Aspirin Inhibition and Acetylation of the Plant Cytochrome P450, Allene Oxide Synthase, Resembles that of Animal Prostaglandin Endoperoxide H Synthase*

The enzymatic reactions leading to octadecanoid lipid signaling intermediates in plants are similar to those of animals and are inhibited by nonsteroidal anti-inflammatory drugs (NSAIDs) such as salicylic acid and aspirin. In animals, NSAIDs inhibit the cyclooxygenase (COX) activity of prostaglandin endoperoxide H synthase, which ultimately blocks the formation of prostaglandins. In plants, NSAIDs block the formation of 12-oxo-phytodienoic acid and jasmonates, which are the equivalent signaling compounds. In this study we show that NSAIDs act as competitive inhibitors of allene oxide synthase (AOS), the cytochrome P450 that initiates plant oxylipin synthesis. We also show that aspirin causes the time-dependent inhibition and acetylation of AOS, which leads the irreversible inactivation of this enzyme. This inhibition and acetylation superficially resembles that observed for the inactivation of COX in animals. In AOS, aspirin acetylates three serine residues near the C-terminal region that appear to be highly conserved among AOS sequences from other plants but are not conserved among “classical” type P450s. The role of these serine residues is unclear. Unlike animal COX, where acetylation of a single serine residue within the substrate channel leads to inactivation of prostaglandin endoperoxide H synthase, the three serine residues in AOS are not thought to line the putative substrate channel. Thus, inhibition by aspirin may be by a different mechanism. It is possible that aspirin and related NSAIDs could inhibit other P450s that have motifs similar to AOS and consequently serve as potential biochemical targets for this class of drugs.

apeutic drugs in the world. They act by interfering with the cyclooxygenase activity of PGHS (1), which ultimately inhibits prostaglandin formation and its associated physiological effects. The widespread use and importance of NSAIDs has fostered the development of new anti-inflammatory drugs that are designed to selectively inhibit different isoforms of PGHS (2)(3)(4). However, NSAIDs also affect other physiological processes that are not linked solely to the suppression of PGHS activity. These include inhibitory effects on transcription factors (5), cellular kinases (6), and nitric oxide synthase (7). In this study we show that NSAIDs also inhibit AOS, the plant cytochrome P450 that serves as the equivalent of PGHS in animals.
The oxylipin pathway leading to prostaglandin synthesis in animals is mimicked in plants by a similar pathway (Fig. 1) that leads to the synthesis of two prostanoid-like fatty acids, 12-oxo-phytodienoic acid (12-oxo-PDA) and 7-iso-jasmonic acid (JA). In animals, synthesis begins with the C 20 polyunsaturated fatty acid, arachidonic acid. Plants, in contrast, use C 18 polyunsaturated linolenic or linoleic acids (8 -11). Although animals and plants synthesize similar pentacyclic fatty acids for signaling, they employ mechanistically different enzymes and reactions to achieve their final product. Animals utilize PGHS, a single, multifunctional enzyme that has both cyclooxygenase and hydroperoxidase activities (Fig. 1). The first step, using cyclooxygenase, adds two molecules of dioxygen to arachidonic acid to form the pentacyclic fatty acid, prostaglandin G 2 . The second hydroperoxidase step converts prostaglandin G 2 to prostaglandin H 2 (Fig. 1). In plants, this analogous step is carried out by two separate enzymes (10). The first uses lipoxygenase to convert linolenic or linoleic acids into their corresponding hydroperoxides. The second, catalyzed by AOS, is the committed step in plant oxylipin synthesis (10,12) and converts hydroperoxide fatty acids into unstable fatty acid epoxides (13,14). After this reaction these allene epoxides can undergo two fates depending on the absence or presence of a third enzyme, allene oxide cyclase, in plant tissue. In the absence of allene oxide cyclase, nonenzymatic reactions generate a minute quantity of 12-oxo-PDA, the first pentacyclic derivative of plants and a preponderance of ␣ and ␥ ketols, which arise by the spontaneous decomposition of allene epoxide. These nonenzymatic reactions are distinguished by the production of 9(R, S) and 13(R, S) racemic side chains in 12-oxo-PDA. In the presence of allene oxide cyclase, the "natural" form of 9(S),13(S)-12-oxo-PDA is biosynthesized (15)(16)(17). Modifications and rearrangements of pentacyclic derivatives then lead to the formation of prostaglandins, prostacyclin, and thromboxane (8,10) in animals and 12-oxo-PDA, JA, methyl JA, and its derivatives (8, 10, 18 -21) in plants. These pentacyclic derivatives then activate numerous physiological roles in their respective organism.
