Plasma platelet-activating factor acetylhydrolase is a secreted phospholipase A2 with a catalytic triad.

Platelet-activating factor (PAF) is a potent pro-inflammatory autacoid with diverse physiological and pathological actions. These actions are modulated by PAF acetylhydrolase, which hydrolyzes the sn-2 ester bond to yield the biologically inactive lyso-PAF. In contrast to most secreted phospholipase A2s, plasma PAF acetylhydrolase is calcium-independent and contains a GXSXG motif that is characteristic of the neutral lipases and serine esterases. In this study we tested whether the serine in this motif is part of the active site of plasma PAF acetylhydrolase and, if so, what the other components of the active site are. Using site-directed mutagenesis, we demonstrated that Ser-273 (of the GXSXG motif), Asp-296, and His-351 are essential for catalysis. These residues were conserved in PAF acetylhydrolase sequences isolated from bovine, dog, mouse, and chicken. The linear orientation and spacing of these catalytic residues are consistent with the α/β hydrolase conformation of other lipases and esterases. In support of this model, analysis of systematic truncations of PAF acetylhydrolase revealed that deletions beyond 54 amino acids from the NH2 terminus and 21 from the COOH terminus resulted in a loss of enzyme activity. These observations demonstrate that although plasma PAF acetylhydrolase is a phospholipase A2 it has structural properties characteristic of the neutral lipases and esterases.

Platelet activating factor (PAF) 1 is a potent pro-inflammatory phospholipid produced by activated platelets, leukocytes, and endothelial cells (1). Its production is tightly regulated, both at the synthetic and degradative levels. PAF is synthesized in activated cells primarily through the remodeling pathway that invokes at least two enzymes, an arachidonatespecific phospholipase A 2 (PLA 2 ) and acetyl-CoA:lyso-PAF acetyltransferase, both of which are activated by phosphorylation (2)(3)(4)(5)(6). PAF is inactivated when the sn-2 ester bond is hydrolyzed by PAF acetylhydrolase(s), of which there are both intracellular and extracellular forms (7)(8)(9)(10). The cytoplasmic enzyme may help regulate PAF production (e.g. in differentiated macrophages (11)), whereas the plasma form is believed to regulate baseline circulating PAF levels and may be critical in resolving inflammation (12).
We recently cloned the plasma PAF acetylhydrolase cDNA (13), and it encodes a 441-amino acid protein including a predicted signal peptide. Recombinant enzyme exhibited a marked preference for phospholipids with short chain moieties at the sn-2 position and blocked PAF-induced inflammation. The deduced amino acid sequence is unique except for the Gly-Xaa-Ser-Xaa-Gly (GXSXG) motif found in most serine esterases and lipases (14 -16). The central Ser of this motif serves as the active site nucleophile of these enzymes (reviewed in Ref. 17). The GXSXG motif in PAF acetylhydrolase is consistent with observations that the activity can be blocked by the active site Ser-specific inhibitor, diisopropyl fluorophosphate (7,13).
PAF acetylhydrolase is a PLA 2 because it hydrolyzes phospholipids at the sn-2 position. More than 60 secretory PLA 2 s and at least two classes of intracellular PLA 2 s have been described (18,19). The secretory PLA 2 s are small (119 -143 amino acids), have a high disulfide bond content (10 or 14 cysteines), and require calcium for catalysis. PAF acetylhydrolase is larger than these PLA 2 s, has fewer cysteines (five), and functions in the absence of calcium. The intracellular PLA 2 s are larger than the secreted ones and do not utilize calcium as part of the catalytic site (20 -23). However, the intracellular localization and the preference of these enzymes for phospholipids with arachidonate or other long chain fatty acids at the sn-2 position distinguish them from the plasma PAF acetylhydrolase. Thus, the unique functional properties of plasma PAF acetylhydrolase likely reflect a molecular structure that is distinct from that of other known PLA 2 s.
