Purification and Characterization of Phosphoinositide 3-Kinase from Rat Liver*

Phosphoinositide 3-kinase was purified 27,000-fold from rat liver. The enzyme was purified by acid precipitation of the cytosol followed by chromatography on DEAE-Sepharose, S-Sepharose, hydroxylapatite, Mono-Q, and Mono-S columns. When analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, the purified phosphoinositide 3-kinase preparation contained an 85-kDa protein and a protein doublet of approximately 110 kDa. The 85- and 110-kDa proteins focus together on native isoelectric focusing gels and are cross-linked by dithiobis(succinylamide propionate), showing that the 110- and 85-kDa proteins are a complex. The apparent size of the native enzyme, as determined by gel filtration, is 190 kDa. The 85-kDa subunit is the same protein previously shown to associate with polyoma virus middle T antigen and the platelet-derived growth factor receptor (Kaplan, D. R., Whitman, M., Schaffhausen, B., Pallas, D. C., White, M., Cantley, L., and Roberts, T. M. (1987) Cell 50, 1021-1029). The two proteins co-migrate on two-dimensional gels; and, using a Western blotting procedure, 32P-labeled middle T antigen specifically blots the 85-kDa protein. The purified enzyme phosphorylates phosphatidylinositol, phosphatidylinositol 4-phosphate, and phosphatidylinositol 4,5-bisphosphate. The apparent Km values for ATP were found to be 60 microM with phosphatidylinositol 4-phosphate or phosphatidylinositol 4,5-bisphosphate as the substrate. The apparent Km for phosphatidyinositol is 60 microM, for phosphatidylinositol 4-phosphate is 9 microM, and for phosphatidylinositol 4,5-bisphosphate is 4 microM. The maximum specific activity using phosphatidylinositol as the substrate is 0.8 mumol/mg/min. The enzyme requires Mg2+ with an optimum of 5 mM. Substitution of Mn2+ for Mg2+ results in only approximately 10% of the Mg2(+)-dependent activity. Physiological calcium concentrations have no effect on the enzyme activity. Phosphoinositide 3-kinase has a broad pH optimum around 7.

The enzyme was purified by acid precipitation of the cytosol followed by chromatography on DEAE-Sepharose, S-Sepharose, hydroxylapatite, Mono-Q, and Mono-S columns.
When analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, the purified phosphoinositide 3-kinase preparation contained an 85-kDa protein and a protein doublet of -110 kDa. The 85-and llO-kDa proteins focus together on native isoelectric focusing gels and are cross-linked by dithiobis(succinylamide propionate), showing that the llO-and 85-kDa proteins are a complex. The apparent size of the native enzyme, as determined by gel filtration, is 190 kDa. The 85-kDa subunit is the same protein previously shown to associate with polyoma virus middle T antigen and the plateletderived growth factor receptor (Kaplan, D. R., Whitman, M., Schaffhausen, B., Pallas, D. C., White, M., Cantley, L., and Roberts, T. M. (1987) Cell 50, 1021-1029).
The two proteins co-migrate on two-dimensional gels; and, using a Western blotting procedure, 32P-labeled middle T antigen specifically blots the 85-kDa protein.
The apparent K,,, values for ATP were found to be 60 PM with phosphatidylinositol as the substrate and 30 pM with phosphatidylinositol 4-phosphate or phosphati-dylinositol4,5-bisphosphate as the substrate. The apparent K,,, for phosphatidyinositol is 60 pM, for phos-phatidylinositol4-phosphate is 9 FM, and for phosphatidylinositol 4,5-bisphosphate is 4 pM. The maximum specific activity using phosphatidylinositol as the substrate is 0.8 pmol/mg/min. The enzyme requires Mg2+ with an optimum of 5 mM. Substitution of Mn2+ for Mg2+ results in only -10% of the Mg'+-dependent activity.
Physiological calcium concentrations have no effect on the enzyme activity.
Phosphoinositide 3-kinase has a broad pH optimum around 7.