Earlier work indicated that exogenous applications of aspirin and SA blocked the wound response of plants (22). Salicylates were also shown to inhibit JA formation and the JA-induced expression of defense-related genes in plants (23,24). Because AOS was vital for JA biosynthesis, it was proposed that aspirin and SA acted at this step of the pathway (23). However, reduced levels of JA could also have occurred if allene oxide cyclase activity was inhibited. Thus, the enzymatic step affected by NSAIDs on plant oxylipins was unclear.
To characterize these possible effects of SA and aspirin, we used recombinant AOS prepared from a complementary DNA to AOS from the desert guayule plant (25). This recombinant enzyme was produced in high yields in Escherichia coli using the pGEX-KG expression vector and showed high specific activity for three different fatty acid hydroperoxide substrates. The enzyme was also inhibited by a number of NSAIDs including SA and aspirin. Moreover, aspirin led to the time-dependent, irreversible inhibition of AOS that was similar to its effect on animal cyclooxygenase and also caused the acetylation of three serine residues in the C-terminal region of AOS. This acetylation may have been responsible for the observed irreversible inhibition of AOS.

EXPERIMENTAL PROCEDURES
Construction of pGEX-AOS Vector and Expression of AOS in E. coli -The EcoRI-XhoI fragment of pRPP30 containing the AOS open reading frame (25) was regenerated by polymerase chain reaction using primers that contained an NcoI restriction site in the sense direction (5Ј-GCGCCATGGCACCCATCGTCTAAACCCCT-3Ј) and a SacI site in the antisense direction (5Ј-CGCCGAGCTCATATACTAGCTCTCTTC-AGGAACG-3Ј). The polymerase chain reaction product was digested with NcoI and SacI and ligated into the pGEX-KG (26) prokaryotic expression vector to generate the pGEX-AOS expression plasmid. Sequencing verified that no unwanted changes had been introduced by the polymerase chain reactions. The pGEX-AOS plasmid ( Fig. 2A) was used to transform E. coli strain JM105, which became the source of recombinant AOS for enzyme studies. Recombinant AOS protein was purified from cultures of JM105/pGEX-AOS using previous methods (27) and analyzed on a 10% SDS-PAGE gel (28) stained with Coomassie Brilliant Blue R (Fig. 2B). Heme staining of a companion gel indicated that heme was only associated with the 53 kDa band (data not shown).
Inhibition of AOS with SA, Aspirin, and Related NSAIDs-Assorted NSAIDs, including phenylpropionic acid and indole forms, were used in the presence of 50 M 13S-HPODE as the substrate. AOS activity was measured in the absence or presence of 10 M SA, aspirin, ibuprofen, indomethacin, or piroxicam using 40 ng of pure AOS in 2 ml of 0.1 mM 13S-HPODE. Data was collected for 4 replicates Ϯ S.D. When aspirin was incubated with AOS for 3 min before adding the 13S-HPODE substrate, Lineweaver-Burk analysis initially indicated that it acted as a noncompetitive inhibitor with a K i of 11 M. In a separate experiment, purified AOS (40 ng) was preincubated with 1 ml of 10 M aspirin for 0, 1, 2, 5, and 10 min and then was combined with 1 ml of 0.1 mM 13S-HPODE in a reaction chamber and assayed for activity. These were compared with preparations where aspirin was omitted. Decay of AOS activity by aspirin showed that it acted as a competitive inhibitor, which irreversibly inhibited activity after several min. In a separate experiment, radioactively labeled aspirin [acetyl-1-14 C]acetylsalicylic acid (55 mCi/mmol, American Radiochemical, St. Louis, MO) was used (see below), and samples of 14 C-labeled AOS were removed after 0, 1, 2, 5, and 10 min and counted by liquid scintillation to determine whether acetylation of AOS by aspirin coincided with the observed loss of enzyme activity.
