Structural aspects of interfacial adsorption. A crystallographic and site-directed mutagenesis study of the phospholipase A2 from the venom of Agkistrodon piscivorus piscivorus.

Recent genetic and structural studies have shed considerable light on the mechanism by which secretory phospholipases A2 interact with substrate aggregates. Electrostatic forces play an essential role in optimizing interfacial catalysis. Efficient and productive adsorption of the Class I bovine pancreatic phospholipase A2 to anionic interfaces is dependent upon the presence of two nonconserved lysine residues at sequence positions 56 and 116, implying that critical components of the adsorption surface differ among enzyme species (Dua, R., Wu, S.-K., and Cho, W. (1995) J. Biol. Chem. 270, 263-268). In an effort to further characterize the protein residues involved in interfacial catalysis, we have determined the high resolution (1.7 Å) x-ray structure of the Class II Asp-49 phospholipase A2 from the venom of Agkistrodon piscivorus piscivorus Correlation of the three-dimensional coordinates with kinetic data derived from site-directed mutations near the amino terminus (E6R, K7E, K10E, K11E, and K16E) and the active site (K54E and K69Y) defines much of the interface topography. Lysine residues at sequence positions 7 and 10 mediate the adsorption of A. p. piscivorus phospholipase A2 to anionic interfaces but play little role in the enzyme's interaction with electrically neutral surfaces or in substrate binding. Compared to the native enzyme, the mutant proteins K7E and K10E demonstrate comparable (20-fold) decreases in affinity and catalysis on polymerized mixed liposomes of 1-hexadecanoyl-2-(1-pyrenedecanoyl)-sn-glycero-3-phosphocholine and 1,2-bis[12-(lipoyloxy)dodecanoyl]-sn-glycero-3-phosphoglycerol, while the double mutant, K7E/K10E, shows a more dramatic 500-fold decrease in catalysis and interfacial adsorption. The calculated contributions of Lys-7 and Lys-10 to the free energy of binding of A. p. piscivorus phospholipase A2 to anionic liposomes (−1.8 kcal/mol at 25°C per lysine) are additive (i.e. −3.7 kcal/mol) and together represent nearly half of the total binding energy. Although both lysine side chains lie exposed at the edge of the proposed interfacial adsorption surface, they are geographically remote from the corresponding interfacial determinants for the bovine enzyme. Our results confirm that interfacial adsorption is largely driven by electrostatic forces and demonstrate that the arrangement of the critical charges (e.g. lysines) is species-specific. This variability in the topography of the adsorption surface suggests a corresponding flexibility in the orientation of the active enzyme at the substrate interface.

sponding interfacial determinants for the bovine enzyme. Our results confirm that interfacial adsorption is largely driven by electrostatic forces and demonstrate that the arrangement of the critical charges (e.g. lysines) is species-specific. This variability in the topography of the adsorption surface suggests a corresponding flexibility in the orientation of the active enzyme at the substrate interface.
Phospholipases A 2 (PLA 2 ; EC 3.1.1.4) 1 catalyze the hydrolysis of the fatty acid ester in the 2-position of 3-sn-phospholipids and are found both in intracellular and secreted forms (for recent reviews, see Refs. [1][2][3][4]. The enzymes act at the lipidwater interface with a preference for organized lipids (micelles and vesicles) that is often orders of magnitudes greater than that shown for dispersed substrate. Calcium-mediated catalysis appears to involve two kinetically and structurally distinct steps (5,6). Adsorption of PLA 2 to the interface precedes and is independent of substrate binding to the active site. Secretory PLA 2 s are small proteins (14 kDa) that can be classified into at least four groups based on structural differences (7,8). Class I (exocrine pancreas and elapidae and hydrophodiae snake venoms) and Class II (mammalian nonpancreatic and crotalidae and viperidae snake venoms) enzymes are highly homologous (see Fig. 1) and have been extensively studied. Crystallographic and NMR analyses of multiple secretory PLA 2 s have implicated a common interfacial adsorption surface that is structurally distinct from the active site (2,9). The proposed interfacial adsorption site is located on a flat external surface which incorporates a ring of hydrophobic and cationic residues. Recent genetic studies have shown that the adsorption of the Class I bovine pancreatic PLA 2 to anionic interfaces is driven primarily by a few cationic residues, particularly Lys-56 and Lys-116 (10). These residues are not conserved among PLA 2 s, indicating that many of the key determinants of interfacial adsorption differ among enzyme species.