In contrast, the neutral lipases share the calcium-independence of PAF acetylhydrolase and contain the conserved GXSXG motif. The three-dimensional structure of several of these lipases from species as divergent as mammals, fungi, and bacteria have been defined and found to have a core ␣/␤ hydrolase fold structure that is characteristic of a variety of esterases and hydrolases (17,24). The ␣/␤ hydrolase structure consists of alternating ␣-helices and ␤-strands; the ␤-strands are arranged in a parallel fashion and form the central core of the molecule (17,24). The active site contains a catalytic triad comprised of the nucleophilic Ser (GXSXG), an acidic residue that is usually Asp (Glu in Candida rugosa and Geotrichum candidum lipases (25-28)), and an invariant His. In most lipases, the triad lies within a hydrophobic pocket that is exposed upon contact with the substrate (interfacial activation) which results in an enhanced reaction rate. Analytically, the property of interfacial activation is manifested in surface dilu-* This work was supported by ICOS Corporation, American Heart Association Award 92023050, and National Institutes of Health Grants HL35828 and HL50153. 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.
The nucleotide sequences of the cDNAs encoding the protein sequences reported in this paper have been submitted to the GenBank TM /EMBL Data Bank with accession numbers U34247 (bovine), U34246 (dog), U34277 (mouse), and U34278 (chicken).
Clearly, PAF acetylhydrolase exhibits characteristics of both the PLA 2 s and the neutral lipases. However, the fact that PAF acetylhydrolase contains the characteristic serine esterase GXSXG motif and is inhibited by the active site serine modifier, diisopropyl fluorophosphate, suggests that mechanistically, the enzyme is more closely related to the neutral lipases and other esterases than it is to the classical PLA 2 s. To better understand the structural basis for the functional characteristics of PAF acetylhydrolase, we used site-directed mutagenesis and deletion analyses to identify potential catalytic residues and the minimal active portion of the molecule.

EXPERIMENTAL PROCEDURES
Cloning of PAF Acetylhydrolase from Multiple Species-PAF acetylhydrolase cDNAs were obtained from mouse, dog, bovine (all in ZAPII, Stratagene), and chicken (gt10, Clontech) spleen cDNA libraries by low stringency hybridization to the human cDNA (13). Approximately 5 ϫ 10 5 to 1 ϫ 10 6 phage were blotted onto nitrocellulose and screened in 20% formamide, 0.75 M sodium chloride, 75 mM sodium citrate, 50 mM sodium phosphate (pH 6.5), 1% polyvinyl pyrrolidine, 1% Ficoll, 1% bovine serum albumin, and 50 ng/ml sonicated salmon sperm DNA. The hybridization probe was a 1-kilobase HindIII fragment from the human PAF acetylhydrolase cDNA labeled with 32 P by hexamer random priming. After overnight hybridization at 42°C, blots were washed extensively in 30 mM sodium chloride, 3 mM sodium citrate, 0.1% SDS at 42°C. Following a secondary screen under identical conditions, individual hybridizing plaques were selected for DNA purification and the nucleotide sequence of both strands of the inserts was determined by the automated dideoxy chain termination method.
Generation of PAF Acetylhydrolase Mutants-Site-directed mutants were generated using the polymerase chain reaction (PCR) (29,30). The most 5Ј sense primer (Start 1) included an XbaI cloning site and a translation initiation codon in-frame with the codon encoding Ile-42, the NH 2 -terminal amino acid of the mature PAF acetylhydrolase initially purified from human plasma (13). At the 3Ј end of the cDNA, the initial primer used (Stop 2) included the native translation termination codon followed by an EcoRV cloning site. This primer was used to generate the wild-type control construct as well as the mutants S108A, S273A, and H395A,H399A. All other site-directed mutants were generated using a second 3Ј primer (Stop 4) that contained 2 translation termination codons followed by a KpnI restriction site for improved cloning efficiency. Internal PCR primers specific for the residues to be mutated are listed below. All of the site-directed mutants were cloned into pUC19 downstream of the trp promoter for expression in Escherichia coli as described previously (13).
Deletion constructs were generated using exonuclease III (Promega) to create a series of nested deletions from both the 5Ј and 3Ј ends. The parent constructs were engineered by PCR using primers with specific restriction sites compatible with the exonuclease III system. For 5Ј deletions, a sense primer was designed to allow cloning the product in-frame with the E. coli thioredoxin gene (31) under transcriptional control of the tac promoter (32) in pUC19. This primer (Start 3) contained XbaI, SphI, and XhoI restriction sites immediately upstream of the codon encoding Ile-42. This primer was paired with the Stop 3 antisense 3Ј primer to generate the parent construct. For 3Ј deletions, a 3Ј antisense primer (Stop 3) containing translation termination codons in all 3 reading frames preceded by XbaI, SacI, and EcoRV restriction sites was paired with Start 6 to generate a parent construct which was cloned into the same pUC19-trp vector used for the sitedirected mutants.
