Identification and absolute configuration of dihydroxy-arachidonic acids formed by oxygenation of 5S-HETE by native and aspirin-acetylated COX-2.

Biosynthesis of the prostaglandin endoperoxide by the cyclooxygenase (COX) enzymes is accompanied by formation of a small amount of 11R-hydroxyeicosatetraenoic acid (HETE), 15R-HETE, and 15S-HETE as by-products. Acetylation of COX-2 by aspirin abrogates prostaglandin synthesis and triggers formation of 15R-HETE as the sole product of oxygenation of arachidonic acid. Here, we investigated the formation of by-products of the transformation of 5S-HETE by native COX-2 and by aspirin-acetylated COX-2 using HPLC-ultraviolet, GC-MS, and LC-MS analysis. 5S,15S- dihydroxy (di)HETE, 5S,15R-diHETE, and 5S,11R-diHETE were identified as by-products of native COX-2, in addition to the previously described di-endoperoxide (5S,15S-dihydroxy-9S,11R,8S,12S-diperoxy-6E,13E-eicosadienoic acid) as the major oxygenation product. 5S,15R-diHETE was the only product formed by aspirin-acetylated COX-2. Both 5,15-diHETE and 5,11-diHETE were detected in CT26 mouse colon carcinoma cells as well as in lipopolysaccharide-activated RAW264.7 cells incubated with 5S-HETE, and their formation was attenuated in the presence of the COX-2 specific inhibitor, NS-398. Aspirin-treated CT26 cells gave 5,15-diHETE as the most prominent product formed from 5S-HETE. 5S,15S-diHETE has been described as a product of the cross-over of 5-lipoxygenase (5-LOX) and 15-LOX activities in elicited rat mononuclear cells and human leukocytes, and our studies implicate cross-over of the 5-LOX and COX-2 pathways as an additional biosynthetic route.

Identifi cation and absolute confi guration of dihydroxy-arachidonic acids formed by oxygenation of 5 S -HETE by native and aspirin-acetylated COX-2 each as described ( 12 ). The products were extracted using a Waters HLB cartridge and analyzed by RP-HPLC using a Waters Symmetry C18 5-m column (4.6 × 250 mm) eluted with a gradient of acetonitrile/water/acetic acid programmed from 20/80/0.01 (by vol) to 70/30/0.01 (by vol) within 20 min at 1 ml/min fl ow rate. The elution profi le was monitored using an Agilent 1200 diode array detector coupled on-line to a Packard Radiomatic A100 Flo-one radioactive detector. The by-products eluting at 19.8 and 20.5 min retention time were collected, extracted from HPLC solvent, and stored in methanol at Ϫ 20°C until further analysis.