The venom of eastern cottonmouth, Agkistrodon piscivorus piscivorus, contains two cationic monomeric PLA 2 s, Asp-49 PLA 2 (App-D49) and Lys-49 PLA 2 , and a dimeric PLA 2 . Among these PLA 2 s, App-D49 has served as the prototype for a Class * The work at the University of Illinois at Chicago was supported by a Biomedical Science Grant from the Arthritis Foundation, Grant-inaid AHA 95006280 from the American Heart Association, and National Institutes of Health Grant GM52598. The work at Yale University was supported by United States Public Health Service Grants GM24324 and NS25867 and by the Howard Hughes Medical Institute. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Construction of Mutant Phospholipase A 2 Genes-The
App-D49 gene, synthesized based on the corrected amino acid sequence of App-D49 (21), was a generous gift from Dr. Gordon Rules of the University of Virginia. The coding region for App-D49 was subcloned into a pET-21-a vector (Novagen; Madison, WI) and designated as pSH-app. This synthetic gene carries the Asn to Ser mutation (N1S) at the amino terminus to facilitate the removal of the initiator Met by an endogenous methionine aminopeptidase (22). The mutant proteins were created using a Sculptor in vitro mutagenesis kit (Amersham Corp.) according to the method of Nakamaye and Eckstein (23) with modifications. In this method, a phagemid DNA prepared from the pSH-app vector in the presence of the helper phage R408 (Promega; Madison, WI) served as a template for mutagenesis, and the mutagenesis protocol was slightly modified to optimize the mutation efficiency. The oligonucleotides used for the construction of the mutants were 5Ј-CTT GAT TAA TTT CCT  AAA CTG GAA CAG-3Ј (E6R), 5Ј-TT CTT GAT TAA TTC CTC AAA  CTG GAA CAG-3Ј (K7E), 5Ј-CC AGT CAT TTT CTC GAT TAA TTT  CTC-3Ј (K10E), 5Ј-T ACC AGT CAT TTC CTT GAT TAA TTT-3Ј (K11E),  5Ј-G CAT TCC GGA CTC ACC AGT CAT TTT-3Ј (K16E), 5Ј-ACA GCC  GGT AAC CTC ACC GTA GCA GCA-3Ј (K54E), and 5Ј-GTA GAT ATC CAT ATA CGG GTT ACA GCC-3Ј (K69Y), respectively, in which the underlined bases indicate the location of the mutation. The K7E/K10E double mutation was performed sequentially; i.e. the K7E mutation followed by the K10E mutation. After the DNA sequences of the entire coding regions of mutants were verified by sequence analysis (24) using a Sequenase 2.0 kit (Amersham), individual recombinant pSH-app vectors were transformed into Escherichia coli strain BL21(DE3) (Novagen) for protein expression.
Protein Expression and Purification-Lyophilized venom from A. p. piscivorus was purchased from the Miami Serpentarium (Salt Lake City, UT). The dimeric Asp-49 PLA 2 , the monomeric Lys-49 PLA 2 , and App-D49 were isolated as described elsewhere (11). The pooled App-D49 was rechromatographed on a CM-Sephadex C-50 column at pH 3.5 and developed with a linear gradient of 50 to 400 mM ammonium formate. This latter step removed several minor impurities which interfered with crystallization of the major enzyme species. The chromatographic and catalytic behaviors of the native App-D49 and the recombinant N1S were indistinguishable (22). Thus, the recombinant N1S will be referred to as wild type hereafter. Wild type and mutant proteins were expressed in BL21(DE3) cells harboring the corresponding pSH-app vectors. A 2-liter culture (Luria broth containing 50 g/mL ampicillin) was grown at 37°C, and protein expression was induced by the addition of 0.4 mM isopropyl ␤-D-thiogalactopyranoside when the absorbance of the medium reached 1.0 ( ϭ 600 nm). After the cells were incubated for an additional 4 h at 37°C, they were harvested, resuspended in 50 ml of 50 mM Tris-HCl buffer, pH 8.0, containing 20 mM EDTA and 0.1% Triton X-100, and incubated for 10 min on ice. Pulse-mode sonication was performed using a Sonifier 450 (Branson) for 10 times at 15 s each. The inclusion body pellet obtained by centrifugation for 20 min at 10,000 ϫ g/4°C was resuspended in 5 ml of 0.2 M Tris-HCl buffer, pH 8.0, containing 4 M guanidinium isothiocyanate and 3 mM EDTA. The insoluble matter was removed by centrifugation at 35,000 ϫ g for 1 h at room temperature, and the inclusion body protein was sulfonated using 2-nitro-5-(sulfothio)-benzoate (10). The reaction mixture was then loaded onto a Sephadex G-25 column (2.5 ϫ 30 cm) equilibrated with 50 mM Tris-HCl buffer, pH 8.0, containing 2 M guanidinium chloride and 3 mM EDTA, and the major protein peak was collected. The sulfonated protein was refolded by slowly adding to the protein solution 200 ml of 25 mM Tris-HCl buffer, pH 8.0, containing 15 mM CaCl 2 , 1 mM EDTA, 2.5 mM reduced glutathion, 0.5 mM oxidized glutathion, and 5 mM dodecyl sucrose. The solution was kept at room temperature for 40 h, at which point the protein solution was dialyzed overnight against 4 liters of 25 mM Tris-HCl buffer, pH 8.0. The solution was filtered through a small (2.5 ϫ 2 cm) Sephadex G-25 column equilibrated with 25 mM HEPES buffer, pH 8.0, to remove any precipitates and to exchange buffer solutions. The clear solution was loaded onto a Pharmacia CM-Sepharose column (2.5 ϫ 20 cm) equilibrated with 25 mM HEPES buffer, pH 8.0. The column was washed extensively with 25 mM HEPES buffer, pH 8.0, until no further peaks emerged, and then the folded protein was eluted with a gradient of 0 to 1 M NaCl in the same buffer (the native App-D49 eluted with 0.7 M NaCl). The collected protein peak was dialyzed against water, lyophilized, and stored at Ϫ20°C. Protein purity was confirmed both by SDS-polyacrylamide electrophoresis and isoelectric focusing. Protein concentrations were determined by the micro bicinchoninic acid method (Pierce). The circular dichroism spectra of proteins were measured in 10 mM phosphate buffer, pH 7.4, at 25°C, using a Jasco J-600 spectropolarimeter. Each spectrum was obtained at wavelengths between 195 and 300 nm and averaged from 10 separate scans.