Aspirin Acetylation and Analysis of AOS-To test for the ability of aspirin to acetylate AOS, acetylation reactions were performed by a modified procedure (30) using [acetyl-1-14 C]acetylsalicylic acid (55 mCi/ mmol, American Radiochemical). Thirty g of purified AOS were incubated with 1 mM 14 C-aspirin in a volume of 180 l. Samples of 60 l each were removed after 0-, 1-, and 2-h incubations at 37°C and terminated by adding 600 l of ice-cold acetone. The precipitate was collected by centrifugation and washed three times with acetone. The protein pellet then was dried under reduced pressure and suspended in loading buffer and subjected to SDS-PAGE. After electrophoresis, the gel was dried, sprayed with EN 3 HANCE TM (NEN Life Science Products) and exposed to x-ray film for 50 days at Ϫ80°C before development. The 53-kDa AOS band was specifically labeled after a 1-or 2-h incubation with 14 C-aspirin. To identify the amino acid residues that were acetylated, 14 C-aspirinlabeled enzyme was digested with Lys-C proteinase (Promega) using a protein to protease ratio of 10:1 in the presence of 50 mM ammonium bicarbonate buffer, pH 7.8, containing 0.05% SDS for 4 h at 37°C. After lyophilization, the resulting endoproteinase digests were analyzed by reverse phase HPLC using a micro Bond Pak C18 column and a linear gradient (0 to 50% B) of solvents A (0.1% trifluoroacetic acid in water) and B (0.08% trifluoroacetic acid in acetonitrile). Absorbance of peaks were monitored at 214 nm, peaks were collected manually, and radioactivity was determined by scintillation counting. Radiolabeled peptides were sequenced by Edman degradation using an Applied Biosystems model 477A equipped with an on-line phenylthiohydantoin-amino acid analyzer. Samples (30%) of the phenylthiohydantoin derivatives released after each Edman cycle were measured for radioactivity by liquid scintillation counting.

RESULTS
Purified, recombinant AOS from E. coli exhibited high turnover rates with each of three different substrates used (Fig. 3). Lineweaver-Burk analysis yielded k cat values of 4700 s Ϫ1 for linoleic hydroperoxide (13S-HPODE), 3700 s Ϫ1 for linolenic hydroperoxide (13S-HPOTE), and 1400 s Ϫ1 for arachidonic hydroperoxide (15S-HPETE) (Fig. 3). AOS also showed higher affinity (K m ϭ 27 M) for 15S-HPETE than it did for 13S-  1 and 2 show the isolated AOS from E. coli as two major bands representing an oxidized (53 kDa) and reduced (60 kDa) form of the enzyme along with a small quantity (80 kDa) of fusion protein (AOS-glutathione S-transferase (GST)) that eluted from the affinity column. Lane 1 shows the reduced protein sample treated with 50 mM Tris-Cl, pH 6.8, 2% SDS, 10% glycerol, and 10 mM ␤-mercaptoethanol, which has the 60-kDa band in higher abundance than the 53-kDa band. Lane 2 is from an identical protein sample where ␤-mercaptoethanol was omitted. Reduction with ␤-mercaptoethanol causes a conformational change in AOS that leads to its higher apparent molecular weight and the resultant loss of its heme. Heme staining of a companion gel (not shown) indicated that heme was only associated with the 53-kDa band.  (Fig. 3).
Inhibitor studies using 13S-HPODE as the substrate showed that SA was a competitive inhibitor of AOS with a K i of 238 M (Fig. 4A). Other NSAIDs such as indomethacin, ibuprofen, and piroxicam yielded similar results (Fig. 4B). When aspirin was incubated with the enzyme for more than 3 min, it appeared to act as a noncompetitive inhibitor of AOS, with a K i of 11 M (Fig. 4C). However, when aspirin was exposed to AOS for periods less than 10 min, it acted as a competitive inhibitor that irreversibly blocked AOS activity in a time-dependent manner (Fig. 4D). This is comparable with the effect aspirin has on PGHS, where covalent acetylation of the protein occurs within 15 min of exposure to aspirin (31)(32)(33) and suggests that the irreversible inhibition observed with AOS might also be because of a covalent modification. To test this, quantities of purified AOS were incubated with 14 C-acetate-labeled aspirin for 1 or 2 h. AOS was then precipitated, repeatedly washed, and subjected to SDS-PAGE autoradiography. Results indicated that the 53-kDa band representing AOS became radioactively labeled because of aspirin acetylation (Fig. 4D and Fig. 5). When 14 C-labeled AOS was treated with KOH, it hydrolyzed the labeled acetate group to yield unlabeled protein, verifying that the protein was acetylated by aspirin.
