods described previously for the respective nonradioactive counterparts (19,20). All synthetic inositol lipids were di-C 16

Assay of PI(3,4,5)P 3 -metabolizing or PI-PLD Activity
During purification, PI(3,4,5)P 3 -metabolizing or PI-PLD activity in all enzyme preparations was assayed by monitoring the liberation of the 3 H-labeled phosphoinositol head group from [1-3 H]PI(3,4,5)P 3 into the medium. [1-3 H]PI(3,4,5)P 3 (0.8 g; total radioactivity, 0.2 Ci), PE (40 g), and PS (5 g) were suspended in 1 ml of 20 mM Hepes, pH 7, containing 120 mM KCl, 10 mM NaCl, 2 mM EGTA, and 0.8 mM sodium deoxycholate. Various amounts of CaCl 2 were added to the mixture before assays, and the free Ca 2ϩ concentration was calculated by a computer program developed by Karl-Josef Foehr and programmed by Wokciech Warchei (1990, version 2.1). The suspension was sonicated in a water bath-type sonicator for 5 min, and mixed vigorously with a vortex mixer before assays. Various enzyme preparations (10 l) were incubated with 40 l of the aforementioned Hepes buffer. The reaction was initiated by adding 50 l of the phospholipid mixture, incubated at 37°C for 30 min, and stopped by adding 200 l of 10% trichloroacetic acid and 150 l of 10 mg/ml bovine serum albumin. The mixture was centrifuged at 15,000 ϫ g for 5 min, and the radioactivity in the supernatant was measured by liquid scintillation. The composition of the control was identical to that mentioned above except that the enzyme preparation was replaced by an equal amount of distilled H 2 O.

Partial Purification of PI-PLD Isozymes from Rat Brain Cytosol
Rat brain was minced and suspended in 8 volumes of ice-cold 10 mM Tris/HCl, pH 7.4, containing 0.25 M sucrose, 1 mM dithiothreitol, 20 g/ml leupeptin, 2 g/ml pepstatin A, and 1 mM AEBSF. The small pieces were homogenized in a Dounce homogenizer with six strokes up and down. The homogenate was centrifuged at 1,000 ϫ g to remove cell debris and intact nuclei. The supernatant was centrifuged at 100,000 ϫ g for 1 h to prepare the cytosolic fraction.
Step 1. DEAE Chromatography-After being dialyzed against 50 mM Tris/HCl, pH 7.4, containing 1 mM dithiothreitol (buffer A) for 12 h, the cytosolic fraction (1.8 g of protein; 0.039 nmol/mg/min PIP 3 -metabolizing activity) was loaded onto a Toyopearl DEAE-650M column (3 ϫ 10 cm) previously equilibrated with buffer A. The column was washed with 400 ml of buffer A followed by 300 ml of a linear NaCl gradient of 0 -200 mM in the same buffer. Fractions of 2.25 ml were collected after the gradient started. PI(3,4,5)P 3 -metabolizing activity was analyzed as described above at two different free Ca 2ϩ concentrations, 0.3 and 20 M, respectively. Two different activity profiles were noted under these two Ca 2ϩ levels ( Fig. 1), indicating the presence of more than one PI(3,4,5)P 3 -metabolizing enzyme. Fractions 271-289 and 295-312, designated as D1 and D2, were pooled separately, concentrated over a molecular weight 10,000 cutoff ultrafiltration membrane (Filtron Omega unit), and dialyzed against buffer A for 12 h.
Step 2. Heparin Chromatography-The dialyzed D1 and D2 samples (76.7 and 79.7 mg of protein; 0.6 and 0.58 nmol of PIP 3 /min/mg specific activity, respectively) from step 1 were individually applied to a heparin-agarose column (1 ϫ 10 cm) equilibrated with buffer A. The column was washed with 85 ml of buffer A, and the adsorbed proteins were eluted with 150 ml of a linear gradient of 0 -400 mM NaCl. Fractions of 1 ml were collected. As shown in Fig. 2, A and B, both D1 and D2 were resolved into two PI(3,4,5)P 3 -metabolizing activity peaks. For D1, fractions 145-160 (designated as D1/H1) and fractions 177-192 (designated as D1/H2) were pooled separately and concentrated by ultrafiltration.

Identification and Quantitation of the Radiolabeled Inositol Phosphates by HPLC
For the substrate specificity study, the identity and quantity of the radioactive phosphoinositols liberated by the enzymatic hydrolysis of  (1,5,20,40, and 60 min) by extracting the mixture with 200 l of HClO4/CHCl 3 . After a brief centrifugation, the two phases were separated. The aqueous phase was treated as described above for the HPLC analysis of Ins(3,4,5)P 3 . The organic phase was transferred to a new vial and dried by a stream of N 2 . The residue was dissolved in CHCl 3 , spotted onto 1% oxalic acid-treated TLC plate, and developed with n-propyl alcohol and 2 M acetic acid (13: 7) overnight. After drying, spots were located by autoradiography and compared with standards. The autoradiograms were scanned by a photodyne image system. The spots corresponding to [ 14 C]PA and [ 14 C]PI(3,4,5)P 3 were scraped off the plate, and the associated radioactivity was measured by liquid scintillation.

