A Structure-Function Study of the C2 Domain of Cytosolic Phospholipase A2

The C2 domain of cytosolic phospholipase A2 (cPLA2) is involved in the Ca2+-dependent membrane binding of this protein. To identify protein residues in the C2 domain of cPLA2 essential for its Ca2+ and membrane binding, we selectively mutated Ca2+ ligands and putative membrane-binding residues of cPLA2 and measured the effects of mutations on its enzyme activity, membrane binding affinity, and monolayer penetration. The mutations of five Ca2+ ligands (D40N, D43N, N65A, D93N, N95A) show differential effects on the membrane binding and activation of cPLA2, indicating that two calcium ions bound to the C2 domain have differential roles. The mutations of hydrophobic residues (F35A, M38A, L39A, Y96A, Y97A, M98A) in the calcium binding loops show that the membrane binding of cPLA2 is largely driven by hydrophobic interactions resulting from the penetration of these residues into the hydrophobic core of the membrane. Leu39 and Val97 are fully inserted into the membrane, whereas Phe35 and Tyr96 are partially inserted. Finally, the mutations of four cationic residues in a β-strand (R57E/K58E/R59E/R61E) have modest and negligible effects on the binding of cPLA2 to zwitterionic and anionic membranes, respectively, indicating that they are not directly involved in membrane binding. In conjunction with our previous study on the C2 domain of protein kinase C-α (Medkova, M., and Cho, W. (1998) J. Biol. Chem. 273, 17544–17552), these results demonstrate that C2 domains are not only a membrane docking unit but also a module that triggers membrane penetration of protein and that individual Ca2+ ions bound to the calcium binding loops play differential roles in the membrane binding and activation of their parent proteins.

The C2 domain of cytosolic phospholipase A 2 (cPLA 2 ) is involved in the Ca 2؉ -dependent membrane binding of this protein. To identify protein residues in the C2 domain of cPLA 2 essential for its Ca 2؉ and membrane binding, we selectively mutated Ca 2؉ ligands and putative membrane-binding residues of cPLA 2 and measured the effects of mutations on its enzyme activity, membrane binding affinity, and monolayer penetration. The mutations of five Ca 2؉ ligands (D40N, D43N, N65A, D93N, N95A) show differential effects on the membrane binding and activation of cPLA 2 , indicating that two calcium ions bound to the C2 domain have differential roles. The mutations of hydrophobic residues (F35A, M38A, L39A, Y96A, Y97A, M98A) in the calcium binding loops show that the membrane binding of cPLA 2 is largely driven by hydrophobic interactions resulting from the penetration of these residues into the hydrophobic core of the membrane. Leu 39 and Val 97 are fully inserted into the membrane, whereas Phe 35 and Tyr 96 are partially inserted. Finally, the mutations of four cationic residues in a ␤-strand (R57E/K58E/R59E/R61E) have modest and negligible effects on the binding of cPLA 2 to zwitterionic and anionic membranes, respectively, indicating that they are not directly involved in membrane binding. In conjunction with our previous study on the C2 domain of protein kinase C-␣ (Medkova, M., and Cho, W. (1998) J. Biol. Chem. 273, 17544 -17552), these results demonstrate that C2 domains are not only a membrane docking unit but also a module that triggers membrane penetration of protein and that individual Ca 2؉ ions bound to the calcium binding loops play differential roles in the membrane binding and activation of their parent proteins.
Phospholipases A 2 (PLA 2 s 1 ; EC 3.1.1.4) are a large family of lipolytic enzymes that catalyze the hydrolysis of the fatty acid ester at the sn-2-position of phospholipids (1,2). Among various mammalian PLA 2 s, 85-kDa cytosolic PLA 2 (cPLA 2 ) can selectively liberate arachidonic acid from membrane phospholipids, which can be converted to potent inflammatory lipid mediators, prostaglandins, thromboxanes, leukotrienes, and lipoxins, collectively known as eicosanoids (3). Recent genetic studies showed that the deletion of the cPLA 2 gene results in loss of lipid mediator biosynthesis (4,5). cPLA 2 is therefore an attractive target for developing specific inhibitors that can be used as novel anti-inflammatory drugs. cPLA 2 is translocated to endoplasmic reticulum membranes and nuclear envelopes in response to a rise in intracellular Ca 2ϩ (6,7). This membrane translocation of cPLA 2 is mediated by its amino-terminal C2 domain, which contains calcium and membrane binding sites (8,9). However, the mechanism by which the C2 domain of cPLA 2 drives its Ca 2ϩ -dependent membrane targeting is not fully understood. Our recent structure-function study of the C2 domain of protein kinase C-␣ showed that the C2 domain not only brings the protein to the membrane surface but also triggers the membrane penetration of protein (10). Also, the study revealed that individual Ca 2ϩ ions bound to the C2 domain play differential roles in membrane binding and activation. Tertiary structures of the C2 domains of protein kinase C (11) and cPLA 2 (12,13), despite their overall similarity, have noticeable differences; i.e. they have different ␤-strand connectivity (14), and, more importantly, the C2 domain of cPLA 2 contains two clusters of exposed hydrophobic residues in the calcium binding loops (see Fig. 1A) (12). Consistent with this unique structure, both cPLA 2 (15) and its isolated C2 domain (16) have been shown to be able to penetrate into the membrane. In this study, we performed an extensive structurefunction analysis of the C2 domain of cPLA 2 to assess the roles of six hydrophobic residues and five Ca 2ϩ ligands located in the calcium binding loops in the membrane binding and activation of cPLA 2 . We also determined the role of a cluster of cationic residues in a ␤-strand. In conjunction with our previous study on protein kinase C-␣, this study reveals interesting similarities and differences between the two C2 domains.