To identify the amino acid residue(s) involved, 14 C-labeled protein was proteolytically cleaved, and the peptides were subjected to HPLC fractionation, sequencing, and scintillation counting. Three labeled peptide fragments were identified (Fig.  6). Peptide fragment #1, with the sequence (S*)VVYESLRI-EPPVPPQYGK, was labeled in cpm as follows (bkg, background): (2000*)/100/75/65/50/40/bkg/bkg/bkg/bkg/bkg/bkg/ bkg/bkg/bkg/bkg/bkg/bkg/bkg (Fig. 6). Only the first of the two serine residues in this peptide fragment was labeled, which represented Ser 332 in guayule AOS. Protein sequence comparisons with AOS from other species revealed that this serine was conserved among other plants and corresponded to Ser 395 in flaxseed AOS (34), Ser 375 in Arabidopsis AOS (35), and to Ser 347 in pepper hydroperoxide lyase (HPL) (36) (Fig. 7). The second labeled peptide fragment, #2, with the sequence SNFTIE(S*)HDATFEVKK, was labeled in cpm to a lesser degree as follows: 55/75/40/60/50/(750*)/40/50/bkg/bkg/bkg/bkg/ bkg/bkg/bkg/bkg (Fig. 6). Only the second of the two serine residues in this peptide was labeled. It represented Ser 359 in guayule AOS, Ser 422 in flaxseed AOS, Ser 402 in Arabidopsis AOS, and Ser 373 in pepper HPL (Fig. 7). The third labeled peptide fragment, #3, with the sequence YVWW(S*)NG PEETESPTVEN, was also weakly labeled in cpm, as follows: 60/50/70/(300*)/75/45/40/bkg/bkg/bkg/bkg/bkg/bkg/bkg/bkg/ bkg/bkg/bkg (Fig. 6). Only the first of the two serine residues in this peptide was labeled and represented Ser 411 in guayule AOS, Ser 474 in flaxseed AOS, Ser 455 in Arabidopsis AOS, and Ser 427 in pepper HPL (Fig. 7). In all cases only serine residues of AOS were acetylated. Of these, Ser 332 of guayule AOS was most heavily labeled at 2000 cpm, whereas Ser 359 and Ser 411 were labeled to a lesser degree. All three of these serines reside in the C-terminal region of AOS, and all are conserved among the known AOS and HPL sequences so far isolated (Fig. 7). DISCUSSION AOS belongs to a group of NADPH-and O 2 -independent nonclassical P450s (37) that utilize fatty acid hydroperoxide substrates. Other P450s in this group include thromboxane synthase and prostacyclin synthase (37,38). These heme peroxidases are noted for not accepting electrons from a proteinaceous reducing partner. This gives them uncommonly high turnover rates compared with classical P450s. Recombinant AOS from guayule exhibits these same high turnover rates, with k cat values in excess of 1000 s Ϫ1 for each of the three substrates tested, which is consistent with this class of cytochrome P450s. It is surprising, however, that AOS shows higher affinity for 15S-HPETE than it does for 13S-HPODE or 13S-HPOTE. These latter two substrates are normal plant metabolites, whereas 15S-HPETE, being the predominant lipid derivative in animals, is rarely encountered in plants (19,10). However, these data appear to agree with early results that show that flaxseed AOS could metabolize a wide variety of lipid hydroperoxides, including 15S-HPETE (39).
Our results show that NSAIDs act as competitive inhibitors of AOS. This supports earlier findings (23,24) that AOS is the possible target of these inhibitors in blocking oxylipin synthesis. It is noteworthy that HPL from sunflower has also been reported to be inhibited 35% by SA (40). This is because it too, is another form of a nonclassical P450 that happens to share 40% homology with AOS (36). Thus, AOS from plants and its closely related enzyme, HPL, appear to possess features that result in inhibition by salicylates. One exception to this observation may be a form of AOS from coral that did not appear to be inhibited by indomethacin (41).
NSAIDs are not generally viewed as inhibitors of cytochrome P450s (42). However, nitric oxide synthase, a heme thiolate enzyme related to cytochrome P450, is reported to be inhibited by aspirin and related NSAIDs (7). Indomethacin can also inhibit at least one classical P450 monooxygenase that is involved in arachidonic acid metabolism (43,44). Thus NSAIDs such as aspirin, which are known to disrupt the arachidonic cascade, may do so by affecting enzymes other than PGHS (43).