Reaction of aspirin-acetylated COX-2 with 5 S -HETE
Recombinant human COX-2 (0.5 M fi nal concentration) was diluted in 1 ml of 100 mM Tris-HCl buffer pH 8.0 and treated with 2 mM aspirin in a 37°C water bath for 30 min ( 15 ). A control reaction incubated with arachidonic acid and analyzed by LC-MS before and after treatment showed >95% inhibition of PG formation. The buffer was supplemented with hematin (1 M) and phenol (500 M), and 30 g of 5 S -HETE were added. After 5 min at room temperature, 25 l of methanol were added, the mixture was acidifi ed to pH 4 with glacial acetic acid, and loaded onto a preconditioned Waters Oasis HLB cartridge. After washing with water, the products were eluted with methanol. The 5,15-diHETE product was isolated using RP-HPLC as described above for the native enzyme. 5 S ,15 S -diHETE was synthesized by reaction of 5 S -HETE with the LOX-1 isozyme from soybean seeds ( 16 ). To 3 ml of 100 mM K 2 HPO 4 buffer pH 10 were added 100 g of 5 S -HETE and 1 l of soybean lipoxygenase solution ( ‫ف‬ 25,000 units; Sigma). 5 R ,15 S -diHETE was synthesized using 5 R -HETE (100 g) as substrate and 3 l of soybean lipoxygenase solution. After 1 min reaction time, the solution was acidifi ed (pH 4) and extracted with methylene chloride. The organic extract was evaporated, dissolved in methanol, and treated with 200 g of triphenylphosphine (TPP) for 15 min at room temperature. 5 S ,15 S -DiHETE and 5 R ,15 S -diHETE were isolated by RP-HPLC using a Waters Symmetry C18 column (4.6 × 250 mm) eluted with a solvent of methanol/water/acetic acid (80/20/0.01, by vol) at 1 ml/min fl ow rate and UV detection at 235 nm. 5 S ,15 R -DiHETE was synthesized by reaction of 15 R -HETE with recombinant human 5-LOX. For the enzymatic transformation, a pellet of Sf9 insect cells expressing 5-LOX ( ‫ف‬ 300 l) was sonicated and transferred to 1 ml of PBS containing 2 mM CaCl 2 and 1 mM ATP. 15 R -HETE (50 g) was added and the reaction was allowed to proceed for 15 min at room temperature. The reaction was terminated by the addition of 250 l of methanol and 10 mg of NaBH 4 . After 15 min at room temperature, the pH was adjusted to 3 using 1 N HCl, and the products were extracted using methylene chloride. 5 S ,15 R -DiHETE was purifi ed by RP-HPLC as described above for the 5 S ,15 S -diastereomer. 5 S ,11 R -DiHETE was synthesized by reaction of 5 S -HETE with the recombinant linoleic acid 9 R -LOX from Anabaena sp. PCC7120 expressed in Escherichia coli , a gift from Alan R. Brash at Vanderbilt University ( 17 ). The substrate, 5 S -HETE (100 g), was added to 3 ml of 100 mM Tris-HCl pH 7.5 containing 150 mM NaCl and 0.01% CHAPS. The reaction was initiated by addition of 1 l of the purifi ed Anabaena 9 R -LOX. After 3 min reaction time, the solution was acidifi ed to pH 4 using 1 N HCl, and the products were extracted using a 30 mg Waters HLB cartridge and eluted with methanol. After evaporation of the solvent, the residue was dissolved in 100 l of methanol, and the products were reduced with 150 g TPP at room temperature for 15 min. activity following treatment with aspirin, acetylated COX-2 gains a novel catalytic activity and forms 15 R -HETE as the sole product ( 10,11 ).

Synthesis and isolation of diHETE reference compounds
The major product formed by oxygenation of the 5-lipoxygenase product, 5 S -HETE, with COX-2 is a bicyclic di-endoperoxide with structural similarities to the arachidonic acid derived PGH 2 ( 12 ). The most significant difference between the two endoperoxides is that the typical cyclopentyl ring of PGH 2 , comprised of carbons 8 through 12, is extended to a seven-membered ring by insertion of a peroxide bridge from C8 to C12 in the 5-HETE derived di-endoperoxide. In addition, the di-endoperoxide contains two hydroxy groups, one at carbon 5 stemming from the 5 S -HETE substrate and the other at C15, equivalent to the 15-hydroxy in the prostaglandins. The relative and absolute stereochemistries of carbons 9, 11, and 15 are the same in PGH 2 and the di-endoperoxide, i.e., 9 S , 11 R , and 15 S ( 13 ).
Here, we report the structural identifi cation and absolute confi guration of two by-products of the COX-2 reaction with 5 S -HETE. In addition, we analyzed the reaction of acetylated recombinant COX-2 with 5 S -HETE. Finally, formation of diHETEs from exogenous 5 S -HETE was confi rmed to be dependent on COX-2 in two mouse cell lines, RAW264.7 and CT26.

Materials
Arachidonic acid was purchased from NuChek Prep, Inc. (Elysian, MN), lipopolysaccharide (LPS) (serotype 0111:B4) was from Calbiochem, and RAW264.7 and CT26 cells were obtained from ATCC (Manassas, VA). 5 S -HETE was prepared by chemical synthesis from arachidonic acid as described ( 13 ). 15 R -HETE and 11 R -HETE were prepared through vitamin E-controlled autoxidation of arachidonic acid methyl ester and purifi ed by consecutive RP-, straight phase-, and chiral phase HPLC [Chiralpak AD ( 14 )], and a fi nal step of mild hydrolysis of the methyl ester using KOH.