EXPERIMENTAL PROCEDURES
Materials-1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) and 1-palmitoyl-2-hydroxy-sn-glycero-3-phosphocholine were purchased from Avanti Polar Lipids (Alabaster, AL). 1,2-Di-O-hexadecyl-sn-glycero-3-phosphocholine (DHPC) and 1,2-sn-dioleoylglycerol were from Sigma. All lipids were used without further purification. 1,2-Di-O-hexadecyl-sn-glycero-3-phosphoglycerol (DHPG) was prepared from DHPC by phospholipase D-catalyzed transphosphatidylation as described by Comfurius and Zwaal (17). Tritiated POPC ([ 3 H]POPC) was prepared from 1-palmitoyl-2-hydroxy-sn-glycero-3phosphocholine and [9,10 -3 H]oleic acid (American Radiolabeled Chemicals) using rat liver microsomes as described (18,19). Phospholipid concentrations were determined by phosphate analysis (20 Construction of Expression Vectors and Mutagenesis-Baculovirus transfer vectors encoding the cDNA of cPLA 2 with appropriate C2 domain mutations were generated by the overlap extension polymerase chain reaction (21) using pVL1393-cPLA 2 plasmid as a template (15). Briefly, appropriate complementary synthetic oligonucleotides introducing the desired mutation and two other primers at the 5Ј-and 3Ј-ends of the cPLA 2 gene, respectively, were used as primers for polymerase chain reactions, which were performed in a DNA thermal cycler (Perkin Elmer) using Pfu DNA polymerase (Stratagene, La Jolla, CA). The method consisted of two steps. In the first step, two DNA fragments overlapping at the mutation site were generated and purified on an agarose gel. Then these two fragments were annealed and extended to generate a full-length mutated cPLA 2 gene, which was further amplified by polymerase chain reaction. The product was subsequently purified on an agarose gel, digested with NotI and BglII, and subcloned into the pVL1393 plasmid. The mutagenesis was verified by DNA sequencing of the cPLA 2 gene using a Sequenase 2.0 kit (Amersham Pharmacia Biotech). The isolated C2 domain constructs (wild type and mutants) containing amino-terminal 137 residues were generated by polymerase chain reaction using corresponding pVL1393-cPLA 2 vectors as templates. An NdeI site was incorporated at the 5Ј-end of the coding sequence, and a stop codon followed by a HindIII site was introduced after Leu 137 at the 3Ј-end. These constructs were subcloned into pET28a vector (Novagen, Madison, WI) using NdeI and HindIII sites. These vectors also encode a 20-residue amino-terminal tail (MGSSHHHHHHSSGLVPRGSH) containing a His 6 tag for affinity purification and a thrombin cleavage site. The identity of constructs was verified by restriction digest and DNA sequencing.
Expression of cPLA 2 and Mutants in Baculovirus-infected Sf9 Cells-Wild type cPLA 2 and mutants were expressed in baculovirus-infected Sf9 cells (Invitrogen, La Jolla, CA). Transfection of Sf9 cells with mutant pVL1393-cPLA 2 constructs was performed using the Baculo-Gold TM transfection kit from Pharmingen (San Diego, CA). Plasmid DNA for transfection was prepared using an EndoFree Plasmid Maxi kit (Qiagen, Valencia, CA) to avoid possible endotoxin contamination. A detailed protocol for expression and purification of cPLA 2 from Sf9 cells is described elsewhere (15). Protein concentration was determined by the bicinchoninic acid method (Pierce).