Nucleotide sequence of the PCR primers follows (all sequences are written in the 5Ј-3Ј direction; parenthetical S and A indicate sense or E. coli Expression and Analysis-Competent E. coli (strain XL-1 Blue) were transformed with the expression constructs, plated on L broth agar containing 50 mg/ml carbenicillin, and incubated overnight at 37°C. The following day, transformed colonies were transferred to 3 ml of L broth containing carbenicillin and incubated for 16 -18 h at 37°C. During incubation, tryptophan in the medium becomes depleted and the trp promoter is activated. Under these conditions, cells typically reach an absorbance of 1-2 at 600 nm wavelength (A 600 ). Cells from 500 l of culture were pelleted by centrifugation at 12,000 rpm for 2 min. The pellet was resuspended in 100 l of lysis buffer containing 25 mM Tris (pH 7.5), 50 mM NaCl, 10 mM CHAPS (Sigma), 1 mM EDTA, and 0.1 mg/ml lysozyme and incubated on ice for 1 h. The lysate was sonicated and PAF acetylhydrolase expression levels were assessed by enzyme activity assay, Western blotting, and enzyme-linked immunosorbant assay (ELISA).
Determination of PAF Acetylhydrolase Activity-E. coli lysates were diluted in 25 mM Tris (pH 7.5), 10 mM CHAPS, 500 mM NaCl, and 1 mM EDTA prior to assay. The hydrolysis of PAF was assayed as described by Ref. 33. Specific activity was measured in units/mg, where 1 unit ϭ 1 mol of PAF hydrolyzed per ml/h; the lower limit of activity detection was 10 nmol/ml/h. Highly purified recombinant PAF acetylhydrolase has a specific activity of 5000 units/mg. 2 The quantity of each recombinant derivative was determined by electrophoresis of E. coli lysates, Western blotting (as described below), and scanning densitometry of antibody reactivity.
Generation of Antibodies-Polyclonal antisera were raised in rabbits by 3 monthly immunizations with 100 g of purified recombinant human PAF acetylhydrolase in Freund's adjuvant. Monoclonal PAF acetylhydrolase-specific antibodies were generated in BALB/c mice (Charles River). The mice received an initial injection of 50 g of recombinant PAF acetylhydrolase in complete Freund's adjuvant followed by two 50-g boosts in incomplete Freund's adjuvant at 3-week intervals. Four days prior to splenocyte harvest, mice received a final boost intraperitoneally of 50 g of enzyme in phosphate-buffered saline. Splenocytes were harvested and fused to NS-1 myeloma cells using standard procedures. Eight days after fusion, hybridomas were screened for production of PAF acetylhydrolase-specific antibodies by ELISA. Immulon 4 (Dynatech) 96-well microtiter plates were coated with 100 ng/well of recombinant PAF acetylhydrolase. Each coated well received 50 l of hybridoma culture supernatant and PAF acetylhydrolase-binding antibodies were detected with horseradish peroxidaseconjugated goat anti-mouse IgG (Jackson ImmunoResearch, West Grove, PA). Clonal hybridomas were obtained by limiting dilution.
Western Analysis-To standardize the quantity of E. coli lysate loaded on each lane of the gel for Western blotting, the number of cells in each induced culture was quantitated by determining the optical density at 600 nm. The number of cells were then equalized prior to lysis. Equivalent quantities of each lysate were electrophoresed and the equivalent loading of each sample was confirmed by Coomassie staining. E. coli lysates were then diluted (usually 1:100) in distilled water, mixed with an equal volume of sample buffer containing 125 mM Tris (pH 6.8), 4% SDS, 100 mM dithiothreitol, and 0.05% bromphenol blue, and boiled for 5 min prior to loading onto a 12% SDS-polyacrylamide gel (Novex). Following electrophoresis at 40 mA, proteins were electrotransferred onto a polyvinylidene fluoride membrane (Pierce) for 1 h at 125 V in 192 mM glycine, 25 mM Tris base, 20% methanol, and 0.01% SDS. The membrane was incubated in 20 mM Tris, 100 mM NaCl (TBS) containing 5% bovine serum albumin (Sigma) overnight at 4°C. The blot was incubated 1 h at room temperature with rabbit polyclonal antisera diluted 1/8000 in TBS containing 5% bovine serum albumin, and then washed with TBS and incubated with alkaline phosphataseconjugated goat anti-mouse IgG in TBS containing 5% bovine serum albumin for 1 h at room temperature. The blot was again washed with TBS then incubated with 0.02% 5-bromo-4-chloro-3-indolyl phosphate and 0.03% nitro blue tetrazolium in 100 mM Tris-HCl (pH 9.5), 100 mM NaCl, and 5 mM MgCl 2 . The reaction was stopped with repeated water rinses. By these analyses, all lysates presented a single band of the expected size except for a few of the point mutants which displayed a doublet of bands. This may be due to variable susceptibility of the mutant proteins to proteolytic degradation of 10 -12 residues from the COOH terminus, a condition which has no effect on specific activity (data not shown).
