Acetylation of Human Prostaglandin Endoperoxide Synthase-2 (Cyclooxygenase-2) by Aspirin"

Aspirin (acetylsalicylate) treatment of human (h) prostaglandin endoperoxide H synthase (PGHS)-l expressed in cos-1 cells caused a time-dependent inactiva- tion of oxygenase activity. Aspirin treatment of hPGHS-2 produced an enzyme which retained oxygenase activity but formed exclusively 15-hydroxy-5,8,11,13-eicosatet- raenoic acid (1CHETE) instead of PGH,. The 16-HETE was exclusively of the 1SR configuration. The K,,, values for arachidonate of native and aspirin-treated hPGHS-2 were about the same suggesting that arachidonate binds to both aspirin-treated and native hPGHS-2 in a similar manner. If, as expected, the formation of 1SR-HETE pro- ceeds through abstraction of the 13proS hydrogen from arachidonate, 0, insertion must occur from the same side as the hydrogen abstraction; with all other lipoxy- genases and cyclooxygenases, 0, addition is antarafacial. When microsomal hPGHS-2 was incubated with [u~etyZ-'~C]aspirin, the enzyme was acetylated. An S516A mutant of hPGHS-2, which retains enzyme activity, was not acetylated. This indicates that Ser-616 is the site of aspirin acetylation of hPGHS-2; this residue is homolo-

PGHS-1 and -2 have quite different patterns of expression.
The cyclooxygenase activity of PGHS-1 had been thought to be the pharmacological target of aspirin and related nonsteroidal anti-inflammatory drugs (NSAIDs) (28,29). However, PGHS-2 is induced under conditions of inflammation (13, 25) and recent evidence suggests that this second isozyme may be the target of NSAIDs acting in their anti-inflammatory capacities (20, 30,31). Accordingly, it is important to understand how various NSAIDs interact with PGHS-2. Previous studies have shown that aspirin and related NSAIDs inhibit the cyclooxygenase but not the peroxidase activities of both PGHS-1 and -2 (3,3234). However, the two isozymes interact differently with different NSAIDs (3,341. The situation with aspirin is particularly intriguing. Aspirin causes a time-dependent inhibition of murine PGHS-2 converting this enzyme to a form which catalyzes the production of 15-hydroxy-eicosatetraenoic acid (15-HETE) instead of PGH, from arachidonate (34, 35). In this study we have determined the stereochemistry of the 15-HETE which is produced. We have also characterized the biochemical alteration induced by treatment of hPGHS-2 with aspirin. EXPERIMENTAL PROCEDURES Materials-Dulbecco's modified Eagle's medium was from Life Technologies Inc. Fetal calf serum and calf serum were obtained from Hyclone. Chloroquine, bovine hemoglobin, hematin, guaiacol, acetylsalicylic acid, and penicillin G were from Sigma. Arachidonic acid, prostaglandin D,, E,, and F,,, and 15-HETE were purchased from Cayman Chemical Co. [l-'4ClArachidonic acid (40- Bio-Rad. DEAE-dextran was purchased from Pharmacia Biotech Inc. CsCl was from Var-Lac-Oid Chemical Co. The oligonucleotides used as primers for preparing the mutants of human PGHS-2 and for DNA sequencing were prepared by the Michigan State University Macromolecular and Sequencing Facility. The mammalian expression vector pOSML (3) was derived from pMT2 (36) as described previously (3). pcDNA vectors containing sequences for human PGHS-1 and -2 were from Timothy Hla of the American Red Cross, Bethesda, MD (14). Different plasmid constructions used for these experiments included pSVT7-PGHS-l0, (37), in which ovine PGHS-1 cDNAwas subcloned into the SalI site of the polylinker region of pSVT7; and pOSML-hPGHS-1 and pOSML-hPGHS-2, in which human PGHS-1 and human PGHS-2 cDNAs, respectively, were subcloned into the SalI site of pOSML (3). Preparation of PGHS-2 Mutants-The cDNAs containing the coding regions for each of the human PGH synthase isozymes (-1.8 kilobase) with unique SalI restriction sites in both the 5'and 3"untranslated sequences were as described previously (3). Mutants of human PGHS-2 (S516A, S516N, S516Q, and S516 M) were prepared starting with M13 mpl9-human PGHS-2, which contains a 1.8-kilobase SalI fragment encoding the native human PGHS-2, according to the method of Kunkel et al. (38) using a Bio-Rad kit essentially as described previously (8, 39,40). Table I shows the oligonucleotides used for the preparation of each of these mutants. Phage samples were sequenced using the dideoxy method (41) to identify mutants. The 1.8-kilobase insert from the replicative form of M13 mpl9-human PGHS-2 containing the desired mutation was isolated after digestion with SalI restriction enzyme and subcloned into the SalI site of pOSML (3). The correct orientation of the insert was determined by restriction digestion with PstI. Plasmids used for transfections were purified by CsCl gradient ultracentrifugation.