Polyphosphoinositides and their metabolites are crucial intracellular signals in the responses to a number of hormones and growth factors (1,2). The recent discovery of a phosphatidylinositol kinase that phosphorylates PI' at the D-3 position of the inositol ring uncovered a new pathway of PI metabolism and potential intracellular signals (3). This pathway is distinct from the pathway which leads from PI to PI-4,5-P2 and then to diacylglycerol and inositol 1,4,5-trisphosphate through the action of phospholipase C.
Phosphoinositide 3-kinase was first discovered in pp60'~"" immunoprecipitates from transformed cells (4). It was subsequently found in middle T immunoprecipitates from polyoma middle T antigen-transformed cells (5) and more recently has been found to coimmunoprecipitate with PDGF (6-8), colony-stimulating factor 1 (9), and insulin receptors (10) in ligand-stimulated cells. Analysis of mutants has closely linked the PI 3-kinase pathway to transformation by several oncogene products and to the mitogenic response to PDGF. Mutants of middle T antigen with which PI 3-kinase does not associate are nontransforming, and PDGF receptor mutants that do not associate with PI 3-kinase do not have a mitogenic response to [11][12][13][14][15][16].
Although these studies indicate that PI 3-kinase produces a crucial second messenger, the signal(s) from this pathway have not yet been identified. In addition to producing PI-3-P, immunoprecipitates of middle T antigen or the PDGF receptor also have enzymatic activities that phosphorylate PI-4-P and PI-4,5-P* to produce PI-3,4-PZ and PIP3 (probably phosphatidylinositol 3,4,5-trisphosphate) (12, 17). Quiescent smooth muscle cells and fibroblasts have detectable levels of PI-3-P, but the levels of this lipid do not change appreciably with stimulation by PDGF or transformation by middle T antigen. PI-3,4P2 and PIP3 are not detectable prior to stimulation or transformation, but reach significant levels after stimulation with PDGF or expression of middle T antigen (12,17). Similarly, Chinese hamster ovary cells, transfected with the human insulin receptor, show elevated levels of PI-3,4-P2 and PIP, in response to insulin (10 C has been found which will cleave any of these phosphoinositides (18,19), suggesting that PI-3,4-P2 and/or PIP,, themselves are the crucial signals emanating from this pathway, rather than inositol polyphosphates.
Although there is a strong correlation between PI 3-kinase activity and cell growth, D-3 phosphorylated phosphoinositides have also been found in nongrowing cells. PIP:* (probably phosphatidylinositol 3,4,5trisphosphate) is found in neutrophils stimulated by fMet-Let-Phe (20, 21). Platelets also contain PI 3-kinase activity' and form PI-3,4-Pe when stimulated by thrombin (22). PI-3,4-P2 has been found to appear in a Leydig tumor cell line after stimulation with epidermal growth factor, which acts as a differentiating, rather than mitogenic, factor for these cells (23). These findings suggest that D-3 phosphorylated polyphosphoinositides are also involved in nonmitotic act,ivation of terminally differentiated cells.
Previous work (6, 7, 13, 16) has implicated an 85-kDa phosphoprotein as the PI 3-kinase because of the correlation of its presence with PI kinase activity in middle T antigen and PDGF receptor mutants. This 85kDa protein is phosphorylated on serine, threonine, and tyrosine in response to PDGF stimulation or middle T antigen transformation (6). In this paper, we show that the PI 3-kinase purified from rat liver cytosol is a heterodimer of the previously described 85 kDa protein and a IlO-kDa protein.

MATERIALS
AND METHODS AND RESULTS" 85-and llO-kDa Proteins in Purified PI 3-Kinase-SDS-PAGE analysis of fractions from the Mono-S column revealed an 85-kDa protein and a protein doublet at 110 kDa that both peaked with PI 3-kinase activity. A silver-stained SDS-polyacrylamide gel of the middle two fractions from the activity peak of the Mono-S column eluate is shown in Fig. 2 (upper). Early fractions of PI kinase activity peak from the Mono-S column contained primarily the lower IlO-kDa protein, whereas the later fractions contained primarily the upper band. PI kinase activity and the amount of 85-kDa protein present both correlated with the sum of the amount of protein in the lower and upper IlO-kDa bands. In some preparations, trace amounts of protein were seen at 65,55, and 45 kDa that seem to be proteolytic fragments of the 85-kDa protein (see below). Only the llO-and 85-kDa bands reproducibly correlate with PI 3-kinase activity.
