Group IV Cytosolic Phospholipase A2 Binds with High Affinity and Specificity to Phosphatidylinositol 4,5-Bisphosphate Resulting in Dramatic Increases in Activity*

The group IV cytosolic phospholipase A2 (cPLA2) exhibits a potent and specific increase in affinity for lipid surfaces containing phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P2) at physiologically relevant concentrations. Specifically, the presence of 1 mol % PtdIns(4,5)P2 in phosphatidylcholine vesicles results in a 20-fold increase in the binding affinity of cPLA2. This increased affinity is accompanied by an increase in substrate hydrolysis of a similar magnitude. The binding studies and kinetic analysis indicate that PtdIns(4,5)P2 binds to cPLA2 in a 1:1 stoichiometry. The magnitude of the effect of PtdIns(4,5)P2 is unique among anionic phospholipids and larger than that for other polyphosphate phosphatidylinositols. The effect of PtdIns(4,5)P2 on the activity of cPLA2 is at least an order of magnitude larger than the concomitant changes in the fraction of the enzyme associated with lipid membranes. Striking parallels between the interaction of cPLA2 with PtdIns(4,5)P2 and the interaction of the pleckstrin homology domain of phospholipase Cδ1 with PtdIns(4,5)2 combined with sequence analysis of cPLA2 lead us to propose the existence and location of a pleckstrin homology domain in cPLA2. We further show that the very nature of the interaction of proteins such as cPLA2 with multiple ligands incorporated into membranes follows a specific model which necessitates the use of an experimental methodology suitable for a membrane interface to allow for a meaningful analysis of the data.

Cytosolic phospholipase A 2 (cPLA 2 ) 1 has unique structural and regulatory properties within the PLA 2 superfamily (1,2). Current widespread interest in the cytosolic phospholipase A 2 stems from its putative role as a key enzyme involved in the inflammatory response, a provider of arachidonic acid, the precursor for prostaglandins and leukotrienes (3,4). This enzyme was also implicated in the regulation of several other processes such as platelet activation, cell proliferation, and the generation of several second messengers (3)(4)(5). A mobilization of intracellular Ca 2ϩ was implicated in the translocation of cPLA 2 to cellular membranes resulting in the subsequent specific liberation of arachidonic acid from the sn-2 position of phospholipids (6,7). This scenario of activation of cPLA 2 was validated by the discovery that the N terminus of this enzyme contains an autonomous calcium and lipid binding domain, CaLB (7), homologous to the C2 domain (8) of protein kinase C and several other proteins shown to associate with lipid membranes in a Ca 2ϩ -dependent fashion (7,9). Both the native enzyme and its CaLB domain were shown to associate with fragments of cellular membranes and synthetic lipid vesicles at physiologically relevant Ca 2ϩ concentrations (6,7,10,11). Thus far, the CaLB domain is the only recognized regulatory domain of cPLA 2 (5).
It was noted early on (12) that several anionic phospholipids including phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P 2 ) activated cPLA 2 . This stimulatory effect was hypothesized to result from an enhancement of the partitioning of cPLA 2 into lipid membranes caused in general by all anionic lipids. We have now employed an approach that was specifically designed for the characterization of proteins that bind to membranes through multiple attachment points to demonstrate that cPLA 2 binds in a 1:1 stoichiometry with high affinity and specificity to PtdIns(4,5)P 2 in lipid vesicles, and the effect is quite distinct from that of other anionic lipids. Furthermore, the resulting increase in membrane affinity is accompanied by a quantitative increase in enzymatic activity. In addition, an apparent functional similarity between cPLA 2 and phospholipase C␦ 1 (PLC␦ 1 ) now allows us to propose the location of a pleckstrin homology (PH) domain in cPLA 2 .
It is now recognized that many of the key proteins involved in signal transduction possess multimodular structures (13). Already hundreds of multimodular proteins have been identified that contain domains such as the PH (14,15), C1 (16), C2 (8,16), src homology-2 (13), and src homology-3 (13) domains. These modules can specifically bind certain membrane proteins (src homology-2, src homology-3, C2, and PH domains) and specific membrane lipids (C1, C2, and PH domains). This fact highlights the need for studies of multimodular membranebinding proteins that take into account the specific properties of the interactions occurring within the two-dimensional surface of cellular membranes (17) as demonstrated here for cPLA 2 .
