Formation of prostaglandins E2 and D2 via the isoprostane pathway: a mechanism for the generation of bioactive prostaglandins independent of cyclooxygenase.

It has heretofore been assumed that the cyclooxygenases (COXs) are solely responsible for peostaglandin (PG) synthesis in vivo. An important structural feature of PGH2 formed by COX is the trans-configuration of side chains relative to the prostane ring. Previously, we reported that a series of PG-like compounds termed isoprostanes (IsoPs) are formed in vivo in humans from the free radical-catalyzed peroxidation of arachidonate independent of COX. A major difference between these compounds and PGs is that IsoPs are formed from endoperoxide intermediates, the vast majority of which contain side chains that are cis relative to the prostane ring. In addition, unlike the formation of eicosanoids from COX, IsoPs are formed as racemic mixtures because they are generated nonenzymatically. IsoPs containing E- and D-type prostane rings (E2/D2-IsoPs) are one class of IsoPs formed, and we have reported previously that one of the major IsoPs generated is 15-E2t-IsoP (8-iso-PGE2). Unlike PGE2, 15-E2t-IsoP is significantly more unstable in buffered solutions in vitro and undergoes epimerization to PGE2. Analogously, the D-ring IsoP (15-D2c-IsoP) would be predicted to rearrange to PGD2. We now report that compounds identical in all respects to PGE2 and PGD2 and their respective enantiomers are generated in vivo via the IsoP pathway, presumably by epimerization of racemic 15-E2t-IsoP and 15-D2c-IsoP, respectively. Racemic PGE2 and PGD2 were present esterified in phospholipids derived from liver tissue from rats exposed to oxidant stress at levels of 24 +/- 16 and 37 +/- 12 ng/g of tissue, respectively. In addition, racemic PGs, particularly PGD2, were present unesterified in urine from normal animals and humans and represented up to 10% of the total PG detected. Levels of racemic PGD2 increased 35-fold after treatment of rats with carbon tetrachloride to induce oxidant stress. In this setting, PGD2 and its enantiomer generated by the IsoP pathway represented approximately 30% of the total PGD2 present in urine. These findings strongly support the contention that a second pathway exists for the formation of bioactive PGs in vivo that is independent of COX.

We have previously reported that a series of PG-like compounds termed isoprostanes (IsoPs) are formed in vivo from the free radical-catalyzed peroxidation of arachidonate independent of COX (6). Analogous to PGs, we have determined that IsoPs contain E/D-, F-, and thromboxane-type prostane rings (7). Although the structures of these compounds are very similar to COX-derived PGs, an important distinction between IsoPs and PGs is that IsoP bicycloendoperoxide intermediates contain side chains that are predominantly (Ͼ90%) oriented cis in relation to the prostane ring because the generation of these intermediates is favored kinetically (4,7,8). Indeed, we have previously reported that two IsoPs that are formed in abundance in vivo are 15-F 2t -IsoP (8-iso-PGF 2␣ ) and 15-E 2t -IsoP (8-iso-PGE 2 ), which are generated from the endoperoxide intermediate 15-H 2t -IsoP (8-iso-PGH 2 ) ( Fig. 1B) (9,10). Although not reported, it would also be predicted that 15-H 2c -IsoP (12iso-PGH 2 ) is formed in abundance and can rearrange to the analogous D-ring IsoP 15-D 2c -IsoP (12-iso-PGD 2 ) (Fig. 1C).
In contrast to other types of prostanoids, E 2 /D 2 -IsoPs are ␤-hydroxyketone-containing compounds that can undergo reversible keto-enol tautomerization under both acidic and basic conditions, allowing rearrangement of the side chains that are initially cis to the more stable trans-configuration. That the trans-configuration is highly favored has been demonstrated by the finding that, when PGE 2 is subjected to conditions that induce keto-enol tautomerism, Ͻ10% of the compound rearranges to the cis-side chain isomer 15-E 2t -IsoP (11). In addition, attempts to synthesize 15-D 2c -IsoP have been unsuccessful because epimerization at C-12 readily occurs during synthesis to yield PGD 2 (12). Furthermore, facile epimerization of a number of other PG-like compounds containing side chains cis to the prostane ring has been reported (13,14).