To determine whether the llO-and 85-kDa proteins copurified because they are in a complex or simply because they behave similarly through the purification steps, we did twodimensional gels with native isoelectric focusing in the first dimension and SDS-PAGE in the second dimension. The llO-and 85-kDa proteins focused together in two spots on the native isoelectric focusing gel (Fig. 2, center). The resolution of the second dimension did not give a clear indication of whether the upper, lower, or both IlO-kDa bands were present. When isoelectric focusing was done under denaturing conditions, the llO-and 85-kDa proteins focused at different positions ( Fig. 2, lower). The 85-kDa protein has a pI of -5, and the IlO-kDa bands have similar p1 values, centered around 6.5. The lower band of the llO-kDa doublet is more acidic than the upper band, consistent with the protein found " C. L. Carpenter, unpublished results. " Portions of this paper (including "Materials and Methods," part of "Results," Figs. 1, 4, and 6-10, and Tables I and II)  in the lower band eluting first from the Mono-S cationexchange column (Fig. 2,upper). Since the 85-and IlO-kDa proteins focused together under native, but not denaturing, isoelectric focusing conditions, they appear to exist as a complex in the native state. Additional evidence that the proteins are a complex is provided by cross-linking studies. When the purified material was incubated with dithiobis(succinylamide propionate), a reversible cross-linking agent, a protein complex of 200 kDa was seen when analyzed by one-dimensional SDS-PAGE.
When cross-linked material, separated in a first dimension tube gel, was separated in a second dimension slab gel with /&mercaptoethanol to reverse the cross-linking, it was evident that the 85-and IlO-kDa proteins were crosslinked in the same complex (Fig. 3). When cross-linking was done on less pure preparations of PI 3-kinase, other proteins present did not cross-link into the complex with the 85-and IlO-kDa proteins.
The apparent molecular mass of PI 3-kinase by gel filtration is -190 kDa (Fig. 4), which is consistent with a heterodimer of the llO-and 85-kDa proteins.
On silver-stained SDS-PAGE of the gel filtration column fractions, the llO-and 85-kDa proteins co-migrated with PI 3-kinase activity (data not shown).
Purified 85-kDa Protein Is the Same 85-kDa Protein That Associates with Middle T Antigen and PDGF Receptor-An 85-kDa phosphoprotein has been implicated as a component of PI 3-kinase because its presence correlates with PI 3-kinase activity in middle T antigen and PDGF receptor immunopre- Cross-linking was done as described under "Materials and Methods." The proteins were separated in a tube gel under nonreducing conditions. The proteins in the tube gel were then separated in a second dimension under reducing conditions to break the cross-linking. The 200.kDa marker is based on the molecular size of the cross-linked complex run on one-dimensional nonreducmg SDS-PAGE. protein that purifies with PI 3-kinase activity is the same protein that associates with middle T antigen.
