Localization of structural elements of bee venom phospholipase A2 involved in N-type receptor binding and neurotoxicity.

We have shown previously that neurotoxic venom secretory phospholipases A2 (sPLA2s) have specific receptors in brain membranes called N-type receptors that are likely to play a role in the molecular events leading to neurotoxicity of these proteins. The sPLA2 found in honey bee venom is neurotoxic and binds to this receptor with high affinity. In this paper, we have used a number of mutants of bee venom sPLA2 produced in Escherichia coli to determine the structural elements of this protein that are involved in its binding to N-type receptors. Mutations in the interfacial binding surface, in the Ca2+-binding loop and in the hydrophobic channel lead to a dramatic decrease in binding to N-type receptors, whereas mutations of surface residues localized in other parts of the sPLA2 structure do not significantly modify the binding properties. Neurotoxicity experiments show that mutants with low affinity for N-type receptors are devoid of neurotoxic properties, even though some of them retain high enzymatic activity. These results provide further evidence for the involvement of N-type receptors in neurotoxic processes associated with venom sPLA2s and identify the surface region surrounding the hydrophobic channel of bee venom sPLA2 as the N-type receptor recognition domain.

Secretory phospholipases A 2 (sPLA 2 s, 1 14 kDa) catalyze the hydrolysis of the sn-2 position of glycerophospholipids to yield fatty acids and lysophospholipids (1)(2)(3). They are found in mammalian tissues and in the venoms of a wide range of organisms (insects, reptiles, amphibians, arachnids, coelenter-ates). Most sPLA 2 s require millimolar concentrations of Ca 2ϩ as a catalytic cofactor and show broad specificity for phospholipids with different polar head groups and fatty acyl chains (2,3). sPLA 2 s have been classified into three main groups according to their primary structure, including the location of their disulfide bridges (2,4,5). The mammalian pancreatic sPLA 2 , the mammalian non-pancreatic inflammatory sPLA 2 , and the sPLA 2 purified from honey bee venom (denoted bvPLA 2 ) are prototypes of group I, II, and III sPLA 2 s, respectively. The structures of type I and II enzymes are very similar, but the structure of bvPLA 2 is similar to those of the other sPLA 2 only in its active site (6). sPLA 2 s are implicated in a diversity of biological functions. Mammalian group II sPLA 2 is involved in inflammatory processes (7)(8)(9)(10)(11), while pancreatic group I sPLA 2 is involved not only in digestion but is also endowed with other cellular functions such as the induction of cell proliferation (12) and smooth muscle contraction (13,14). On the other hand, most sPLA 2 s from venom are toxic enzymes: they can act as neurotoxins, myotoxins, anticoagulants, and they can also cause inflammation (15)(16)(17)(18).
The existence of specific receptors for sPLA 2 s has been demonstrated recently (19,20). Using OS 1 and OS 2 , two sPLA 2 s purified from Taipan snake Oxyuranus scutellatus scutellatus venom, two types of receptors have been characterized. The M-type receptor is a 180-kDa protein that was first characterized in rabbit skeletal muscle (20,21). This receptor binds various venom sPLA 2 s, including OS 1 and OS 2 , but does not bind bvPLA 2 . It also binds pancreatic type and inflammatory type sPLA 2 s with high affinities (1-10 nM), suggesting that these sPLA 2 s are probably the endogenous ligands of this sPLA 2 receptor (21). The M-type receptor has been cloned recently from various animal species (21)(22)(23)(24), and its molecular properties have been analyzed in detail (23,(25)(26)(27)(28)(29). Notably, the interaction between the M-type receptor and sPLA 2 has been studied both from the receptor side (25,29) and from the sPLA 2 side to show that the Ca 2ϩ -binding loop of the sPLA 2 ligand plays a central role in binding to M-type receptor (28). Although the biological role of the M-type receptors remains to be clearly elucidated, this receptor has been proposed to be involved in many biological effects of the pancreatic sPLA 2 (30).
Another type of PLA 2 receptor has been identified in rat brain using OS 2 as a ligand and called the N (for neuronal)-type sPLA 2 receptor (19). N-type receptors are highly expressed in brain, but are also present in heart, skeletal muscle, kidney, lung, liver, pancreas, and smooth muscle (31). They have been shown to consist of proteins of 36 -51 and 85 kDa (19). N-type receptors have a pharmacological binding profile distinct from that of M-type receptors. They bind several neurotoxic sPLA 2 s, including OS 2 and bvPLA 2 , with high affinity, but unlike M-type receptors they do not bind pancreatic and inflammatory type sPLA 2 s (19,32). These N-type receptors are thought to mediate some of the physiological, pathophysiological, and toxic effects of sPLA 2 s. The normal biological role(s) of N-type receptors is presently unknown. Based on the fact that these latter receptors bind neurotoxic sPLA 2 s but not non neurotoxic sPLA 2 s with very high affinities, we have suggested that Ntype receptors are involved in the neurotoxicity of venom sPLA 2 s after intracerebroventricular injection (19,33).