ELISA-Immulon 4 plates were coated with 100 ng/well of an anti-PAF acetylhydrolase monoclonal antibody, 90G11D, overnight at 4°C. Prior to addition of sample, wells were blocked for 1 h at room temperature with 0.5% fish skin gelatin (Sigma) then washed with phosphatebuffered saline containing 0.05% Tween 20 (PBST). E. coli lysates diluted in phosphate-buffered saline containing 15 mM CHAPS were added to the coated wells and incubated 1 h at room temperature. Following four washes with PBST, 250 ng of a biotinylated second, non-competitive monoclonal antibody, 90F2D, was added to each well and incubated for 1 h at room temperature. The wells were then washed three times with PBST and treated with 100 ng of peroxidase-conjugated ExtraAvidin (Sigma) in PBST for 1 h at room temperature. PAF acetylhydrolase-specific binding of the antibodies was detected by the addition to each well of 1 mg/ml o-phenylenediamine (Sigma) in 100 mM citrate buffer (pH 4.5) containing 0.3% H 2 O 2 .

RESULTS
Cloning of PAF Acetylhydrolase cDNA from Multiple Species-A valuable test of the importance of any specific amino acid sequence within a protein is its conservation within homologs from other species. Therefore we used low stringency hybridization with a human probe to obtain full-length plasma PAF acetylhydrolase clones from mouse, bovine, dog, and chicken (Fig. 1). Multiple cDNA clones were isolated from each species; our results are consistent with a single gene encoding PAF acetylhydrolase in all five species. As shown in Table I, the more evolutionarily related species have more similar PAF acetylhydrolase sequences. About 38% of the residues are completely conserved in all five sequences, suggesting that many of these amino acids are important for structure and function. The most divergent regions are at the amino-terminal end (containing the signal sequence) and the carboxyl-terminal end; as shown below, these regions are not critical for activity. Importantly, the GXSXG motif is conserved in all of the species.
Identification of Potential Catalytic Triad Residues-Our general approach was to use site-directed mutagenesis of candidate amino acids to test whether enzymatic activity was altered. In addition to enzyme activity assays, we measured the level of expression by Western blotting of lysates from transformed E. coli using a polyclonal antibody. Each mutant protein was also assayed in a sandwich ELISA using conformation-specific monoclonal antibodies to assess the consequences of each mutation on the gross structure of the enzyme. To identify candidate residues of a catalytic triad in human PAF acetylhydrolase, we considered three separate parameters: (i) sequence similarities with other lipases or serine esterases, (ii) location of the residues relative to each other in the linear sequence, and (iii) conservation of potential catalytic residues across multiple species as described above. As reported (13), the only significant homology found between PAF acetylhydrolase and other sequences spans the GXSXG motif (Ser-273), which contains the active site serine in other lipases and serine esterases. Thus, Ser-273 is a likely candidate for the active site nucleophile in plasma PAF acetylhydrolase. To test this, Ser-273 was mutated to Ala (S273A). This mutation reduced the  enzyme activity to below the level of detection of the assay. In this assay, the lower limit of detection is 10 nmol/ml/h; thus the S273A mutation reduced the activity by more than 4 orders of magnitude (Fig. 2). In contrast, mutation of another conserved Ser (S108A) had no deleterious effect on enzyme activity. Both recombinant proteins were readily detected in a sandwich ELISA using two distinct monoclonal antibodies specific for PAF acetylhydrolase (Table II). This suggests that the mutations did not cause gross conformational changes in the enzyme and supports the notion that Ser-273 is the active site nucleophile of PAF acetylhydrolase. Identifying the position of the active site Ser provided a landmark from which the positions of the Asp and His components of a catalytic triad might be predicted. We compared the linear orientation of triad residues in lipases with defined structures (Table III). Although these lipases are from phylogenetically diverse organisms, they all have the basic ␣/␤ hydrolase core structure with the catalytic triad formed at one edge of the central ␤-sheet (24). In each case, the linear order of the active site residues is Ser, acidic residue, His. In these enzymes, the Ser and the acidic residue are generally separated by either one or two ␤-strands and a variable number of ␣-helices, and the complexity of this region correlates with the distance between the two residues. Specifically, in both human pancreatic lipase and human lipoprotein lipase, the catalytic Asp lies 24 residues downstream of the Ser. The sequence between these residues contains an ␣-helix followed by a ␤-strand (34,35). The structural constraints required to maintain the ␣/␤ organization and to position the Ser and Asp in close proximity may dictate a minimal distance of approximately 24 amino acids between the two residues. In most of the other lipases shown in Table III, the greater distance between the Ser and Asp correlates with the presence of two ␤-strands and a variable number of ␣-helices. These generalized observations were used to predict the location of possible active site Asp residues in plasma PAF acetylhydrolase. The structural characteristics of the region between the active site Asp and His residues of lipases are highly variable and consequently the spacing between these residues has less predictive value.