Acetylation of PGHS Isozymes by ~a~etyl-14ClAspirin-Microsomal preparations of various PGHS isozymes (250 pl) were incubated with 1 m~ [a~etyl-l-'~C]aspirin dissolved in ethanol (1% final concentration) and 80-111 aliquots were removed after 0,l-, or 2-h incubations at 37 "C. Reactions were terminated by the addition of 2 ml of ice-cold acetone, protein was collected by centrifugation, and washed twice with cold acetone. After evaporation of residual acetone under a stream of N, gas, the protein pellet was solubilized in 150 pl of electrophoresis loading buffer (0.2 M Tris-HC1, pH 6.8, containing 2.3% SDS, 5% 2-mercaptoethanol, and 10% glycerol). Aliquots (10-20 pl) were subjected to SDS-PAGE on 10% polyacrylamide gels containing 0.1% SDS, with a 5% stacking gel and the discontinuous buffer system of Laemmli (42). Radiolabeled bands were detected after fluorography of the gel using EN3HANCE@ according to the specifications of the manufacturer. Autoradiography of the dried fluor-impregnated gel was performed using Kodak XAR-5 film (preflashed in order to reach an optical density of 0.15 above the background (43)) at -80 "C for periods up to 40 days. Assay of Cyclooxygenase Actiuity-The cyclooxygenase activities were assayed on microsomal preparations made the same day and were measured by monitoring the initial rate of 0, uptake at 37 "C (200 p Oq) using an 0, electrode as described previously (44). A typical assay mxture contained 3 ml of 0.1 M Tris-HC1 buffer, pH 8.0,l m~ phenol, 85 pg of hemoglobin (as a source of heme), 100 p~ arachidonate, and 250 pg of microsomal protein. Reactions were initiated by adding the microsomal enzyme preparations in a volume of 50 pl. K,,, values for arachidonate were determined using concentrations of arachidonate ranging from 1 to 100 p~.
Assay of Peroxidase Actiuity-The peroxidase activities of microsomal preparations were measured spectrophotometrically, at room temperature, essentially as described by Marnett et al. (45). The reaction mixture contained 0.1 M Tris-HC1 buffer, pH 7.2, 5.6 m~ guaiacol, 50-250 pg of microsomal protein, 1 p hematin (added in dimethyl sulfoxide, 1.5% final concentration), and 0.4 m~ H,O, in a total volume of 0.9 ml. The reactions were initiated by the addition of H,O, and the progress of the oxidation of guaiacol to 3,3'-dimethoxy diphenol-4,4'-quinone was monitored at 436 nm (45).