The purpose of this study is to identify the region of sPLA 2 s that is responsible for both their interaction with N-type receptors and their neurotoxicity. These studies were carried out using a series of mutants of the neurotoxic bvPLA 2 , a very specific ligand of the N-type receptors (32)(33)(34) and one in which its tridimensional structure has been obtained at high resolution (35).
Construction of the Recombinant bvPLA 2 Mutants-All mutants were made using a PCR-based method. All primers used for mutagenesis are listed in Table I. Some of the primers contain a mixture of nucleotides so that several mutations at a single site were introduced in a single mutagenesis experiment. One-hundred l of PCR reactions consist of 1 ng of template p6xHis-Kall-BV-PLA 2 -#2 (36), 20 pmol of mutant primer, 20 pmol of T7 primer (Stratagene, La Jolla, CA), 0.2 mM of each deoxyribonucleotide triphosphate, 0.1 g of bovine serum albumin (New England Biolabs), 1-4 units of either Vent DNA polymerase (New England Biolabs or Boehringer Mannheim) or Taq DNA polymerase (New England Biolabs or Boehringer Mannheim) and the appro-priate polymerase buffer as supplied by the manufacturer. PCR reaction conditions were set at 96°C for 30 s, 45°C for 30 s (if no PCR product is detected, the annealing temperature was varied in the range 37-55°C), 2 min primer extension at 72°C, and this cycle was repeated 35 times. At the end of the 35th cycle, 10 min of extension at 72°C was used. PCR products were analyzed and purified by gel electrophoresis (1% agarose) to remove the unreacted primers and the template. The PCR product band was cut from the gel, and the DNA was recovered using the Geneclean kit. After restriction digestion, the PCR fragment was ligated into p6xHis-Kall-BV-PLA 2 -#1 (36) that had been digested previously with the appropriate restriction enzymes. The constructs were used to transform XL1-Blue (Stratagene). The insert in p6xHis-Kall-BV-PLA 2 -#1 lacks the BamHI site that is present in the PCR product (36), and restriction analysis with BamHI was used to identify the recombinant p6xHis-Kall-BV-PLA 2 -#1 that contains the PCR fragment. In many cases, the PCR mutagenesis primers also contain a second mutation that either adds or eliminates a restriction site to facilitate the identification of recombinant clones. After restriction mapping, plasmids from candidate clones were purified using the QIA-GEN plasmid mini-kit, and the DNA was sequenced using the DyeDeoxy terminator sequencing kit on a model 373A DNA Sequencer (Applied Biosystems, Foster City, CA) using the primers described previously (36). Finally, the plasmid was transferred into M15/pRep4 cell line (36) for protein expression.
Preparation of Recombinant bvPLA 2 Mutant Proteins and Analysis of Their Enzymatic Activities-For the production of recombinant bvPLA 2 mutants, cells were grown and induced as described (36). The fusion proteins were obtained by chromatography on Ni 2ϩ -nitrilotriacetic acidagarose as described (36) except that the sample before application to the column was clarified by centrifugation at 100,000 ϫ g for 40 min at 4°C instead of by filtration. The denatured fusion protein was refolded as described, and the protein was purified by chromatography on SP-Sephadex C-25 (Pharmacia Biotech Inc.) as described (36) except that the column was equilibrated and washed in 50 mM Tris, pH 7.5, and eluted with a linear gradient from 50 mM Tris, pH 7.5, to 1 M Tris, pH 9.0. The fractions containing protein were pooled and dialyzed against 5 mM Tris, pH 7.5. After dialysis, the solution was concentrated up to 4 mg of protein/ml in a Speed-Vac concentrator (Savant Instruments, Farmidale, NY). Mutant proteins were stored at Ϫ80°C. The protein concentrations of recombinant bvPLA 2 solutions were obtained from the absorbance at 280 nm (38) and by the Bradford assay (Bio-Rad) using commercial bvPLA 2 (Boehringer Mannheim) as a standard (both methods gave similar results). The enzymatic activities of the mutants were assayed using a spectrofluorimeter (Jasco 821-FP) with the fluorescent substrate 1-palmitoyl-2[10-(1-pyrene)decanoyl]-sn-glycerol-3-phosphomethanol lithium salt (39). Mutants in which a bvPLA 2 residue was  changed to cysteine require special comment. Analysis of these mutants with Ellman's reagent after refolding them in the presence of the cysteine/cystine redox couple buffer showed that in all cases these proteins lack free SH groups. This suggests that not only have all the 10 cysteines of wild-type bvPLA 2 formed disulfides, but the added cysteines is engaged in a disulfide bond with free cysteine from the buffer. This result was confirmed by analysis of of the mutants by mass spectrometry and by the fact that treatment of these mutants with DTT under conditions that do not cleave the disulfide of wild-type enzyme results in the appearance of 0.7-1.0 equivalents of SH/mol of enzyme. The full details of these analyses will be reported elsewhere.