Binding Assay-The association of cPLA 2 with large unilamellar vesicles (LUVs) (pore diameter 0.2 m) was quantitated in a sucroseloaded vesicle assay originally devised for the study of PLC␦ 1 (21). In brief, LUVs were produced by extrusion (22) in 170 mM sucrose, 20 mM PIPES, pH 6.8. The lipid composition varied depending on the experiment. Subsequently, the LUVs were suspended in 5 volumes of solution containing 100 mM KCl, 20 mM PIPES, pH 6.8 (referred to hereafter as the "standard binding solution"), and centrifuged at 150,000 ϫ g for 15 min at 25°C. The supernatant was discarded, and the pellet was resuspended in KCl/PIPES buffer. Enzyme (0.5 g/ml) and lipid vesicles (0.1 to 1 mM total lipid) were suspended in KCl/PIPES buffer with 0.1 mg/ml BSA, 0.5 mM EGTA and CaCl 2 at a concentration required to achieve the desired concentration of free ion, calculated according to a published algorithm (23). This suspension was allowed to equilibrate for 10 min and was then centrifuged again for 15 min at 25°C at 150,000 ϫ g. The contents of each tube were separated into a top (75% total volume) and a bottom fraction. To ensure identical conditions during deposition of the sample on the Immobilon-P membrane, both fractions were restored to the original volume (0.75 ml) and the composition of the original sample, including the lipid but not the enzyme content. Equal aliquots of these fractions, as well as of several protein suspensions of known concentration, 0.30 ml each, were spotted on an Immobilon-P membrane and subsequently immunoblotted and detected according to a standard enhanced chemiluminescence protocol (Amersham Corp.). The signal was quantitated by scanning densitometry and converted into protein mass using cPLA 2 standards included in the same Immobilon-P membrane. The amount of protein associated with vesicles was calculated as described elsewhere (24). As quantitated by liquid scintillation counting of 14 C-labeled PAPC, routinely included in all lipid mixtures, approximately 95% of the lipid was found in the bottom fraction. The apparent membrane affinity of cPLA 2 , K app , used to quantitate the association of this enzyme with vesicles is defined as the reciprocal of the total lipid concentration at which 50% of the protein is associated with vesicles. Due to a large excess of lipid over protein in the experiments reported here, this quantity can be treated as a molar partition coefficient of the protein into lipid (25).
Activity Assay-The activity of cPLA 2 toward PAPC was quantitated in a modified Dole assay (26) as described previously (27). LUVs were prepared by extrusion, using the same procedure as described in the binding assay, using either standard binding solution (with the addition of BSA at 1 mg/ml) or "standard micelle assay solution" depending on the comparisons necessary; however, they contained a larger fraction of 14 C-labeled PAPC, to ensure approximately 200,000 cpm per assay sample.
Triton X-100/phospholipid mixed micelles were prepared by adding an appropriate aliquot of the detergent to a suspension of multilamellar lipid vesicles to achieve the final total (lipid ϩ detergent) concentration of 4 mM in the assay. Unless indicated otherwise, the reaction was carried out at 40°C in a solution containing 100 mM KCl, 0.2 mM CaCl 2 , 1 mg/ml BSA, and 1 mM dithiothreitol (Cleland's reagent), 20 mM HEPES, pH 7.5, the standard micelle assay solution. The dependence of the activity on the cPLA 2 concentration was found to be linear in both vesicles and micelles.
The kinetics of cPLA 2 activity were analyzed using the dual phospholipid kinetic model, Equation 1 (17,28), where X s is the molar fraction of the substrate; S is the bulk concentration of phospholipid, K s is the dissociation constant from the inter-face, calculated per phospholipid only; V max is the maximal rate of hydrolysis, and K m is the interfacial Michaelis constant. When the concentration of the substrate in the interface is far below saturation (X s Ͻ Ͻ K m ) this model can be simplified to Equation 2, where the enzymatic efficiency E ff ϭ V max /K m and the fraction of the enzyme associated with micelles, X a ϭ S/(K s ϩ S). Secondary Structure Prediction-The secondary structure was predicted using three different algorithms incorporated into the program DNAsis (Hitachi Software Engineering America, Ltd., San Francisco, CA) or using the service offered by the Protein Design Group, EMBL (Heidelberg, Germany).
Data Analysis-Unless indicated otherwise, experiments were performed in two independent sets, each in duplicate. Results are presented as the mean Ϯ S.D. (not shown when the bar size is smaller than the symbol). Kinetic and binding models were fitted to the data with a weighted least-squares procedure. All other lines shown were smoothed curves through the data to allow for ready visual comparisons. Although all results are internally consistent for each figure and table, the range of PtdIns(4,5)P 2 activation of cPLA 2 varies between Tables II and III and Fig. 6 for several reasons. (i) The intrinsic activity of cPLA 2 on PAPC varies between pure PAPC vesicles and TX-100 mixed micelles. (ii) The enzyme undergoes a gradual decrease in specific activity over storage time. (iii) PtdIns(4,5)P 2 also degrades over storage time and is light-sensitive. (iv) Some experiments resulted in as high as 25% substrate depletion due to the dramatic activation of PtdIns(4,5)P 2 which tended to lower the apparent fold activation.