The third parameter used to evaluate all potential Asp and His catalytic site residues was whether they were conserved in PAF acetylhydrolase from several species. Downstream from Ser-273 in the human sequence lie five His residues which could potentially be part of the catalytic triad. Three of these (His-367, His-399, and His-428) are not conserved in the other species (Fig. 1) and are therefore unlikely candidates. In contrast, there are nine conserved Asp residues carboxyl to Ser-273. We focused first on identifying the essential His then examined the Asp residues found between that His and Ser-273.
Both of the conserved His residues downstream of Ser-273 were mutated to Ala. The H351A mutation was expressed as an independent construct; however, because of the close proximity of His-395 and His-399 and concerns of potential substitution effects, mutations of both residues were expressed in a single construct (H395A,H399A). As shown in Fig. 2, H351A lacked any detectable PAF acetylhydrolase activity while H395A,H399A produced more than wild-type levels of activity. As with the Ser mutations, both His mutations reacted well with the monoclonal antibodies, suggesting that these mutations did not grossly alter the conformation of the enzyme (Table II). The distance of 78 amino acids between His-351 and Ser-273 is less than that of any of the lipases in Table III but   histidine residues of a Pseudomonas dienlactone hydrolase, a thiol esterase with an ␣/␤ core and a catalytic triad (36). From these results, we conclude that His-351 is the active site histidine in plasma PAF acetylhydrolase.
If PAF acetylhydrolase has the prototypical lipase structure, the active site Asp would be expected to lie between Ser-273 and His-351. Within this region are four conserved or semiconserved Asp residues (Asp-286 is Glu in chicken; Asp-304 is Glu in mouse). Each was mutated to Ala and evaluated for an effect on PAF acetylhydrolase activity. The two mutations most distal to Ser-273, D304A and D338A, had no adverse effect on the activity of the enzyme (Fig. 2). In contrast, the D296A mutation reduced enzyme activity by more than 99% and D286A reduced the activity by more than 96% (Fig. 2 and Table  II). Since this loss of activity could arise from either an absolute requirement for both residues for catalysis or simply from a conformational perturbation introduced by one or both of the mutations, a more conservative mutation, Asp to Asn, was introduced at both positions. Again, mutation of Asp-296 (D296N) greatly reduced enzyme activity (by more than 99%). The D286N mutation, however, restored activity to within 70% of wild-type levels ( Fig. 2 and Table II). This suggests that the Asn mutations were less disruptive to conformation than the Ala mutations. This suggestion is supported by the observation that the Asp to Ala mutations rendered the proteins unrecognizable to the monoclonal antibodies but the Asp to Asn mutations restored the ability of the antibodies to bind the proteins (Table II). Together, these data support the conclusion that Asp-296 is the acidic component of the catalytic triad. Interestingly, the distance of 23 amino acids between the active site Ser-273 and Asp-296 is similar to that of the pancreatic lipase family (24 amino acids).