Preparation of Deglycosylated bvPLA 2 -Natural purified bvPLA 2 (300 g) was incubated with or without 1 unit N-glycosidase F (Boehringer Mannheim, catalog number 913 782) in 100 mM sodium phosphate buffer (pH 7.4) containing 10 mM EDTA for 18 h at 37°C. After digestion, the mixtures were chromatographed on a C18 Waters column (4.6 ϫ 250 mm) prewarmed at 30°C with a Beckman system gold apparatus. Elution was performed using an acetonitrile linear gradient in 0.5% trifluoroacetic acid, 0.9% triethylamine, 0.001% ␤-mercaptoethanol, 27-40% acetonitrile for 35 min at 1.4 ml/min. The eluted fractions were lyophilized, resuspended in water, and assayed for protein concentration by measurement of the absorbance at 280 nm. 1.5 g of the samples were then analyzed by high resolution SDS/Tris/Tricine gels (16.5%) under reducing conditions followed by Coomassie Blue staining. Protein molecular mass markers were from Promega.
Binding Experiments-All N-type receptor binding experiments were performed as described (19). Briefly, membranes, 125 I-OS 2 , and competitors were incubated at 20°C in 0.5 or 1 ml of buffer (140 mM NaCl, 0.1 mM CaCl 2 , 20 mM Tris, pH 7.4, and 0.1% bovine serum albumin). Incubations were started by addition of rat brain membranes and filtered after 90 min of incubation through GF/C glass fiber filters presoaked in 0.5% polyethyleneimine.
Toxicity Experiments-Neurotoxic properties of the various recombinant sPLA 2 s have been determined by intracerebroventricular injections in adult male Balb/C mice (average weight, 20 g). Animals were anesthetized using a mixture of ketamine hydrochloride (100 mg/kg, Rhône Mérieux) and xylazine hydrochloride (12 mg/kg, Bayer Pharma). Injections of sPLA 2 s in a volume of 5 l of saline solution were made into the right ventricle according to stereotaxic coordinates (40). Lethality was checked 24 h after injection.

Binding of Recombinant bvPLA 2 to N-type Receptors-A
panel of mutants of recombinant bvPLA 2 have been used to identify the region of the enzyme that is involved in binding to N-type receptors. Binding properties of these mutants and of wild-type bvPLA 2 have been determined by competition binding experiments with 125 I-OS 2 , which is the canonical labeled ligand for N-type receptors (19).
As a first step toward the identification of the binding domain of sPLA 2 involved in the interaction with N-type receptors, we have analyzed the binding properties of recombinant bvPLA 2 expressed either as a fusion protein or after the removal of the N-terminal His-tag fusion peptide and compared their binding properties with those of the natural bvPLA 2 . As shown in Fig. 1 and Table II, the K 0.5 values obtained for the recombinant forms of bvPLA 2 (4.6 and 2.0 nM for recombinant cleaved bvPLA 2 and recombinant fusion protein bvPLA 2 , respectively) are close to the K 0.5 value measured for the natural purified bvPLA 2 protein (0.79 nM). These results establish that recombinant forms of bvPLA 2 expressed either as a fusion protein or as a free protein are good tools for the study of the molecular determinants of sPLA 2 that are implicated in binding to N-type receptors. Furthermore, as both uncleaved fusion protein and cleaved recombinant bvPLA 2 bind nearly equally well to N-type receptors, the different mutants have been used as recombinant proteins containing the N-terminal peptide extension. Fig. 1 indicates a 3-6-fold difference in affinity between recombinant and natural bvPLA 2 . Natural bvPLA 2 is heavily glycosylated on Asn-13 (41, 42), but the recombinant forms of bvPLA 2 have been produced in Escherichia coli where glycosylation does not occur. This suggests that differences in affinity between recombinant and natural bvPLA 2 could be due to the presence of the carbohydrate moiety in the native enzyme. To address this point, we have enzymatically deglycosylated the natural bvPLA 2 with N-glycosidase F and analyzed the binding properties of the purified deglycosylated enzyme. After treatment with N-glycosidase F, deglycosylated bvPLA 2 was purified by reverse phase high performance liquid chromatography ( Fig. 2A), and the resulting peaks were analyzed by SDS-Tricine gel electrophoresis (Fig. 2B). A shift in mobility of about 3 kDa was observed between the two main peaks that is consistent with the molecular mass of the carbohydrate moiety of bvPLA 2 made of 14 N-glycans (42). This suggests the major peak (peak B) effectively corresponds to deglycosylated bv-PLA 2 . Furthermore, the migration position of the faster moving protein is identical to that of the non-glycosylated fraction that is present in natural bee venom (36). The identification of peak B as deglycosylated bvPLA 2 was further confirmed by N-terminal protein sequencing, indicating that Asn-13 has been fully deglycosylated (not shown). Results from competition binding experiments with deglycosylated bvPLA 2 for binding to N-type receptors are shown in Fig. 2C. Interestingly, deglycosylated bvPLA 2 displays a K 0.5 value nearly identical to that of the cleaved recombinant bvPLA 2 (Fig. 1). This K 0.5 value is about five times higher than that of the natural glycosylated bvPLA 2 . Taken together, these results indicate that the carbohydrate moiety of bvPLA 2 , although not crucially involved in the interaction of bvPLA 2 with N-type receptors, increases the affinity by a factor of about 5.