Cell culture
RAW264.7 cells were cultured in DMEM and grown at 37°C in an atmosphere of 5% CO 2 . Cells of passages 5 and 6 only were used. Cells were stimulated by treatment with 100 ng/ml LPS and 10 units/ml of IFN-␥ for 6 h to induce expression of COX-2. CT26 cells were cultured in RPMI 1640 medium. 5 S -HETE, 5 g dissolved in 1 l of ethanol, was added to ‫ف‬ 70% confl uent cells in 100 mm dishes, and after 10 min at 37°C, the culture medium was removed, acidifi ed to pH 4, and extracted using a 30 mg Waters HLB cartridge. Products were eluted from the cartridge with methanol, evaporated, and dissolved in 50 l of LC-MS solvent A. CT26 and RAW264.7 control cells were not treated with LPS. In some experiments, CT26 and activated RAW264.7 cells were treated with 2 mM aspirin, respectively (from a 40 mM stock solution in DMSO), or with 10 M NS-398 (from a 10 mM stock solution in ethanol) 30 min prior to incubation with 5 S -HETE.

Reaction of recombinant COX-2 with 5 S -HETE
The reaction of 5 S -HETE (120 g total; containing 300,000 cpm of [1-14 C]5 S -HETE) with recombinant human COX-2 was performed in four separate 2 ml reactions with 30 g substrate erated in the negative ion mode. User modifi ed parameters of sheath and auxiliary gas pressures, temperature, and voltage settings were optimized using direct infusion of a solution of PGD 2 . A Waters Symmetry Shield C18 3.5 m-column (2.1 × 150 mm) was eluted with a linear gradient of acetonitrile/water, 10 mM NH 4 OAc (5/95, by vol; solvent A) to acetonitrile/water, 10 mM NH 4 OAc (95/5, by vol) at a fl ow rate of 0.2 ml/min within 10 min. Negative ion collision-induced dissociation (CID) mass spectra of the standards of PGD 2 , 5 S -HETE, 5 S ,15 S -diHETE, and 5 S ,11 R -diHETE were obtained. The fragmentation patterns were used to establish ion transitions for analyses in the selected reaction monitoring (SRM) mode. The following transitions were monitored: for PGD 2 and PGE 2 : m/z 351 → 271; 5-HETE: m/z 319 → 115; 5,15-diHETE: m/z 335 → 201; and 5,11-diHETE: m/z 335 → 183. Relative levels of prostaglandins and diHETEs between treatments were calculated using peak areas of the signals in the SRM chromatograms.

Reaction of native and acetylated COX-2 with 5 S -HETE
RP-HPLC analysis of the transformation of [1-14 C]5 S -HETE by recombinant COX-2 shows one main product that was identifi ed previously as a highly oxygenated diendoperoxide ( 12 ), in addition to two minor, less polar peaks designated I and II representing by-products of the reaction ( Fig. 1A ). When COX-2 was treated with aspirin prior to incubation with 5 S -HETE, one major product (III) was formed with retention time similar to peak I in the untreated enzyme ( Fig. 1B ). Both I and III had a characteristic UV spectrum with a max at 243 nm that was readily identifi ed as 5,15-diHETE ( Fig. 1C ) ( 18 ). The UV spectrum of peak II had a maximum at 238 nm with shoulders around 228 nm and 247 nm ( Fig. 1C ). The retention time and UV spectrum of II implicated that the product also contained two hydroxy groups and conjugated diene moieties.
The products I and II were isolated using RP-HPLC and further purifi ed as the methyl ester derivatives using SP-HPLC. GC-MS analysis in the EI mode (70 eV) of the hydrogenated, TMS-ether derivatives confi rmed the identifi cation of the fi rst peak I as 5,15-diHETE. Characteristic ␣ -cleavage fragments were found at m/z 203 (55% relative intensity) and m/z 311 [after loss of O-trimethylsilyl (OTMS); 9%] for the 5-hydroxy, and at m/z 173 and 341 (after loss of OTMS) (56% and 7%, respectively) for the 15-hydroxy group; the base peak was at m/z 73. Peak III from the aspirin-acetylated COX-2 reaction was identifi ed as 5,15-diHETE based on identical UV spectra and retention times on RP-HPLC, and in addition to subsequent experimental evidence as described below.
Product II gave a very weak [M + ] ( m/z 502) and [M-CH 3 + ] ( m/z 487) ion, with characteristic ␣ -cleavage fragments at m/z 203 (42%) and m/z 311 (after loss of OTMS; 4%) indicating a 5-hydroxy group, and at m/z 229 (38% relative intensity) and m/z 285 (after loss of OTMS) (5%) indicative of a 11-hydroxy group. The LC-ESI mass spectrum confi rmed the molecular weight as 336 and also gave a major fragment at m/z 183 and a minor fragment at m/z 115, compatible with two hydroxyls at carbons 5 and 11 5 S ,11 R -DiHETE was isolated using RP-HPLC conditions as described above for 5 S ,15 S -diHETE. HPLC-purifi ed 5 S ,11 R -diHETE was dissolved in CDCl 3 for NMR analysis using a Bruker AV-II 600 MHz spectrometer equipped with a cryoprobe. Chemical shifts are reported relative to the signal for residual CHCl 3 at ␦ 7.25 ppm.
A mixture of 5,11-diHETE diastereomers was synthesized by autoxidation of racemic 11-HETE. Three 200 g aliquots of 11-HETE were evaporated in small plastic tubes and placed in an oven at 37°C. After 2 h, the samples were dissolved in 50 l of methanol, treated with triphenylphosphine (TPP), and analyzed using RP-HPLC. The diastereomers eluted as a single peak and purifi cation was performed as described for 5 S ,15 S -diHETE.