RESULTS
To investigate the metabolic fate of D-3 phosphoinositides, [1-3 H]PI(3,4,5)P 3 was synthesized and exposed to the cell lysate of rat brain. The incubation mixture was extracted with CHCl 3 /CH 3 OH to isolate the phosphoinositide metabolites. However, substantial radioactivity appeared in the aqueous phase in a time-and protein concentration-dependent manner, suggesting that the 3 H-labeled head group was released from [1-3 H]PI(3,4,5)P 3 via phospholipase hydrolysis. HPLC analysis of the aqueous fraction revealed that the liberated radioactivity was associated with free inositol and trace amounts of inositol mono-and bisphosphates (data not shown). This result showed that the inositol phosphate generated from PI(3,4,5)P 3 was rapidly metabolized in the crude extract. Other rat tissues examined including the liver, the kidney, and platelets also contained such PI(3,4,5)P 3 -metabolizing activity.
As part of our effort to verify the identity of the PI(3,4,5)P 3metabolizing enzyme(s), purified phospholipase preparations from different sources were examined for the activity toward [1-3 H]PI-(3,4,5)P 3 . These included porcine pancreas PLA 2 , recombinant PLC-␥1, recombinant PLC-␦1, B. cerus PI-specific PLC, recombinant PLD1, and cabbage PLD. However, none of these enzymes showed appreciable hydrolysis of [1-3 H]PI(3,4,5)P 3 . With regard to enzyme inhibition, neomycin (1 mM), which binds phosphoinositides with high affinity (22), completely blocked the hydrolysis by rat brain extracts. Other known phospholipase inhibitors such as aristolochic acid for PLA 2 , ET-18-OCH 3 for PI-PLC, and dihydro-D-erythro-sphingosine for PLD gave no or only partial inhibition of the enzyme activity even at concentrations 20 times over the corresponding IC 50 values (data not shown).
These data prompted us to identify the enzyme(s) responsible for [ 3 H]PI(3,4,5)P 3 hydrolysis. Subcellular fractionation indicated that more than 85% of the [ 3 H]PI(3,4,5)P 3 -metabolizing activity resided in the cytosolic fraction (100,000 ϫ g supernatant). The rest of the activity was associated with various membrane fractions including the plasma membrane and microsomes. Thus, the rat brain cytosol was subjected to chromatographic purification by DEAE ion exchange, heparin-agarose, and Sephacryl S-200-HR columns, in tandem. To explore the possibility that there might exist multiple responsible enzymes with different Ca 2ϩ requirements, PI(3,4,5)P 3 -metabolizing activity was monitored at low and high Ca 2ϩ concentrations, represented by 0.3 and 20 M, respectively, throughout the purification. As shown in the DEAE chromatographic profile (Fig. 1), a broad peak exhibiting PI(3,4,5)P 3 -metabolizing activity was detected at each Ca 2ϩ level (0.3 M, closed triangles; 20 M, open squares).
Nevertheless, the activity peaks detected at 0.3 and 20 M Ca 2ϩ showed different migration patterns in the elution profile. This finding suggested that there were two or more PI(3,4,5)P 3metabolizing enzymes with different Ca 2ϩ dependence among these active fractions. Thus, fractions 271-289 and 295-312, designated as D1 and D2, were pooled separately. The subsequent heparin chromatography provided definite evidence that D1 and D2 each contained two PI(3,4,5)P 3 -metabolizing activities (Fig. 2, A and B).
As shown, fraction D1 gave two well resolved activity peaks between fractions 150 and 210 ( Fig. 2A). Among these active fractions, the enzyme activity of the first peak increased with higher [Ca 2ϩ ] (0.3 M, closed triangles; 20 M, open squares), whereas the activity of the second peak was unaffected by Ca 2ϩ change. Accordingly, fractions 145-160 and 177-192 were pooled separately and were designated as D1/H1 and D1/H2, respectively. Similarly, two PI(3,4,5)P 3 -metabolizing activity peaks were also noted for fraction D2 after the heparin column ( Fig. 2B; 0.3 M, closed triangles; 20 M, open squares). In contrast, the first peak exhibited a higher activity at low Ca 2ϩ compared with high Ca 2ϩ , whereas the second, smaller peak showed no significant difference in the Ca 2ϩ requirement. Considering that fractions D1 and D2 were juxtaposed in the DEAE elution profile, it was possible that the second peak from fraction D2 was identical to fraction D1/H2 in light of their elution times and Ca 2ϩ dependence. Thus, only the first activity peak was collected, which was designated as D2/H. Fractions D1/H1, D1/H2, and D2/H were chromatographed further on a Sephacryl S-200-HR column (Fig. 3, A-C).