Analysis of Mutations in the Putative Interfacial Recognition Surfaces of bvPLA 2 -The x-ray crystal structures of group I-III sPLA 2 s containing a bound phospholipid analog reveal a common constellation of catalytic residues lying about two-thirds down the length of the active site slot where the substrate binds (6). Despite these common features, the global topology of bvPLA 2 is very different than those of group I and II sPLA 2 s. It is currently thought that the surface of bvPLA 2 that contacts the membrane bilayer, the interfacial recognition surface (IRS), surrounds the opening of the active site slot (Figs. 3 and 4). This putative IRS of bvPLA 2 is composed of residues that lie on one face of an ␣-helix (composed of residues 76 -91), of residues in the N-terminal portion of the protein, and of a portion of the Ca 2ϩ -binding loop (35,43). K 0.5 values for the inhibition of 125 I-OS 2 binding to N-type receptors by bvPLA 2 proteins containing mutations in the IRS are listed in Table II 5. For these studies, a number of mutants were made in which the wild-type residue was replaced with cysteine. During in vitro refolding of mutants, cysteine and cystine are present in the buffer to provide a redox couple that favors formation of the five disulfide bridges that are present in natural bvPLA 2 . After refolding of all mutant proteins containing the desired additional cysteines, the cysteines end up in a disulfide linkage to a free cysteine from the redox buffer. This is useful for initial studies aimed at mapping the receptor recognition surface of bvPLA 2 , since the wild-type residue is replaced by a fairly large residue that is also zwitterionic at physiological pH and thus very hydrophilic.
Replacement of Ile-1 and Ile-2 by cysteine (I1CC and I2CC, respectively, where CC denotes the protein's cysteine disulfide linked to a free cysteine) results in a 20 -25-fold increase in K 0.5 value (i.e. 20 -25-fold decrease in receptor affinity) as compared with that of wild-type bvPLA 2 , suggesting that these two positions play a role in recognition of N-type receptors (Table II). Mutations of surface residues located within the 76 -91 helix also lead to reduced binding properties. Mutations of Lys-85 into cysteine (K85CC) or glutamic acid (K85E) reduce the affinity by a factor of about 40 (Table II). Binding properties of double and triple mutants involving K85E have also been analyzed. When Lys-85 is mutated together with either Lys-133 (K85E/K133E), Arg-23 and Lys-133 (R23E/K85E/K133E), or with Lys-94 and Lys-133 (K85E/K94E/K133E), a dramatic decrease in receptor binding was observed, especially for triple mutants (Table II). Mutations of hydrophobic surface residues of the 76 -91 helix also lead to a significant decrease in the binding affinity of bvPLA 2 to N-type receptors. Mutation of Ile-91 into phenylalanine (I91F) leads to a mutant with a K 0.5 of 98 nM. Furthermore, mutation of Ile-78 and Phe-82 to cysteine (I78CC and F82CC, respectively) dramatically decreases the affinity of bvPLA 2 (K 0.5 values of 490 and 280 nM, respectively) (Fig. 5A). Conversely, mutation of Ile-78 to tyrosine has no dramatic effect on the affinity of bvPLA 2 to N-type receptors (Table II), highlighting the importance of the hydrophobic nature of the residue at position 78 for receptor binding. Finally, mutations of Tyr-87 to phenylalanine (Y87F) and of Asp-92 to cysteine (D92CC) have only a minor effect on the binding properties of bvPLA 2 , probably because the side chains of these residues are pointing toward the center of the protein (Table II  and Fig. 3). Altogether, these data suggest that the N-terminal region and the 76 -91 helix of bvPLA 2 are involved in the interaction with N-type receptors.
Analysis of Mutations in the Ca 2ϩ -binding Loop-The Ca 2ϩbinding loops of sPLA 2 s anchor a single Ca 2ϩ ion in the active site where it functions to bind one of the oxygens of the substrate's phosphate group and may also serve as a Lewis acid to activate the carbonyl of the enzyme susceptible ester (6). The a All of the mutants were also assayed using the well established pH-stat method with vesicles of the anionic phospholipid 1,2-dimyristoylphosphatidylmethanol (36). The results are essentially the same, within 20%, as those obtained with the fluorimetric assay. In both assays, all of the mutants are tightly bound to vesicles, and thus the decrease in catalytic turnover is due to an increase in the interfacial K m , a decrease in the interfacial k cat , or a combination of both.

FIG. 2. Preparation and binding properties of deglycosylated bvPLA 2 .