Bacterial Expression and Purification of Isolated C2 Domains of cPLA 2 and Mutants-Escherichia coli strain BL21(DE3) (Novagen) was used as a host for protein expression. Five hundred ml of Luria broth supplemented with 50 g/ml kanamycin was inoculated with 1 ml of overnight culture grown at 37°C. Cells were grown at 22°C until their absorbance at 600 nm reached ϳ0.6, and the protein expression was then induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside (Research Products, Mount Prospect, IL). After overnight incubation, cells were harvested by centrifugation at 5000 ϫ g and 4°C for 10 min. Cells were resuspended in 50 ml of 50 mM Tris-HCl buffer, pH 8.0, containing 50 mM NaCl, 2 mM EDTA, 0.4% (v/v) Triton X-100, 0.4% (w/v) sodium deoxycholate, 1 mM phenylmethylsulfonyl fluoride. After the suspension was sonicated, the inclusion body pellet was obtained by centrifugation at 50,000 ϫ g for 15 min at 4°C. The pellet was resuspended in 50 ml of 50 mM Tris-HCl buffer, pH 8.0, containing 50 mM NaCl, 2 mM EDTA, 0.8% (v/v) Triton X-100, 0.8% (w/v) sodium deoxycholate. After centrifugation at 50,000 ϫ g for 15 min at 4°C, the pellet was resuspended in 50 ml of 50 mM Tris-HCl buffer, pH 8.0, containing 5 M urea and 5 mM EDTA. The pellet was stirred at room temperature for 15 min, and then the suspension was centrifuged at 100,000 ϫ g for 10 min at 4°C. The washed inclusion body was resuspended in 10 ml of 50 mM Tris-HCl buffer, pH 8.0, containing 8 M guanidinium chloride. Insoluble matter was removed by centrifugation at 100,000 ϫ g for 10 min at 4°C, and the supernatant was loaded onto a Sephadex G-25 column (2.5 ϫ 35 cm) equilibrated with 50 mM Tris-HCl buffer, pH 8.0, containing 5 M urea and 5 mM EDTA. The first major peak was collected (35 ml) and dialyzed against 50 mM Tris-HCl, pH 8.0, 1.5 M urea and then against 50 mM Tris-HCl, pH 8.0. The refolded C2 domain was purified using a Ni-NTA column (Qiagen) according to the manufacturer's instructions. Purity of protein samples was higher than 90% electrophoretically. Aliquots of purified proteins were stored at Ϫ20°C.
Determination of cPLA 2 Activity-Activity of cPLA 2 was assayed by measuring the initial rate of [ 14 C]SAPC hydrolysis as described (22).
Assay mixtures contained small unilamellar vesicles of [ 14 C]SAPC (10 M), 16 M BSA, 0.1 M KCl, and varying concentrations of CaCl 2 in 60 l of 20 mM HEPES, pH 8.0. Free calcium concentration was adjusted using a mixture of EGTA and CaCl 2 according to the method of Bers (23). 12 mol % of 1,2-sn-dioleoylglycerol was added to [ 14 C]-SAPC vesicles when the free enzyme concentration was determined by cPLA 2 activity assay in vesicle binding measurements (see below). Reactions were started by adding cPLA 2 to the mixture (to a final concentration of ϳ20 nM) and quenched by adding 370 l of chloroform/methanol/HCl (2:1:0.01) solution after a given period of incubation (3-5 min) at room temperature. Liberated [ 14 C]arachidonic acid was separated from the reaction mixture on small silica gel columns as described by Ghomashchi et al. (22). Radioactivity of hydrolyzed arachidonic acid was measured by liquid scintillation counting. cPLA 2 -Phospholipid Binding-The binding of cPLA 2 to phospholipids was measured by a centrifugation assay using phospholipid-coated hydrophobic styrene-divinylbenzene beads. The beads were coated with DHPC or DHPG as described elsewhere (19). The concentration of phospholipids coated on beads was determined by measuring the radioactivity of a trace of where [PL] o represents total phospholipid concentration. This equation assumes that each enzyme binds independently to a site on the interface composed of n phospholipids with a dissociation constant of K d . To determine the Ca 2ϩ dependence of binding, cPLA 2 (ϳ60 nM) was incubated for 15 min with phospholipid-coated beads (60 M) and 1 M BSA in 150 l of 10 mM HEPES (pH 7.0) containing 100 mM KCl and varying concentrations of Ca 2ϩ (or EGTA) (see Fig. 2). After centrifugation (12,000 ϫ g for 2 min), [E] b was determined by the activity assay (see above). The binding of Ca 2ϩ ions to the isolated C2 domain was shown to be consistent with the cooperative Hill model (24). The concentration of Ca 2ϩ giving rise to half-maximal binding (or activity) ([Ca 2ϩ ]1 ⁄2 ) was thus determined from curve fitting of data to a Hill equation, where y, a, h, and [Ca 2ϩ ] are relative binding (or activity), arbitrary normalization constant, Hill coefficient, and free Ca 2ϩ concentration, respectively. The binding of isolated C2 domains to POPC vesicles was measured by a centrifugation assay using sucrose-loaded unilamellar vesicles (100-nm diameter) (25). Sucrose-loaded vesicles were prepared as follows. The lipid solution of POPC was added to a round-bottomed flask, and organic solvent was removed by rotary evaporation. The lipid film was suspended in 20 mM Tris-HCl buffer, pH 7.5, containing 0.2 M sucrose (to a final lipid concentration of ϳ20 mM) and vortexed vigorously. Unilamellar vesicles were prepared by multiple extrusion through a 0.1-m polycarbonate filter (Millipore Corp.) in a Liposofast microextruder (Avestin; Ottawa, Ontario). For binding measurements, the vesicle solution was diluted into 20 mM Tris-HCl buffer, pH 7.5, 0.1 M KCl, and 0.5 mM Ca 2ϩ (or EGTA) to yield a final phospholipid concentration of 0.5 mM. After adding wild type or a mutant C2 domain (final concentration of 0.5 M), binding mixtures were incubated for 15 min at room temperature and centrifuged at 100,000 ϫ g for 30 min at 25°C. Supernatants were decanted, and pellets were redissolved in 15 l of 10 mM HEPES buffer, pH 7.0, containing 0.1 M KCl and 0.5 mM EGTA. Resuspended pellets were loaded on 14% polyacrylamide gels, and C2 domains were separated by SDS-polyacrylamide gel electrophoresis. The amount of protein in each band was quantified using IS-1000 Digital Imaging System (Key Scientific, Mt. Prospect, IL). To convert the protein band density to the protein concentration, a standard curve was constructed from density values of varying amounts of C2 domain protein samples (1-5 g).