In the course of studies to characterize various E/D-ring IsoPs formed in vitro and in vivo from the peroxidation of * This work was supported in part by National Institutes of Health Grants DK48831, GM42056, CA77839, HD12304, and HL46296. 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  arachidonic acid, analysis of oxidation products by gas chromatography (GC)/mass spectrometry (MS) disclosed the generation of significant amounts of compounds that had retention times and molecular weights identical to those of PGE 2 and PGD 2 . Because IsoPs are formed nonenzymatically, compounds generated by this pathway would be predicted to be racemic (6,7). Using a variety of chromatographic and mass spectrometric approaches, we present evidence that compounds identical in all respects to COX-derived PGE 2 and PGD 2 and their respective enantiomers are formed in vitro and in vivo via the IsoP pathway. A proposed mechanism by which the formation of PGE 2 and PGD 2 occurs from 15-E 2t -IsoP and 15-D 2c -IsoP, respectively, via base-catalyzed isomerization is shown in Fig. 2  (A and B). Generation of PGE 2 and PGD 2 from 15-E 2t -IsoP and 15-D 2c -IsoP, respectively, would also be predicted to occur via acid catalysis (11).
These findings strongly support the contention that a second pathway exists for the formation of bioactive PGs in vivo that is independent of COX. This finding is of potential physiological and pharmacological importance because it would be predicted that the generation of PGs via this mechanism would not be inhibited by aspirin or other COX inhibitors. For purposes of discussion hereafter, PGs possessing a structure identical to those generated by COX are referred to PGE 2 and PGD 2 . Compounds that are enantiomeric to COX-derived PGs are referred to as ent-PGE 2 and ent-PGD 2 . The racemic mixtures are termed rac-PGE 2 and rac-PGD 2 (Fig. 3 (10,16,17).
Oxidation of Arachidonic Acid-Arachidonic acid was oxidized in vitro using an iron/ADP/ascorbate mixture as previously described (16,17).
Isolation of E/D-ring IsoPs and PGs from Rodent and Human Tissue and Urine-A mixture of E/D-ring IsoPs and PGs was isolated from the livers of Sprague-Dawley rats 2 h after intragastric administration of CCl 4 (2 mg/kg) in corn oil (16,17). The animals were anesthetized with pentobarbital (60 mg/kg) intraperitoneally and killed, and the livers were removed. Depending on the experiment, 1-4 g of tissue was immediately extracted to obtain a crude phospholipid extract containing IsoPs and PGs esterified in phospholipids. The lipid extract was then subjected to hydrolysis (30 min) in boronate buffer with A. mellifera venom containing phospholipase A 2 as described (16,17). In control experiments, complete hydrolysis of 2-[ 3 H 7 ]arachidonylphosphatidylcholine was effected during this incubation. In addition, these hydrolysis conditions resulted in Ͻ5% epimerization of 15-E 2t -IsoP to PGE 2 . Subsequently, free IsoPs and PGs were extracted, partially purified using C 18 and silica Sep-Pak cartridges, and subjected to HPLC. For selected experiments, liver tissue was also obtained from day 19 COX-1 Ϫ/Ϫ /COX-2 Ϫ/Ϫ mouse pups harvested in utero as described (18).
In some experiments, 24-h urine samples were collected from rats treated with CCl 4 or from normal humans. Unesterified IsoPs and PGs were extracted using Sep-Pak columns as described (6).