A, elution profile of natural bvPLA 2 treated with N-glycopeptidase F on C18 reverse phase column. Retention time for bvPLA 2 treated without N-glycopeptidase F corresponds to peak A (not shown). B, gel analysis of untreated bvPLA 2 , N-glycopeptidase F/bvPLA 2 reaction mixture, and peaks A and B. C, comparison of binding properties to N-type receptors for natural glycosylated bvPLA 2 (E) and purified deglycosylated bvPLA 2 (q). Binding conditions are as indicated in the legend to Fig. 1.
Ca 2ϩ -binding loop of bvPLA 2 consists of residues 7-14, while in group I and group II sPLA 2 s, it comprises residues 26 -33 (35). Despite this very different localization in the primary sequence, the conformation of the bvPLA 2 Ca 2ϩ -binding loop is strikingly similar to that of group I and group II sPLA 2 s, and it contains the consensus motif X-Cys-Gly-X-Gly found in all catalytically active sPLA 2 s (35).
Mutations of Asn-13 and Ser-15 to cysteine (N13CC and S15CC, respectively) have a weak effect on the receptor binding properties of bvPLA 2 ( Fig. 5B and Table II). Mutation of Lys-14 to glutamic acid (K14E) or to cysteine (K14CC) also have a modest effect on receptor binding, although the double mutant K14E/R23E binds 580-fold less tightly to N-type receptor. Mutations at position 12 have interesting consequences. Replacement of Gly-12 with the hydrophilic residues cysteine (G12CC) or serine (together with a conservative mutation of Lys-14 into arginine, which was inadvertently introduced during PCR) (G12S/K14R) weakens receptor binding only slightly. However, mutation of Gly-12 to the hydrophobic residue alanine (G12A) results in a protein that does not detectably bind to N-type receptors, suggesting that the N-type receptor donates a residue that forms a specific contact with residue 12 (Table II). The weak effects observed with the Ser-15 mutant is in accordance with the fact that its side chain is pointing toward the center of the protein and thus is not likely to contact the N-type receptor. Taken together, these results indicate that the Ca 2ϩ -binding loop is also involved in the interaction with N-type receptors.
Analysis of Mutations in the Active Site Slot-Mutations were also made on residues located in the active site slot region of bvPLA 2 (Fig. 3). Among these residues, Thr-57 forms a portion of the wall of the active site slot and is known to hydrogen bond directly with the sn-3 phosphate of bound phospholipid substrate (43). This residue appears very important for binding of bvPLA 2 to N-type receptors, since mutation to phenylalanine (T57F), valine (T57V), leucine (T57L), and aspartic or glutamic acid (T57D and T57E, respectively) strongly decreases or abolishes the binding of bvPLA 2 to the receptor ( Fig. 5C and Table II).
We have also analyzed the effects of mutation of Asp-64, which is directly involved in the catalytic mechanism (35,44).
This residue corresponds to the active site residue Asp-99 of group I and group II sPLA 2 s, which is conserved in all catalytically active enzymes including both neurotoxic and nontoxic sPLA 2 s (5). Mutation of Asp-64 to alanine (D64A) results in a dramatic decrease in receptor affinity (K 0.5 Ͼ 1 M). However, more conservative substitutions of Asp-64 by glutamic acid (D64E) or by asparagine (D64N) do not lead to a dramatic decrease in receptor affinity (Fig. 5C). Finally, a double mutant where Asp-64 is replaced by a glutamic acid (D64E) together with a mutation at position 66 (D64E/K66CC, D64E was introduced inadvertently during the preparation of K66CC by PCR) also retains a high affinity for N-type receptors (Table II).
Analysis of Deletions and Other Mutations-bvPLA 2 is a basic enzyme (pI ϭ 8.04) that has a number of lysine and arginine residues on its surface that may interact with N-type receptors. This protein also contains several solvent-exposed threonine and serine residues that might also have a role in receptor binding. Mutation of lysines 66 and 72, which are located on the side of bvPLA 2 opposite of the IRS, to either cysteine (K66CC and K72CC) or glutamic acid (K66E) has no effect on the K 0.5 (Table II), suggesting that this region of the molecule is not involved in binding to N-type receptors. Similarly, mutation of Lys-122 and Lys-133 to glutamine (K122Q) or glutamic acid (K133E) produces only modest effects on receptor binding (Table II). Conversely, mutation of Lys-94 to glutamic acid leads to a moderate but significative decrease in the binding affinity of the mutant bvPLA 2 relative to the wildtype enzyme (Table II). Similarly, mutation of Arg-58 into alanine (R58A) also results in a decrease of the K 0.5 value, the corresponding K 0.5 value is more than 35 times higher than that of the wild-type enzyme. In addition, a synergistic effect was observed with the double mutant R23E/R58A in that this protein did not detectably bind to N-type receptors (Table II). This result is not surprising since Lys-94 and Arg-58 are located in the vicinity of the 76 -91 helix, and residues of this helix have already been shown to be involved in receptor binding (see above). Threonines 51 and 53 are located on the surface of bvPLA 2 , but far away from the IRS (Fig. 3). As expected, mutation of these residues into cysteine (T51CC and T53CC) fail to produce drastic effects on receptor binding (Table II). The view is such that the methyl groups at the ends of the alkyl chains of the active site phospholipid analog inhibitor (unfilled bonds) are pointing toward the reader. The front face is probably the IRS. The position of those mutants that increase K 0.5 for binding to N-type receptors by at least 100-fold are shown as large red spheres (12,57,64,78,82), those that increase K 0.5 by less than 10-fold are shown as small blue spheres (13,15,25,51,53,66,72,87,92,122,133), and those that increase K 0.5 by an intermediate amount are shown as medium yellow spheres (1,2,14,23,24,58,85,91,94). The image was created with the MolScript program (56).