RESULTS
Membrane Association-We investigated the effect of PtdIns(4,5)P 2 , phosphatidylethanolamine (PE), phosphatidylinositol (PI), and phosphatidylserine (PS) on the association of recombinant human cPLA 2 with LUVs composed primarily of PAPC. From among this group of lipids, PtdIns(4,5)P 2 proved to be particularly effective at increasing vesicle association. To avoid alterations in the chemical composition of the vesicles over the time course of the binding experiment, binding studies were performed using a Ser-228 3 Ala mutant of cPLA 2 , which is devoid of enzymatic activity (19). The affinity of this mutant for PAPC vesicles with or without 0.3 mol % PtdIns(4,5)P 2 was the same as the native protein (data not shown). Fig. 1 demonstrates that as little as 0.1 mol % PtdIns(4,5)P 2 was sufficient to cause a measurable increase in the association of the enzyme with LUVs. Furthermore, at higher mol % PtdIns(4,5)P 2 , the ratio of membrane-bound to free enzyme increased linearly with the mole fraction of PtdIns(4,5)P 2 in the membrane, i.e. 1 order of magnitude increase in mol % PtdIns(4,5)P 2 was accompanied by a numerically identical increase in the membrane affinity. This suggests that the enzyme and PtdIns(4,5)P 2 form a 1:1 complex at the membrane interface. The simplest model of binding to a specific ligand at the interface (25,29,30) is shown in Equation 3, where K 1 is the experimentally determined value of the membrane association constant in the absence of PtdIns(4,5)P 2 , X p is the mol % PtdIns(4,5)P 2 in the membrane, and K d is the interfacial dissociation constant for the PtdIns(4,5)P 2 /enzyme complex. A fit of Equation 3 for to the dependence of membrane association of cPLA 2 on mol % PtdIns(4,5)P 2 ( Fig. 1) yielded a K d of 0.05 Ϯ 0.02 mol %. In other words, just 1 molecule of PtdIns(4,5)P 2 per 2,000 lipid molecules in the membrane (ϳ360,000 lipid molecules per LUV) is sufficient to double the affinity of cPLA 2 for the lipid bilayer. Inositol (1,4,5)-trisphosphate, a polar moiety of PtdIns(4,5)P 2 liberated in the cell by PLC-catalyzed hydrolysis (31), had no apparent effect on the PtdIns(4,5)P 2 -induced increase in membrane affinity for up to a 10-fold molar excess over this phospholipid (data not shown). As illustrated in Table I, the high affinity binding of cPLA 2 to PtdIns(4,5)P 2 was sufficient to cause a measurable association with PAPC/PtdIns(4,5)P 2 vesicles in the presence of EGTA ([Ca 2ϩ ] Ͻ2 nM). As in the presence of Ca 2ϩ , the affinity of cPLA 2 for PAPC/PtdIns(4,5)P 2 vesicles increased linearly with mol % PtdIns(4,5)P 2 . In contrast to PtdIns(4,5)P 2 , PI displayed little effect on the association of the enzyme with lipid vesicles even at a 10-fold higher mol % than PtdIns(4,5)P 2 ( Table I). Lack of any appreciable effect of a mono-anionic lipid, PI, on the Ca 2ϩ -induced association of cPLA 2 with lipid membrane was consistent with the lack of the effect of another anionic lipid, PS, on the Ca 2ϩ dependence of cPLA 2 binding to lipid vesicles reported earlier (11). We have decided to compare the effect of PS on the association of cPLA 2 under experimental conditions identical to those used in characterization of the interaction of this phospholipid with the Ca 2ϩ -dependent protein kinase C (24), another enzyme that associates with lipid membranes in a Ca 2ϩdependent manner (9,32). Both of these proteins share a homologous CaLB/C2 domain (7,33) shown to associate with lipid membranes in the presence of Ca 2ϩ (11).
To allow for a direct comparison, we studied cPLA 2 under conditions that were identical to those employed with the Ca 2ϩdependent protein kinase C, i.e. using vesicles composed of POPC:PS mixtures. As illustrated in Fig. 2, the fraction of vesicle-bound enzyme declined with an increasing mol % PS. Thus, in contrast to the case with PtdIns(4,5)P 2 , the monoanionic lipids, PI and PS, displayed little or even an opposite effect on the association of cPLA 2 with lipid vesicles. Surprisingly, under the same experimental conditions, the membrane affinity of cPLA 2 for vesicles composed of a 1:1 mixture of PAPC:PE was about 6-fold that of vesicles composed entirely of PAPC (Table I).