HPLC Separation of Racemic E/D-ring PGs and IsoPs-Depending on the experiment, incubations of oxidized arachidonic acid, partially purified tissue extracts, or urine samples were analyzed for rac-PGE 2 , rac-PGD 2 , or rac-15-E 2t -IsoP (9, 10). To the biological sample was added ϳ0. 5  maximize purification and resolution of each compound, we used HPLC procedures that yielded relatively long retention volumes for each compound (ϳ15-40 ml), and each solvent was run isocratically. In pilot experiments, it was also shown that the three compounds readily separated from one another under the HPLC conditions utilized. In addition, radiolabeled PGE 2 , PGD 2 , and 15-E 2t -IsoP separated to a significant extent (1-2.5 ml) from unlabeled compounds due to the fact that the radiolabeled compounds contained either six or seven tritium atoms. Thus, for each HPLC step, fractions corresponding to those containing both labeled and unlabeled PGs or IsoPs were collected and pooled for further purification. For rac-PGE 2 and rac-15-E 2t -IsoP, the first HPLC step was normal-phase using a Econosil SI column (25 cm ϫ 4.6 mm, 5-m particles; Alltech Associates Inc., Deerfield, IL). The solvent system was 88:12:0.1 (v/v/v) hexane/isopropyl alcohol/acetic acid at a flow rate of 1 ml/min. The second HPLC step was reversed-phase using an Econosil C 18 column (25 cm ϫ 4.6 mm, 5 m; Alltech Associates Inc.). The solvent system was 30:70:0.1 (v/v/v) acetonitrile/water/ acetic acid at a flow rate of 1 ml/min. For the third and fourth HPLC steps, IsoPs or PGs were converted to PFB esters and rechromatographed on normal-and reversed-phase HPLC columns. A solvent system of 92:8 (v/v) hexane/isopropyl alcohol was used for the third HPLC step, and 51:49 (v/v) acetonitrile/water was used for the fourth HPLC step, both at a flow rate of 1 ml/min. For the purification of rac-PGD 2 , the same columns were utilized, but the solvent systems varied. For the first HPLC step, the solvent system was 93:7:0.1 (v/v/v) hexane/isopropyl alcohol/acetic acid; the second HPLC solvent system was 33:67:0.1 (v/v/v) acetonitrile/water/acetic acid; the third HPLC solvent system was 95:5 (v/v) hexane/isopropyl alcohol; and the fourth HPLC solvent system was 58:42 (v/v) acetonitrile/water. Chiral HPLC Separation of rac-PGE 2 and rac-PGD 2 -Racemic PGs purified by the methods described above were subsequently subjected to chiral HPLC to separate enantiomers using a Chiralpak AD column (25 cm ϫ 4.6 mm, 5 m; Chiral Technologies, Exton, PA). To separate rac-PGE 2 , a solvent system of 93:7 (v/v) hexane/isopropyl alcohol was utilized; and for rac-PGD 2 , a solvent system of 95:5 (v/v) hexane/isopropyl alcohol was employed.
Analysis of PGs and IsoPs by GC/MS-Quantification of E/D-ring PGs and IsoPs in partially purified biological extracts and throughout subsequent HPLC purification procedures was performed by analyzing aliquots by selected ion monitoring GC/negative ion chemical ionization MS using either [ 2 H 4 ]PGE 2 or [ 2 H 4 ]PGD 2 as an internal standard. Compounds were quantified as O-methyloxime, PFB ester, trimethylsilyl ether derivatives by monitoring the M-PFB (M Ϫ 181) ions at m/z 524 for endogenous compounds and at m/z 528 for the deuterated standards (10,19).

Epimerization of 15-E 2t -IsoP in Phosphate
Buffer-We initially determined the extent to which 15-E 2t -IsoP undergoes epimerization to PGE 2 in a buffered solution at physiological pH (50 mM KPO 4 , pH 7.4). Products that were quantified included the starting material 15-E 2t -IsoP and PGE 2 and their respective dehydration products, 15-A 2t -IsoP and PGA 2 . Amounts are expressed as percent of total PG at a particular time point. The identification of PGE 2 was also confirmed by NMR comparison with a chemically pure PGE 2 standard. The results are shown in Fig. 4. As shown, 15-E 2t -IsoP epimerized in a time-dependent manner to PGE 2 . The half-life for this conversion under the conditions noted was ϳ2 h. In addition, small amounts of 15-A 2t -IsoP and PGA 2 were formed, presumably as a result of dehydration of 15-E 2t -IsoP and PGE 2 , respectively. These findings support the hypothesis that E 2 /D 2 -IsoPs can readily rearrange to E/D-ring PGs in aqueous environments.