Western Tkansfer BlottingSolubilized microsomal membranes were resolved by one-dimensional SDS-PAGE and transferred electrophoretically to BA85 nitrocellulose filters (0.45 pm) essentially as described previously (40,46). For detection of human PGHS-2 by enhanced chemiluminescence, filters were incubated for 1-2 h with a 1:2000 dilution of a monospecific rabbit anti-human PGHS-2 serum (47). The filters were washed and incubated with a 1:2000 dilution of goat anti-rabbit IgG horseradish peroxidase. The filters were again washed and incubated for 1-5 min with Amersham ECL Western blotting detection reagents.
The filters were immediately blot-dried and exposed to XAR-5 film.
Assay of Prostaglandins Synthesized by Tkansfected Cos-1 Cells-Forty hours following transfection, cos-1 cells were treated in the presence or absence of aspirin (500 p~) added directly to the medium bathing the cells. Following a 40-min incubation, the cells were harvested in phosphate-buffered saline (3 mudish) using a rubber policeman, collected by centrifugation at 1000 x g for 5 min, and resuspended in Dulbecco's modified Eagle's medium without serum (0.5 ml of Dulbecco's modified Eagle's mediudthree dishes) before incubation with l-["C]ar- Aspirin 1500 p~i was addrd ( + I or not (-) dirrctly to the culture medium of cos-1 cells following transfection with a n expression vector containing either human I'GHS-1 or -2 or following transfection without any DNA (sham 1. After incuhation with aspirin, cells wrrr harvested and washed hrfore incuhation with I1'Clarachidonate (60 pw) for 15 min a t 37 "C. as descrihrd under "Experimen-  achidonate (60 p~) for 15 min a t 37 "C. After incuhation. cells were removed by centrifugation, and the radioactive products present in the supernatant were separated and analyzed by thin layer chromatogra-

Differential Effects of Aspirin on ProstaRlnnrlin Enrloprr-
oxide Formation by Human PGH Svnthaws-1 and -2 " H u m a n ( h ) PGHS-1 and PGHS-2 w e r r e x p r e s s e d t r a n s i r n t l y i n cos-1 cells and microsomal membranes isolated from t h e t r a n s f r c t e d cells w e r e a s s a y e d for t h e i r s e n s i t i v i t y to inhihition by aspirin hibited the cyclooxygenase activity of hPGHS-1 following a 1-h incubation at 37 "C ( Fig. 1, panel A ); in the absence of aspirin the cyclooxygenase activity of hPGHS-1 was stable. When hPGHS-2 was treated with aspirin (1 mM), the initial rate of oxygenation of arachidonate dropped by about 508 after a 1-h incubation; however, a simple control incubation of hPGHS-2 at 37 "C for 10 min also caused the oxygenase activity to fall by 30%. In some experiments, the inhibition of hPGHS-2 was more pronounced, but there was always detectable oxygenase activity after aspirin treatment. The profiles of arachidonatederived products made by intact cos-1 cells transfected with one of the two PGHS isoforms and treated with or without aspirin (500 p~) for 40 min at 37 "C is shown in Fig. 2. Sham-transfected cos-1 cells did not transform arachidonate to products. Cos-1 cells transfected with either hPGHS-1 or hPGHS-2 synthesized essentially the same products; the major radiolabeled bands comigrated with standard PGD,, PGE,, and PGF2, (Fig.   2). After treatment of the cells expressing hPGHS-1 with aspirin, the synthesis of prostaglandins was almost completely abolished. Aspirin treatment of cells expressing hPGHS-2 resulted in almost complete inhibition of prostaglandin synthesis, but there was a substantial increase in the synthesis of a product which comigrated with 15-HETE. Analysis of the stereochemistry of the 15-HETE by chiral HPLC column indicated that 297% of the 15-HETE was in the R configuration (Fig. 3).

