Functional Analysis of the Purified Anandamide-generating Phospholipase D as a Member of the Metallo-β-lactamase Family*

In animal tissues, bioactive N-acylethanolamines including the endocannabinoid anandamide are formed from their corresponding N-acylphosphatidylethanolamines (NAPEs) by the catalysis of a specific phospholipase D (NAPE-PLD) that belongs to the metallo-β-lactamase family. Despite its potential physiological importance, NAPE-PLD has not yet been characterized with a purified enzyme preparation. In the present study we expressed a recombinant NAPE-PLD in Escherichia coli and highly purified it. The purified enzyme was remarkably activated in a dose-dependent manner by millimolar concentrations of Mg2+ as well as Ca2+ and, hence, appeared to be constitutively active. The enzyme showed extremely high specificity for NAPEs among various glycerophospholipids but did not reveal obvious selectivity for different long chain or medium chain N-acyl species of NAPEs. These results suggested the ability of NAPE-PLD to degrade different NAPEs without damaging other membrane phospholipids. Metal analysis revealed the presence of catalytically important zinc in NAPE-PLD. In addition, site-directed mutagenesis studies were addressed to several histidine and aspartic acid residues of NAPE-PLD that are highly conserved within the metallo-β-lactamase family. Single mutations of Asp-147, His-185, His-187, Asp-189, His-190, His-253, Asp-284, and His-321 caused abolishment or remarkable reduction of the catalytic activity. Moreover, when six cysteine residues were individually mutated to serine, only C224S showed a considerably reduced activity. The activities of L207F and H380R found as single nucleotide polymorphisms were also low. Thus, NAPE-PLD appeared to function through a mechanism similar to those of the well characterized members of this family but play a unique role in the lipid metabolism of animal tissues.

N-Acylethanolamines (NAEs) 2 are ethanolamides of long chain fatty acids and exist in various organisms including animals and plants (1,2).
Among different NAEs, anandamide (N-arachidonoylethanolamine) has been most extensively studied and is well known as an endogenous ligand of cannabinoid receptors and transient receptor potential vanilloid 1 channel (3,4). As such, anandamide shows a variety of central and peripheral activities (5) and has recently received much attention due to its role in the regulation of female and male fertility in mammals (6,7). On the other hand, cannabinoid receptor-inactive N-palmitoylethanolamine is known to be an anti-inflammatory substance (8,9), N-oleoylethanolamine is known as an anorexic mediator (10), and N-stearoylethanolamine is known as a pro-apoptotic (11) and anorexic mediator (12). Furthermore, unsaturated C18 NAEs were recently reported to activate transient receptor potential vanilloid 1 (13). Noticeably, NAEs markedly increase in a variety of animal models of tissue degeneration (2, 14 -16).
It is generally accepted that in animal tissues NAEs are principally biosynthesized from membrane phospholipids by two steps of enzyme reactions (1,2,14,17). In the first reaction, N-acylphosphatidylethanolamine (NAPE) is produced from phosphatidylethanolamine (PE) by calcium-dependent N-acyltransferase, and in the second reaction the resultant NAPE is hydrolyzed to NAE and phosphatidic acid by a phosphodiesterase of the phospholipase D (PLD) type, generally referred to as NAPE-PLD.
Recently we cloned cDNA of NAPE-PLD from mouse, rat, and human (18,19). The deduced primary structure of NAPE-PLD showed no homology with those of other known PLDs but revealed that the enzyme belongs to the metallo-␤-lactamase family. The recombinant NAPE-PLD expressed in COS-7 cells produced long chain NAEs including anandamide from their corresponding NAPEs. However, the enzyme did not hydrolyze phosphatidylcholine (PC) or PE and lacked the ability to catalyze transphosphatidylation. We also observed that stable expression of NAPE-PLD in mammalian cells caused a decrease in the endogenous levels of NAPEs and a concomitant increase in those of NAEs without showing obvious selectivity for N-acyl species (20). These results substantially agreed with earlier findings with crude preparations (21)(22)(23)(24)(25)(26)(27) and confirmed that NAPE-PLD is structurally and catalytically distinguishable from the known PLDs. The activity, mRNA, and protein of NAPE-PLD were detected in various mouse organs such as brain, testis, and kidney (18) and all the brain regions of rat (28), suggesting that the enzyme plays the central role in the formation of anandamide and other NAEs in animal tissues.
To elucidate physiological and pathophysiological significance of an enzyme, catalytic properties of the purified enzyme must be clarified in detail. However, NAPE-PLD has not yet been characterized with the purified enzyme preparation. The first purpose of the present study is, therefore, to fully analyze the activation mechanism and substrate specificity of NAPE-PLD with a highly purified recombinant enzyme. In addition, although NAPE-PLD was classified into the metallo-␤-lacta-mase family based on its primary structure, it remained unclear whether or not NAPE-PLD functions through a catalytic mechanism similar to those of the well characterized members of this family. Thus, we attempted to identify catalytically important amino acid residues of NAPE-PLD. For this second purpose, we extensively performed sitedirected mutagenesis of the enzyme for the first time.