CD spectroscopy
Aliquots of ‫ف‬ 20 g each of 5 S -HETE, 15 S -HETE, 15 R -HETE, 11 S -HETE, 11 R -HETE, and the enzymatically synthesized standards of 5 S ,15 S -diHETE, 5 R ,15 S -diHETE, and 5 S ,11 R -diHETE were treated with ethereal diazomethane for 30 s, evaporated, and dissolved in 50 l of dry acetonitrile. To the solution was added 1 l of 1,8-diazabicyclo [5.4.0]undec-7-ene (DBU), and a few grains each of 4-dimethylaminopyridine (DMAP) and 2-naphthoylchloride. The reaction was carried out at room temperature overnight, the solvent was evaporated, and the residue was dissolved in methylene chloride and washed with water twice. Purification of the methyl ester, 2-naphthoyl derivatives of the HETEs and diHETEs was achieved by RP-HPLC using a Waters Symmetry C18 column (4.6 × 250 mm) eluted with a solvent of methanol/water/acetic acid (95/5/0.01, by vol) at 1 ml/min fl ow rate and UV detection at 235 nm. Samples were extracted from HPLC solvent using methylene chloride and dissolved in acetonitrile to an optical density (OD) of 0.75 absorbance unit (AU) for HETE derivatives and OD 1.5 AU for diHETE derivatives (when possible), respectively. Circular dichroism (CD) spectra were recorded using an Aviv Model 215 CD spectrometer at room temperature in a 1 cm pathlength cuvette scanning from 350-200 nm. The 1 H NMR spectrum (600 MHz) of the 2-naphthoate derivatized 5 S ,11 R -diHETE methyl ester was recorded in CD 3 CN, ␦ 1.93 ppm.