Activity-We have also tested the effect of PtdIns(4,5)P 2 , PI, PS, and PE on the activity of the enzyme toward PAPC in the same vesicles as used in the binding assay (Fig. 3). The presence of modest mol % PtdIns(4,5)P 2 in lipid vesicles resulted in a substantial increase of enzymatic activity of cPLA 2 (Fig. 3). Notably, this increase was much higher than could result solely from a PtdIns(4,5)P 2 -induced increase in the fraction of the enzyme associated with lipid vesicles. The percentage of membrane-associated enzyme rose from about 30% to about 95% when the mol % PtdIns(4,5)P 2 rose from 0 to 3%. In contrast, the activity rose by a factor of 55 (Fig. 3). At Ca 2ϩ levels approximating those of a quiescent cell, 70 nM, 1 mol % PtdIns(4,5)P 2 caused an increase in cPLA 2 activity from 7 Ϯ 2 to 260 Ϯ 80 nmol/min/mg. This increase was equally dramatic as that seen at 2 M Ca 2ϩ (Fig. 3). The presence of 1 mol % PtdIns(4,5)P 2 was sufficient to elicit measurable enzymatic FIG. 1. PtdIns(4,5)P 2 increases the affinity of cPLA 2 for lipid vesicles. The membrane affinity was quantitated using the apparent membrane association constant, K app , defined as the reciprocal of the total lipid concentration required for half-maximal association of the protein with the membrane. The experiment was performed in the standard binding solution at 2 M free Ca 2ϩ under conditions described under "Experimental Procedures." Large unilamellar vesicles were composed of PAPC and indicated mol % PtdIns(4,5)P 2 . Each point represents the average of at least two independent determinations at two different phospholipid concentrations (1 and 0.2 mM, except for the point at 3 mol % PtdIns(4,5)P 2 which used 0.5 and 0.1 mM phospholipid). Due to the large excess of lipid over protein, the ratio of membranebound to free enzyme was proportional to the total lipid concentration for all lipid compositions used. The line represents a fit of Equation 3 to the data. The interfacial dissociation constant for the PtdIns(4,5)P 2enzyme complex, K d ϭ 0.05 Ϯ 0.02 mol %, was determined from the fit of K app to the mol % PtdIns(4,5)P 2 in the membrane, X p . The experimentally determined value of the membrane association constant in the absence of PtdIns(4,5)P 2 , K 1 ϭ 550 Ϯ 80 M Ϫ1 , was used in this fit.   (Fig. 4A). The activity under these conditions was linear with time. In contrast, at 2 M free Ca 2ϩ the linear character of the activity was sustained only for the first few minutes of the reaction (Fig. 4B). Note, however, that as was the case for PtdIns(4,5)P 2 , the Ca 2ϩinduced increase in enzymatic activity of cPLA 2 (Fig. 4, A versus B) was about 10-fold larger than the increase in the percentage of membrane-bound enzyme, from about 10 to just over 90% (Table I and Fig. 1). The same time course was observed at a 20 times lower enzyme concentration (data not shown). Interestingly, the rate of hydrolysis did not diminish continuously over time. For times longer than 10 min, the rate appeared to achieve another steady value at approximately 1/3 that of the rate in the first 2 to 3 min. Nevertheless, the experimental condition-dependent activity made it difficult to study the kinetics and lipid specificity in the activation of cPLA 2 . Fortunately, as demonstrated in Fig. 5, the time course was linear even at high activity conditions when the phospholipids were dissolved in Triton X-100 to form detergent/lipid mixed micelles. This linearity suggests also that cPLA 2 does not preferentially hydrolyze PtdIns(4,5)P 2 over PAPC. Fig. 6A demonstrates that in the presence of approximately 1 molecule of PtdIns(4,5)P 2 per micelle (0.75 mol % of total lipid ϩ detergent), the initial rate of hydrolysis of PAPC by cPLA 2 was essentially linear with the molar percentage of PAPC and about 40 times higher than in the absence of PtdIns(4,5)P 2 (Fig.  6B). In contrast, a similar percentage of PI displayed little effect on the hydrolysis of PAPC (Fig. 6B). The non-linear character of the kinetics observed at the lower percentage of PAPC and in the absence of PtdIns(4,5)P 2 (Fig. 6B) is consistent with the dual phospholipid kinetic model (17,28). An increase in the mol % substrate increases both the fraction of the enzyme associated with the micelles and the rate of catalysis, since the enzyme is far from being saturated by the substrate. In contrast, PtdIns(4,5)P 2 causes most of the enzyme to associate with micelles even at the lowest mol % PAPC. Thus, the linear kinetics shown in A reflects only the change in the interfacial concentration of the substrate.
The dissociation constant of the enzyme from the PAPC/ Triton X-100 micelles, 0.7 Ϯ 0.2 mM, obtained from the fit of the dual phospholipid kinetic model (17,28), is very similar to that obtained from direct measurements on PAPC vesicles, 0.5 Ϯ 0.1 mM, under otherwise identical conditions. Hence, even in the absence of PtdIns(4,5)P 2 , approximately two-thirds of the enzyme was already associated with PAPC vesicles or Triton X-100 mixed micelles. The mixed micelle system proved to be valuable for investigating the specificity of the PtdIns(4,5)P 2 -cPLA 2 interaction compared with other phosphatidylinositol polyphosphates. The effects of the structurally homologous PI, phosphatidylinositol 4-phosphate (PtdIns(4)P), PtdIns(3)P, PtdIns(3,4)P 2 , PtdIns(4,5)P 2 , PtdIns(3,4,5)P 3 were all tested, and the results are summarized in Table II. Significantly, none of these other phosphatidylinositol polyphosphates activated cPLA 2 higher than PtdIns(4,5)P 2 emphasizing the importance of the 4-and 5-phosphate together to achieve the highest activation.