Expression and Purification of Recombinant Rat NAPE-PLD in E. coli-The full-length rat NAPE-PLD cDNA with SalI and NotI sites at the 5Ј-and 3Ј-ends was generated by PCR from rat NAPE-PLD-pcDNA3.1(ϩ) using the forward primer, 5Ј-GTCGACATGGATGAAAATGAGAA-CAGCCAG-3Ј, and the reverse primer, 5Ј-GCGGCCGCTCATGTT-TCCTCAAAGGCTTTGTC-3Ј and ligated into pGEX6P-1 with the aid of SalI and NotI. pGEX6P-1 enables the product to be expressed as a GST fusion protein with a PreScission protease cleavage site that allows removal of the GST tag. The construct NAPE-PLD-pGEX6P-1 was confirmed by sequencing in both directions. E. coli BL21 cells were transformed with this plasmid together with the chaperone plasmid pGro7. Cultures of a positive clone were induced with 0.1 mM isopropyl-␤-Dthiogalactopyranoside at an A 600 of 0.7, allowed to grow at 22°C for 16 h, and pelleted at 6000 ϫ g for 15 min at 4°C. After freezing and thawing, the cells were resuspended in 1 ⁄ 5 of the original culture volume of 100 mM triethanolamine-HCl (pH 7.4) containing 150 mM NaCl, 1 mM DTT, and 1 mM PMSF and lysed by sonication on ice 10 times each for 20 s with an interval of 1 min. The lysate was solubilized with 1% CHAPS, and the insoluble fraction was removed by centrifugation at 15,000 ϫ g at 4°C for 30 min. The supernatant was diluted 3-fold in 100 mM triethanolamine-HCl (pH 7.4) containing 150 mM NaCl, 20 mM MgCl 2 , 50 mM KCl, 1% CHAPS, 10 mM ATP, 1 mM DTT, and 1 mM PMSF (buffer A). Because the co-expressed GroEL (a gene product of pGro7) formed a complex with the NAPE-PLD-GST fusion protein, denatured proteins from bacterial lysate, which can bind to GroEL, were also added as described previously (33) except that the denatured proteins were precipitated with trichloroacetic acid. After incubation at 37°C for 20 min, the sample was centrifuged at 15,000 ϫ g for 30 min at 4°C, and to the resultant supernatant derived from 100 ml of the original culture, 2 ml of glutathione-Sepharose 4B beads pre-equilibrated with buffer A was added. The mixture was then incubated at 4°C overnight with gently mixing to allow the GST-NAPE-PLD fusion protein to bind to the beads. The beads were then packed into a column and washed twice with 10 ml of buffer A and once with 10 ml of buffer A in which the ATP concentration was reduced to 5 mM. The fusion protein was eluted from the column with 6 ml of 50 mM Tris-HCl (pH 8.0) containing 10 mM glutathione, 1% CHAPS, 1 mM DTT, and 1 mM PMSF. After removal of glutathione by extensive dialysis, the fusion protein was subjected to digestion with PreScission protease at 4°C for 10 h to release the GST tag. The sample was loaded onto a glutathione-Sepharose 4B column (bed volume, 2 ml) again to remove the cleaved GST tag, and the GST-free NAPE-PLD, which passed through the column, was collected. The above-mentioned purification procedure was performed six times, and the purified enzyme was pooled and finally applied onto a Bio-Gel HTP hydroxyapatite column (1 ml). After washing the column with 10 ml of 50 mM Tris-HCl (pH 7.4) containing 0.1% CHAPS, 1 mM DTT, and 1 mM PMSF (buffer B) and then with 15 ml of buffer B containing 50 mM potassium phosphate, the enzyme was eluted with 6 ml of buffer B containing 200 mM potassium phosphate. All the purification procedures were performed at 4°C. The purified enzyme was stored in the presence of 1% octyl glucoside at Ϫ80°C until use. Protein concentration was determined by the method of Bradford (34) with bovine serum albumin as standard.
Mutagenesis-Single mutations were introduced into the mammalian expression vector pcDNA3.1(ϩ) harboring rat or human NAPE-PLD by PCR with the aid of a QuikChange site-directed mutagenesis kit. A series of deletion mutants was constructed by PCR using mouse NAPE-PLD cDNA as a template, and the PCR products were inserted into pcDNA3.1(Ϫ)-HisA using EcoRV and BamHI sites of the multicloning sites. The oligonucleotides used as PCR primers are listed in Table 1. All the constructs were subsequently sequenced to check the introduction of the desired mutations.