GC-MS and LC-MS analysis
For GC-MS analysis, 5,15-diHETE and 5,11-diHETE formed by reaction of COX-2 with 5 S -HETE were purifi ed using RP-and SP-HPLC and methylated using ethereal diazomethane. Hydrogenation was performed in 100 l of ethanol in the presence of palladium/carbon and bubbling with hydrogen gas for 5 min. Trimethylsilyl ethers were prepared using bis (trimethylsilyl)trifl uoroacetamide at room temperature for 1 h. The reagents were evaporated and the samples were dissolved in hexane. GC-MS analysis was carried out in the EI mode (70 eV) using a Thermo-Finnigan DSQ mass spectrometer equipped with a 5 m SPB-1 column (0.1 mm i.d., fi lm thickness 0.25 m) and a temperature program from 100°C, hold 2 min, and then increased to 260°C at 20°C/min. LC-MS was performed using a ThermoFinnigan Quantum Access instrument equipped with an electrospray interface and op- ( Fig. 1D ). Based on UV, GC-MS, and LC-MS analyses, product II was identifi ed as 5,11-diHETE. 1 H NMR and H,H COSY data for product II were recorded using a chromatographically and spectroscopically (UV, LC-MS/MS) identical standard of 5 S ,11 R -diHETE that was prepared as described below. The 1 H NMR spectrum showed eight signals in the double bond region that appeared as a pair of two similar motives of four protons each comprised of the two conjugated cis , trans -dienes (H7: ␦ 6.57 ppm, dd, J = 15.1 Hz/11.0 Hz; H8: ␦ 6.13, dd, J = 11.0; H6: ␦ 5.70, dd, J = 14.9 Hz/6.3 Hz, H9: ␦ 5.55, m; and H13: ␦ 6.51, dd, J = 14.9 Hz/11.4 Hz; H14: 5.96, dd, J = 11.0 Hz; H12: ␦ 5.67, m; H15: ␦ 5.46, m). Two protons attached to carbons bearing a hydroxyl group were located at 4.25 ppm (H11: ␦ 4.25, dt, J = 6.3 Hz/6.1 Hz) and 4.17 ppm (H5: ␦ 4.17, dt, J = 6.2 Hz/6.0Hz). H4 was detected as a cross-peak from H5 in the H,H-COSY spectrum at 1.57 ppm, H3 was a multiplet (1.70 ppm) and was coupled to the triplet signal of H2 at 2.34 ppm ( J = 7.4 Hz). Both protons of H10 were detected as a multiplet at 2.47 ppm, and H16 was a dt signal at 2.17 ppm ( J = 7.6 Hz/7.2 Hz).
The confi guration of C-15 in the 5,15-diHETE products (I and III) and of C-11 in the 5,11-diHETE (II) was established by coelution with corresponding diHETE diastereomers of known confi guration. The confi guration of the 5-hydroxy group in all diHETE products was expected to be unchanged from the starting substrate, 5 S -HETE. Table 1 gives an overview of the diHETE standards prepared as reference compounds. Authentic 5 S ,15 S -diHETE was prepared by reaction of soybean LOX-1 with 5 S -HETE. Synthesis of 5 S ,15 R -diHETE by reaction of 15 R -HETE with the recombinant human 5-LOX gave only a minor yield of product, albeit it was suffi cient to determine the retention times on RP-and SP-HPLC. In addition, the enantiomer 5 R ,15 S -diHETE was prepared by reaction of 5 R -HETE with the lipoxygenase from soybean seeds. 5 S ,15 R -diHETE and 5 R ,15 S -diHETE have indistinguishable retention times on RP-and SP-HPLC.

Synthesis of standards of diastereomeric diHETEs
An authentic standard of 5 S ,11 R -diHETE was prepared by reaction of 5 S -HETE with the recombinant 9 R -LOX from Anabaena sp PCC7120. A mixture of the 5,11-diHETE diastereomers was prepared by thin-fi lm autoxidation of racemic 11-HETE. Initial attempts to prepare 5 S ,11 S -and 5 S ,11 R -diHETEs by reaction of 11 S -HETE and 11 R -HETE, respectively, with the recombinant human 5-LOX did not yield a signifi cant amount of either 5,11-diHETE diastereomer. The assignment of the absolute confi guration of the hydroxy groups in the diHETE standards was confi rmed using CD spectroscopy (see below).

CD-spectroscopy of diHETE standards
We used the exciton-coupled circular dichroism method in order to confi rm assignment of the absolute confi guration of the hydroxy groups in the diHETE standards. This method uses the coupling of two chromophores attached to the chiral center in circular polarized light in order to determine the absolute confi guration from the sign of the