Monolayer Measurements-Surface pressure () of solution in a circular Teflon trough was measured using a Wilhelmy plate attached to a computer-controlled tensiometer (Nima Technology, Coventry, United Kingdom). The trough (4-cm diameter ϫ 1-cm depth) has a 0.5-cm-deep well for a magnetic stir bar and a small hole drilled at an angle through the wall to allow an addition of protein solution. Five to ten ml of phospholipid solution in chloroform was spread onto 10 ml of subphase (10 mM HEPES, pH 7.0, 0.1 M KCl, and 0.1 mM of free Ca 2ϩ ) to form a monolayer with a given initial surface pressure ( 0 ). The subphase was continuously stirred at 60 rpm with a magnetic stir bar. Once the surface pressure reading of monolayer had been stabilized (after ϳ5 min), the protein solution (typically 50 l) was injected to the subphase, and the change in surface pressure (⌬) was measured as a function of time at 23°C. Typically, the ⌬ value reached a maximum after 20 min. The maximal ⌬ value depended on the protein concentration at the low concentration range and reached a saturation when the protein concentration was higher than 2 g/ml. Protein concentrations were therefore maintained above 3 g/ml for cPLA 2 and 6 g/ml for C2 domains to ensure that the observed ⌬ represented a maximal value. The critical surface pressure ( c ) was determined by extrapolating the ⌬ versus 0 plot to the x axis (26).

Properties of Calcium Ligand
Mutants-According to the structure of the C2 domain of cPLA 2 shown in Fig. 1B, five Ca 2ϩ -binding residues can be classified into three groups: Asn 65 , which primarily coordinates CA1; Asp 93 and Asn 95 , which mainly coordinate CA2; and two CBR1 aspartates, Asp 40 and Asp 43 , which coordinate both Ca 2ϩ ions. Based on this Ca 2ϩ binding geometry, the mutations of Asn 65 and Asp 93 / Asn 95 would selectively affect the binding of CA1 and CA2, respectively, whereas the mutations of Asp 40 /Asp 43 would have effects on the binding of both Ca 2ϩ ions. Thus, it is possible to systematically analyze the roles of the two Ca 2ϩ ions in the membrane binding and activation of cPLA 2 by selectively deactivating their ligands and separately measuring the effects of mutations on membrane binding and activity. For this purpose, we made five mutants that contain either Asp to Asn (e.g. D40N) or Asn to Ala (e.g. N65A) mutations. All five mutants exhibited full membrane binding affinity and enzyme activity at saturating Ca 2ϩ concentrations (see Figs. 2 and 3), demonstrating the lack of deleterious conformational changes. We first measured the binding of wild type and mutants to hydrophobic beads coated with DHPC, which is a nonhydrolyzable ether analog of phospholipid, as a function of Ca 2ϩ concentration. Phospholipid-coated hydrophobic beads have been shown to be useful in determining the membrane affinity of PLA 2 s (19). In particular, this model membrane allows for rapid and accurate measurement of PC affinity, which is normally difficult to achieve with PC vesicles due to their low pelleting efficiency at low concentrations. 2 As shown in Fig. 2, all mutants exhibited increased Ca 2ϩ requirements for PC binding, and [Ca 2ϩ ]1 ⁄2 values varied from 40 M for N65A (7-fold increase) to 3.5 mM for D43N (614-fold increase) (see Table I). In general, Asp to Asn mutations had much larger effects than Asn to Ala mutations, because aspartates provide charged bidentate Ca 2ϩ ligands, whereas asparagines provide neutral unidentate ones. Comparison among three Asp to Asn mutations and between two Asn to Ala mutations revealed a common theme. First, the mutational effect of Asp 93 that only coordinates CA2 is comparable with those of Asp 40 and Asp 43 that coordinate both Ca 2ϩ ions. This implies that mutational effects of Asp 40 and Asp 43 on membrane binding derive mainly from its coordination to CA2. Second, between the two Asn to Ala mutants, N95A has slightly higher [Ca 2ϩ ]1 ⁄2 value than N65A, again suggesting that CA2 is more important for membrane binding of cPLA 2 than is CA1.