Basic residues are also found on the surface of the bvPLA 2 , in a loop located between the Ca 2ϩ binding site and the ␣-helix containing the active site residue His-34 (Fig. 3). Relative to the IRS, this loop lies toward the center of the globular protein, at the left edge of the structure, and thus the amino acid side chains of this loop are relatively far from the Ca 2ϩ loop and from the IRS. When Lys-25 of this loop was mutated to glutamine, the corresponding mutant (K25Q) retains binding properties (K 0.5 ϭ 2.7 nM) very similar to that of wild-type recombinant bvPLA 2 , indicating that this position is not involved in binding to N-type receptors (Fig. 5D). Mutations of the neighboring Arg-23 and Phe-24, which are located closer to the IRS than Lys-23 (Fig. 3), into either glutamic acid (R23E and F24E, respectively) or cysteine (R23CC and F24CC) result in a significative decrease in receptor binding ( Fig. 5D and Table II), consistent with the results obtained with the double mutant K14E/R23E (see above).
We have also analyzed the role of the ␤-sheet-like structure composed of residues 99 -118 of bvPLA 2 (Fig. 3) in receptor binding. Interestingly, a similar structure is also found in group I and group II sPLA 2 s (2, 5), but no role has been proposed for this domain in supporting the enzymatic activity or toxicity of sPLA 2 s. When this ␤-sheet domain is deleted (⌬99 -118), the mutant bvPLA 2 binds to N-type receptors with an affinity (K 0.5 ϭ 7.6 nM) that is similar to that of wild-type protein. This result clearly indicates that the ␤-sheet-like structure does not interact with N-type receptors.
Is bvPLA 2 Enzymatic Activity Required for Binding to N-type Receptors?-At first glance, Table II shows that most of the mutants having weak receptor binding properties also have low enzymatic activity. However, a detailed comparison of the data indicates that receptor binding and enzymatic activity are two independent molecular events. Indeed, a clear dissociation between binding and catalytic activity can be found with various mutants of the Ca 2ϩ -binding loop and the IRS. For example, G12A has no measurable affinity for N-type receptors, whereas it retained 37% enzymatic activity as compared with wild-type enzyme. K85E/K133E binds to N-type receptors 180-fold weaker than wild-type enzyme, but this mutant retains 69% enzymatic activity. K14CC and K85CC have reduced binding properties with K 0.5 values of 42 and 69 nM, respectively, while they still have high enzymatic activity (29 and 42% relative to wild-type, respectively). The F82CC mutant also shows a lack of correlation in receptor binding and enzymatic activity, but to a lesser extent (280 nM and 7% enzymatic activity). These results indicate that binding to N-type receptors and catalytic activity are independent events, although several residues of the Ca 2ϩbinding loop and of the IRS are required for both activities.