Absolute confi guration of 5,11-diHETEs
Standards for the 5 S ,11 R -and 5 S ,11 S -diHETE diastereomers were prepared by thin-fi lm autoxidation of racemic 11-HETE followed by reduction with triphenylphosphine. 5,11-DiHETE was the almost exclusive diHETE formed, and the diastereomers eluted as a single peak when analyzed by RP-HPLC. Using SP-HPLC, satisfactory resolution of the methyl ester derivatives was achieved ( Fig. 4 ). The fi rst peak comprised of the 5 S ,11 S -and 5 R ,11 R -diastereomers eluted at 17.8 min and the second peak (5 S ,11 R -and 5 R ,11 S -diastereomers) eluted at 18.3 min. The authentic standard of 5 S ,11 R -diHETE prepared using the Anabaena LOX coeluted with the second peak on SP-HPLC and established the elution order. SP-HPLC analysis of 5,11-diHETE from recombinant human COX-2 showed that the confi guration was >98% 5 S ,11 R -diHETE.  Fig. 1A, B . b , c , d diHETEs with the same superscript letter are enantiomers. e This reaction gave a very low yield. f Although 5 R -HETE was readily converted by the Anabaena LOX formation of 5 R ,11 R -diHETE was not observed.
h Agilent Zorbax RX-SIL column (250 x 4.6 mm) eluted with hexane/isopropanol/acetic acid 95/5/0.1 at 1 ml/min fl ow rate. Retention times for the 5,11-diHETEs are for the methyl ester derivatives. The confi guration of C11 in 5 S ,11 R -diHETE formed by reaction of the Anabaena 9-LOX with 5 S -HETE was confi rmed using the same approach. The individual CD spectra of 2-naphthoate-derivatized 5 S -HETE and 11 R -HETE are mirror images of each other, but unexpectedly, the CD spectrum of 5 S ,11 R -diHETE was not a fl at line ( Fig. 6 ). The spectrum showed CEs at 245 nm ( ⌬ +7.6) and 228 Cotton effects (CEs). One of the chromophores in the (di)HETEs is present as the conjugated diene system, and in order to introduce the second chromophore, the hydroxy groups were derivatized to the 2-naphthoate ester ( 19 ). Mono-and di-2-naphthoate derivatives, of the following methyl ester fatty acids were prepared, respectively: 5 S -HETE, 11 S -HETE, 11 R -HETE, 15 S -HETE, 15 R -HETE, 5 S ,15 S -diHETE, 5 R ,15 S -diHETE, and 5 S ,11 R -diHETE.
As expected, pairs of derivatized enantiomers, e.g., 15 S -HETE and 15 R -HETE, gave mirror-image CD spectra ( Fig. 5 ). Furthermore, the CD spectra of all S -confi guration HETEs (5 S -HETE, 11 S -HETE, and 15 S -HETE) gave a positive fi rst Cotton effect (i.e., the CE at the higher wavelength) and a negative second CE. Due to the additive nature of the absorbance in the CD spectrum, the two chiral centers in the diHETEs were expected to result in a CD spectrum that represents the mathematical sum of the spectra obtained for the two individual chiral centers. The spectrum of 5 S ,15 S -diHETE showed increased intensities for the two CEs, although the ⌬ intensities were not doubled when compared with 15 S -HETE, which was likely due to saturation effects at the high concentration measured (1.5 AU in the UV) ( Fig. 5 ). In contrast, for the 5 R ,15 S -diHETE diastereomer, the CEs cancelled each other out and the resulting CD spectrum was an almost fl at line. The CD spectra of the diastereomeric 5,15-diHETEs confi rmed the assignment of the absolute confi guration of the two chiral centers in the 5 S ,15 S -diHETE standard.  Fig. 2 were used. All chromatograms were recorded at UV 235 nm using a diode array detector. can be taken as a measure for the alignment of the chromophores within the conformer, however, gave essentially equivalent values, i.e, 6.7 Hz and 6.8 Hz, respectively. Because SP-HPLC has confi rmed the relative confi guration of the 5 S ,11 R -diHETE ( Fig. 4D ), the question why the corresponding CD spectrum showed slight predominance of the S -confi gurated chiral center remains unexplained.