Expression of NAPE-PLD in COS-7 Cells-COS-7 cells were grown at 37°C to 70% confluency in a 100-mm dish containing Dulbecco's modified Eagle's medium with 10% (v/v) fetal calf serum in a humidified 5% CO 2 , 95% air incubator. The cells were then treated with the plasmid Enzyme Assay-NAPE-PLD was incubated with 25 M 14 C-labeled substrates (2500 cpm/2.5 nmol) in 100 l of 50 mM Tris-HCl (pH 7.4) containing 10 mM MgCl 2 and 0.1% octyl glucoside at 37°C for 10 -120 min unless otherwise noted. S. chromofuscus PLD was incubated with 25 M 14 C-labeled substrates (2500 cpm/2.5 nmol) in 100 l of 50 mM Tris-HCl (pH 8.0) containing 10 mM CaCl 2 and 1% Triton X-100 at 37°C for 10 -120 min. For radioactive substrates with 14 C in the N-acyl group, the enzyme reactions were terminated by the addition of 0.3 ml of a mixture of chloroform/methanol (2:1, v/v) containing 5 mM 3(2)-t-butyl-4-hydroxyanisole. For those with 14 C in the sn-2 acyl group, a mixture of chloroform, methanol, 36% HCl (2:1:0.01, v/v) containing 5 mM 3(2)-t-butyl-4-hydroxyanisole was added. As for sn-glycero-3-phospho(N-[ 14 C]palmitoyl)ethanolamine, the reaction was terminated with a mixture of 1.5 ml of chloroform/methanol (2:1, v/v) containing 5 mM 3(2)-t-butyl-4-hydroxyanisole followed by extraction with 0.2 ml of 2 M KCl containing 0.1 M EDTA according to the method of Folch et al. (35). After the termination of enzyme reaction, 100 l of the organic phase was spotted on a silica gel thin-layer plate (10-or 20-cm height) and developed in chloroform, methanol, 28% ammonium hydroxide (40:10:1, v/v) (for N-[ 14  Distribution of radioactivity on the plate was quantified by a BAS1500 bioimaging analyzer (Fujix, Tokyo, Japan). When PI was used as a substrate, the enzyme reaction was performed with 25 M L-␣-[myoinositol-2-3 H(N)]PI (25,000 cpm/2.5 nmol). After the reaction, 1 ml of chloroform, 0.5 ml of methanol, and 0.2 ml of water were added, and the produced [ 3 H]inositol in a 400-l aliquot of the upper phase was measured by liquid scintillation counting (36). All the enzyme assays were performed in triplicate. K m and V max were determined according to the method of Lineweaver and Burk (37).
Western Blotting-After separation by SDS-PAGE with a 10% gel, proteins were electrotransferred to a hydrophobic polyvinylidene difluoride membrane (Hybond P). The membrane was blocked with phosphate-buffered saline containing 5% dried milk and 0.1% Tween 20 (buffer C) and then incubated with anti-NAPE-PLD antiserum (1:200 dilution) or anti-hexahistidine antibody (1:10,000 dilution) in buffer C at room temperature for 1 h followed by incubation with the horseradish peroxidase-labeled secondary antibody (1:2000 dilution) in buffer C at room temperature for 1 h. Finally, NAPE-PLD protein was visualized using ECL plus kit and analyzed by a LAS1000plus lumino-imaging analyzer (Fujix, Tokyo, Japan).
Atomic Absorption Spectrometry-Zinc content was determined by a Shimadzu atomic absorption/flame spectrometer model AA-630 -01. Zinc standard solution was diluted with 10 mM Tris-HCl (pH 7.4) containing 40 mM potassium phosphate and 0.1% CHAPS in a range of 0 -1.0 ppm to establish a standard calibration curve. Zinc values determined were based on comparison with the standard curve. Each of the purified NAPE-PLD preparations was analyzed in triplicate.

RESULTS
Purification of Recombinant NAPE-PLD-We constructed a prokaryotic expression vector to generate a fusion protein of rat NAPE-PLD and GST. This fusion protein was expressed in E. coli together with the chaperone protein GroEL and cochaperonin GroES, which assist in the folding of a large number of proteins in E. coli (38). The co-expression resulted in the increase in the amount of the fusion protein solubilized with 1% CHAPS up to 10-fold in terms of the specific activity of NAPE-PLD (data not shown). As shown in Table 2 and Fig. 1, the solubilized fusion protein was purified by glutathione-Sepharose 4B chromatography and was then digested with PreScission protease to detach its GST tag. After removal of the released GST tag by the second cycle of glutathione affinity chromatography, the GST-free NAPE-PLD was further purified by hydroxyapatite chromatography. When the final preparation was analyzed by SDS-PAGE, a major protein band was seen at a position of 46 kDa corresponding with the molecular mass of NAPE-PLD (Fig. 1, lane 6). Through this purification procedure, we could reproducibly prepare recombinant NAPE-PLD with a specific activity of about 2.0 mol/min/mg of protein with N-palmitoyl-PE as the substrate. The purified enzyme was unstable, and one cycle of freezing and thawing caused loss of the enzyme activity up to 50%. Because 1% (w/v) octyl glucoside was found to improve the stability of the purified   (26). However, the stimulatory effects of divalent cations on the purified NAPE-PLD remained unclear. When the purified recombinant enzyme was allowed to react with N-palmitoyl-PE in the presence of increasing concentrations of MgCl 2 or CaCl 2 , the activity was dose-dependently enhanced up to 15-30-fold ( Fig. 2A). The EC 50 values of Mg 2ϩ and Ca 2ϩ were 2.3 and 1.4 mM, respectively. These results indicated that the purified NAPE-PLD was also markedly stimulated by millimolar concentrations of Mg 2ϩ and Ca 2ϩ . Therefore, 10 mM MgCl 2 was used in all the following assays. The K m value of the enzyme stimulated with 10 mM Mg 2ϩ was a little lower than that of the Mg 2ϩ -free enzyme (1.7 versus 5.9 M), and V max of the former was much higher than that of the latter (1833 versus 148 nmol/min/mg of protein) (Fig. 2B). The effect of 10 mM Mg 2ϩ could be replaced not only by Ca 2ϩ but also by other divalent cations such as Co 2ϩ , Mn 2ϩ , Ba 2ϩ , and Sr 2ϩ of the same concentration, although their stimulatory effects varied (Fig. 2C). In contrast, Fe 2ϩ , Cu 2ϩ , Hg 2ϩ , and Zn 2ϩ were inhibitory. We also contained 0.1% octyl glucoside in the standard assay mixture as a weak activator showing a synergistic effect with Mg 2ϩ (data not shown).