Structure Determination-Large, single crystals of App-D49 suitable for diffraction work were grown by the vapor diffusion method. Ten-l droplets containing 30 g/l App-D49, 20% v/v ethanol, and 0.1 M Tris-HCl, pH 7.5, were plated onto plastic coverslips and inverted over 1-ml reservoirs of 40% v/v ethanol and 0.2 M Tris-HCl, pH 7.5. Crystals, which nucleated within 2 weeks, grew to their maximum size of 0.7 ϫ 0.2 ϫ 0.2 mm after a variable period ranging from 3 to 6 weeks. Diffraction data for the App-D49 structure were collected from a single crystal with a dual panel San Diego Multiwire System detector at 4°C using graphite monochromated CuK ␣ emission from a RU-300 x-ray generator. The space group was determined to be P1 with unit cell dimensions of a ϭ 34.71 Å, b ϭ 39.89 Å, c ϭ 42.85 Å, ␣ ϭ 94.17°, ␤ ϭ 79.10°, and ␥ ϭ 106.27°. The collected data were processed with SCALZO (25) and yielded a final data set that was 82% complete at 1.5 Å ( Table I). Because of significant degradation of the data at high resolution, however, a nominal cutoff of 1.7 Å was chosen for refinement purposes. The calcium-free crystal structure of App-D49 was solved by molecular replacement with a modified version of the computer program Merlot 1.5. The search model was created by pruning nonhomologous side chains from one subunit of the refined coordinates of the App-D49 dimer (80% sequence homology). The rotation function yielded two strong peaks consistent with the expected presence of two independent molecules in the crystallographic asymmetric unit. The intermolecular translation function was also sharp generating a solution with excellent packing. Refinement with the computer programs PROFFT and X-PLOR rapidly reduced the R factor while maintaining reasonable stereochemistry (Table II). When necessary, the coordinates were manually rebuilt with the graphics program FRODO. Two hundred and sixty-six water molecules were assigned based on density in the Fourier difference maps, hydrogen-bonding potentials, and refined temperature factors of less than 45 Å 2 at full occupancy. A second group of 52 less well defined water molecules refined to temperature factors ranging from 46 to 65 Å 2 . The distal portions of solvent-exposed side chains (e.g. lysines) accounted for the majority of the protein disorder.
Kinetic Measurements-PLA 2 -catalyzed hydrolysis of polymerized mixed liposomes was performed at 37°C in 2 ml of 10 mM HEPES buffer, pH 7.4, containing 0.1 M pyrene-containing phospholipids (1 mol %) inserted into 9.9 M BLPG, 2 M bovine serum albumin, 0.16 M NaCl, and 10 mM CaCl 2 . Enzyme concentrations were adjusted to keep the half-life of the reaction below 5 min (e.g. [wild type] ϭ 1-10 nM and [K7E/K10E] ϭ 0.1-1 M). Progress of the reaction was monitored as an increase in fluorescence emission at 378 nm using a Hitachi F4500 fluorescence spectrometer with an excitation wavelength of 345 nm. The spectral bandwidth was set at 5 nm for both excitation and emission. The pseudo-first-order rate constant was determined from the nonlinear least squares analysis of the reaction progress curves (10,18,19). This was divided by the enzyme concentration to obtain the apparent second-order constant, (k cat /K m ) app . The kinetics of PLA 2 -catalyzed hydrolysis of diC 8 PC and diC 8 PE micelles were measured in the pres-ence of 0.5 mM phospholipid, 0.16 M NaCl, and 10 mM CaCl 2 with enzyme concentrations between 1-10 nM. The time course of the phospholipid hydrolysis was monitored with a computer-controlled Metrom pH-stat (Brinkmann) in a thermostated vessel. Under these conditions, the hydrolysis of diC 8 PC (and diC 8 PE) followed first-order kinetics, since the substrate concentrations remained lower than the apparent K m values.
