Synthesis of Prostaglandin E2 Ethanolamide from Anandamide by Cyclooxygenase-2*

Because of its structural similarity to polyunsaturated fatty acids, anandamide could serve as substrate for enzymes such as lipoxygenases and cyclooxygenases, which metabolize polyunsaturated fatty acids to potent bioactive metabolites. Here the ability of recombinant human cyclooxygenase-1 (hCOX-1) and cyclooxygenase-2 (hCOX-2) to metabolize anandamide was studied. Baculovirus-expressed and -purified hCOX-2, but not hCOX-1, effectively oxygenated anandamide. Reverse phase high pressure liquid chromatography analysis of the products derived from 1-14C-labeled anandamide showed that the products formed are similar to those formed with arachidonic acid as substrate. The major prostanoid product derived from anandamide was determined by mass spectrometry to be prostaglandin E2 ethanolamide. Incubation of anandamide with lysates and the intact cell line expressing COX-2 but not that of COX-1 produced prostaglandin E2 ethanolamide. These results demonstrate the existence of a COX-2-mediated pathway for anandamide metabolism, and the metabolites formed represent a novel class of prostaglandins.

pholipase D in response to various stimuli (7). The AEA thus released inside the cell could participate in signal transduction as a second messenger and display some of its immunomodulatory and anti-inflammatory activities independent of its interaction with the CB receptors. In fact, AEA has been shown to antagonize CB2 receptors, and it is not clear how this antagonism results in immunomodulatory activities observed in cells only expressing CB2 receptors (8). It is possible that a metabolite of AEA rather than AEA itself could account for all or some of these properties. Furthermore, because of its structural similarities to polyunsaturated fatty acids, endogenously released AEA could serve as substrate for lipoxygenases and cyclooxygenases (COX) that metabolize polyunsaturated fatty acids to potent bioactive molecules. It has been demonstrated that lipoxygenase could metabolize AEA, and the metabolites have potent biological activities (9,10). However, it is not known whether COX can metabolize AEA. Arachidonic acid (AA) is the substrate for both COX-1 and COX-2. AEA is structurally identical to AA except that the carboxylic acid group of AA is replaced by an ethanolamide group in AEA.
Here, we report for the first time the ability of hCOX-2 but not hCOX-1 to bind and oxidize AEA. We further demonstrate that the products of AEA metabolism by COX-2 are unique and represent a novel class of eicosanoids.
Recombinant COX-1 and COX-2 Enzymes-Recombinant human COX-1 and COX-2 enzymes were expressed in a baculovirus system and partially purified as described previously (17). The specific activity of the enzymes used were between 13,000 and 21,000 units (1 unit ϭ 1 nmol of oxygen consumed/mg of protein/min) with a purity of Ϸ60% as judged by using sodium dodecyl sulfate-polyacrylamide gel electrophoresis and silver staining.
Enzymatic Assays-The purified COX enzymes were reconstituted with 2 mM phenol and 1 M hematin (Sigma), and the cyclooxygenase activity was measured either by the oxygen uptake assay or by the radiometric assay exactly as described previously (17). K m values were determined radiometrically using 2-200 M 1-14 C-labeled AA or AEA (specific activity, 3-4 ϫ 10 5 cpm/nmol). For characterization of the metabolites by RP-HPLC, reconstituted hCOX-1 and hCOX-2 enzymes were incubated in a total volume of 100 l with 40 M [1-14 C]AA or [1-14 C]AEA (specific activity, 3-4 ϫ 10 5 cpm/nmol) for 5 min at room temperature. The reaction was stopped by the addition of 10 l of 2 N HCl and 1 volume of methanol. An aliquot (50 -75 l) was directly injected and analyzed by HPLC as described below. For characterization of the metabolites by HPLC-MS, unlabeled AA and AEA were used and processed as above for RP-HPLC studies.
Cell Culture-Human promonocytic THP cells were grown in RPMI medium (Life Technologies, Inc.), washed, and suspended in fresh RPMI medium. COX-1 was induced by treating cells with 0.1 M phorbol 12-myristate 13-acetate (Sigma) for 40 -48 h at 37°C. Primary cultures of human foreskin fibroblasts (HFF) were obtained from ATCC * 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.