Substrate Specificity of Purified Recombinant NAPE-PLD-We were interested to know the precise substrate specificity of the purified NAPE-PLD with a variety of glycerophospholipids and related compounds. We first examined the reactivity of the pure enzyme toward various NAPEs with N-acyl groups consisting of different carbon numbers (C1 to C20) (Fig. 3). All the tested NAPEs with C4 or longer N-acyl chains, including precursors of bioactive NAEs (N-arachidonoyl-PE, N-stearoyl-PE, N-oleoyl-PE, and N-palmitoyl-PE), were found to be highly active substrates showing specific activities of 1088 -2420 nmol/ min/mg of protein. The highest specific activity was observed with N-lauroyl-PE. However, N-acetyl-PE and N-formyl-PE were much less active (305 and 64 nmol/min/mg of protein, respectively). These results clarified that the purified enzyme does not have an obvious preference regarding carbon numbers of long chain or medium chain N-acyl species of NAPEs, but such an N-acyl group is indispensable to serve as a substrate of NAPE-PLD.
To examine the role in the substrate specificity of O-acyl chains of the sn-1 and sn-2 positions and glycerol structure of NAPE, we next tested several compounds prepared by partial digestion of N-palmitoyl-PE (Table 3). Although the enzyme generated N-palmitoylethanolamine from N-palmitoyl-lyso-PE and glycerophospho(N-palmitoyl)ethanolamine, the specific activities with these compounds were only 4 and 1% of that with N-palmitoyl-PE. Furthermore, N-palmitoylethanolamine phosphate was totally inactive even with a large amount of the enzyme. Under the same assay conditions, we also tested the PLD-type hydrolyzing activity of the membrane fraction of rat brain toward these compounds (Table 3). The NAPE-PLD activity with N-palmitoyl-PE was detected as reported previously (23,25), and the membranes also showed the PLD-type activity toward N-palmitoyl-lyso-PE, glycerophospho(Npalmitoyl)ethanolamine, and N-palmitoylethanolamine phosphate, which was much higher than that toward N-palmitoyl-PE. Based on these results, it appeared that phosphatases other than NAPE-PLD are mostly responsible in rat brain for the PLD-type hydrolysis of the compounds partially digested from NAPE.  We also examined the reactivity of the purified NAPE-PLD with major glycerophospholipids existing in biomembranes (PC, PE, PI, and PS). Commercially available S. chromofuscus PLD, which is known to have a NAPE-hydrolyzing activity (32), hydrolyzed not only NAPE but also PC, PE, PI, and PS, although the reaction rates were largely varied among these substrates (Table 4). This broad substrate specificity was in agreement with a previous report (39). On the other hand, NAPE-PLD hydrolyzed PE at an extremely low rate (0.04% of that of N-palmitoyl-PE) and was totally inactive with PC, PI, and PS (Table 4). These results demonstrated that NAPE-PLD hardly hydrolyzes major glycerophospholipids of biomembranes, confirming its high specificity for NAPE in contrast to S. chromofuscus PLD. In addition, other phospholipids structurally related to NAPE (N-palmitoyl-PS, PEt, and phosphatidylbutanol) were hydrolyzed at much lower rates (less than 0.4%) as compared with N-palmitoyl-PE (Table 4), suggesting the importance of the N-acylethanolamine moiety of NAPE for the recognition by NAPE-PLD.
Furthermore, we investigated the effects on the purified recombinant NAPE-PLD of various phosphate compounds including ATP, ADP, AMP, cAMP, and diphosphoric acid. When contained in the reaction mixture at 1 mM, these compounds did not affect the catalytic activity of NAPE-PLD toward N-palmitoyl-PE (data not shown). The results suggested that these compounds do not function as substrates or activators.