Liposome-Protein Binding Measurements-The adsorption of the wild type and mutant enzymes to polymerized liposomes was examined fluorometrically. Typically, the fluorescence emission (excitation at ϭ 280 nm) of the protein solution (0.5 M) at 345 nm (F 345 ) was measured at equilibrium in 2 ml of 10 mM HEPES buffer, pH 7.4, containing 0.1 to 100 M BLPG polymerized liposomes and varying concentrations of NaCl and CaCl 2 . An initial decrease in protein fluorescence was observed due to the adsorption of the protein to the cuvette wall (less than 5% of the total signal) (26). Protein concentrations were, therefore, corrected for this loss and the liposome solution was added only after the protein signal stabilized. To minimize potential artifacts due to both the inner filter effect and the scattering by BLPG polymerized liposomes, each observed F 345 value was background-corrected against the F 345 value observed in the presence of the same concentration of BLPC polymerized liposomes. BLPC polymerized liposomes have the same spectroscopic properties as BLPG polymerized liposomes but App-D49 has an extremely low affinity for them (apparent dissociation constant Ͼ 1 mM) (14). The relative fluorescence change (F rel ) for each phospholipid concentration was calculated as (F max Ϫ F 345 )/(F max Ϫ F min ) where F max and F min represent the protein fluorescence at 345 nm in the absence of BLPG and in the presence of an excess amount of BLPG (e.g. 0.2 mM), respectively. Values of n and K d were determined by the nonlinear least squares analysis of the F rel versus [BLPG] o data using Equation 1, where [E] o and [BLPG] o indicate total enzyme and BLPG concentrations, respectively. This equation assumes that each phospholipid molecule binds independently with a dissociation constant of K d to n equivalent sites on a PLA 2 molecule.

Crystal Structure of App-D49 and the Design of App-D49
Mutants-App-D49 is a typical Class II PLA 2 with high sequence homology to other members of this class (Fig. 1). The enzyme was crystallized in the absence of added calcium ion and, as expected, both independent molecules in the crystallographic asymmetric unit lack discernible metal ions (Fig. 2). Instead, a single water molecule replaces each of the primary calcium ions and makes comparable hydrogen bonds to a carboxylate oxygen of Asp-49 and three backbone carbonyls (Tyr-28, Gly-30, and Gly-32). Least squares superimposition of the ␣-carbon traces of App-D49 and its dimeric counterpart from A.  p. piscivorus venom reveals the anticipated conservation of the homologous core of tertiary structure (Fig. 3). The root-mean square deviation of the homologous cores of App-D49 and the dimer is 0.66 Å, whereas the comparable value for the non-core regions is 1.42 Å. The differences between the enzymes, as well as the differences that may be attributed to the effects of pH and crystallization media on the structure of App-D49, are largely confined to those regions whose conformations have been previously noted to be highly variable (e.g. the calcium binding loop (residues 26 -34), the ␤-wing (residues 74 -84), and the carboxyl terminus (residues 123-131)). The residues focused on in this study (Glu-6, Lys-7, Lys-10, Lys-11, Lys-16, Lys-54, and Lys-69) are all well resolved in the electron density map as they reside in stable regions with little protein disorder. The side chains of Lys-7, Lys-10, Lys-16, and Lys-54 can be positioned to their respective C␦ atoms, whereas Arg-6, Lys-11, and Lys-69 can be visualized in their entirety. The dimensions of the crystalline App-D49 are approximately 22 ϫ 30 ϫ 42 Å with 50% of the structure ␣-helix and 10% ␤-sheet. The two long anti-parallel ␣-helices that form the backbone of the molecule (residues 37-54 and 90 -109) are securely braced by a series of disulfide bridges. The active site residues (His-48, Asp-49, Tyr-73, and Asp-99) occupy the base of a short internal hydrophobic cavity or cleft that opens onto the external surface. The disposition of productively bound substrate within the active site has been shown through the co-crystallization of several PLA 2 s with transition state and substrate analogs (17,(27)(28)(29). The structural work (9, 30), along with extensive biochemical (31, 32), genetic (10, 33-36), and electrostatic data (37), suggests that the interfacial adsorption surface surrounds and incorporates the external opening of the hydrophobic channel.
All of the mutant proteins were expressed and refolded as efficiently as the wild type (data not shown), indicating similar thermodynamic stability. Retention times for the refolded wild type App-D49 and the native enzyme were identical on a CM-Sepharose column. The mutant enzymes, including the double mutant (K7E/K10E), migrated as expected for their altered surface charges. The CD spectra of the native, wild type, and the mutant proteins were indistinguishable (Fig. 4) as were estimates of ␣-helical content based on 222 values (ϳ50%), indicating that the mutations did not induce any gross structural change in protein conformation. The preservation of tertiary structure is consistent with the surface locations of the altered side chains.