We then measured the cPLA 2 activity of wild type and mutants using [ 14 C]SAPC vesicles as a function of Ca 2ϩ concentration (Fig. 3). In general, increases in [Ca 2ϩ ]1 ⁄2 for mutants from activity measurements are uniformly (ϳ3-fold) smaller than those from membrane binding measurements (Table I). This might simply reflect the structural differences between DHPC-coated beads and SAPC vesicles used for binding and activity assays, respectively. Note that, however, Ca 2ϩ dependences of both binding and activity show essentially the same trend for those mutants that are involved in coordination to CA2; i.e. the degree of increase in [Ca 2ϩ ]1 ⁄2 is in the order of D43N Ͼ D40N Ϸ D93N Ͼ Ͼ N95A, and their relative [Ca 2ϩ ]1 ⁄2 values were nearly identical for binding and activity. In striking contrast, the mutation of CA1 ligand Asn 65 to Ala exhibited 2 L. Bittova and W. Cho, unpublished observation. a larger effect on activity than on binding. This unique effect is best illustrated by comparing the [Ca 2ϩ ]1 ⁄2 values of N65A and N95A for binding and activity. The [Ca 2ϩ ]1 ⁄2 value of N65A is about half of that of N95A for binding but three times larger for activity. These results imply that CA1 might be more important for cPLA 2 activity than CA2. Taken together, it appears that the two calcium ions bound to the C2 domain of cPLA 2 appear to play differential roles in its membrane binding and activation.
Properties of Mutants of Membrane Binding Residues-Our previous study showed that cPLA 2 has a unique ability to partially penetrate into membranes, which is essential for its membrane binding and interfacial catalysis (15). The Ca 2ϩ binding loops of cPLA 2 , CBR1, and CBR3 have a cluster of hydrophobic residues on the tips of their protruded structures (see Fig. 1A), suggesting that these residues might be involved in membrane penetration. We therefore mutated six hydrophobic residues in these regions to Ala: F35A, M38A, and L39A in CBR1 and Y96A, Y97A, and M98A in CBR3. To assess the contribution of electrostatic interactions to membrane binding of the C2 domain, we also mutated four cationic residues (R57E/K58E/R59E/R61E) that form a cationic patch along a ␤-strand and are implicated in interaction with anionic phospholipid head groups. First, we measured the relative enzyme activity of these mutants using [ 14 C]SAPC vesicles as a substrate. The relative activity determined from the initial rates of hydrolysis is listed in Table II. All mutants have modestly lower activity than does wild type with two mutants in the CBR3, Y96A and V97A, showing the lowest activity. This indicates that some of these residues are important in interfacial catalysis of cPLA 2 but that none is critical. Since mutations of C2 domain residues would not significantly affect the catalytic efficiency of the enzyme, the relative activity of mutants should reflect their relative membrane binding affinity.
To quantitatively assess the effects of mutations on membrane binding, we measured dissociation constants for the binding of cPLA 2 to DHPC-coated beads. Binding isotherms for cPLA 2 and selected mutants are illustrated in Fig. 4, and n and K d values determined from the curve fitting are summarized in Table II. Although there are several different ways to analyze the binding isotherms, we used a simple Langmuir-type equation (Equation 1) assuming the presence of the enzyme binding sites on the interface consisting of n phospholipids. Since K d is expressed in terms of molarity of these binding sites, nK d is the dissociation constant in terms of molarity of lipid molecules, and the relative binding affinity can be best described in terms of relative values of (1/nK d ) ( Table II). The lateral mobility of phospholipids is not taken into account in this model, because little is known about the lateral mobility of phospholipids coated on hydrophobic beads. Since K d was measured under conditions of enzyme crowding on beads, and because of possible protein-protein interaction, the value obtained might be somewhat different from the K d obtained under conditions of low bead coverage.
In general, the relative affinity of mutants for DHPC-coated beads is consistent with their relative activity toward [ 14 C]SAPC vesicles. For some mutants, there are changes in both n and K d values, indicating that the mutations affect both their membrane affinity and binding mode. This effect is pronounced for those mutants with lower relative affinity, including F35A, L39A, and Y96A. Overall, all of the hydrophobic residues except for Met 38 and Met 98 appear to have significant and comparable contributions to hydrophobic membrane binding. Interestingly, the quadruple mutant of the cationic patch (R57E/K58E/R59E/R61E) exhibits lower activity and affinity for PC vesicles, suggesting that these cationic residues are somehow involved in membrane binding. Because it was unclear how the cationic residues participate in binding to the zwitterionic membrane surface, we measured the binding of wild type and mutants including R57E/K58E/R59E/R61E to anionic DHPG-coated beads. It was previously shown that both cPLA 2 (27) and its isolated C2 domain (28) have lower affinity for monoanionic phospholipids than for zwitterionic phospholipids. Consistent with these findings, cPLA 2 showed ϳ7.5-fold lower affinity for anionic DHPG-coated beads than for DHPCcoated beads. Also, most of the mutations of hydrophobic residues decreased the binding affinity for DHPG-coated beads less significantly than that for DHPC-coated ones, suggesting that cPLA 2 might bind to phosphatidylglycerol membranes in a slightly different mode. Qualitatively, however, similar trends were observed for binding of the mutants to DHPC-and DHPGcoated beads. Surprisingly, R57E/K58E/R59E/R61E had a slightly higher affinity for DHPG-coated beads than the wild type did, demonstrating that these residues are not involved in binding to anionic membranes. This also suggests that the low affinity of the quadruple mutant for PC-coated beads might derive from a secondary effect unrelated to membrane binding.