Formation of diHETEs in RAW264.7 and CT26 cells
RAW264.7 were treated in four different ways and incubated with 4 M 5 S -HETE. We used nonstimulated cells, cells stimulated with LPS only, and LPS-stimulated cells treated with NS-398 or aspirin, respectively. Formation of diHETEs was analyzed using negative ion LC-ESI-MS in the SRM mode ( Fig. 7A ). Both 5,15-diHETE and 5,11-diHETE were detected in RAW264.7 cells activated with LPS and IFN-␥ ( Fig. 7A , upper panel), and their concentration was reduced to 0.5% and 3%, respectively, by incubation nm ( ⌬ -6.7), resembling a weaker version of the CD spectrum of 5 S -HETE with slightly shifted maxima. We hypothesized that the transition moment of the chromophores at C5 gave a stronger spectrum due to more optimal alignment of the two chromophores, thereby overcompensating the spectrum for the chiral center at C11. 1 H NMR determination of the J 5,6 and J 11,12 coupling constants that  respectively ( Fig. 7A , lower panel). Levels of PGD 2 and PGE 2 were reduced to 4% and 30% by NS-398 and aspirin, respectively. PGD 2 and PGE 2 are not completely absent in the inhibitor treated cells because a fraction was formed of the cells with the COX-2 inhibitor NS-398 (10 M) prior to the addition of 5 S -HETE ( Fig. 7A , middle panel). Treatment of RAW264.7 cells with 2 mM aspirin led to about 60% and 50% reduction in 5,15-diHETE and 5,11-diHETE, ( Fig. 8 ) ( 5,26 ). In contrast to the reaction with arachidonic acid, 5 S -HETE reacts only with COX-2; the COX-1 isozyme is inactive with 5 S -HETE ( 12 ). The confi guration of C-15 of 5,15-diHETE was a ‫ف‬ 3.5:1 mixture of 15 S and 15 R , and a similar mixture of 15 S and 15 R confi guration is found in the 15-HETE formed by COX-1 and COX-2 ( 4 ). The confi guration of C-11 in 5,11-diHETE is >98% 11 R , identical to the near exclusive 11 R -confi guration of 11-HETE formed by COX-1 and COX-2 ( 2,3,27 ). We can conclude that the modes of binding of arachidonic acid and of 5 S -HETE in the cyclooxygenase active site must be very similar. There is precise control over the C-11 and C-15 oxygenations in the formation of the prostaglandin endoperoxide and the di-endoperoxide as well as in the formation of the 11 R -HETE and 5 S ,11 R -diHETE by-products, respectively ( 28 ). There is less control of the oxygen insertion and/or less tight binding of the -tail of the fatty acid substrate in the case of the formation of the 15-HETE or 5,15-diHETE by-products. The major difference in catalytic outcome with 5 S -HETE, however, is the insertion of another molecule of oxygen in place of the C8-C12 carbon bond resulting in the formation of two endoperoxide rings. from (endogenous) arachidonic substrate already before addition of the drugs. Furthermore, aspirin showed only modest effi cacy in reducing eicosanoid formation, consistent with previous fi ndings that a high cellular redox state in activated RAW264.7 cells impedes aspirin's ability to covalently modify COX enzymes ( 20 ).
Formation of diHETEs was also analyzed in CT26 mouse colon carcinoma cells incubated with 5 S -HETE. CT26 cells incubated with 5 S -HETE showed robust formation of 5,15-diHETE and 5,11-diHETE, and the levels of both were reduced >100-fold by preincubation with 10 M NS-398 ( Fig. 7B ). Pretreatment with 2 mM aspirin inhibited the formation of 5,11-diHETE by about 90%, and 5,15-di-HETE was reduced to only 23% compared with the cells not treated with aspirin, refl ecting enhanced formation of 5 S ,15 R -diHETE.