Kinetic Properties of App-D49 and Mutants-A major advantage of the polymerized mixed liposome system over other PLA 2 kinetic systems is that it allows an unambiguous distinction between phospholipids interacting with the active site of PLA 2 (i.e., hydrolyzable inserts) and ones interacting with the interfacial binding site of PLA 2 (i.e. polymerized matrix) (18). Thus, it is possible to separately determine the head group specificity of the interfacial adsorption surface and the active site. Furthermore, it is also possible to separately analyze the effects of protein and phospholipid modification on interfacial adsorption and substrate binding (10,18,33). Using this approach, we measured the kinetic properties of wild type and mutants toward a wide variety of polymerized mixed liposomes. Two zwitterionic phospholipids, pyrene-PC and pyrene-PE, and anionic pyrene-PG were used as inserts in the anionic BLPG polymerized matrix to determine the phospholipid head group specificity of the active site of each enzyme. Apparent second-order constants, (k cat /K m ) app , determined for these proteins are summarized in Table III. Although (k cat /K m ) app has no obvious physical meaning, it is a useful parameter for comparing overall interfacial activity of wild type and mutants (10,33). In agreement with a previous report (22), recombinant wild type App-D49 showed essentially the same activity and head group specificity toward polymerized mixed liposomes as native App-D49. The active sites of both enzymes prefer anionic phospholipids to zwitterionic ones by ϳ3.5-fold. This is a modest value when compared to the high (up to 5,000-fold) adsorption preference of App-D49 for anionic interfaces (18). Of note, App-D49 has comparable activities toward pyrene-PC and pyrene-PE substrates indicating that the enzyme cannot distinguish between the two zwitterionic head groups. Thus, the ratio of enzyme activity (pyrene-PG/pyrene-PC) can be used to describe the anionic phospholipid preference of each protein (see Table III).
The cluster of lysines lying adjacent to the carboxyl end of the amino-terminal ␣-helix is geographically remote from the catalytic site (e.g., Ͼ15 Å from His-48). Mutation of these residues to Glu had no appreciable effect on the enzyme's head group preference. These mutations did, however, significantly decrease adsorption and catalysis on polymerized mixed liposomes. When compared to the wild type, K7E and K10E demonstrated an approximately 20-fold drop in activity toward pyrene-PC/BLPG polymerized mixed liposomes. In contrast, K11E and K16E displayed modest 3.8-and 2.6-fold decreases, respectively. The rate decrease for the double mutant (K7E/ K10E) was 500-fold on polymerized mixed liposomes and approximated the product of the decreases for the individual mutations. To investigate the effect of mutations on the binding of App-D49 to electrically neutral interfaces, we measured the activities of wild type and mutants using zwitterionic diC 8 PC micelles. These micelles were chosen because they provide neutral interfaces and yet are a good substrate for App-D49 due to their loose interfacial packing density. When assessed using these micellar substrates, none of the amino-terminal mutations caused a significant drop in activity. Even the K7E/K10E mutant showed 83% of wild type activity. This suggests that Lys-7 and Lys-10 play an essential role in binding to anionic interfaces but are dispensable in adsorbing to electrically neutral interfaces or in substrate binding.
Many secretory PLA 2 s contain a cationic residue at position 6 in addition to those at position 7 and 10. If the enzymatic activity of App-D49 toward substrate aggregates is simply proportional to its electrostatic affinity for the interface, one would expect that the E6R mutant with additional positive charge in the vicinity of Lys-7 would be more active on anionic polymerized mixed liposomes. Surprisingly, the E6R enzyme displayed a 4-fold decrease in activity toward all polymerized mixed liposomes. This inhibition was not due to a disruption of the catalytic apparatus, since the activity of E6R toward electrically neutral diC 8 PC micelles was only modestly affected (i.e. a 13% decrease). The decrease in activity suggests, however, that this mutant binds to the BLPG polymerized mixed liposomes in The view of the enzyme shown here is similar to that used in previous publications to illustrate the location of a co-crystallized transition-state analog (9,17,28). The active site lies at the base of the central cavity formed from the amino-terminal helix, residues 19 -23, portions of the calcium-binding loop, and the side chain of Lys-69 and is indicated by the side chain of His-48 (in black). The plane of the putative interfacial adsorption surface lies perpendicular to the hydrophobic channel and incorporates residues surrounding the external opening of the channel. In the present study, specific lysine residues (Lys-7, Lys-10, Lys-11, Lys-16, Lys-54, and Lys-69) were changed into glutamates and tyrosine (Lys-69) in an effort to characterize the structural determinants of interfacial adsorption. a suboptimal mode. The activity of PLA 2 toward anionic interfaces is, therefore, not merely proportional to the net charge of the interfacial adsorption surface.