Acetylation of Human PGH S-ynthases-1 and -2 hy Aspirin-
Previous studies on purified PGHS-1 obtained from sheep vesicular glands have demonstrated that aspirin inhibits the cyclooxygenase activity of this enzyme via irreversible acetylation of a unique serine residue, the "active site" serine located at position 530 ( 8 ) . There is an homologous serine residue (Ser-516) in hPGHS-2 in a region which shares about 70% identity TAIII.F: 11

NI). not detrrmined.
with hPGHS-I. First, to drtrrminr if hPGHS-I and -2 nrr acetylated by aspirin, microsomal prrparatinns of c o s -] crlls expressing either ovinc or human PGHS-I or human P(;flS-2 were incubated with [ncFbl-l-"C]aspirin ( 1 mv 1, and thr rndiolabeled proteins werc srparatrd hy SDS-PAGE a n d a n n l y z r d by fluorography (Fig. 4). iVith microsomrs from crlls transfected with ovine P(;HS-l (as compnrrd to sham-transfrctcd cells), one unique radiolahclrd protrin with an appnrcnt molecular mass of 70 kDa was prrscnt as rxprctrd for acrtylation of the ovine PGHS-I f4J. Similar rrsults wrrr ohtaincd with microsomes from cells transfrctrd with hPGHS-I; a singlc radiolabeled protein hand was prrsrnt which comipatrd with thr one observed with ovinr PGHS-I. With microsomrs from crlls transfected with hPGHS-2, thrrr radiolahclrd protrin hands with apparent molrcular massrs of 70. 72, and 74 k I h wrrr observed. Western transfrr hlotting of hPCHS-2 also yicldrd a triplet of immunorrnctivr protrins of thr samc mohilitirs; prctreatment of the sample with endoglycosidasc H prior to SIX-PAGE yielded a single immunorcactivr protcin with M, 65,000 corresponding to t h r expcctcd mohility of drglvcosylntcd hPGHS-2.2 Our results indicatr that hPGHS-2 is ncrtvlntd hv aspirin.
To determine if the sitr of acrtylation of hPGHS-2 hy aspirin involved the serine rrsidue homologous to that acrtylatcd in PGHS-1 we suhstituted an alaninr residuc for Srr-5lfi in hPGHS-2 by site-directed mutagenesis and thrn comparcd thr acetylation of native hPGHS-2 and thr S516A mutant of hPGHS-2 by ~~c r~~v l -l -~~C l a s p i r i n ( F i g . 5 ) . With mcmhrnnrs prepared from cos-1 cells transfrctcd with nativr hPGHS-2. w r observed the expected three proteins with rstimatrd molrcular masses of 70, 72, and 74 kDa as comparrd to sham-transfrctcd cells. In contrast, the aspirin laheling pattern ohsrrvcd with the S5lGA mutant of hPGHS-2 was indistinmishahlr from thc one observed with sham-transfrctrd crlls. This rrsult rstahlishes that the S516A mutant cannot hr acctylatcd hy aspirin and suggests that Ser-516 is t h r single sitc of acrtylntion of hPGHS-2 by aspirin. We next mrasurcd the activity of t h r S5lGA hPGHS-2 to determine if its insensitivity toward aspirin acetylation was simply the result of a dramatic stnlrtrlrnl change. Fig. 6 shows an autoradiogram of the products synthesized from ['4CJarachidonate by cos-1 cells transfected with native hPGHS-1 and with the S516A mutant. The synthesis of prostaglandins by cells expressing native hPGHS-2 is almost completely abolished by aspirin and 15R-HETE is produced as the major product. Cells expressing the S516A mutant of hPGHS-2 do synthesize about the same amount and type of products as native hPGHS-2. However, aspirin had no effect on product formation by the S516A mutant of hPGHS-2. When assayed using an 0, electrode the S516A mutant was found to exhibit cyclooxygenase activity comparable to native hPGHS-2 (Table 11); moreover, the K,,, for arachidonate for S516A hPGHS-2 was 2.9 p~, a value comparable to that obtained with native hPGHS-2 (5.4 p~) (3). Finally, the peroxidase activities of hPGHS-2 and S516A hPGHS-2 were found to be quite similar to one another.