Although it is clear that PtdIns(4,5)P 2 enhances the binding of cPLA 2 to substrate-containing surfaces, it was not known if there was a way to down-regulate this interaction as is the case for PLC␦ 1 . Ins(1,4,5)P 3 , the soluble head group of PtdIns(4,5)P 2 , binds tightly to the PH domain of PLC␦ 1 and thus down-regulates the enzymatic activation caused by binding to PtdIns(4,5)P 2 as mediated through the PH domain. To determine if this were also the case for cPLA 2 , the effects of the soluble head groups of both PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 were investigated on cPLA 2 activity and its PtdIns(4,5)P 2 activation. Soluble Ins(1,4,5)P 3 and Ins(1,3,4,5)P 4 were added at a 10-and 5-fold molar excess, respectively, relative to PtdIns(4,5)P 2 . The results (summarized in Table III) showed that neither species affected the PtdIns(4,5)P 2 activation of cPLA 2 . Additionally the results showed that neither appreciably activated cPLA 2 themselves.
In addition to PtdIns(4,5)P 2 other phospholipids, especially the anionic phospholipids, were reported to activate cPLA 2 (12). Table II clearly shows that while the anionic phospholipids do activate cPLA 2 , it is by several orders of magnitude less than the effect of PtdIns(4,5)P 2 . Interestingly, PE which appeared to increase the affinity of cPLA 2 for PE:PAPC vesicles actually decreased the rate of hydrolysis of PAPC in the phospholipid/Triton X-100 micelles (Fig. 7). However, the progress of hydrolysis of PAPC was about 5-fold faster in 1:1 PAPC:PE vesicles than in vesicles composed entirely of PAPC. Consistent with the effect of Ca 2ϩ and PtdIns(4,5)P 2 , the PE-induced increase in the activity of cPLA 2 was severalfold higher than the accompanying increase in the association of the enzyme with lipid vesicles, from approximately 70 to just over 90% of the enzyme bound to vesicles under identical experimental conditions.
The stimulatory effects of PtdIns(4,5)P 2 and PE on the activity of cPLA 2 were largely independent of one another, con-sistent with the previously published data (12). Substitution of a quarter of PAPC by PE in vesicles already containing 1 mol % PtdIns(4,5)P 2 caused a further 5-fold increase in the activity of cPLA 2 toward PAPC (Fig. 8), similar to that seen in the absence of PtdIns(4,5)P 2 (Fig. 7). This increase is consistent with a 6-fold higher affinity of cPLA 2 for membranes containing 50 mol % (Table I). In contrast, replacing half of the substrate by unlabeled PS reduced the rate of hydrolysis by a factor of 2.

DISCUSSION
Our results demonstrate that cPLA 2 binds with high affinity and specificity to PtdIns(4,5)P 2 in a 1:1 stoichiometry. The PtdIns(4,5)P 2 -induced increase in membrane affinity is accompanied by an equally large increase in the enzymatic activity of cPLA 2 . Importantly, this increase in the activity is at least an order of magnitude larger than the concomitant increase in the fraction of membrane-bound enzyme. The interaction of cPLA 2 with PtdIns(4,5)P 2 is sufficiently strong to cause measurable effects both on the membrane association and the activity of this enzyme in the absence of exogenous Ca 2ϩ . cPLA 2 appears to contain a region resembling a PH domain. We demonstrate a striking difference in the requirements for mono-anionic lipids in Ca 2ϩ -dependent association with lipid membranes between cPLA 2 and other proteins that contain CaLB/C2 domains. We also discuss implications of this study for the investigation of other multimodular signal transduction proteins.
PtdIns(4,5)P 2 Binding-The analysis of PtdIns(4,5)P 2 -induced association of cPLA 2 with lipid membranes revealed that those two molecules form a 1:1 complex at the interface with a dissociation constant of 0.05 mol %. This binding appears to arise from a specific interaction rather than a nonspecific electrostatic attraction. The presence of mono-anionic lipids at   7. PE enhances the activity of cPLA 2 in large unilamellar vesicles but has no effect in Triton X-100/phospholipid mixed micelles. cPLA 2 activity over time was measured in the presence of vesicles composed of 1:0 (f) or 1:1 (q) PAPC:PE or mixed micelles composed of 1:0:3 (Ⅺ) or 3:1:12 (E) PAPC:PE:Triton X-100 using the standard micelle assay solution at 1 mM total lipid. concentrations assuring much higher charge density than that produced by the mol % PtdIns(4,5)P 2 employed here failed to increase the affinity of the enzyme for lipid vesicles. Consistent with the specific interaction was the fact that the PtdIns(4,5)P 2 -stimulated increase in affinity for the interface was also seen in Triton X-100/phospholipid mixed micelles. This observation makes it unlikely that the PtdIns(4,5)P 2 -induced increase in the affinity of cPLA 2 for lipid vesicles has its origin in the altered physical properties of the lipid bilayer.