To identify the origin of App-D49's small, but definite, preference for anionic head groups, two lysines (Lys-54 and Lys-69) that are located close to the active site were mutated, and the resulting proteins' activities were measured on polymerized mixed liposomes and micelles. If the active site of App-D49 prefers pyrene-PG to pyrene-PC because of electrostatic repulsion between the lysine side chain(s) and the choline head group, one would expect that K54E and K69Y would have enhanced activities toward pyrene-PC while showing the same activities toward pyrene-PG as the wild type. On the other hand, if the preference for pyrene-PG derives from favorable interactions (e.g. hydrogen bonding) between the lysine(s) and the glycerol head group, the mutants should show reduced activity on pyrene-PG but not on pyrene-PC. As it turned out, the K54E mutation had little effect on the enzyme activity toward pyrene-PC (15% decrease), pyrene-PG (20% decrease), or micellar diC 8 PC (17% decrease). This mutation did, however, result in a 2-fold enhancement of activity on pyrene-PE in both polymerized mixed liposomes and micellar diC 8 PE. In the case of diC 8 PE, the (k cat /K m ) app values for wild type and the K54E mutant were 4.5 ϫ 10 5 M Ϫ1 s Ϫ1 and 1.0 ϫ 10 6 M Ϫ1 s Ϫ1 , respectively. These results suggest that Lys-54 does not directly interact with productively bound phospholipids, but a glutamate side chain at this position might potentially form a favorable electrostatic contact with the ammonium ion of phosphatidylethanolamine. The activity of the K69Y mutant toward both pyrene-PC and pyrene-PE lipids in polymerized mixed liposomes was identical to the wild type, indicating that the Lys-69 side chain has a negligible effect on the productive binding of zwitterionic head groups. The mutation did, however, decrease the enzyme activity toward pyrene-PG/BLPG by approximately 3-fold, rendering K69Y nonselective for the phospholipid head group. The modest nature of these changes underlines the equivalent abilities of the tyrosine and lysine side chains in coordinating the sn-3 phosphate of a productively bound phospholipid. At the same time, the decrease in activity toward the pyrene-PG substrate suggests that the ⑀-ammonium group of Lys-69 may form hydrogen bond(s) with the glycerol head group which are not achievable with the phenolic hydroxyl of Tyr-69 (Fig. 5).
Binding of App-D49 and Mutants to BLPG Polymerized Liposomes-To delineate the relationship between the enzyme activity and the interfacial adsorption affinity of mutants, we measured their binding to BLPG polymerized liposomes. We previously showed that nonhydrolyzable BLPG polymerized liposomes could be used to determine the dissociation constant for PLA 2 -liposome complex (18,19). The n and K d values (Table  IV) determined for the wild type and mutant enzymes were derived from curve fittings of the respective isotherms (Fig. 6). All of the proteins showed analogous saturation binding curves, indicating similar binding modes. Under the conditions in which the kinetic measurements were performed (10 mM Ca 2ϩ and 0.16 M NaCl), the wild type enzyme showed a high affinity for anionic BLPG liposomes comparable to that reported for the native App-D49 (n⅐K d ϭ 0.7 M) (19). Calcium ion had no effect on this binding (Fig. 6). When compared to the wild type, the mutant enzymes displayed a wide range of K d values with only minor variation in n (n Ϸ 40). In general, the relative binding affinity of mutants to BLPG polymerized liposomes paralleled their relative enzymatic activity toward pyrene-PC/BLPG polymerized mixed liposomes. For example, K7E, K10E, and K7E/K10E had 50-, 45-, and 500-fold increases in K d , respectively. In the case of K7E/K10E, for which the binding saturation could not be reached with BLPG concentration up to 200 M, the K d value was determined from a curve fitting assuming n ϭ 40. Also, K11E and K16E showed 2-fold increases in K d , whereas K54E and K69Y had essentially the same K d as wild type. For mutants of amino-terminal lysines, these results indicate that their reduced activities are primarily due to the decreases in interfacial adsorption affinities and again underscore the essential roles of Lys-7 and Lys-10 in the interfacial adsorption of App-D49. For K54E and K69Y, the binding data confirm that their altered substrate selectivity is solely due to changes in substrate binding. From these kinetic and binding data, it appears that wild type and mutants all bind to anionic BLPG-based polymerized mixed liposomes in the same productive mode and that tighter interfacial binding will lead to higher interfacial activity of enzyme. A notable exception was E6R which showed 3-fold increase in binding affinity despite its 4-fold decrease in activity. For this mutant, the enhanced binding to anionic interfaces is consistent with an increase in net positive charge in the vicinity of the critical Lys-7 and Lys-10. Therefore, the decrease in enzyme activity for this mutant again suggests that it might bind to the BLPG polymerized mixed liposomes in a suboptimal mode. Finally, we measured the effect of ionic strength of the medium on the adsorption of wild type and mutants to BLPG polymerized liposomes. As summarized in Table IV, the binding of wild type was significantly diminished in the presence of 1 M NaCl, whereas the mutants with lower interfacial binding affinities were less sensitive to the ionic strength. In the presence of 1 M NaCl, the wild type protein and the K7E and K10E mutants showed comparable binding affinities for BLPG polymerized liposomes. Furthermore, the n⅐K d values for these enzymes were smaller than that of the K7E/K10E mutant only by an order of magnitude, demonstrating the electrostatic nature of the interfacial adsorption of App-D49 to anionic BLPG polymerized liposomes.