Analysis of E 2 /D 2 -IsoPs from the Oxidation of Arachidonic Acid in Vitro and in Vivo- Fig. 5A shows the selected ion current chromatograms for E 2 /D 2 -IsoPs obtained from the analysis of arachidonic acid with iron/ADP/ascorbate for 2 h. Compounds were analyzed as O-methyloxime, PFB ester, trimethylsilyl ether derivatives. In the lower m/z 528 chromatogram are two peaks representing the syn-and anti-O-methyloxime isomers of the [ 2 H 4 ]PGE 2 internal standard. In the upper m/z 524 chromatogram are a series of peaks representing various E 2 /D 2 -IsoPs. The peaks indicated by asterisks represent compounds that co-chromatographed upon GC with the O-methyloxime isomers of chemically synthesized PGE 2 . In addition, the peaks denoted by plus signs co-chromatographed with the O-methyloxime isomers of chemically pure PGD 2 . The total E 2 /D 2 -IsoPs present were ϳ1500 ng/g of arachidonic acid. The materials designated by the peaks denoted by asterisks and plus signs each represent ϳ20% of the total E 2 /D 2 -IsoPs in the mixture.
In addition to the analysis of oxidized arachidonic acid in vitro, Fig. 5B shows the selected ion current chromatograms obtained from the hydrolysis of rat liver phospholipids after administration of CCl 4 to animals to induce oxidant stress. As shown, a similar pattern of peaks was present as shown in Fig.  5A. The analyses in Fig. 5 (A and B) were performed on separate days, accounting for differences in GC retention time. Again, the peaks indicated by asterisks represent compounds that co-chromatographed upon GC with chemically synthesized PGE 2 , whereas those denoted by plus signs co-chromato-graphed with PGD 2 . The total E 2 /D 2 -IsoPs present in this sample were ϳ400 ng/g of liver tissue. The materials designated by the peaks denoted by the asterisks and plus signs each repre-  Fig. 5B. Details regarding solvent systems used are described under "Experimental Procedures." Tritiated PGE 2 was added to the mixture at the beginning, and aliquots of eluted fractions were assayed for radioactivity (q). Aliquots were also assayed and quantified by GC/MS for the presence of E 2 /D 2 -IsoP peaks with the same retention time as authentic PGE 2 (OE). All HPLC purifications were carried out isocratically. A, normal-phase HPLC as free acids of the initial mixture of E 2 /D 2 -IsoPs as shown in Fig. 5B. The plus sign denotes fractions in which chemically pure unlabeled PGE 2 eluted using this solvent system. B, reversed-phase HPLC as free acids of the material that eluted at the retention volume between 14.5 and 18.5 ml in A. C, normal-phase HPLC as PFB esters of the material that eluted at the retention volume between 28.5 and 33 ml in B. D, reversed-phase HPLC as PFB esters of the material that eluted at the retention volume between 27.5 and 34 ml in C. sent ϳ15-20% of the total E 2 /D 2 -IsoPs in the mixture. A very similar pattern of peaks representing E 2 /D 2 -IsoPs was obtained from the livers of both COX-1 Ϫ/Ϫ /COX-2 Ϫ/Ϫ and control fetal mice without induction of oxidant stress, although total E 2 /D 2 -IsoP levels were ϳ5-10 ng/g of liver tissue (data not shown). In addition, a pattern of peaks virtually identical to those shown in the chromatograms in Fig. 5 was obtained after hydrolysis of rat liver phospholipids that had been treated with methyloxime HCl prior to hydrolysis. These latter findings suggest that epimerization of E 2 /D 2 -IsoPs occurs while compounds are esterified in phospholipids. Taken together, these data support the contention that significant amounts of compounds that coelute upon GC with PGE 2 and PGD 2 are generated from the peroxidation of arachidonic acid in vitro and in vivo.