Functional Analysis of Deletion Mutants of NAPE-PLD-As we reported previously (18), NAPE-PLD belongs to the metallo-␤-lactamase family, a superfamily including a wide variety of hydrolases such as B. cereus ␤-lactamase, human glyoxalase II, arylsulfatase, and cAMP phosphodiesterase (40,41). Members of this family have the metallo-␤lactamase domain that is highly conserved and is suggested to be catalytically important. The region spanning Asp-147-His-331 of NAPE-PLD corresponds to this domain (Fig. 4) (18). Although many members of the metallo-␤-lactamase family are soluble proteins (41), NAPE-PLD is a membrane-bound protein (28). Therefore, we were interested in examining whether the N-terminal or C-terminal region outside the metallo-␤-lactamase domain of NAPE-PLD is responsible for the membrane association. We constructed six mutants in which either the N-terminal region or C-terminal region was variably deleted (Table 5 and Fig. 4) and tested the membrane association and activity of the mutant proteins overexpressed in COS-7 cells. Successful expression of all the mutants was confirmed by detecting the C-terminal c-Myc-His 6 tag by Western blotting with anti-hexahistidine antibody. When the cell homogenates were subjected to ultracentrifugation, Western blotting revealed that all the deletion mutants as well as the wild-type enzyme were mostly recovered in the membrane fraction rather than the cytosol, suggesting that the N-terminal region or C-terminal region is not essential for the membrane binding (data not shown). After solubilization from the membrane fractions with 1% octyl glucoside, the contents of all the deletion mutants in the solubilized proteins were similar to that of the wild-type (Fig. 5A). As shown in Table 5, the NAPE-PLD activity in the solubilized proteins was completely abrogated with four deletion mutants (⌬N85, ⌬N138, 367stop, and 377stop). A moderate decrease in the V max values without an obvious change in K m was observed with ⌬N55 and 387stop. These results provided evidence for an essential role of the N-terminal and C-terminal regions in maintaining the catalytic activity.
Functional Analysis of Single Mutants of NAPE-PLD-Because the aspartic acid and histidine residues highly conserved in the metallo-␤lactamase domain have been presumed to be involved in binding and processing of substrates (40 -42), we were interested in examining whether their corresponding residues of NAPE-PLD actually contribute to the catalytic activity. Three aspartic acid residues (Asp-147, Asp-189, and Asp-284) and five histidine residues (His-185, His-187, His-190, His-253, and His-331) of NAPE-PLD were presumed to be such conserved residues and were completely conserved among rat, mouse, and human NAPE-PLDs (Fig. 4) (18). These residues of rat NAPE-PLD were separately changed to asparagine by site-directed mutagenesis, and the mutants were expressed in COS-7 cells. As analyzed by Western blotting, the expression levels of the mutants were similar to that of the wild type (Fig. 5B). Table 6 indicates that the N-palmitoyl-PE-hydrolyzing activity was abrogated with D147N, H253N, and D284N or remarkably reduced (less than 0.1% that of the wild type) with H185N, H187N,  D189N, and H190N, suggesting that these seven residues are necessary for the catalysis of NAPE-PLD. The activity of H331N was also low but was higher than those of the other mutants (4% that of the wild type). The K m value of H331N (21.1 Ϯ 6.2 M) was higher than that of the wild type (4.7 Ϯ 2.6 M). In consideration of the significant activity of H331N, it was likely that a neighboring histidine residue plays an important role in place of His-331. Thus, we prepared single mutants of His-321, His-343, and His-353, which were conserved among rat, mouse, and human NAPE-PLDs. As shown in Table 6, the catalytic activity was abrogated with H321N, whereas the K m and V max values of H343N (3.9 Ϯ 0.4 M and 259.4 Ϯ 33.1 nmol/min/mg of protein) and H353N (5.5 Ϯ 1.4 M and 260.9 Ϯ 17.8 nmol/min/mg of protein) were similar to those of the wild type.
Several of the highly conserved aspartic acid and histidine residues have also been suggested to be related to the metal binding, and zinc is known as a representative metal contained in the members of this protein family (40 -42). Therefore, we analyzed the purified NAPE-PLD for zinc content by atomic absorption spectrometry. E. coli alkaline phos-

TABLE 3 The PLD-type activity of the purified recombinant NAPE-PLD and rat brain membranes toward compounds prepared by partial digestion of N-palmitoyl-PE
The purified recombinant rat NAPE-PLD and the 105,000-g pellet (membrane fraction) from homogenates of the brain of adult Wistar rats (28) were allowed to react with 25 M concentrations of the indicated substrates in the presence of 10 mM MgCl 2 and 0.1% octyl glucoside at 37°C for 10 min, and the produced N-palmitoylethanolamine was quantified. Mean values Ϯ S.D. are shown (n ϭ 3). GP(NP)E, glycerophospho(Npalmitoyl)ethanolamine; NPE-P, N-palmitoylethanolamine phosphate.  phatase (2 g atoms of zinc per mol of enzyme) (43) and soybean lipoxidase (1 g atom of iron, but no zinc, per mol of enzyme) (44) were used as a positive control and a negative control, respectively, and their zinc contents were determined to be 1.66 Ϯ 0.15 and 0.00 Ϯ 0.00 g atom per mol of enzyme (mean Ϯ S.D., n ϭ 3). When four different preparations of the purified NAPE-PLD, which exhibited a single protein band on SDS-PAGE, were analyzed under the same conditions, their zinc contents were 0.34 Ϯ 0.02, 0.25 Ϯ 0.00, 0.16 Ϯ 0.01, and 0.14 Ϯ 0.02 g atom per mol of enzyme. We noticed that the zinc content of each preparation was well correlated with its specific enzyme activity (4020 Ϯ 225, 2891 Ϯ 115, 1922 Ϯ 52, and 1690 Ϯ 77 nmol/min/mg of protein, respectively). These results suggested that NAPE-PLD contains zinc, which is catalytically important. However, stoichiometry between zinc and the enzyme protein was not clarified.