Properties of Isolated C2 Domain and Its
Mutants-For hydrophobic residues to contribute to the energetics of membrane binding, they must penetrate into the hydrophobic core of the membrane. Thus, mutants of essential hydrophobic residues of the C2 domain are expected to have significantly altered membrane penetrating ability. Phospholipid monolayers at the airwater interface serve as a highly sensitive tool to measure the membrane penetrating ability of protein (10,15,29). Monolayer measurements of mutant proteins were, however, hampered by generally low protein expression yields of cPLA 2 and mutants from our baculovirus-infected insect cells. To overcome this difficulty, we prepared the isolated C2 domain of cPLA 2 (residues 1-137) and its mutants. Since these isolated C2 domains can be expressed in E. coli as inclusion bodies and readily refolded with high yields (ϳ5 mg/liter of culture), we were able to obtain proteins sufficient for monolayer measurements. This also allowed us to compare the membrane penetrating ability of the C2 domain versus the whole cPLA 2 molecule. We first measured relative membrane binding affinity of isolated C2 domain mutants to see if these mutants behave similarly to their parent proteins. For these measurements, we used sucrose-loaded POPC vesicles at a high enough concentration (i.e. 0.5 mM) to ensure complete pelleting, and the relative affinity was estimated from the amount of protein bound to POPC vesicles. As listed in Table III, relative binding affinity of isolated C2 domain mutants is comparable with that of cPLA 2 mutants (see Table II), indicating that the mutational effects are essentially the same for the isolated C2 domain and the whole protein. Then we measured the penetration of these C2 domains into POPC monolayers. In these studies, a POPC monolayer of a given 0 was spread at constant area, and the ⌬ was monitored after the injection of the protein into the subphase. In general, ⌬ is inversely proportional to 0 of the phospholipid monolayer, and an extrapolation of the ⌬ versus 0 plot yields the c , which specifies an upper limit of 0 of a monolayer that a protein can penetrate into (26). Fig. 5 shows that the isolated C2 domain of cPLA 2 has high intrinsic ability to penetrate into densely packed monolayers with c of 34 dyne/cm. In fact, the isolated C2 domain has much higher monolayer penetrating power than the whole cPLA 2 molecule ( c ϭ 21 dyne/cm). The isolated C2 domain, however, shows no detectable penetration in the absence of Ca 2ϩ , demonstrating that the penetration is a Ca 2ϩ -dependent process. All three mutations of hydrophobic residues in CBR 1 significantly lower ⌬ values at a given 0 . As a result, c values decrease by 5-7 dyne/cm, with F35A and L39A having slightly lower c values than M38A, which is consistent with their relative membrane binding affinity. Differential effects of the mutations on the  2 Relative activity toward SAPC vesicles was calculated from initial velocities of hydrolysis. Specific activity for wild type is 5.2 Ϯ 0.5 mol/min/mg. Values of n and K d (best fit values Ϯ SD) were determined from nonlinear least-squares analysis of data using Equation 1. Relative affinity is the ratio of the 1/nK d value of wild type cPLA 2 to that of a mutant. ⌬A is calculated from the accessible surface area of each amino acid determined from a tripeptide model by Chothia (32). ⌬⌬G 0 is calculated from the relative affinity of mutants using the equation, ⌬⌬G 0 ϭ ϪRT ln(relative affinity).  monolayer penetration are more pronounced for CBR3 mutants illustrated in Fig. 6. Y96A, with the lowest binding affinity for PC-coated beads and PC vesicles, shows a dramatic decrease in c . The order of decrease in c (Y96A Ͼ V97A Ͼ M98A) is again consistent with their relative affinity for PC membranes. In contrast to all hydrophobic mutants, the quadruple mutant of cationic residues has essentially no effect on monolayer penetration, thereby demonstrating the specific nature of reduced monolayer penetrating power seen for the mutations of hydrophobic residues.