Purification of Putative rac-PGE 2 from Rat Liver Hydrolysates by HPLC-We subsequently sought to determine whether the compounds from rat liver hydrolysates that coeluted upon GC with PGE 2 and PGD 2 were, in fact, structurally identical to PGE 2 and PGD 2 and their respective enantiomers. If PGE 2 and PGD 2 are formed via the IsoP pathway, it would be predicted that they would be racemic mixtures because they would be formed from the epimerization of rac-15-E 2t -IsoP and rac-15-D 2c -IsoP, respectively (6,7). Of note, enantiomers of PGE 2 and PGD 2 would not be expected to separate using standard nonchiral HPLC methods.
The formation of rac-PGE 2 was assessed initially. For these studies, ϳ2000 ng of E 2 /D 2 -IsoPs from rat liver containing 3 Ci of [ 3 H 7 ]PGE 2 was subjected to four successive HPLC purification steps. The first HPLC step was normal-phase using a solvent system of 88:12:0.1 (v/v/v) hexane/isopropyl alcohol/ acetic acid. Aliquots of fractions that eluted from the HPLC column were then analyzed for E 2 /D 2 -IsoPs by GC/MS and for radioactivity (Fig. 6A). Radiolabeled PGE 2 eluted in this system between 16 and 18.5 min. Compounds representing endogenous E 2 /D 2 -IsoPs were present that had the same retention time upon GC as PGE 2 , but that eluted with different retention volumes compared with PGE 2 upon HPLC (10 -12, 18 -21, and 22.5-25 ml). Radiolabeled PGE 2 eluted at a volume of ϳ1.0 -1.5 ml after unlabeled PGE 2 using this HPLC solvent system. Significantly, as shown in Fig. 6A, an endogenous E 2 /D 2 -IsoP peak (indicated by the plus sign) was detected that coeluted with unlabeled PGE 2 , suggesting that this compound is endogenously derived rac-PGE 2 .
The material that eluted from the HPLC column between 14.5 and 18.5 ml in Fig. 6A was subsequently subjected to reversed-phase HPLC using an isocratic solvent system of 30: 70:0.1 (v/v/v) acetonitrile/water/acetic acid. Aliquots of fractions collected were again analyzed for endogenous E 2 /D 2 -IsoPs by GC/MS and for radioactivity (Fig. 6B). Radiolabeled PGE 2 eluted from the HPLC column with a retention volume of 28.5-32.5 ml. Analysis of aliquots of the eluted fractions by GC/MS showed that almost all of the unlabeled E 2 /D 2 -IsoP material detected in the chromatogram eluted at the retention volume of unlabeled PGE 2 (29.5-33 ml), except for a small amount of additional material that eluted at 37-39 ml.
Altering the polarity of a compound by derivatization and rechromatography of the compound can provide a powerful approach for purification and separation of biomolecules (9). Thus, the material that eluted from the HPLC column between 28.5 and 33 ml in Fig. 6B was converted to a PFB ester and rechromatographed on a normal-phase HPLC column using a solvent system of 92:8 (v/v) hexane/isopropyl alcohol. Fig. 6C shows the result of this HPLC step. Radiolabeled PGE 2 eluted between 29.5 and 33.5 ml. A large peak representing endoge-nous E 2 /D 2 -IsoPs that coeluted with the PFB ester of unlabeled PGE 2 was detected (27.5-31.5 ml).
Compounds that eluted from the HPLC column between 27.5 and 34 ml in Fig. 6C were then pooled. This material was subjected to further purification by reversed-phase HPLC using a solvent system of 51:49 (v/v) acetonitrile/water. The results of the analyses for radioactivity and endogenous E 2 /D 2 -IsoPs in the eluted fractions are shown in Fig. 6D. A single E 2 /D 2 -IsoP peak presumably representing rac-PGE 2 was present that coeluted exactly with the PFB ester of unlabeled PGE 2 (42-44.5 ml). Radiolabeled PGE 2 eluted slightly before the endogenous E 2 /D 2 -IsoP compound. Virtually identical results were obtained when putative rac-PGE 2 generated from arachidonic acid oxidized in vitro was analyzed by the HPLC protocols described above.