We conclude from these results that hPGHS-2 is acetylated by aspirin at Ser-516.
We made three additional substitutions of Ser-516 in hPGHS-2 replacing the serine with asparagine, glutamine, and methionine residues. The purpose of these experiments was to determine the effect of the size of the side chain at position 516 of hPGHS-2 on cyclooxygenase and peroxidase activities ( Table  11). The S516N mutant, which contains a side chain roughly isosteric with an acetylated serine, retains activity and the products made from arachidonate by the cells expressing this mutant are the same as for native hPGHS-2 (Fig. 7). The K,,, of this mutant for arachidonate was determined to be 3.6 p~. The S516Q mutant which has a side chain one methylene group larger than the S516N mutant lacked cyclooxygenase activity but retained peroxidase activity (see Table 11 expressing the S516Q mutant failed to synthesize prostaglandin or hydroxy acid products from [l-"Clarachidonate (see Fig. A).
An S516M mutant which also has a side chain significantly larger than an acetylated serine was active but exhibited only 22% of the oxygenase activity of the native enzyme (see Table  11). The K,,, value of this mutant for arachidonate is -600 VM. which is roughly 100 times that observed for the native enzyme. An unexpected result was that the S5l6M mutant enzyme, when expressed in cos-1 cells, synthesized 15-HETE ( Fig. 9) which, like that formed by the aspirin-acetylated hPGHS-2, was of the 15R configuration (data not shown). There were some differences in the nature and the relative abundance of the products made by acetylated hPGHS-2 w r s u s the S516M mutant of PGHS-2. The relative amount of 15R-HETE produced by the acetylated hPGHS-2 was greater than that observed for the S516M mutant, and PGF,, was not produced in the case of the S516M mutant. All the mutants we made at position 516 retained peroxidase activity (Table 11). Western blot analysis with rabbit monospecific IgGs directed against human PGHS-2 revealed three bands of 70. 72. and 74 kDa which comigrated with the ones observed for the native PGHS-2 (data not shown), indicating that mutations at position 516 did not significantly alter the structure of the enzyme as compared to the native enzyme.

DISCUSSION
In the present study, we have demonstrated that aspirin acetylates hPGHS-2 expressed in cos-1 cells. Three radiolabeled proteins not observed in sham-transfected cells were observed when microsomal membranes from cos-1 cells ex- pressing hPGHS-2 were treated with [ocefyl-'"C]aspirin; moreover, the radiolabeled proteins had the same mobility as immunoreactive hPGHS-2 expressed in cos-l cells.' Aspirin was unable to acetylate the S516A mutant of hPGHS-2. S516A hPGHS-2 exhibits cyclooxygenase (and peroxidase) activity and catalytically active forms of PGHS-1 are known to be acetylated by aspirin. Accordingly, studies with the S516A mutant of hPGHS-2 support the concept that native hPGHS-2 is acetylated at a single site by aspirin-Ser-516. Moreover, the Ser-516 of hPGHS-2 is homologous to the ovine PGHS-1 active site Ser-530 which is known to be acetylated by aspirin (8,49). In contrast to the results with ovine and human PGHS-1 (Fig. 3), acetylation of human PGHS-2 by aspirin produces three labeled proteins at 74, 72, and 70 kDa instead of one band at 70 kDa. These three bands are also detected by Western blot analysis with monospecific rabbit IgGs directed against hPGHS-2 and represent forms of hPGHS-2 which are variously glycosylated.' The fact that each of these bands is reactive with aspirin suggests that each of these three forms of hPGHS-2 retains cyclooxygenase activity.