Substrate
Recently, we presented the inhibitory effect of p-chloromercuribenzoic acid on recombinant NAPE-PLD with an IC 50

TABLE 5 Catalytic activity of deletion mutants of NAPE-PLD
The octyl glucoside-solubilized proteins (0.018 -42 g) of COS-7 cells overexpressing wild type or deletion mutants of mouse NAPE-PLD were allowed to react with various concentrations of N-͓ 14 C͔palmitoyl-PE in the presence of 10 mM MgCl 2 and 0.1% octyl glucoside at 37°C for 10 min. Assays were repeated three to five times for each sample, and mean values Ϯ S.D. are shown. rected mutagenesis, and the NAPE-PLD activity of the mutants transiently expressed in COS-7 cells were investigated. Western blot analysis revealed their expression levels comparable with that of the wild type (Fig. 5B). As shown in Table 7, the K m values (4.2-6.1 M) and V max values (257-269 nmol/min/mg of protein) of C222S, C237S, C255S, and C288S were similar to those of the wild type. A moderate decrease in the activity was seen with C170S. Notably, a considerable decrease in the activity was observed with C224S (K m , 14.3 Ϯ 5.5 M; V max , 23.3 Ϯ 5.3 nmol/min/mg of protein), suggesting the importance of Cys-224 in the catalysis. Functional Analysis of Single Nucleotide Polymorphisms of NAPE-PLD-We examined a possible presence of single nucleotide polymorphism (SNP) of NAPE-PLD by the use of the SNP human data base of NCBI and found four SNPs in the open reading frame of the human NAPE-PLD gene. Although these SNPs did not appear to link to any diseases according to the data base, they caused the substitution of an amino acid residue: S152A (point mutation at nucleotide 454T3 G), L207F (621G3 C), H380R (1139A3 G), and D389N (1165G3 A). Because all of these amino acid residues were conserved among NAPE-PLDs of human, rat, and mouse, the residues might be catalytically important. We prepared these four mutants of the human enzyme by site-directed mutagenesis and successfully expressed them in COS-7 cells as revealed by Western blot analysis (Fig. 5C). Table 8 indicates that the K m values (2.9 -3.7 M) and the V max values (21.9 -22.7 nmol/ min/mg of protein) of S152A and D389N were similar to those of the wild type. In contrast, the NAPE-PLD activity was not detected with L207F and H380R. Because we noticed that the wild type of human NAPE-PLD was much less active than that of the rat enzyme, we also prepared L207F and H380R of rat NAPE-PLD. The activities of these rat mutants were detectable but much lower than that of the wild type of the rat enzyme (less than 5%) ( Table 8). These results suggested that L207F and H380R as SNPs of human NAPE-PLD have pathological significance.

DISCUSSION
Recent cDNA cloning of NAPE-PLD by our group enabled us to analyze this enzyme by molecular biological approaches (18). We performed initial characterization of recombinant NAPE-PLD with crude preparations of COS-7 cells transiently expressing this enzyme and showed that the enzyme is catalytically distinguishable from PLDs of the HKD/phosphatidyltransferase family in terms of the reactivity specific for NAPE and the lack of the transphosphatidylation activity (18). We further overexpressed recombinant NAPE-PLD in E. coli by a conventional method and attempted to purify it. However, most of the expressed enzyme protein was insoluble and inactive, and we purified the inactive enzyme only for the purpose of preparation of anti-NAPE-PLD antibody (18). Thus, the purified NAPE-PLD has not yet been characterized.
In the present study we highly purified an active recombinant NAPE-PLD for the first time. This was achieved by expressing an NAPE-PLD-GST fusion protein together with chaperon proteins in E. coli. By the purification procedure developed, the specific activity of the enzyme reached about 2 mol/min/mg of protein with N-palmitoyl-PE as a substrate. This value was 5-fold higher than that of the enzyme purified from rat heart (18) but was not as high as that we expected on the basis of the high purity of the final preparation revealed by SDS-PAGE (Fig. 1,  lane 6). It was likely that some of the recombinant NAPE-PLD protein was expressed as an inactive form in E. coli and/or was denatured during the purification procedure that takes 6 days.