Analysis of Endogenous Putative rac-PGE 2 by GC/MS-The material that eluted between 40.5 and 44.5 ml upon the fourth HPLC step was then analyzed by GC/MS. As shown in Fig. 7, two E 2 /D 2 -IsoP peaks were present in the m/z 524 chromatogram, representing the syn-and anti-O-methyloxime isomers of putative rac-PGE 2 . The amount of putative rac-PGE 2 present in this rat liver hydrolysate was ϳ35 ng/1000 ng of total E 2 /D 2 -IsoP based upon losses of [ 3 H 7 ]PGE 2 that occurred with the four HPLC purification steps. When the material indicated by peaks in the m/z 524 chromatogram in Fig. 7 was mixed with an equivalent amount of derivatized synthetic PGE 2 , the two compounds co-chromatographed perfectly upon capillary GC without any suggestion of a shoulder on the GC peaks (data not shown). Additional experiments were subsequently undertaken to confirm the identification of the compound in Fig. 7 as PGE 2 . First, analysis of the material as a deuterated O-methyloxime derivative disclosed the presence of one carbonyl group. Second, analysis as a deuterated trimethylsilyl ether derivative revealed that the compound had two hydroxyl groups. Third, catalytic hydrogenation showed two double bonds (16). Finally, treatment of putative rac-PGE 2 with 15% methanolic KOH for 30 min converted it to a compound with a molecular weight and retention time identical to those of PGB 2 when analyzed by GC/MS (Fig. 8) (20). Taken together, these findings strongly support the contention that the material represented in the m/z 524 chromatogram in Fig. 7 is rac-PGE 2 .
Analysis of Putative rac-PGE 2 by Chiral HPLC-As noted, it is predicted that PGE 2 generated by the IsoP pathway should be racemic. The HPLC steps utilized above to purify putative rac-PGE 2 will not separate enantiomers. Thus, the compounds represented in the chromatogram in Fig. 7 were subjected to chiral column chromatography, and fractions that eluted from the HPLC column were analyzed by GC/MS. Fig. 9 shows the results of the analysis. The peak indicated by the asterisk co-chromatographed under these HPLC conditions with chemically synthesized PGE 2 , whereas the peak denoted by the plus sign co-chromatographed with chemically synthesized ent-PGE 2 . The material from each peak co-chromatographed perfectly with both PGE 2 and ent-PGE 2 upon GC and was indistinguishable upon MS analysis. Approximately equal amounts of the compounds were present, as would be expected. Furthermore, the ratio of methyloxime isomers of ent-PGE 2 was essentially identical to that of PGE 2 . Taken together, these studies provide compelling evidence that PGE 2 and ent-PGE 2 are generated in vivo in significant quantities from the IsoP pathway. Essentially identical results were obtained from the analysis of putative rac-PGE 2 formed from the peroxidation of arachidonate in vitro.
Purification of Putative rac-PGD 2 from Rat Liver Hydrolysates by HPLC-As noted in Fig. 5 (A and B), chromatographic peaks were present that coeluted not only with PGE 2 , but also with PGD 2 . We thus employed similar approaches as those used to obtain evidence for the formation of rac-PGE 2 in vitro and in vivo to determine whether rac-PGD 2 is also generated. Table I shows the HPLC conditions utilized to purify putative rac-PGD 2 and the retention time of the compound at each step. Fig. 10 illustrates the results from GC/MS analysis of the material that eluted between 21 and 24 ml, where PGD 2 eluted, upon the fourth HPLC step. As shown, two E 2 /D 2 -IsoP peaks were present in the m/z 524 chromatogram, representing the syn-and anti-O-methyloxime isomers of putative rac-PGD 2 . The amount of putative rac-PGD 2 present in the rat liver hydrolysate from this analysis was ϳ55 ng/1000 ng of total E 2 /D 2 -IsoP, based upon losses of [ 3 H 7 ]PGD 2 that occurred with the four HPLC purification steps. When the material denoted by the peaks in the m/z 524 chromatogram in Fig. 10 was mixed with an equivalent amount of derivatized synthetic PGD 2 , the two compounds co-chromatographed perfectly upon capillary GC without any suggestion of a shoulder on the GC peaks (data not shown). Additional experiments confirmed the identification of the compound in Fig. 10 as PGD 2 . First, analysis of the material as a deuterated O-methyloxime derivative disclosed the presence of one carbonyl group. Second, analysis as a deuterated trimethylsilyl ether derivative revealed that the compound had two hydroxyl groups. Third, catalytic hydrogenation showed two double bonds.