Relationships between Neurotoxicity, Binding to N-type Receptors, and Enzymatic Activity-The neurotoxicity of venom sPLA 2 s, including bvPLA 2 , has been previously suggested to be associated with their binding to N-type receptors (19,33). This is based on the observation that the neurotoxicity of a collection of sPLA 2 s correlates reasonably well with their affinity for N-type receptors. The conclusions from these previous studies would be strengthened by carrying out neurotoxicity studies with point mutant bvPLA 2 s that retain high enzymatic activity but fail to bind to N-type receptors and those proteins with the vice versa properties. Studies along these lines are summarized in Table III. For these studies, we have chosen bvPLA 2 mutants that show a clear dissociation between enzymatic activity and receptor affinity. The data in Table III indicate that recombinant wild-type bvPLA 2 expressed in bacteria as a fusion protein is lethal to mice at very low doses of 25-125 nmol/kg when injected by the intracerebroventricular route. The behavioral symptoms occurring after injection were periods of catalepsy while the mice showed breathing problems, and these events were separated by periods of violent convulsions lasting from 10 to 60 s. These symptoms are very similar to those observed after injection of natural bvPLA 2 into rat brain (33). These symptoms were also observed with all of the mutants showing neurotoxicity. K14CC and K85CC are mutants that retain a high degree of bvPLA 2 enzymatic activity (within 4-fold of wild-type enzyme), whereas their K 0.5 values for binding to N-type receptors are increased by a factor of 20 -30 relative to that of the wild-type protein. The data in Table III show that these mutants are still lethal, although the effective doses needed to kill mice are much higher. Indeed, 125 nmol/kg of K14CC killed only three out of six mice, while identical or lower quantities of wild-type bvPLA 2 killed all or almost all of FIG. 4. Molecular surface of bv-PLA 2 . The view is the same as in Fig. 3 and was created with the program GRASP. The positions of mutants that affect the binding to N-type receptors by a large amount, an intermediate amount, and a small amount are shown in red, yellow, and blue, respectively. The inhibitor is shown with blue sticks. the injected mice (Table III). Similarly, 250 nmol/kg of K85CC is needed to kill four out of nine injected mice, while injections of a 5-fold lower amount of wild-type enzyme killed five out of six mice. The neurotoxic properties of mutants with very weak or no measurable affinity for N-type receptors were also analyzed. Importantly, some of these mutants still display a high enzymatic activity relative to wild-type bvPLA 2 (Table III). None of these mutants were lethal in mice, even at the highest doses that are possible to inject (125 nmol/kg for K85E/K133E and 250 nmol/kg for G12A and F82CC). Finally, as a control, we checked for the lethal properties of a double mutant (D64E/ K66CC) that has both high enzymatic activity and high affinity for N-type receptors (Table III). This mutant displays a toxicity very similar to that of the wild-type recombinant bvPLA 2 , suggesting that this region of the molecule is not required to invoke lethality of bvPLA 2 in mice. Altogether, these results demonstrate that the neurotoxicity of bvPLA 2 is related to its ability to bind to N-type receptors and not to its ability to hydrolyze phospholipids. DISCUSSION The aim of this paper was 2-fold: to determine the region(s) of bvPLA 2 involved in binding to N-type receptors and to determine if binding of bvPLA 2 to N-type receptors and/or the enzymatic activity of this protein are required for the neurotoxic effects of bvPLA 2 . As a first approach to identify the region(s) of sPLA 2 involved in binding to N-type receptors, we performed sequence comparisons between sPLA 2 s that bind or do not bind to these receptors. Two different sets of sequence comparisons were made. In the first set, we aligned a large number of group I and group II sPLA 2 s, including OS 1 and OS 2 , which have a relatively high level of sequence identity but have distinct receptor binding properties (28). A second set of sequence alignments was made with group III sPLA 2 s, since bvPLA 2 is a typical member of this group and since it also binds with very high affinity and specificity to N-type receptors. Other members of group III sPLA 2 s have been purified from the venom of the Gila monster lizard H. suspectum (45,46) and of the Mexican beaded lizard Heloderma horridum horridum (47) as well as from the mediterranean medusa Rhopilema nomadica (48). Of interest for our study are the two sPLA 2 s, namely Pa2 and Pa5, which are major components of Gila monster (H. suspectum) venom and which have been entirely sequenced (46). These two sPLA 2 s display an overall identity of 38% with the bvPLA 2 , the level of identity increasing to 58% when the Ca 2ϩ -binding loop and the active site region were compared. Competition experiments between labeled OS 2 and Pa2 and Pa5 for binding to N-type receptors have indicated that these H. suspectum sPLA 2 s do not bind to N-type receptors (not shown), suggesting that the molecular determinants for binding to these receptors are located within residues that are distinct in the structure of sPLA 2 s from Heloderma and bee venoms. Despite a detailed analysis of all these sequence alignments, it was essentially impossible to predict with any accuracy the region(s) of sPLA 2 s that is most probably involved in the interaction with N-type receptors. This led us to the present studies using site-specific mutants of bvPLA 2 ; because this protein is conveniently expressed in functional form in bacteria (36), its high resolution x-ray structure is known (35), and it binds with high affinity to N-type receptors (19). It may be noted that when the sequence of Pa5 is overlaid onto the structure of bvPLA 2 , it is apparent that the putative IRS of the Gila monster enzyme lacks most of the basic residues found on the IRS of bvPLA 2 . This may be the reason that Pa5 fails to bind to N-type receptors, since results of the present study indicate that basic residues on the IRS of bvPLA 2 are important for receptor binding.
A particular feature of the bvPLA 2 is the presence of a high-mannose carbohydrate motif N-linked to Asn-13 (42). A possible role of this motif in binding to N-type receptors was suggested by comparison of the binding data obtained with recombinant and natural glycosylated bvPLA 2 . The five times lower affinity observed with fully deglycosylated bvPLA 2 indicates that the carbohydrate moiety contributes to N-type receptor interaction, but only to a minor extent. Therefore, elements of recognition of N-type receptors are clearly within the bvPLA 2 protein sequence.