NH 2 -and COOH-terminal Deletion Analysis-The NH 2 terminus of circulating plasma PAF acetylhydrolase appears to be heterogeneous. We reported Ile-42 to be the NH 2 -terminal residue (13). However, in a subsequent purification of the enzyme from fresh human plasma, two additional NH 2 termini were encountered, Ser-35 and Lys-55 (Fig. 3). Subsequent re-analysis of NH 2 -terminal sequence data from the first purification revealed that a small amount of protein beginning with Lys-55 was also detectable. The heterogeneity may be the natural state of the enzyme in plasma or may occur during purification. In any case, the specific activity of the purified material is very close to that of recombinant PAF acetylhydrolase (not shown). To better understand how much of the enzyme is required for full enzymatic activity, we systematically deleted short segments from either end of the molecule. The NH 2 -terminal deletion mutants were expressed as fusion proteins with the bacterial thioredoxin gene (31) which ensured consistent high level expression of recombinant protein. COOH-terminal dele-tion mutants were expressed directly. In either case, only a few residues could be deleted without inhibiting catalysis (Fig. 4). From the NH 2 terminus (defined as beginning with Met-1), deletion of 53 amino acids appeared to enhance enzyme activity but removal of 60 residues reduced activity by 92% relative to the wild-type control and removal of 67 residues reduced enzyme activity to undetectable levels. At the COOH terminus, removal of 21 amino acids caused a slight loss of activity but the 30-residue deletion reduced catalysis to below the limit of detection. These results suggest that more than 83% of the primary sequence is required for full catalytic activity. The expendability of a few amino acids from either end of the protein reflects the sequence heterogeneity found in those regions in the different species (Fig. 1). Overall, these results are consistent with the ␣/␤ hydrolase prototype in which most of the protein is engaged in forming the core structure.

DISCUSSION
Plasma PAF acetylhydrolase is a unique extracellular PLA 2 that abolishes the diverse, potent actions of PAF and oxidized phospholipids. The enzyme is intriguing from a biochemical viewpoint for several reasons. First, unlike the other secreted PLA 2 s, PAF acetylhydrolase does not require calcium for activity (7). Another reason is the marked specificity the enzyme has for short chain sn-2 moieties (37). This specificity has the important advantage of allowing the enzyme to circulate in a completely active form, unlike the plasma proteases, which circulate as inactive zymogens. The basis for the specificity is unknown, but suggests that PAF acetylhydrolase has a substrate-binding region that excludes longer acyl chains. The structures of several extracellular PLA 2 s have been solved at high resolution and they have been shown to have a hydrophobic pocket into which the acyl chains of the substrate phospholipid fit, positioning the sn-2 ester at the active site (19,38). The structure of PAF acetylhydrolase is expected to be quite distinct because of its strict specificity for the short sn-2 residues. The observations reported here are consistent with this notion and illuminate another intriguing feature of the enzyme: mechanistically, plasma PAF acetylhydrolase appears to utilize a Ser-Asp-His catalytic triad and may assume the classical ␣/␤ hydrolase conformation of the neutral lipases. This enzyme is the first PLA 2 purported to exhibit these characteristics. The catalytic site of the secretory PLA 2 s contains an essential Asp and a His but no Ser. Rather, a calcium ion is used to stabilize the transition state (18). In contrast, a cloned intracellular PLA 2 does not require calcium for catalysis but neither does it have the conserved pentapeptide, GXSXG, that is characteristic of the neutral lipases (39 -41). Thus, plasma PAF acetylhydrolase is functionally a PLA 2 but is structurally and mechanistically distinct from the other known PLA 2 s. It is possible that another PAF-specific PLA 2 , the distinct, cytoplasmically active PAF acetylhydrolase described by Hattori et al. (9,10), has similar structural features since it also contains a (semi)conserved GXSXG motif. The evidence for the presence of a catalytic triad in PAF acetylhydrolase is compelling. Previous experiments have demonstrated that activity of the enzyme is inhibited by diisopropyl fluorophosphate, an agent that modifies active site Ser residues (7). Cloning of the plasma PAF acetylhydrolase gene revealed that the enzyme has the characteristic lipase/serine protease consensus pentapeptide, GXSXG (13), which generally contains the active site Ser. In the current report we have demonstrated by site-directed mutagenesis that the Ser within this motif is required for activity. The presence of a catalytically essential His was first suggested by experiments in which a modifier of active site His residues, p-bromophenacylbro-mide, was used to inhibit the activity of PAF acetylhydrolase (42). Based upon the linear configuration of catalytic residues in other lipases, the His as well as the acidic components of a triad would be expected to lie carboxyl to Ser-273. There are only two conserved His residues within this region; mutation of one, His-351, destroyed enzyme activity while mutation of the other did not affect function of the enzyme. Identification of the acidic triad component was complicated by the fact that between Ser-273 and His-351 lie four conserved Asp residues, but multiple analyses revealed that only Asp-296 was essential for hydrolytic activity. The distance of 23 amino acids between the essential Ser-273 and Asp-296 is interesting because it is very similar to the spacing (24 amino acids) between the analogous residues in the pancreatic lipase family (Table III). Within this region in pancreatic lipase as well as in other lipases, the GXSXG motif forms a tight "nucleophilic elbow" at the immediate carboxyl end of a ␤-strand (24). This is followed by one ␣-helix, another ␤-strand, and then the catalytic Asp. This distinguishes the secondary structure of the pancreatic lipase family from that of other lipases or hydrolases in which at least two ␣-helix/␤-strand pairs separate the Ser and acidic residues. Therefore, by extrapolation, we predict that the secondary structure of this region in plasma PAF acetylhydrolase may resemble that of the pancreatic lipase family.