Notably, the affinity of cPLA 2 for PtdIns(4,5)P 2 appears to be quantitatively identical to that of PLC␦ 1 for this same lipid. Inclusion of 0.5 mol % PtdIns(4,5)P 2 in lipid vesicles increased their affinity for PLC␦ 1 by a factor of 11 (21). With double the concentration (1 mol %) of PtdIns(4,5)P 2 in PAPC vesicles, the affinity of cPLA 2 for the membrane increased by a factor of 20 ( Fig. 1), also double the factor seen for PLC␦ 1 . Thus, it is safe to assume that the interfacial dissociation constant of cPLA 2 and PLC␦ 1 from PtdIns(4,5)P 2 incorporated into lipid membranes is about 0.05 mol % as shown in Fig. 1 for cPLA 2 . However, in strong contrast with the case for PLC␦ 1 (34), Ins(1,4,5)P 3 has no effect on the PtdIns(4,5)P 2 -induced association of cPLA 2 with phospholipid vesicles.
In contrast to protein kinase C with whom cPLA 2 shares a homologous CaLB/C2 domain (7,33), its requirement for monoanionic lipids in Ca 2ϩ -dependent membrane association is strikingly different. A PS-dependent decrease in the membrane affinity of cPLA 2 caused by PS contrasts very strongly with the 1,000-fold increase in the affinity of protein kinase C for vesicles whose PS content was raised from 10 to 50 mol % (35), under otherwise identical experimental conditions to those employed here. The binding of cPLA 2 to PS vesicles is hardly detectable, and the membrane-bound/free cPLA 2 ratio increases essentially in linear fashion with the mol % of POPC. Thus, the enzyme appears to associate with a single molecule of PC. Whether this interaction occurs through the CaLB or the catalytic domain of cPLA 2 remains to be resolved.
The increased affinity of cPLA 2 for membranes containing PE likely originates in the change of physical properties of lipid membranes induced by this phospholipid (36). This conclusion is corroborated by the fact that in Triton X-100 micelles PE acts as a neutral diluent (Fig. 7). Interestingly, diacylglycerols that share with PE its ability to induce the same change in the physical properties in the lipid bilayers (37) were also shown to increase the activity of cPLA 2 when constituting a large fraction of the lipid mixture (12).
Activity-The PtdIns(4,5)P 2 -induced increase in the mem-brane affinity was accompanied by an equally large increase in the enzymatic activity of cPLA 2 . Interestingly, the effect of PtdIns(4,5)P 2 on the enzymatic activity of cPLA 2 , shown here, appears to be much larger than the effect on the activity of PLC␦ 1 (38), an enzyme with essentially identical affinity for PtdIns(4,5)P 2 (21). The stimulatory effect of PtdIns(4,5)P 2 was much larger than that previously reported (12), presumably due to the earlier use of partially pure cPLA 2 on uncharacterized sonicated liposomes. Notably, the linear kinetics of cPLA 2 , as shown in Fig. 6, clearly demonstrate that the enzyme interacts with only a single molecule of substrate. As demonstrated elsewhere (39), the nonlinear kinetics seen in phosphatidylmethanol:PAPC mixed vesicles and interpreted as the evidence for cooperativity in cPLA 2 /PAPC interaction (40) has its origin in the limited miscibility of phosphatidylmethanol and PAPC.
The PtdIns(4,5)P 2 -stimulated increase in enzymatic activity was quite specific for this phosphatidylinositol. The structurally similar PtdIns(3,4,5)P 3 activated cPLA 2 to roughly 60% the level of PtdIns(4,5)P 2 (Table II), so it seems there is some tolerance for the addition of the 3-phosphate into the binding site. The fact that PtdIns(3,4)P 2 activated cPLA 2 to only 63% that by PtdIns(4,5)P 2 (Table II) and that PtdIns(3)P activated cPLA 2 to only 32% that by PtdIns(4)P (Table II) demonstrates the stereoselectivity of cPLA 2 for PtdIns(4,5)P 2 and tends to rule out the involvement of PI-3-kinase in the regulation of cPLA 2 .
It is clear that the stimulatory effect of other anionic lipids is much smaller than the PtdIns(4,5)P 2 stimulation (Table II). This contrasts with an earlier report (12) where the effects of PtdIns(4,5)P 2 , PI, PS, and phosphatidic acid were well within the same order of magnitude.
The stimulatory effects of PtdIns(4,5)P 2 and PE, reported earlier (12), were suggested to arise from an increase in the fraction in the enzyme associated with lipid vesicles. As demonstrated here, the observed stimulatory effect of PtdIns(4,5)P 2 , Ca 2ϩ , and PE on the enzymatic activity of cPLA 2 are at least an order magnitude larger than could possibly arise from just the changes in the fraction of enzyme associated with either vesicles or micelles. Thus, the PtdIns(4,5)P 2 -induced increase in the enzymatic efficiency of cPLA 2 might occur through allosteric effects. However, the fact that Ca 2ϩ and PE act in the same manner as PtdIns(4,5)P 2 does, makes this explanation less plausible. The fact that Ins(1,4,5)P 3 and Ins(1,3,4,5)P 4 had little effect on either the membrane affinity or the enzymatic activity of cPLA 2 (Table III) casts further doubt on the presence of a PtdIns(4,5)P 2 -induced allosteric activation.