Elucidation of the activation mechanism and substrate specificity is an important step to understand physiological and pathophysiological roles of an enzyme. In the present study we carefully examined the activation mechanism and substrate specificity of NAPE-PLD with the purified recombinant enzyme. In animal tissues NAPEs generally exist much abundantly than NAEs (2), suggesting the presence of a regulatory mechanism of the NAPE-PLD activity. In previous work we found that Ca 2ϩ potently activated the enzyme partially purified from rat heart (26). However, millimolar concentrations of Ca 2ϩ was required to cause significant activation of this enzyme, and Ca 2ϩ could be replaced by several other inorganic divalent cations including Mg 2ϩ . On the other hand, the recombinant enzyme expressed in COS-7 cells was stimulated only 2-fold by 10 mM Ca 2ϩ and Mg 2ϩ (18). In the present study we

Site-directed mutagenesis of NAPE-PLD addressed to highly conserved aspartic acids and histidines
The octyl glucoside-solubilized proteins (0.017-19 g) of COS-7 cells overexpressing wild type or mutants of rat NAPE-PLD were allowed to react with 100 M N-͓ 14

Functional Analysis of Anandamide-generating Phospholipase D
showed that the pure NAPE-PLD can also be stimulated markedly not only by Ca 2ϩ but also by Mg 2ϩ and other divalent cations. Although EC 50 of Mg 2ϩ was as high as 2.3 mM, in consideration of intracellular Mg 2ϩ present at 20 mM the enzyme appeared to be constitutively active in vivo.
If NAPE-PLD constitutively exists as an active phospholipase in the cell, high substrate specificity should be critical to minimize damage of membrane phospholipids that might be caused by undesirable side reactions of this enzyme. Our present studies revealed that the pure enzyme is almost inactive with major glycerophospholipids in biomembranes (PC, PE, PS, and PI). Moreover, the enzyme hardly hydrolyzed phospholipids structurally related to NAPE such as N-acyl-PS, PEt, and phosphatidylbutanol. As shown in Table 4, such a high specificity of NAPE-PLD for NAPE was largely different from the wide substrate specificity of S. chromofuscus PLD, which does not belong to the metallo-␤-lactamase family (45).
The crude preparations of NAPE-PLD were previously reported not to show selectivity with respect to long chain N-acyl groups of NAPE (18,23). Our results with the pure recombinant enzyme confirmed the earlier findings and demonstrated a lack of the preference for N-arachidonoyl-PE as the anandamide precursor. Interestingly, NAPE-PLD also hydrolyzed NAPEs with medium chain (C4-C14) N-acyl groups. Thus, NAPE-PLD appeared to be responsible for the degradation of a variety of NAPEs with different N-acyl groups in vivo.
Because Natarajan et al. (22) reported earlier that microsomes of dog brain had the PLD-type activity for N-acyl-lyso-PE and glycerophospho(N-acyl)ethanolamine, it was likely that NAPE-PLD can hydrolyze compounds prepared by partial digestion of N-palmitoyl-PE (N-palmitoyl-lyso-PE, glycerophospho(N-palmitoyl)ethanolamine, and N-palmitoylethanolamine phosphate). In the present study we also showed high reactivity of membrane fraction of rat brain with these phosphate compounds, whereas the pure NAPE-PLD revealed relatively low or no reactivities with the compounds ( Table 3), suggesting that the brain tissues have phosphatases other than NAPE-PLD that can catalyze the PLD-type hydrolysis reactions. Thus, it was likely that NAE can be formed from NAPE by phosphatases once NAPE is hydrolyzed by phospholipase A 1 , A 2 , or C. In agreement with this finding, we recently suggested that sequential reactions by secretory phospholipase A 2 and a lysophospholipase D distinct from NAPE-PLD formed NAE from NAPE via N-acyl-lyso-PE in rat tissues (29). Liu et al. (46) also suggested that in RAW264.7 cells anandamide could be formed from N-arachidonoyl-PE by a combination of phospholipase C and a phosphatase that hydrolyzed anandamide phosphate (46).
Site-directed mutagenesis study on NAPE-PLD has not yet been performed. By this method we first examined the role of the N-terminal and C-terminal regions of NAPE-PLD, which are not conserved within the metallo-␤-lactamase family, in contrast to the conserved metallo-␤lactamase domain. Our present results with deletion mutants suggested that the N-terminal or C-terminal region is not essential for membrane association. However, we could not rule out a possibility that both of the regions are involved in it. When examined using the SOSUI (47), TMPRED (48), and TopPred program (49), any predicted transmembrane domains were not found in the whole primary structure of NAPE-PLD. Moreover, using PSORT II program (50), subcellular localization signals such as endoplasmic reticulum retention signals (KKXX motif at the C terminus and XXRR motif at the N terminus) were not detected. Thus, mechanism for the membrane association remains unclear. On the other hand, the NAPE-hydrolyzing activity was reduced or abrogated with all the deletion mutants. In agreement with the results with the C-terminal deletion mutants (377stop and 387stop), single mutation of His-380, but not Asp-389, resulted in a remarkable decrease in the catalytic activity (Table 8).