Metabolism of AA and AEA by THP and HFF Cells-COX-1-expressing THP cells and COX-2-expressing HFF cells were washed with phosphate-buffered saline, and the medium was replaced with Tris-HCl buffer (50 mM Tris, pH 7.8) containing 2 mM EDTA. Cells (5-8 ϫ 10 4 cells) were incubated with 50 -100 M of either [1-14 C]AA or [1-14 C]AEA (specific activity, 3-4 ϫ 10 5 cpm/nmol) for 30 min at 30°C. The reaction was stopped by adding 1 volume of isopropanol:acetic acid (100:1.2, v/v) and 1 volume of chloroform. The tubes were mixed by vortexing and centrifuged (1000 ϫ g, 10 min); the lower chloroform layer was aspirated and concentrated to dryness, and the residue was dissolved in 200 l of methanol and analyzed by HPLC as described below.
Metabolism of AEA by Cell Lysate-Following COX-2 induction, HFF cells were detached by treating with trypsin and washing with phosphate-buffered saline. COX-1-expressing THP cells and COX-2-expressing HFF cells were suspended in 50 mM Tris-HCl buffer containing 2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, 20 g/ml leupeptin, 10 g/ml trypsin inhibitor, and 10 g/ml pepstatin. Cells were lysed by homogenization, and the cell lysate was stored on ice until used. Lysates (200 -400 g) from COX-1-or COX-2-expressing cells were incubated with either [1-14 C]AA or [1-14 C]AEA (40 M) for 20 min at 30°C. The reaction was stopped and processed as described above for intact cells.
HPLC-Products from the reaction of [ 14 C]AA and [ 14 C]AEA with purified hCOX-1 and hCOX-2, intact cells, and the cell lysates were separated using a Shimadzu LC-6A HPLC system. RP-HPLC was carried out on a Waters Bondapak C18 column (300 ϫ 3.9 mm, 10-m particle size) using solvent A (water/trifluoroacetic acid, 100:0.05, v/v) and solvent B (acetonitrile/trifluoroacetic acid, 100:0.05, v/v). The initial settings were 30% solvent B from 0 to 5 min, which was linearly increased to 60% solvent B in the next 30 min, held there for an additional 5 min, linearly increased to 100% solvent B in the next 10 min, and held there for 9 min. The flow rate was 1.5 ml/min, and the total run time was 60 min. The radioactivity in the column effluent was detected using an on-line Beckman 171 radioactivity detector. The signals from the detector were acquired, stored, and analyzed using Ezychrome software (Shimadzu, Pleasanton, CA).
HPLC-Mass Spectrometry-HPLC was performed using a model 140B Applied Biosystems liquid chromatograph with a C-18 Vydac 2.1 ϫ 250-mm column using solvent C (water/acetonitrile/acetic acid, 90:10:0.01, v/v) and solvent D (methanol). The initial settings were 35% solvent B from 0 to 3 min, which was linearly increased to 100% solvent B in the next 57 min and held there for 5 min, and the flow rate was 200 l/min. The injector was a model 7725i Rheodyne with a 100-l injection loop. A total of 20 l of sample was injected on the column. The HPLC flow was sent directly into a Finnigan TSQ 700 triple stage quadrupole mass spectrometer (Finnigan-MAT, San Jose, CA) equipped with an atmospheric pressure chemical ionization source. The mass spectrometer was operated in the positive ion mode with a vaporizer temperature of 400°C and a heated capillary temperature of 150°C. Nitrogen was used as a sheath gas at 75 p.s.i. to assist in nebulization. The instrument was scanned from m/z 100 to 750 in 2 s. Data were acquired on a DEC station 5000 data system.  a Specific activity is the initial rate expressed as mol of oxygen consumed/mg of protein/min and is the average of two determinations (COX-1) or mean Ϯ S.D. of three determinations (COX-2).
b Not determined.