DISCUSSION
After the initial oxygenation of arachidonic acid to form a conjugated diene hydro(pero)xide (H(P)ETE), additional sites remain in the molecule for subsequent reaction with molecular oxygen ( 21 ). This opens the possibility for formation of di-hydroxylated (diHETEs) and tri-hydroxylated derivatives of arachidonic acid. Enzymatic synthesis of diHETEs can be catalyzed via several distinct routes, all of which involve one or more LOX reactions: i ) The consecutive reaction of two separate LOX enzymes is involved in the biosynthesis of 5 S ,15 S -diHETE in elicited rat mononuclear cells and human leukocytes ( 18 ). In this case, arachidonic acid is fi rst oxygenated by 5-LOX followed by 15-LOX or vice versa. ii ) An alternative route to 5 S ,15 S -diHETE is through double oxygenation of arachidonic acid catalyzed by a single LOX enzyme. This possibility is best recognized for the LOX-1 isozyme from soybean seeds in the formation of 5 S ,15 S -diH(P)ETE and 5 S ,8 S -diH(P)ETE by oxygenation of the primary product 15 S -H(P)ETE ( 22,23 ). iii ) The third possibility is exemplifi ed by the biosynthesis of leukotriene (LT)B 4 and 12-epi -LTB 4 (i.e., 5 S ,12 R -diHETE and 5 S ,12 S -diHETE, respectively). In this case, the diHETEs are hydrolysis products of the unstable LTA 4 epoxide. The LTA 4 epoxide is formed by 5-LOX catalyzing a second hydrogen abstraction (at C-10) of its initial 5 S -HPETE product. But rather than inserting a second molecule of oxygen the reaction is completed by dehydration of the existing hydroperoxide to give the epoxide ( 24,25 ). iv ) Finally, a fourth distinct route to diHETEs is implicated by the fi ndings presented in this report. This route involves a cross-over of the activities of the 5-LOX and COX-2 enzymes, with the 5-LOX product 5 S -HETE being oxygenated by COX-2 to form, as by-products, a mixture of 5 S ,15 S -diHETE, 5 S ,15 R -diHETE, and 5 S ,11 R -diHETE.
The formation of 5,15-diHETE and 5,11-diHETE as by-products of the COX-2 catalyzed transformation of 5 S -HETE bears strong resemblance to the formation of 15-HETE and 11-HETE as by-products of the COXcata lyzed transformation of arachidonic acid to PGH 2 Fig. 8. Synopsis of the transformation of 5 S -HETE and arachidonic acid by native and acetylated COX-2, respectively. ment of COX-2 can be implicated if the formation of 5,15-diHETE is attenuated upon application of a COX-2 specifi c inhibitor. Alternatively, a minor amount of 5,11-diHETE, in addition to 5,15-diHETE, could be indicative of COX-2 involvement. Formation of 5 S ,15 R -diHETE as an alternative specific marker of COX-2 involvement is difficult to establish because the 5 S ,15 S -and 5 S ,15 Rdiastereomers do not resolve using standard RP-HPLC conditions.
We thank Dr. David Wright and Jonas Perez for assistance in using the CD spectrometer.
Acetylation of Ser-516 in the COX-2 active site by aspirin has a remarkable effect on its catalytic activity ( 10,11 ). Formation of the prostaglandin endoperoxide is prevented, and instead, a novel catalytic activity is gained, forming 15 R -HETE as the sole enzymatic product. The basis for the complete inversion of the stereochemistry of C15 from 15 S in PGH 2 to 15 R in 15 R -HETE has not been defi nitely elucidated but it likely involves a change in the binding of the -end of arachidonic acid beyond C-13 in the channel above Ser-516 (28)(29)(30)(31)(32). When 5 S -HETE was incubated with acetylated COX-2, formation of the di-endoperoxide was inhibited and 5,15-diHETE was the only product detected. Not too surprisingly, the confi guration of C-15 of the 5,15-diHETE was found to be >95% 15 R .
The formation of diHETEs was analyzed using the RAW264.7 mouse macrophage cells and CT26 mouse colon carcinoma cells. Neither cell type produced detectable amounts of 5 S -HETE, and we were unable to detect 5-LOX protein by Western blotting (data not shown). 5,15-DiHETE and 5,11-diHETE were detected in both cell types upon incubation with exogenous 5 S -HETE. Formation of the diHETEs was dependent on COX-2 because they were absent in nonstimulated cells and in cells treated with the COX-2 inhibitor NS-398. Treatment of CT26 cells with aspirin prior to incubation with 5 S -HETE enhanced the biosynthesis of 5,15-diHETE, consistent with the fi ndings using recombinant COX-2 enzyme. Aspirin, even at a high concentration, did not show this effect in RAW264.7 cells, most likely due to lesser effi cacy for covalent modifi cation of the COX enzyme in cells with a highly oxidative tone ( 20,33 ). 5 S ,11 R -diHETE and 5 S ,15 R -diHETE have not been described as metabolites of arachidonic acid before, although there are two reports published more than 20 years ago that implicate the possibility of COX-dependent biosynthesis of diHETEs in human umbilical arteries ( 34,35 ). Unfortunately, the presumed diHETE metabolites were left uncharacterized, and it is diffi cult to estimate whether the products described could be similar or identical to 5,15-diHETE or 5,11-diHETE. 5 S ,15 S -DiHETE as well as 8 S ,15 S -diHETE and 15 S -HETE were reported to enhance the degranulation of human neutrophils elicited by platelet-activating factor, whereas they had no such effect when the neutrophils were stimulated with the tripeptide formyl-met-leu-phe, phorbol ester, LTB 4 , or calcium ionophore ( 36 ). 5 S ,15 S -DiHETE and 8 S ,15 S -diHETE were also identifi ed as eosinophil-derived eosinophil chemotactic lipids, invoking their participation in a self-sustaining mechanism of eosinophil accumulation ( 37 ). An additional, more potent eosinophil chemotactic eicosanoid was noted at the time and later identifi ed as 5-oxo-15-hydroxy-eicosatetraenoic acid, an oxidation product of 5,15-diHETE ( 38 ).
Our studies invoke the possibility of a previously unrecognized biosynthetic route to 5 S ,15 S -diHETE, and therefore, additional experiments are required to distinguish whether, if detected in vivo, 5,15-diHETE is formed by cross-over of the 5-LOX and 15-LOX pathways, or by cross-over of the 5-LOX and COX-2 pathways. Involve-