Comparability of Dissociation Constants-The comparison between cPLA 2 and PLC␦ 1 was possible because the binding of both cPLA 2 and PLC␦ 1 to PtdIns(4,5)P 2 was measured taking into account that PtdIns(4,5)P 2 is an integral part of the membrane. The values for the dissociation constants for the protein and PtdIns(4,5)P 2 that follow the same convention as for soluble ligands are 0.5 M for cPLA 2 (herein), 1.7 M for PLC␦ 1 (21), 53 M for mSos1 (41), and 30 M for pleckstrin (42). These values although readily used and compared in the literature are, however, valid only for the specific experimental conditions employed, such as the content of PS in the membrane for PLC␦ 1 (21,43) and the concentration of Ca 2ϩ for cPLA 2 , reflecting the effect of the multiple attachment points on the binding of all of these proteins to membranes.
To illustrate the point that the values are not comparable, at 2 M free Ca ϩ2 the apparent dissociation constant for the PtdIns(4,5)P 2 -cPLA 2 -vesicle complex is 0.5 M. However, in the absence of exogenous calcium, and in the presence of 0.5 mM EGTA as well as 30 M PtdIns(4,5)P 2 (3 mol % of 1 mM total phospholipid), only 23 Ϯ 6% protein was associated with vesicles. Thus, under such conditions, the apparent dissociation constant of the PtdIns(4,5)P 2 -cPLA 2 complex would appear to be more than 30 M. (Under the same conditions, but in the absence of PtdIns(4,5)P 2 , the percentage of the enzyme associated with PAPC vesicles did not differ significantly from 0.) This means that not only are these values reported in units of molarity not comparable among any other PtdIns(4,5)P 2 -binding proteins yet reported (except for cPLA 2 and PLC␦ 1 which also present surface concentration units), but also the expression of the dissociation constant in molarity units casts doubt on those numbers and their physiological significance.
As can be seen from Equation 3, the concentration of the specific ligand at which 50% of the protein will be associated with the surface depends on the affinity of other modules for the membrane through general or specific interactions; this is contained within K 1 . One additional caveat for K 1 , rarely taken into account in the studies of domains extracted from multimodular proteins, is that its value is valid only for either the whole protein or a domain and that these numbers are likely to be different. In contrast, the interfacial K d measured in surface concentration units (mol %), calculated from the PtdIns(4,5)P 2induced increase in the membrane affinity of cPLA 2 , does not depend on any other mode of membrane binding, e.g. Ca 2ϩinduced membrane association. This should also hold true for other proteins with multiple modes of interaction with the membrane.
Putative PH Domain-The apparent 1:1 stoichiometry, specificity, and high affinity observed in the binding of cPLA 2 to vesicles containing PtdIns(4,5)P 2 suggest that this enzyme contains a specific binding site for PtdIns(4,5)P 2 . The strong enhancement of enzymatic activity by PtdIns(4,5)P 2 in either lipid vesicles or Triton X-100/phospholipid mixed micelles is consistent with this notion. Notably, the effect of PtdIns(4,5)P 2 on the association of cPLA 2 with lipid vesicles is quantitatively identical to that reported for PLC␦ 1 (21). The high affinity PtdIns(4,5)P 2 -binding site was localized in PLC␦ 1 within a PH domain (43,44). Recently the PH domain, for which structural information exists in only 6 of the more than 100 identified proteins, has been proposed to function as a membrane-localizing domain through specific interactions with PtdIns(4,5)P 2 based on the results of several proteins that have been shown to associate with PtdIns(4,5)P 2 (21,(41)(42)(43); yet only PLC␦ 1 (21) binds with sufficient affinity to be biologically relevant given the bioavailability of PtdIns(4,5)P 2 .
However, a computer-aided search for a PH domain in cPLA 2 was inconclusive, presumably due to the very low homology in the amino acid sequence of this domain as noted previously (14). Nevertheless, we noticed a large degree of similarity between a short stretch of amino acids, 271-283, in cPLA 2 with the sequence in PLC␦ 1 that comprises a part of the PtdIns(4,5)P 2 -binding site. Fig. 9A illustrates that all four key residues in this motif, shown to coordinate PtdIns(4,5)P 2 in the crystal structure of PLC␦ 1 (44), have conservative or identical matches in cPLA 2 . The one other key residue shown by mutational analysis, Arg-37 (38), also readily aligned with cPLA 2 . A characteristic part of this motif, the KXK sequence, was shown to have the two lysines coordinating the 4-and 5-phosphate in the inositol ring in both the regulatory (44) and catalytic (45) PtdIns(4,5)P 2 -binding sites in PLC␦ 1 . A Trp involved in the coordination of the 1-phosphate (44) has several basic residues on a C-flank in both proteins.