Differential Roles of Ca 2ϩ Ions and Their
Ligands-Our previous structure-function study on the C2 domain of protein kinase C-␣ showed that individual Ca 2ϩ ions bound to the C2 domain have different roles in the membrane binding and activation of protein kinase C-␣ (29). Specifically, it was shown that a calcium ion bound close to CBR1 brings the protein to the membrane surface, whereas calcium ions bound to CBR3 induce the membrane penetration of the protein, resulting in enzyme activation. Results from the mutational analysis of the C2 domain of cPLA 2 also suggest differential roles of calcium ions bound to its C2 domain. As far as the specific role of each calcium is concerned, however, a major difference seems to exist between the two C2 domains. It should be noted here that there are considerable differences in stoichiometry and geom-etry of calcium coordination between the two C2 domains. Thus, comparison of the role of a calcium ion in cPLA 2 with that of its counterpart in protein kinase C-␣ should be made with caution. For cPLA 2 , the following results indicate that CA2 bound close to CBR3 is important for membrane binding and that CA1 bound close to CBR1 is involved in cPLA 2 activation. First, mutations of Asp 40 , Asp 43 , Asp 93 , and Asn 95 , which coordinate CA2, have larger effects on membrane binding than that of Asn 65 , which coordinates CA1. Second, the mutation of Asp 43 , which is closest to CA2 (see Fig. 1B), has the largest impact on membrane binding. Third, the mutations of CA2 ligands show comparable effects on membrane binding and enzyme activity, whereas the mutation of Asn 65 affects enzyme activity much more significantly. Notice that this assignment of roles is distinct from that given to calcium ions bound to protein kinase C-␣. This also implies that the two C2 domains interact with membranes in different modes. A recent study of the isolated C2 domain of cPLA 2 showed that two Ca 2ϩ ions bind to the C2 domain sequentially with positive cooperativity, which in turn induces intradomain conformational changes and membrane translocation (24). Since the membrane binding of cPLA 2 precedes its interfacial catalysis, it is likely that CA2 binding takes place first, which then promotes CA1 binding although the reverse order of calcium addition cannot be precluded. The 10-fold increase in [Ca 2ϩ ]1 ⁄2 for the membrane binding of N65A (see Table II) indicates that CA1 binding also contributes to membrane binding, albeit to a lesser degree.
Thus, it appears that the binding of both Ca 2ϩ ions to the C2 domain takes place prior to the membrane binding of cPLA 2 . Since the calcium affinity of cPLA 2 is much higher in the presence of membranes, it is reasonable to assume that binding of both calcium ions occurs in the vicinity of the membrane surface. Thus, it is likely that the binding of two Ca 2ϩ ions and the membrane binding occur essentially simultaneously. In the case of protein kinase C-␣, it was demonstrated that the binding of calcium ion(s) to CBR3 induces the membrane penetration of protein, which leads to activation. Because the measurements of membrane penetration and/or conformational changes for Ca 2ϩ ligand mutants were hampered by low protein expression yields, it is unclear at present how exactly the two Ca 2ϩ ions play their specific roles. Based on the coordination geometry of the two Ca 2ϩ ions (see Fig. 1B) and the membrane binding properties of cPLA 2 , we speculate that CA2  induces the local conformational changes in the calcium binding loops to expose hydrophobic residues for hydrophobic membrane binding (see below), whereas CA1 binds to one or more membrane phospholipids to properly orient the cPLA 2 molecule for interfacial catalysis. This notion is based on the fact that CA2 has six coordinations to both CBR1 and CBR3, whereas CA1 has only four coordinations to CBR1 and CBR2 (see Fig.  1B). Thus, the higher affinity binding of CA2 to its ligands would alter the conformation of essential hydrophobic side chains in CBR1 and CBR3 and possibly promote the lower affinity binding of CA1 (hence the order of addition). On the other hand, CA1, due to its incomplete coordination sphere in the C2 domain, should still be capable of coordinating to membrane phospholipids even after binding to the C2 domain. Since the membrane binding of cPLA 2 is driven mainly by hydrophobic interactions (see below), the binding of CA2, if it indeed exposes all hydrophobic residues, would be more important for membrane binding than that of CA1, which would contribute less to conformational changes and thus primarily contribute to less important electrostatic interactions. The calcium-induced conformational changes of calcium binding loops containing essential hydrophobic residues are supported by a recent NMR study of the isolated C2 domain of cPLA 2 (13). Also, the notion that CA1 binds to phospholipids and thus contributes to electrostatic interactions is supported by our finding that N65A has a considerably higher [Ca 2ϩ ]1 ⁄2 value for anionic DHPGcoated bead binding than for DHPC-coated bead binding, whereas N95A shows the essentially the same calcium dependence for both types of beads (data not shown). Further investigation is needed to elucidate exact roles of the two calcium ions and determine the temporal and spatial sequences of calcium and membrane binding in the interfacial catalysis of cPLA 2 .