RESULTS AND DISCUSSION
There is considerable interest in recent years in understanding the metabolism of AEA as a mechanism of termination of its potent neuromodulatory and immunomodulatory activities. The principal mode of metabolism of AEA appears to be an amidohydrolase-mediated hydrolysis to AA and ethanolamine (19,20). Here we test the ability of COX enzymes to metabolize AEA. Fig. 1 shows the relative initial rates of oxygenation of AA and AEA by recombinant hCOX-1 and hCOX-2. Both en-zymes effectively oxygenated AA. In contrast, COX-2 oxygenated AEA at 60 -85% of the rates observed with AA, whereas COX-1 displayed no detectable activity against this substrate. Using the more sensitive radiometric method (17) no activity was detected for COX-1 with AEA as substrate (data not shown). Furthermore, when 2-5 molar excess unlabeled AEA was added along with [1-14 C]AA, no inhibition of AA metabolism by COX-1 was observed suggesting that not only is AEA not a substrate for COX-1 but it also fails to compete with AA  Fig. 2, except unlabeled AA and AEA were used, and following the reaction the samples were separated and analyzed by LC-MS as described under "Experimental Procedures." Panel A is the base peak profiles of metabolites of AA with hCOX-2, and panel B is the base peak profiles of metabolites of AEA with hCOX-2. Peak numbers correspond to those in Fig. 2. Peaks eluting between 36 and 40 min are due to detergents in the enzyme preparations.
for binding to the enzyme active site (data not shown).
The affinity of hCOX-2 for AEA was consistently lower than that for AA as evident form the mean K m values for the two substrates (Table I). The initial rates of oxygenation of AEA by COX-2 were also consistently lower than that of AA but did not achieve statistical significance because of batch to batch variations in the specific activities of COX-2. Furthermore, as with AA, COX-2 underwent inactivation following oxygenation of AEA (data not shown). With both AA and AEA as substrate, the COX-2 enzyme turned over Ϸ250 -400 times before it inactivated. Thus, it appears that the mechanism involved in the oxidation of AEA by COX-2 is similar to that of AA.
Next, to characterize the metabolites, the reaction products of AEA with COX-2 were analyzed by HPLC. Fig. 2 shows the RP-HPLC profiles of the metabolites of the reaction of hCOX-1 and hCOX-2 with the 14 C-labeled AA and AEA. Once again, it is clear that hCOX-1 was unable to use AEA as substrate (Fig.  2B), whereas it effectively converted AA into one major product ( Fig. 2A, peak 3) that comigrated with the [ 3 H]PGE 2 standard. Small amounts of minor metabolites (peaks 4 -8) were also formed. The concentrations of these minor metabolites varied from assay to assay and represented Ͻ1-5% of the total radioactivity. In contrast to hCOX-1, hCOX-2 effectively oxygenated both AA and AEA as substrates (Fig. 2, C and D). The profiles of the products formed from these two substrates by hCOX-2 were similar except for their retention times (compare Fig. 2C  with 2D). Here again, hCOX-2 produced one major metabolite with AA (Fig. 2C, peak 3) that comigrated with the [ 3 H]PGE 2 FIG. 4. Mass spectra of PGE 2 standard and peak 3. Shown are the atmospheric pressure chemical ionization mass spectra of PGE 2 standard (A) and peak 3 of Fig. 3 (B). standard. It also produced significant amounts of other metabolites (peaks 5-8), which is in agreement with previous reports (21,22). These minor peaks displayed significant UV absorption at 235 nm, and two of these peaks (peak 6 and 8) had the same retention time as 12-HHT and 15-HETE, respectively. The metabolic profile of the products of hCOX-2 with AEA was very much identical to that with AA, except for shorter retention time because of the ethanolamide group and relative abundance of the individual peaks (Fig. 2D).