CJpper, purified PI 3-kinase (50 ng) and a middle T immunoprecipitate from "S-labeled middle T antigen-transformed fibroblasts were run together on a denaturing two-dimensional gel. The top panel is an autoradiograph of the "S-labeled proteins; the bottom panel is a silver stain of the same gel to show the position of the purified 85-kDa protein. b indicates the basic end and a the acidic end of the isoelectric hocusing gel. Lower, samples of rat liver cytosol (lane A) and of nurified PI 3-kinase (lane B) were seDarated bv 5% SDS-PAGE and transferred to nitrocellulose. The nit;ocellulose was probed with 'IPlabeled middle T antigen as described under "Materials and Methods." cipitates (6). To determine whether we have purified the same protein, we compared the position of the two proteins on twodimensional gels. Purified PI 3-kinase was spiked, added to a trace of an anti-middle T immunoprecipitate of [""Slmethionine-labeled middle T antigen-transformed fibroblasts, and the proteins were separated by two-dimensional gel electrophoresis. The middle T antigen-associated 85-kDa protein has been shown previously (26, 29) to resolve into a series of evenly spaced spots upon isoelectric focusing. The more acidic spots seem to be higher phosphorylated forms of the protein since, with phosphatase treatment, the spots resolve to the most basic form. The major silver-stained spot of the 85-kDa subunit of purified PI 3-kinase co-migrated identically to the most basic form of the [""Slmethionine-labeled 85-kDa protein cluster, indicating that it is identical to the nonphosphorylated form of the middle T antigen-associated 85-kDa protein (Fig. 5, upper). In some preparations, a more acidic form of the 85-kDa protein was also detected, which is likely phosphorylated (Fig. 2, lower). Polyoma Middle T Antigen Specifically Associates with 85-KDa Subunit of Purified PI 3-Kinase.-More dramatic evidence that the 85-kDa subunit of purified PI 3-kinase is the protein previously shown to associate with middle T antigen is provided by a middle T antigen blotting experiment.
Cohen et al. (29) previously showed that if middle T antigen, expressed in a baculovirus system along with c-src, is phosphorylated by pp60""' and then purified, middle T antigen will form a tight complex with the 85-kDa protein from cell lysates. In addition, if proteins from total cell lysates are separated by SDS-PAGE and transferred to nitrocellulose, "LP-labeled middle T antigen can be used to blot the proteins, and only an 85-kDa protein is detected. To determine whether the blotted protein is the 85-kDa subunit of PI 3-kinase, this procedure was used with a rat liver homogenate and purified PI 3-kinase. As shown in Fig. 5 (lower), middle T antigen specifically blotted an 85-kDa band in the rat liver homogenate (lane A) that was dramatically enhanced in the purified PI 3-kinase preparation (lane B). This is a remarkable result and indicates that phosphorylated middle T antigen specifically associates with the SDS-denatured 85-kDa protein from cell lysates even though the protein is low in abundance. On some blots of purified PI 3-kinase, weak bands at 65, 55, and 45 kDa were visible. It is likely that these are proteolytic products of the 85-kDa protein. The llO-kDa proteins did not blot, indicating that they are not precursors of the 85-kDa protein.
Characteristics of Enzyme-PI 3-kinase activity was found to be maximal at an Mg'+ concentration of 5 mM (Fig. 6). Activity using Mn2+ instead of Mg2+ was 10% or less of the M$+-dependent activity. Calcium at physiological concentrations (l-100 FM) had no effect on activity. The pH optimum is broad and centered around 7 (Fig. 7).
The Km capp) for ATP is 60 pM when PI is used as the substrate and 30 pM when PI-4-P or PI-4,5-P2 is used as the substrate (Fig. 8). We also determined the dependence of the reaction on phosphoinositide concentration (keeping the molar ratio of phosphoinositide to PS at 1) (Fig. 9). The K,, (n,,Pl for PI is 80 FM, for PIP is 10 FM, and for PIP, is 4 pM. The maximum specific activity (based on the data shown in Fig.  8) for PI 3-kinase activity is 0.8 pmol/mg/min, for PI-4-P kinase activity is 0.13 pmol/mg/min, and for PI-4,5,-P2 kinase activity is 0.22 pmol/mg/min. Table I presents the specific  activities for each step of the purification when all three substrates are present simultaneously and also under conditions of maximal specific activity when only one substrate is present.
The effect of phospholipid composition of the liposomes on substrate utilization was further investigated. When all three substrates are present in equimolar amounts with 40% PS as carrier, the relative rate of formation of product is: PIP, 100%; PIPa -2O%, and PIPS, -40%. PI-4-P and PI-4,5-P* are very poor substrates when used without a carrier lipid such as PS. Unexpectedly, when crude brain phosphoinositides (PS:PI:PIP:PIP2 of -30:30:15:15) were used as substrates. PIP, was the preferred substrate (Table I). Crude brain phosphoinositides (Sigma) seem to contain a compound that specifically enhances PIP2 kinase activity. This phenomenon also seems to be batch-dependent.