PGHS
Previous studies with murine PGHS-2 had shown that this isozyme formed 15-HETE when the enzyme was treated with aspirin. Related studies on a novel form of ovine PGHS, most likely ovine PGHS-2, had also established that aspirin treatment increases 15-HETE formation and that the 15-HETE which was produced had a stereochemical composition of 3 : l of R:S (35). Our studies indicate that all of the 15-HETE formed by aspirin-treated hPGHS-2 is in the R configuration. 15-HETE formed via lipoxygenases is of the S configuration (50). Thus, our results indicate that one can distinguish between

15-HETE products formed via lipoxygcnasc
[ ' c r s u s thosr formed via aspirin-treated PGHS-2 hy determining the strreochemistry. In fact, we speculate that this is probnhlv what occurred in the system examined hy Holtzman and co-workers (35) which prohahlv contained a mixture of 15-lipoxygcnasr and ovine PGHS-2.
The mechanism by which aspirin-acetylated hPGHS-2 forms 15R-HETE is not clear. With cyclooxygenase and lipoxygrnasc O? insertion reactions studied to datr, 0, insertion (rcurs antarafacial to hydrogen ahstraction (50-55,. If antarafacial OL insertion occurs with aspirin-acrtylated hPGHS-2. the 13proii hydrogen would need to be r e m o v d Previous studirs with cyclooxygenase have indicated that it is the l3pro.q hydrogen which is abstracted (51). Recause the K,, for arachidonate for native and aspirin-treated hPGHS-2 arc essrntially the same. we suspect that arachidonatr hinding occurs similarly with both forms of the enzyme; if so. removal of the 1.3pro.q hydroKen would occur. Given this scenario for hydrogen ahstraction a n d the fact that the 0, is inserted in the K configuration, our results indicate that acetylation must alter the orientation of incoming 0, so that insertion occurs from the same side (synfacial) of the carbon chain as hydrogen ahstraction. ,Marcover. aspirin acetylation of PGHS-2 also appears to prevent the oxygenation of arachidonate at C-11.
Studies with ovine PGHS-1 have indicated that Srr-5.30 is uniquely acetylated by aspirin ( 8 ) . An alanine substitution at that position had no effect on the cyclooxygenasr activity of the enzyme, but increasing the sizc of the sidc chain at position 5.70 interfered with arachidonate binding. For example, thr S530N and S530Lmutantsofovine PGHS-1 lacked cyclwxvgenasc hut retained peroxidase activity (8). Our mutational analysis of the Ser-516 residue of hPGHS-2, which is homologous to Scr-530 of ovine PGHS-1, indicates that increasing the size of the side chain at this position can also eliminate cyclooxggenase activity. However, larger groups are required in the case ofhPGHS-2 as compared to ovine PGHS-1. For example, the S516N mutant of hPGHS-2 is active whereas the SSSON mutant of ovine PGHS-1 is not. Elimination of cyclooxygenase activity of hPGHS-2 requires that the larger glutamine group be used to replace Ser-516. The present results obtained with Ser-516 mutants of hPGHS-2 when compared with those obtained earlier with ovine PGHS-1 suggest that the cyclooxygenase active site pocket of PGHS-2 is larger than that of PGHS-1. This view is also consistent with studies of competitive inhibition of human and murine PGHS-1 and PGHS-2 by various NSAIDs other than aspirin (3, 34).
The expected effect of ingestion of an anti-inflammatory dose of aspirin would be the inhibition of prostaglandin synthesis and production of 15R-HETE. A possible anti-inflammatory effect of 15R-HETE needs to be further analyzed, although i t should be noted that other anti-inflammatory drugs do not cause murine PGHS-2 to form 1.W-HETE (34). It is also worth noting that R-HETEs are generally more potent chemotactic agents than their S counterparts (56, 57).
In conclusion, we have shown that human PGHS-2 is acetylated by aspirin on a single serine residue, Ser-516, the residue homologous to the active site serine (Ser-530) of ovine PGHS-1. The acetylation causes the enzyme to make 15R-HETE. Mutational analysis of Ser-516 suggests that the active site pocket of PGHS-2 is larger than that of PGHS-1. The Met-516 mutant is like the acetylated wild-type enzyme in that both make 15R-HETE.