The major metabolite of AEA (Fig. 2D, peak 3a) did not display any UV absorption suggesting that it is unlikely to have conjugated double bonds. To get more structural information, the products of the reaction of hCOX-2 with AA and AEA were subsequently analyzed by LC-MS, and the base peak profiles are shown in Fig. 3. The mass spectra of major peaks of interest are shown in Figs. 4 and 5. Peak 3 (Fig. 3A) displayed a fragmentation ion pattern identical to the synthetic PGE 2 standard (compare Fig. 4A with 4B). The base peak at m/z 317 was due to the loss of a -OH group and a molecule of H 2 O from the molecular ion (M Ϫ 35). Other diagnostic fragment ions included M ϩ H 2 O (at m/z 370), the molecular ion (at m/z 352), and M Ϫ OH (at m/z 335). Similarly, peak 3a (Fig. 3B) displayed a fragmentation ion pattern identical to the synthetic PGE 2 ethanolamide (compare Fig. 5A with 5B). The base peak at m/z 378 was due to the loss of a -OH group from the molecular ion. Other diagnostic ion fragments included MHϩ  /z 342). The fragmentation ion pattern of peak 4a (Fig. 3B) was identical to that of peak 3a (data not shown). Thus, it is likely that peak 4a is an isomer of PGE 2 ethanolamide or PGD 2 ethanolamide. Based on the limited fragmentation ions (data not shown), peaks 6, 6a, 8, and 8a (Fig. 3) were tentatively identified as 12-HHT, 12-HHT-ethanolamide, 15-HETE, and 15-HETE-ethanolamide, respectively. Additional studies are needed to confirm their structural identities.
The study demonstrates for the first time that COX-2 can accommodate a non-carboxylic fatty acyl chain such as AEA into its active site and oxygenate it. Furthermore, it seems that in the active site of the enzyme the substrate AEA assumes the same L-shaped configuration and undergoes similar oxidation and cyclization as has been proposed in the case of AA (23). At present we do not know the stereochemistry of the products formed from AEA, and as a result the exact conformation of AEA within the active site of the enzyme may or may not be similar to that of AA. However, it can be presumed that in the case of AA the carboxylate group is anchored at the bottom of the active site channel of COX-2 (as well as COX-1) by a hydrogen bond network involving arginine 120, tyrosine 355, and glutamic acid 524, which is analogous to the way flurbiprofen and zomepirac bind to COX-1 and COX-2, respectively (12,15). Despite the absence of a free carboxylate group, AEA effectively bound to COX-2 but not COX-1, which suggests that COX-2 has additional binding site(s) at the bottom of the channel to accommodate non-carboxylic fatty acyl substrates such as AEA. In fact, the recent crystal structure of an inhibitorbound hCOX-2 shows an alternate salt bridge formation between Tyr-355, Glu-524, and Arg-513 at the bottom of the channel (12). It is not known whether this network of amino acids helps COX-2 bind AEA.
Next, we wanted to test whether cells expressing COX-1 or COX-2 could utilize AEA as substrate and convert it into ethanolamide prostaglandins. Therefore, we incubated HFF cells expressing COX-1 or COX-2 and THP cells expressing COX-1 with 14 C-labeled AEA, and the products formed were analyzed by RP-HPLC (Fig. 6). Like the purified enzyme, HFF cells expressing COX-2 effectively oxygenated AEA. The profile of the products formed from AEA by the COX-2-expressing cells was identical to that formed by the purified enzyme (Fig. 6B). Similar results were also obtained when [ 14 C]AEA was incubated with cell-free preparations of COX-2-expressing HFF cells. On the other hand, neither the intact cells nor the lysates of cells expressing COX-1 oxygenated [ 14 C]AEA (data not shown). These data clearly demonstrate that AEA is also metabolized by COX-2 in a more physiological environment.
At the present time the physiological significance of the metabolism of AEA by COX-2 is not known. Several possible roles for this pathway can be speculated. First, metabolism of AEA by COX-2 is yet another physiological mechanism by which AEA is inactivated. Second, this could be a pathway for generating novel prostaglandins that in turn might have many important pathophysiological functions. Third, by its ability to compete with AA, AEA might modulate the local prostaglandin tone mediated by COX-2. Our future work should provide answers to some of these questions.