As shown in Figs. 3 and 4, mutations that dramatically reduce the affinity of bvPLA 2 to N-type receptors are segregated on the same face of the protein, namely on the 76 -91 helix (especially positions 78 and 82), the Ca 2ϩ -binding loop, and regions of the N terminus of the protein. These are the same regions that make up the putative IRS that is thought to anchor the enzyme to the membrane interface (35). It is very likely that specific residues of the IRS directly contact the N-type receptor rather than the membrane to which the receptor is embedded. This is based on the observation that several point mutations such as G12A which retain good enzymatic activity in a membrane based assay, and thus bind well to membranes, fail to bind to N-type receptors. The dramatic effect observed on binding to N-type receptors after the single mutation of Gly-12 to Ala-12 suggests that position 12 is crucially involved in the interaction with N-type receptors. Interestingly, Gly-30 of porcine pancreatic sPLA 2 , which is analogous to Gly-12 of bvPLA 2 , is involved in the binding to M-type receptors (28).
The observation that residues of the Ca 2ϩ -binding loop that point away from the protein are important for receptor binding has to be considered along with the previous observation that the binding of OS 2 to N-type receptors requires micromolar amounts of Ca 2ϩ (19). It is not yet clear whether this Ca 2ϩ dependence is due to the binding of Ca 2ϩ to the receptor, to sPLA 2 , or to both. All of the Ca 2ϩ -binding loop mutants examined in the present study retain high enzymatic activity, which implies that they all bind Ca 2ϩ . Thus, the failure of mutants such as G12A to bind to N-type receptors is not due to a lack of binding of Ca 2ϩ to the bvPLA 2 mutant.
In general, mutations in other regions of bvPLA 2 besides the IRS do not drastically affect the binding to N-type receptors. The exceptions are mutations at positions 57 and 64, located in the active site slot near the center of the globular protein and amino acids substitution at these positions dramatically reduce receptor binding. One interpretation of these surprising results is that a portion of the N-type receptor penetrates the active site slot and directly contacts Thr-57 and Asp-64. However, it is also possible, and perhaps more likely, that positions 57 and 64 mutants have structural alterations that propagate out to the surface of the protein. Key observations are obtained with the mutants D64N and D64A. Asp-64 of wild-type bvPLA 2 is hydrogen-bonded to the NH of His-34 that faces away from the active site slot. When this residue is replaced with the iso-structural Asn, the D64N mutant retains good receptor binding affinity. The same result is obtained with D64E/K66CC. However, the D64A mutant fails to bind to N-type receptors almost certainly, because loss of the Asp-64-His-34 hydrogen bond results in nonlocalized structural changes that affect receptor binding.
Results obtained with the deletion mutant ⌬99 -118 are interesting in light of the fact that removal of such a large structural domain of bvPLA 2 has essentially no effect on the enzymatic activity or the receptor binding affinity of the mutant. Furthermore, this ␤-sheet structure, which lies adjacent to the IRS, is found in group I, II, and III sPLA 2 s, suggesting that it has been conserved since the divergence between vertebrates and insects (5). The biological role, if any, of this domain remains to be established.
Since bvPLA 2 is neurotoxic, has high enzymatic activity, and also binds with high affinity to N-type receptors that have been suggested previously to be involved in neurotoxicity of venom sPLA 2 s (19), it was important to probe the relationships between the receptor binding properties of bvPLA 2 mutants and their neurotoxicity (as measured by the evaluation of their lethality in mice). Table III indicates that the residues of bv-PLA 2 that are involved in its lethal properties are located on the surface of the enzyme, namely in the 76 -91 helix and in the Ca 2ϩ -binding loop. Mutations that reduce binding to N-type receptors also reduce the lethal properties, even when the mutation does not significantly decrease the lipolytic enzymatic activity of the mutant. This result indicates that the neurotoxicity of bvPLA 2 is closely related to its affinity for N-type receptors rather than to its enzymatic activity.
Previous structure-function studies using toxic group I and group II sPLA 2 s have suggested the involvement of two distinct regions of sPLA 2 sequences in conferring neurotoxicity to these enzymes. Comparisons of primary structures of a large number of sPLA 2 s have suggested that the N-terminal region of sPLA 2 is involved in neurotoxicity (49). This notion is consistent with the result of the present study. In accordance with this hypothesis, chemical modifications of residues located in the N-terminal region of notexin led to a decrease in its lethal properties (50,51). However, other studies also based on sequence comparisons have suggested that the hydrophobic region located at the C-terminal end (residues 80 -110) of neurotoxic sPLA 2 s is involved in neurotoxicity (52). This second hypothesis is supported by results of chemical modifications of notexin on tryptophan residues (53) and by immunological approaches using antipeptide antibodies directed against the C-terminal part of ammodytoxin A and crotoxin (54,55).
In conclusion, this work has shown that residues of the IRS of bvPLA 2 (Ca 2ϩ -binding loop and 76 -91 helix) are implicated both in binding to N-type receptors and in neurotoxicity, strongly suggesting that toxicity is related to binding to N-type receptors. Whether other venom sPLA 2 s belonging to group I/II also use similar residues to bind to N-type receptors and to exert neurotoxicity remains to be elucidated.
Acknowledgments-We are very grateful to Drs. Thomas Dudler and Robert Annand for constructing some of the mutants used in this study