DISCUSSION
Tertiary Structure of App-D49 and Mutants-The three-dimensional structure of App-D49, crystallized in the absence of calcium ion, shows the structural features typical of other Class II PLA 2 s. The catalytic site, calcium-binding apparatus, and hydrophobic channel are identical to those described pre-viously. Based on analogy with other PLA 2 s, the interfacial adsorption surface of App-D49 surrounds the opening of the hydrophobic channel and includes residues 2 (Leu), 3 (Phe), 6 (Glu), 19 (Met), as well as portions of the calcium-binding loop. App-D49 is notable for its relatively large number of cationic residues with a correspondingly high pI (9.5). These charges are arranged asymmetrically with maximal positive charge potential in the region of the putative interfacial adsorption  N. n. atra (A). Class II PLA 2 s, including App-D49, substitute a lysine residue for the tyrosine at sequence position 69. The K69Y mutant has essentially the same activity as the wild type enzyme toward PC and PE substrates but shows a 3-fold drop in activity toward PG substrate. One explanation for this finding is that the ⑀-ammonium group of Lys-69 forms additional hydrogen bonds with phospholipid head groups, especially with PG whose hydroxyl groups can function as hydrogen bond acceptors (B). Such an interaction would not be achievable by the phenolic oxygen of Tyr-69 or with PC and PE as substrate.  surface (Fig. 7B). This distribution of positive charge potential is in sharp contrast to that observed for bovine pancreatic PLA 2 . The residues chosen for mutation (Glu-6, Lys-7, Lys-10, Lys-11, Lys-16, Lys-54, and Lys-69) lie on the perimeter of the proposed interfacial adsorption surface (Fig. 7A). Mutations at these positions do not significantly affect the tertiary structure as determined by CD. With the exception of Glu-6 and Lys-54, all of the side chains are solvent-exposed lysines facing the same side of the molecule. Lys-7 and Lys-10 are located on the same side of the amino-terminal ␣-helix, whereas Lys-11 and Lys-16 are both located further from the hydrophobic channel with their side chains remote from the plane of the proposed interfacial adsorption surface (Fig. 8).
Interfacial Adsorption of App-D49 -App-D49 normally shows high enzymatic activity toward anionic phospholipid aggregates but extremely low activity toward densely packed zwitterionic phospholipid monolayers and bilayers (12,14,38). As a result, the activation of App-D49 on zwitterionic aggregates generally requires the formation of anionic lipid domains containing a reaction product such as a fatty acid (38). The contributions of Lys-7 and Lys-10 to the free energy of binding of App-D49 to anionic BLPG polymerized liposomes can be estimated using the equation ⌬⌬G 0 ϭ R⅐T ln (n⅐K d for wild type/n⅐K d for mutant). Under standard conditions with the concentration of free phospholipid set at 1 M, each lysine contributes approximately Ϫ1.8 kcal/mol at 25°C. This calculated value compares well with the known value (Ϫ1 to Ϫ3 kcal/mol) for electrostatic interactions between two oppositely charged residues in proteins (39) and with our previous mutational experiments with the bovine pancreatic PLA 2 (10). Based on these structural and functional data, it is probable that the enzyme's helical surface containing Lys-7 and Lys-10 lies parallel to the membrane during interfacial adsorption. Since the contributions of Lys-7 and Lys-10 to the free energy of interfacial binding/adsorption appear additive (Ϫ3.7 kcal/mol at 25°C), there is, at most, limited synergy in the binding of the two lysines to anionic interfaces (40). The strong electrostatic effect on interfacial adsorption is supported by the striking correlation between adsorption and the ionic strength of the medium. Other contributions to the binding energy presumably come from a number of weaker electrostatic and hydrophobic interactions (for instance, see Maliwal et al. (32)). The presence of calcium ion does not appreciably affect the interfacial adsorption of App-D49, consistent with the minor effect that an additional calcium ion has on the protein's global electrostatics (37).