The evidence for an ␣/␤ hydrolase conformation of plasma PAF acetylhydrolase is inferential. The enzyme lacks any sequence homology with other lipases except for the GXSXG motif. In the lipases, the Ser within this motif functions as the essential nucleophilic component of the catalytic triad. The presence of the conserved pentapeptide and other components of an apparent catalytic triad as well as the lipid specificity of PAF acetylhydrolase place it on common ground with other lipases. In addition, the intriguing linear similarity between PAF acetylhydrolase and the pancreatic lipase family within the active site Ser-Asp region is consistent with the enzyme assuming a similar secondary structure. Taken together, our results are consistent with the hypothesis that the enzyme is an ␣/␤ hydrolase with a catalytic triad, but confirmation of this hypothesis awaits crystallization and x-ray structural analysis.
We have demonstrated that plasma PAF acetylhydrolase is a PLA 2 with mechanistic and structural characteristics of the neutral lipases. Also like these lipases, PAF acetylhydrolase displays the property of interfacial activation (7) which, in the lipases, has been tied to the presence of a "lid" comprised of ␣-helices or loops that cover the hydrophobic catalytic pocket when the enzyme is in the inactive state (43)(44)(45). Upon contact with a lipid surface, conformational rearrangements occur that expose the active site to the substrate. Enzymes lacking the lid over the catalytic site are not activated at the lipid-water interface (guinea pig pancreatic phospholipase (46) and fungal cutinase (47)). Given the other similarities between PAF acetylhydrolase and the lipases, the enzyme might also be expected to have a lid-like structure that mediates its interfacial properties.
Acknowledgments-We thank Dina Leviten and Christi Wood for synthesizing oligonucleotides and for cDNA nucleotide sequencing, FIG. 3. Amino-terminal heterogeneity of purified plasma PAF acetylhydrolase. The enzyme was partially purified from human plasma on two separate occasions as described previously (13) and the NH 2 -terminal sequence was determined by Edman degradation. The purification procedures were identical except that fresh plasma was used in purification 2 whereas plasma used in purification 1 was stored at 4°C for up to 1 month.
FIG. 4. PAF acetylhydrolase activity produced in E. coli by enzyme truncated at the NH 2 or COOH termini. A series of nested truncations of the PAF acetylhydrolase cDNA was generated using an exonuclease deletion system. NH 2 -terminal deletions were expressed as fusions with the prokaryotic thioredoxin protein whereas the COOHterminal truncations were directly expressed. The NH 2 -or COOHterminal amino acid of each deletion is indicated below the abscissa. The specific activity values were determined as in Fig. 2 and represent the results of three independent experiments. The deletions are schematically summarized below the graph. Deletions that do not negatively affect enzyme activity are indicated above the bar, all others are below. In addition, the initiating Met (M1) and the NH 2 -terminal residue of the enzyme after cleavage of the putative signal peptide (V18 (13)) are indicated. The shaded region of the bar represents the minimal functional portion of the enzyme as defined by the deletions.
Shawn Housmann for conducting PAF acetylhydrolase activity assays, David Chantry and W. Michael Gallatin for critically reviewing this manuscript, and Louesa Isett for helping prepare the manuscript.