Nonionic detergents inhibit PI 3-kinase when present above the critical micellar concentration.
When used in concentrations below the critical micellar concentration, there is an activation of the enzyme (Fig. 10). We also examined the effects of other lipids on enzyme activity and found that addition of phosphatidylcholine, phosphatidic acid, lysophosphatidic acid, or cardiolipin inhibited activity and that diacylglycerol had little effect (Table II). We found minimal inhibition of the enzyme by adenosine at concentrations up to 100 FM, as previously shown (8). DISCUSSION PI 3-kinase is a heterodimer of llO-and 85-kDa proteins. We have not been able to identify which subunit has PI kinase activity. We have been unable to separate the two proteins under native conditions and have also been unable to renature activity once they have been separated by SDS-PAGE. The 85-kDa component of the heterodimer is the same 85-kDa protein that is found in association with middle T antigen and the PDGF receptor, based on the blot done with labeled middle T antigen and the identical migration of the proteins on two-dimensional gels (Fig. 5). Previous work (6) has shown that the 85-kDa proteins associated with middle T antigen and the PDGF receptor are the same. Whether the llO-kDa subunit of PI 3-kinase associates with middle T antigen and/ or the PDGF receptor along with the 85-kDa subunit is not yet known. Proteins of -110 kDa have been found to associate with the PDGF receptor, however (13).
The two IlO-kDa proteins seem to be very closely related, but not identical proteins. They have similar isoelectric points, and both appear to associate with the 85-kDa protein.
They also have similar but not identical protease V8 fragments (data not shown). Protein sequence analysis of protease V8 fragments reveals what seems to be similar, but not identical amino acid sequences, suggesting that the upper and lower 1 lo-kDa bands are products of different genes.
PI 3-kinase is found in the cytosol and apparently is recruited to the membrane when oncogene products are present or receptors are activated (6,8,29). Recruitment to the membrane may allow modulation of the enzyme's activity or substrate specificity through protein-protein interactions or phosphorylation. At the membrane, PI 3-kinase would be in close proximity to its substrates.
All other PI kinases which have been purified are membrane-associated, and all phosphorylate PI at the D-4 position of the inositol ring (8, [31][32][33][34]. The PIP kinase which has been purified from human erythrocytes phosphorylates PI-4-P at the D-5 position of the inositol ring (35). These enzymes are in the classic phosphatidylinositol pathway which leads to synthesis of PI-4,5-P2.
The physiologically relevant substrate for PI 3-kinase is not known. Undoubtedly, PI-3-P is produced by this enzyme in uiuo. The purified protein phosphorylates PI, PIP, and PIP*, so it is possible that the PI-3,4-P* and PIPS found in uiuo could be produced by phosphorylation of PI-4-P and PIP-4,5-P2 at the D-3 position by PI 3-kinase. It is also possible that some or all of the PI-3,4-P2 found in vivo is synthesized by the phosphorylation of PI-3-P at position 4. Three types of PI kinases have been described (8,31). Type I is PI 3kinase. Types II and III are PI 4-kinases that can be distinguished by detergent and adenosine sensitivity and a specific inhibitory antibody.4 Type II PI kinase from human red blood cells does not phosphorylate PI-3-P.5 This is the predominant PI 4-kinase in most cells, and it seems to function solely in the classic PI turnover pathway. Whether there are PI 4kinases that can phosphorylate PI-3-P remains to be determined. The correlation between the association of PI 3-kinase 4 Endemann, G. C., Graziani, A., and Cantley, L. C. (1990) Biochem. J., in press. "A. Graziani, K. R. Auger, L. A. Serunian, and L. C. Cantley, unpublished results. with the PDGF receptor and the elevation of PI-3,4-PZ and PIP3 (but not PI-3-P) levels in smooth muscle cells suggests that the PDGF receptor enhances the ability of PI 3-kinase to utilize PI-4-P and PI-4,5-PZ as substrates in viuo (17) Deacylatio". deglycerat~c" and HPLC a"alysi* of the products of the phosphollpid kl"a** r*aci~c"s were do"* as prewusly described (17