In animals, salicylates block autacoid formation by inhibiting the cyclooxygenase activity of PGHS and do so by acetylating a single serine residue that resides near the C terminus of the protein. The acetyl group blocks entry of the substrate in the channel leading to the active site, causing irreversible inhibition of the enzyme. In plants, although salicylates are also known to block autacoid formation, it is not known if AOS is inhibited by a similar mechanism. Plants do not possess cyclooxygenase, and AOS has no sequence similarity to it. This is the first known example of a cytochrome P450 to be acetylated and irreversibly inhibited by aspirin. Although aspirin at high quantities is known to acetylate other proteins such as hemoglobin, the fact that low concentrations of aspirin used here led to the irreversible inactivation that coincided with its rate of labeling in vitro argues that acetylation may be the reason for its inactivation. Moreover, the quantities of radiolabeled aspirin used in our studies were identical to those used in the original experiments demonstrating aspirin acetylation of animal PGHS (30). Except for a few isolated examples, classical P450s do not become inhibited by salicylates and are not expected to be acetylated by aspirin. Thus, we do not expect aspirin to acetylate most other classical type P450s.
The best-studied target of aspirin is PGHS. Salicylates are able to compete for the active site of PGHS despite having a molecular structure that is completely different from the normal lipid substrate of the enzyme (1). Inactivation is caused by the acetylation of a serine residue positioned some distance from the active site in the substrate channel. This acetate further blocks the entry of lipid substrates and causes the irreversible inhibition of the enzyme (45)(46)(47). Because AOS responds in a similar manner, it hints that aspirin inhibition might be comparable with PGHS. The kinetics of aspirin inhi- FIG. 6. HPLC analysis of proteolysis products of 14 C-aspirinlabeled AOS. Three radioactively labeled peptide fragments were identified, as indicated by arrows on the HPLC elution profile. These were collected and subjected to amino acid analysis and scintillation counting. The sequence of each fragment is shown, and the most heavily labeled residues from each fragment are indicated in parentheses.
FIG. 8. Comparison of amino acid sequences in the K helix region of assorted cytochrome P450s. The labeled serine residue (*) of guayule AOS is conserved in AOS and HPL from other species. but serines are not found in other classical cytochrome P450s where these residues are typically Ala, Met, or Gln at this position. The highly conserved Glu and Arg residues (arrows) are found in all cytochrome P450s, including AOS and HPL. bition on AOS are similar to PGHS, and the acetylated residues occur on serines found near the C terminus. If similar mechanisms are involved, one would expect that one or more of these acetylated serines would reside in the substrate channel of AOS. Unfortunately, crystal structures of AOS and other eukaryotic P450s are not currently available to verify this. However, three-dimensional structures for prokaryotic P450s (48,49) used to make predictions for structural features of eukaryotic P450s (50) were used in comparing sequence alignments between conventional P450s and for residues Ser 332 , Ser 359 , and Ser 411 of guayule AOS and corresponding positions of AOS and HPL from other plants (Figs. 7 and 8). The most heavily labeled residue, at Ser 332 , resides in the conserved K helix (50), or the so called domain B (51), found in all P450s (Fig. 8). This residue lies 4 and 7 residues downstream of the highly conserved Glu and Arg residues, which are common to all P450s. These Glu and Arg residues are required for H bonding and normal enzyme function. Thus, it is conceivable that acetylation of this serine in AOS could disturb normal H bonding and possibly inactivate the enzyme. In BM3 (49), this residue is occupied by Met 316 , so it would not be affected by acetylation. However, serines occupy this position in AOS and HPL from the other plant species (Fig. 8), which are all thought to be inhibited by salicylates.
The two other serines we observed in guayule AOS at Ser 359 and Ser 411 , which were labeled to a lesser extent, correspond to residues Gly 343 and Ile 385 of BM3. In BM3 these positions are found on the surface of the protein and are, thus, not suspected of affecting enzyme activity in such classical P450s. However, for other plant forms of AOS and HPL, these residues are conserved. Thus we cannot conclude if these other residues contribute to enzyme inactivation. Because NSAIDs appear to act as competitive inhibitors, one would expect that these inhibitors should affect the substrate binding site of AOS. Without a three-dimensional structure, it is difficult to predict if this occurs or to be certain whether aspirin acetylation is responsible for the observed inhibition of AOS.
Salicylates occur naturally in plants and are known to regulate several important cellular processes in them (52)(53)(54). Salicylates are not produced in animals, but animals are greatly affected by them, largely through their effect on PGHS and autacoid synthesis. It is noteworthy that oxylipin synthesis in plants and animals follow parallel pathways, despite the involvement of different enzymes and mechanisms. It is interesting that salicylates act by blocking the same step of the pathway, yet do this by inhibiting two structurally different enzymes. In animals it is the non-P450 cyclooxygenase activity of PGHS (1), whereas in plants it is the peroxidase activity of AOS, a cytochrome P450. It appears that the control of autacoid biosynthesis across a wide spectrum of organisms is affected by salicylates at the same point in the pathway.