Lys-7 and Lys-10 are clearly the dominant residues directing the interfacial adsorption of App-D49. This is in contrast to our previous finding (10) for the bovine pancreatic PLA 2 where Lys-56 and Lys-116 are critical residues confirming that the determinants of interfacial adsorption are specific to the species of PLA 2 being studied. Our results also dispute the notion that the interfacial adsorption of secretory PLA 2 is driven by a large number of weak interactions which cannot be specifically blocked. Whether the variability in interface topology translates into orientational differences at the interface remains  Fig. 2. The proposed interfacial adsorption surface is oriented in the plane of the page with its free surface facing the reader. Most of the side chains undergoing mutation (Lys-7, Lys-10, Lys-11, Lys-16, and Lys-69) can be seen in this view. Anionic residues are colored red, cationic residues blue, hydrophobic residues green, and cysteines yellow. B, electrostatic representations of the surface shown in A. Both representations are contoured similarly. Note the distinct cationic patch at the right base which is composed of Lys-7, Lys-10, Lys-11, and Lys-16. The former two residues are particularly important in the adsorption of App-D49 to anionic surfaces. unclear since hydrophobic groups also play key roles in optimizing the interaction. The introduction of an additional charge to the amino-terminal ␣-helix by the E6R mutation modestly (⌬⌬G 0 ϭ Ϫ1.1 kcal/mol at 25°C) increases the affinity of App-D49 for anionic interfaces. This enhanced adsorption, however, does not facilitate productive-mode substrate binding to the active site since E6R has lower activity toward anionic polymerized mixed liposomes than the wild type. This reduction in activity may stem from electrostatic repulsion between Arg-6 and Lys-7 leading to a nonoptimal orientation of the lysine side chain. Alternatively, the side chain of Arg-6 may mechanically interfere with the proper orientation of the enzyme at the interface.
Roles of Lys-54 and Lys-69 -Residues 53 through 58 lie at the carboxyl terminus of the first anti-parallel ␣-helix. This section of the enzyme lies under the calcium-binding loop and adjacent to the region occupied by the head group of productively bound substrate. The charge and orientation of side chains in this region have been shown to influence the head group selectivity of PLA 2 . For example, Lys-53 of the bovine pancreatic PLA 2 (10) and Arg-53 of the porcine pancreatic PLA 2 (41) appear to be responsible for the anionic phospholipid preferences of these enzymes. Glu-56 of the Class II human nonpancreatic enzyme makes a direct stabilizing contact with the ammonium ion of the PE head group in the x-ray structure of the transition-state analog complex (17). The side chain of App-D49's Lys-54 points toward bulk solvent and, consistent with this orientation, its mutation to Glu did not significantly change the enzyme activity toward PC or PG substrates. The small (2-fold) increase in activity of K54E toward PE substrates either in polymerized mixed liposomes or in micelles indicates that the ␥-carboxylate of Glu-54 may weakly interact with the ethanolamine head group. The calculated decrease in substrate binding energy due to this putative electrostatic interaction is correspondingly small; 0.4 kcal/mol at 25°C (⌬⌬G 0 ϭ ϪR⅐T ln ((k cat /K m ) app for mutant/(k cat /K m ) app for wild type)).
The crystal structures of several transition-state and substrate analog complexes have shown that both the Tyr-69 of Class I PLA 2 s and the Lys-69 of Class II enzymes form a hydrogen bond with the pro-S nonbridging oxygen of the sn-3 phosphate (17, 29) (see Fig. 5). We suggest that the ⑀-ammonium group of Lys-69 may be able to form additional hydrogen bonds with phospholipid head groups, especially with PG whose hydroxyl groups can function as hydrogen bond acceptors (Fig. 5B). Such an interaction would not be achievable by the phenolic oxygen of Tyr-69 or with PC and PE as substrate. This would explain why K69Y has the essentially the same activity as wild type toward PC and PE substrates but shows a 3-fold drop in activity toward PG substrate. Our recent data indicate that the Lys-69 of App-D49 and of the human nonpancreatic PLA 2 also interacts favorably with other anionic phospholipids (e.g. phosphatic acid and phosphatidylserine) that can serve as strong hydrogen bond acceptors (42). Class II secretory PLA 2 s presumably show a modest degree of substrate selectivity due to the ability of Lys-69 to form a hydrogen bond(s) with anionic phospholipid head groups. FIG. 8. A stereoview of the amino-terminal region of the crystalline App-D49. The orientation of the peptide fragment is similar to that shown in Fig. 2. The proposed interfacial surface is oriented in the plane of the page with its free surface facing the reader. The opening of hydrophobic channel lies immediately above the amino-terminal helix with the channel's axis perpendicular to the interfacial adsorption surface. The integrity of the amino-terminal helix is essential for the high catalytic efficiency associated with interfacial catalysis. Residues 2, 5, and 9, which contribute to the inner walls of the hydrophobic channel, are invariant among PLA 2 . Mutation of residues in this region (Glu-6, Lys-7, Lys-10, Lys-11, and Lys-16) demonstrates a critical role for Lys-7 and Lys-10 in the enzyme's interaction with anionic phospholipids.