In contrast, as pointed out earlier (44), pleckstrin (42), spectrin (46), or mSos1 (41) bind PtdIns(4,5)P 2 with lower affinity and specificity than PLC␦ 1 (44) or now cPLA 2 . The former proteins have only one basic residue flanking the equivalent tryptophan and lack the KXK motif.  9. A, the key residues for PtdIns(4,5)P 2 binding were observed within the PLC␦ 1 PH domain, and many of these key residues can also be found in cPLA 2 . Amino acids observed in the crystal structure to contact inositol 1,4,5-trisphosphate for PLC␦ 1 are shown in red (44). Amino acids found by mutagenesis to be critical for PtdIns(4,5)P 2 binding in PLC␦ 1 are underlined (38). The residues from cPLA 2 that match the known key residues in PLC␦ 1 are shown in pink. B, sequence alignment of PtdIns(4,5)P 2 binding PH domains. The sequence-based multiple alignment (14) of PH domains is shown for PLC␦ 1 (residues 22-129), human N-terminal pleckstrin (residues 5-100), and mouse brain ␤-spectrin (residues 2197-2305) with gaps under PLC␦ 1 (Arg-38 and Cys-48) as derived from the structure-based alignment (44). The ␤-strands and ␣-helices of PLC␦ 1 , derived from the crystal structure, are shown as arrows and rectangles, respectively. The highlighted residues indicate conservative matches in the alignment of at least 3 of 4 of the aligned PH domains. These residues are colored red for polar/charged, purple for nonpolar, violet for aromatic residues, and yellow for identical residues in at least the three known PH domains. The Chou-Fasman prediction of secondary structure for cPLA 2 is presented under the sequences with B/b for ␤-strand and A/a for ␣-helix with capital letters as more probable. The boxed residues are emphasized in A.
Starting from this motif, we have attempted to align this part of cPLA 2 with the PH domains of the three proteins that were shown to associate with vesicles containing PtdIns(4,5)P 2 (Fig. 9B). The alignment included the standard constraints for PH domain alignments as follows: (i) the PH domain is presented in six conserved modules; (ii) the six modules have a minimum size separated by variable spacers; (iii) the basic residues in the region from cPLA 2 271-290, the universal Trp of module 6, as well as the polar/non-polar character of the residues were used as the basis for the alignment for all modules (14).
In addition, we checked the secondary structure prediction for this putative PH domain, testing first the available algorithms (DNAsis from Hitachi Software Engineering America, Ltd., San Francisco and service offered by the Protein Design Group, EMBL, Heidelberg, Germany) by comparing the predicted and experimentally determined structure of the PH domain of PLC␦ 1 . The Chou-Fasman method (47) proved to be the most accurate for PLC␦ 1 and was subsequently used to predict the secondary structure of the putative PH domain in cPLA 2 . As illustrated in Fig. 9B, this predicted structure is quite similar to the known structure of the PH domain from PLC␦ 1 (44) and PH domains in general (14).
Considering all these factors, the region of cPLA 2 presented in Fig. 9B has much more similarity to the PH domain of PLC␦ 1 and the N-terminal PH domain of pleckstrin than those from many other proteins included in the PH domain consensus (14,15,48). There are 11 identical residues among the PH domains of both cPLA 2 and pleckstrin and cPLA 2 and PLC␦ 1 . For comparison, 15 residues are identical among the PH domains of PLC␦ 1 and pleckstrin. Thus, it is very likely that the region shown in Fig. 9B constitutes a PH domain. Although the existence of this PH domain in cPLA 2 is also a likely explanation for the specificity, high affinity binding, and activating properties of PtdIns(4,5)P 2 to cPLA 2 , further structural studies and mutagenesis studies are required to identify the site(s) of interaction between PtdIns(4,5)P 2 and cPLA 2 as well as the important chemical moieties involved in the interaction.
Functional Implications and Conclusions-The presence of PtdIns(4,5)P 2 in membranes targeted by cPLA 2 is likely to increase both the fraction of membrane-associated enzyme as well as its specific activity. Several recent studies demonstrated that PtdIns(4,5)P 2 can be synthesized in cellular membranes "on demand" and in a compartmentalized fashion (see Ref. 49 for review). If PtdIns(4,5)P 2 synthesis also occurs on demand in the nuclear and endoplasmic reticulum membranes, which are targeted by cPLA 2 (5), then PtdIns(4,5)P 2 might be involved in the up-regulation of cPLA 2 activity. At Ca 2ϩ levels approximating those of a quiescent cell, PtdIns(4,5)P 2 displays a particularly dramatic effect on the activity of cPLA 2 . That may lead to alternative modes of cPLA 2 activation that do not require Ca 2ϩ mobilization, as suggested recently (50).
In conclusion, we have demonstrated that cPLA 2 binds with high affinity, specificity, and 1:1 stoichiometry to PtdIns(4,5)P 2 in lipid vesicles and that there is a resulting quantitative increase in the enzymatic activity. We demonstrate also that quantitative approaches that take into account the two-dimensional nature of ligand presentation are essential to elucidate the structural and functional relationship in multimodular membrane-binding proteins that play key roles in signal transduction.