Our site-directed mutagenesis addressed to the highly conserved aspartic acid and histidine residues of NAPE-PLD suggested the catalytic importance of Asp-147, His-185, His-187, Asp-189, His-190, His-253, and Asp-284. Recently, site-directed mutagenesis studies on various members of the metallo-␤-lactamase family have been reported. Single mutation of the residues corresponding to His-185, His-187, Asp-189, and His-253 of NAPE-PLD always resulted in remarkable reduction of the catalytic activity with E. coli and Arabidopsis thaliana ribonuclease Z (51, 52), human Artemis (53,54), Bacillus thuringiensis N-acyl-L-homoserine lactone hydrolase (55), and IMP-1 metallo-␤-lactamase (56 -58). Furthermore, the residue corresponding to Asp-147 (Artemis), that corresponding to His-190 (ribonuclease Z and IMP-1), and that corresponding to Asp-284 (ribonuclease Z, Artemis, and N-acyl-L-homoserine lactone hydrolase) were also considered to be catalytically important based on the results of mutagenesis (51)(52)(53)(54)(55)(56)(57)(58). On the other hand, contrary to our speculation, H331N still showed an activity higher than the other mutants ( Table 6), suggesting the presence of another histidine residue as a substitute for His-331. Therefore, we examined His-321, His-343, and His-353 of NAPE-PLD and found that replacement of His-321 with asparagine resulted in the complete loss of the activity. Although we have previously defined His-331 as one of the conserved histidines in the metallo-␤-lactamase domain based on the multiple sequence alignment (18), the present results suggested the catalytic importance of His-321 rather than His-331. It was also reported with ribonuclease Z and phosphorylcholine esterase that mutants of the histidine residue corresponding to His-321 exhibited very low activities (51,52,59). Taken together, the results of our mutagenesis directed to the highly conserved aspartic acid and histidine residues of NAPE-PLD were in good agreement with those of other members of the metallo-␤-lactamase family.
Furthermore, recent crystallographic studies revealed three-dimensional structures of members of the metallo-␤-lactamase family such as B. cereus ␤-lactamase (60), Fluoribacter gormanii Zn-␤-lactamase (61), human glyoxalase II (62), Desulfovibrio gigas rubredoxin oxygen:oxidoreductase (63), Bacillus subtilis and Thermotoga maritima ribonuclease Z (64,65), Streptococcus pneumoniae phosphorylcholine esterase (59), and B. thuringiensis N-acyl-L-homoserine lactone hydrolase (55). The results elucidated that the conserved domain within the family is composed of ␣␤/␤␣ sandwich structure. Two atoms of metal such as zinc and iron were contained, and the metal binding sites were located at one edge of the ␤-sandwich. Our metal analysis also suggested the presence of zinc in NAPE-PLD. Because we noticed that there is a close resemblance between NAPE-PLD and ribonuclease Z in terms of catalytic importance of the conserved aspartic acid and histidine residues, we chose the crystal structure of B. subtilis ribonuclease Z (PDB code 1y44) (64) as a template and performed homology-based protein modeling of the conserved domain spanning L125-D389 of NAPE-PLD with the aid of ModBase (66) and 3D-JIGSAW (67). This model predicted that His-185, His-187, and His-253 of NAPE-PLD bind to the first metal ion, Asp-189, His-190, and His-343 bind to the second metal ion, and Asp-284 forms a bridge between the two metal ions. Because H343N was catalytically active in our assay, further examination will be necessary to clarify the role of His-343 in the metal binding. Although H321N was totally inactive, His-321 appeared not to be involved in the metal binding. Instead, the model suggested that His-321 of NAPE-PLD corresponds to His-247 of ribonuclease Z, which was shown to bind to a phosphate ion (64). Because both NAPE-PLD and ribonuclease Z catalyze phosphodiesterase reaction, His-321 may also be involved in the positioning of the phosphate moiety of substrate. In addition, Asp-147 was assumed not to directly bind to metals but to form salt bridges to the conserved motif backbone, probably stabilizing the conformation of the motif to coordinate optimally the metal ions (53). A definite proof will be given when the three-dimensional structure of NAPE-PLD is available by crystallography.
Further site-directed mutagenesis studies showed that specific mutations of Cys-224, Leu-207, and His-380 in NAPE-PLD resulted in remarkable reduction of the enzyme activity (Tables 7 and 8). Earlier, the brain NAPE-PLD was reported to be inhibited by a sulfhydryl blocking reagent (68). Recently, we showed that p-chloromercuribenzoic acid dose-dependently inhibited recombinant rat NAPE-PLD (28). These observations suggested the presence of a cysteine residue(s) affecting the catalytic activity of NAPE-PLD. In the present study individual mutation of six cysteines conserved among rat, mouse, and human NAPE-PLDs revealed that only C224S caused considerable reduction of the activity. Thus, Cys-224 may be a target for development of inhibitors. In agreement with this finding, a higher concentration of p-chloromercuribenzoic acid was required to inhibit the remaining activity of the C224S (data not shown). Our results also indicated that L207F and H380R revealed a markedly reduced NAPE-hydrolyzing activity. These mutants were found as SNPs but did not appear to link to any diseases according to the data base. Analysis of NAPE-PLD gene knock-out mice may reveal an abnormality caused by deficiency of NAPE-PLD. Recently, the NAPE-PLD gene was suggested to be a candidate of myeloid tumor suppressors (69).
In conclusion, we investigated the activation mechanism and substrate specificity of NAPE-PLD using the purified recombinant enzyme and clarified that NAPE-PLD is a constitutively active phospholipase with extremely high specificity for NAPE. The findings strongly suggested that the major role of NAPE-PLD is the formation of various NAEs including anandamide and other bioactive NAEs from their corresponding NAPEs. Furthermore, our site-directed mutagenesis studies suggested that the catalytic mechanism of NAPE-PLD is similar to those of the well characterized members of the metallo-␤-lactamase family.