Hydrophobic Residues-Both electrostatic and hydrophobic interactions are involved in the membrane binding of most peripheral membrane binding proteins and their relative contributions vary with the type of proteins (see, for example, Refs. 30 -33). This study shows that the binding of C2 domain of cPLA 2 is mainly driven by hydrophobic interactions that are achieved by the penetration of hydrophobic residues in the calcium binding loops (CBR1 and CBR3) into the membrane core. To our knowledge, the C2 domain of cPLA 2 has the highest membrane penetrating power than any other C2 domain studied, including that of protein kinase C-␣, 3 presumably due to its unique structure, containing a number of hydrophobic residues in the calcium binding loops. A combined contribution of six hydrophobic residues to binding energy, which can be estimated using the equation ⌬⌬G o ϭ ϪRT ln(relative affinity) (see Table II), is 9.1 kcal/mol at 25°C, assuming the additivity of individual contributions. This suggests that the hydrophobic interactions are strong enough to drive the membrane binding of cPLA 2 . Our monolayer data show that Ca 2ϩ is essential for the membrane penetration of the C2 domain. As described above, this is mainly because Ca 2ϩ triggers local conformational changes in the calcium binding loops, which exposes hydrophobic residues and thereby elicits their membrane penetration. Mutational effects of hydrophobic residues on activity, membrane binding, and monolayer penetration indicate that four hydrophobic residues, Phe 35 and Leu 39 of CBR1 and Tyr 96 and Val 97 of CBR3, are directly involved in membrane penetration and hydrophobic interactions. Although these data apparently indicate that Tyr 96 makes the largest contribution to the membrane penetration, it should be taken into account that the Tyr to Ala mutation involves a larger change in accessible surface area than any other mutation. To assess more quantitatively the contribution of each amino acid to membrane penetration and hydrophobic interactions, one thus has to determine the change in free energy of binding per change in accessible surface area (⌬⌬G 0 /⌬A) caused by each mutation. These values determined for individual mutations are given in Table II. It was shown that an average ⌬⌬G 0 /⌬A value for transferring an amino acid from ethanol to water is ϳ24 cal/ mol/Å 2 (34). Thus, one can estimate from this value the degree of penetration of each hydrophobic residue in the calcium binding loops. For instance, Leu 39 and Val 97 the mutations of which yield high ⌬⌬G 0 /⌬A values of 30 and 40 cal/mol/Å 2 , respectively, should be fully inserted into the hydrophobic core of membranes, whereas Phe 35 and Tyr 96 with smaller ⌬⌬G 0 /⌬A values would partially penetrate into the membrane during membrane binding. This notion is also consistent with a general observation that aromatic side chains have a high tendency to reside in the membrane-water interface (35). If the membrane binding of cPLA 2 is driven primarily by hydrophobic 3 M. Medkova and W. Cho, manuscript in preparation. The backbone of the C2 domain is shown in a ribbon diagram, and mutated residues are illustrated in a space-filling representation. Leu 39 and Val 97 , whose side chains are fully inserted, are shown in red, and Phe 35 , Met 38 , Tyr 96 , and Met 98 , whose side chains are partially inserted, are shown in pink. Four cationic residues are shown in blue.
interactions, then its affinities for PC and phosphatidylglycerol membranes should be comparable. Thus, the PC preference of cPLA 2 and its C2 domain (28) suggests the presence of either a specific PC binding site(s) or an anionic patch on the membrane binding surface of the C2 domain. While the answer to this question entails further investigation, the PC preference of cPLA 2 makes it less likely that four cationic residues in the ␤3 strand are directly involved in membrane binding of cPLA 2 . This is also consistent with the finding that wild type cPLA 2 and the quadruple mutant (R57E/K58E/R59E/R61E) have comparable affinities for DHPG-coated beads. The modestly lower binding affinity of the mutant for PC membranes might be due to local structural changes that affect PC binding more significantly than phosphatidylglycerol binding. Similar differential effects of mutations on PC and phosphatidylglycerol binding are also seen with mutations of hydrophobic residues (see Table II).
Based on these findings, we propose a membrane binding mode of the C2 domain of cPLA 2 as illustrated in Fig. 7. This model is different from one proposed by Williams and co-workers (12) in that both CBR1 and CRB3 are involved in membrane penetration, and the cationic cluster does not make direct contact with the membrane surface. The finding that the isolated C2 domain of cPLA 2 has higher penetrating power than cPLA 2 indicates the difference in membrane penetrating mode between the isolated C2 domain and the C2 domain in the intact cPLA 2 molecule. Presumably, the carboxyl-terminal domain of cPLA 2 , which contains the active site, interferes with the maximal membrane penetration of the amino-terminal C2 domain. We previously showed that the partial membrane penetration of cPLA 2 is essential for its interfacial catalysis (15). Thus, it is possible that any membrane factors, such as diacylglycerols, which make up for the compromised membrane penetrating power of cPLA 2 may induce its membrane translocation and activation.
In summary, this report represents the second part of our systematic structure-function analysis on the C2 domain of membrane-binding proteins. These studies reveal similarities and differences between the C2 domains of protein kinase C-␣ and cPLA 2 in terms of their mechanisms of membrane binding and activation. The two C2 domains are similar in that they are not only a membrane docking unit but also a module that triggers membrane penetration of protein and in that their individual Ca 2ϩ ions bound to the calcium binding loops play differential roles in membrane binding and activation. On the other hand, individual Ca 2ϩ ions of the two C2 domains have different specific roles, and the membrane penetration of hy-drophobic residues plays a more important and direct role in the membrane binding of the C2 domain of cPLA 2 than in that of protein kinase C.