The resulting active fractions were designated as D1/H1/S, D1/H2/S, and D2/H/S for further discussion. This three-step procedure resulted in 354-, 1,056-and 1,027-fold purification with specific activities of 13.7, 40.9, and 39.8 nmol of PI(3,4,5)P 3 /mg of protein/min ([Ca 2ϩ ] ϭ 2 M) for fractions D1/H1/S, D1/H2/S, and D2/H/S, respectively. Concerning PI(3,4,5)P 3 -metabolizing activity, fraction D2/H/S was highly unstable. Up to 80% of the enzyme activity was lost after storing at 0°C for 2 days. The Ca 2ϩ dependence of these enzyme preparations was examined. Fig. 4 indicates that the PI(3,4,5)P 3 -metabolizing activity of these enzymes was inhibited by EDTA, and the inhibition could be overcome by adding Ca 2ϩ in a concentrationdependent manner. Both D1/H2/S and D2/H/S showed a similar Ca 2ϩ requirement, with maximum PI(3,4,5)P 3 -metabolizing activities at a sub-M range (0.3 and 0.9 M, respectively), whereas D1/H1/S displayed an optimal [Ca 2ϩ ] at 20 M. The difference in the Ca 2ϩ requirement underscored a potential distinction in the roles of these enzymes in PI(3,4,5)P 3 metabolism. Moreover, Mg 2ϩ had no appreciable effect by itself on the activity of or on the Ca 2ϩ dependence for any of these enzymes (data not shown).
In an effort to gain insight into the catalytic behaviors of these enzymes, we also synthesized [1-3 H]PI(3,4)P 2 , [1-3 H]PI-(4,5)P 2 , [1-3 H]PI(3)P, and [1-3 H]PI(4)P for examinations. These phospholipids were exposed to individual enzymes, and the released water-soluble products were analyzed by reversephase HPLC, aiming at both product identification and substrate specificity determination. Representative HPLC profiles of the [1-3 H]phosphoinositol products from incubations of the respective substrates with fraction D1/H2/S are shown in Fig. 5.
These HPLC profiles revealed two important findings. First, D1/H2/S displayed a high degree of substrate specificity for PI(3,4,5)P 3 . The relative potency for the substrates examined was PI(3,4,5)P 3 Ͼ Ͼ PI(4)P Ͼ PI(3)P Ͼ Ͼ PI(4,5)P 2 and PI(3,4)P 2 . The utilization of the latter two, especially PI(3,4)P 2 , accounted for less than 5% of that of PI(3,4,5)P 3 (Fig. 6, panel A). These HPLC results were consistent with those obtained by measuring [ 3 H]phosphoinositol release from the respective substrates by liquid scintillation.    (1,4)P 2 , of which the retention times would differ from the respective experimental data by almost 10 min because of an additional phosphate moiety in the PLC products. Moreover, it is worthy to note that after periodate oxidation/ NaBH 4 reduction, the radioactivity associated with the hydrolysis product of PI(3,4,5)P 3 was completely lost, confirming that the adjacent 1-, 2-, and 6-hydroxyls were unsubstituted (data not shown).
The effect of detergents on the PI-PLD activity was investigated. In all of the aforementioned assays, the reaction mixture contained 0.8 mM sodium deoxycholate. Removal of the detergent or replacement with 0.1-1% Nonidet P-40 or Triton X-100 resulted in substantial loss of enzyme activity for all three isozymes, indicating the stringent requirement of sodium deoxycholate for PI-PLD activity. This dependence might be attributable to the effect of detergent on PI(3,4,5)P 3 packaging in lipid vesicles, which affected the substrate availability and/or enzyme accessibility.
Earlier studies have shown that PC-PLDs were strongly stimulated by PI(4,5)P 2 and PI(3,4,5)P 3 with an equal potency (23,24). Fig. 8 depicts the effect of PI(4,5)P 2 on the PI(3,4,5)P 3metabolizing activity of three PI-PLD isozymes. The individual enzymes were incubated with [ 3 H]PI(3,4,5)P 3 in the presence of increasing amounts of PI(4,5)P 2 without sodium deoxycholate. As shown, PI(4,5)P 2 enhanced the basal enzyme activity up to 2.5-fold. However, this stimulating effect occurred only within a narrow range of PI(4,5)P 2 /PI(3,4,5)P 3 molar ratios between 1:1 and 2:1. Excess amounts of PI(4,5)P 2 either inhibited or had no effect on the PI(3,4,5)P 3 -metabolizing activity. It is plausible that because PI(4,5)P 2 was a poor substrate, it might compete with PI(3,4,5)P 3 for enzyme binding, thereby counteracting its stimulating effect. DISCUSSION PI 3-kinase activation leads to a transient accumulation of PI(3,4,5)P 3 and PI(3,4)P 2 , of which the concentrations rise from 0.05-0.2 M at resting states to 1-2 M upon agonist stimulation (25). The prevailing levels of these PI 3-kinase lipid products are regulated by a delicate balance between its rates of synthesis and metabolism. This study presents the first evidence that there exist at least three distinct cytosolic Ca 2ϩ -dependent PLD isozymes that may take part in PI(3,4,5)P 3 regulation in vivo. Taken