Analysis of Putative rac-PGD 2 by Chiral HPLC-As with PGE 2 generated by the IsoP pathway, it is predicted that PGD 2 should be racemic. Thus, the compounds represented in the m/z 524 chromatogram in Fig. 10 were subjected to chiral column chromatography, and fractions that eluted from the HPLC column were analyzed by GC/MS (Table I). Fig. 11 shows the results of the analysis. The peak indicated by the asterisk co-chromatographed under these HPLC conditions with chem- FIG. 9. Chiral HPLC separation of rac-PGE 2 . The putative rac-PGE 2 purified as described in the legend to Fig. 6 was subjected to chiral column chromatography using the solvent system 93:7 (v/v) hexane/isopropyl alcohol. Aliquots of fractions that eluted from the HPLC column were then analyzed for PGE 2 by GC/MS. The peak indicated by the asterisk co-chromatographed in the HPLC solvent system with chemically pure PGE 2 , whereas the peak denoted by the plus sign coeluted with ent-PGE 2 . a The flow rate those of for all HPLC solvents was 1 ml/min. The retention times of putative rac-PGD 2 were identical to chemically pure PGD 2 in HPLC Steps 1-4. Table I. Only a single set of m/z 524 peaks representing the syn-and anti-O-methyloxime isomers of endogenous putative rac-PGD 2 remained after the four HPLC purification procedures shown in Table I. The peaks in the m/z 528 chromatogram represent the syn-and anti-O-methyloxime isomers of the deuterated PGD 2 internal standard. The amount of putative rac-PGE 2 in the fraction analyzed was ϳ55 ng/1000 ng of total E 2 /D 2 -IsoP.

FIG. 10. Selected ion current chromatogram obtained from the GC/MS analysis of the material that eluted at a retention volume of 21-24 ml after the fourth HPLC step to purify rac-PGD 2 as noted in
ically synthesized PGD 2 , whereas the peak denoted by the plus sign represents ent-PGD 2 . The material from each peak cochromatographed perfectly with PGD 2 upon GC and was indistinguishable upon MS analysis. Approximately equal amounts of compounds were present, as would be expected; and the relative amounts of methyloxime isomers of PGD 2 and ent-PGD 2 were very similar. Taken together, these studies provide strong evidence that PGD 2 and ent-PGD 2 , in addition to PGE 2 and ent-PGE 2 , are generated in vivo in significant quantities from the IsoP pathway. Again, essentially identical results were obtained from the analysis of putative rac-PGD 2 formed from the peroxidation of arachidonate in vitro.
Quantitative Analysis of rac-PGE 2 and rac-PGD 2 Generated in Vitro and in Rat Liver-The above studies provide substantial support for the hypothesis that PGE 2 and PGD 2 and their respective enantiomers can be generated via the IsoP pathway. We subsequently undertook experiments to determine the total amounts of rac-PGD 2 and rac-PGE 2 generated from the oxidation of arachidonate in vitro and in vivo in comparison with other E 2 /D 2 -IsoPs. These determinations are highly important because, based on previous reports (4), the vast majority (Ͼ90%) of endoperoxide intermediates generated by the autoxidation of polyunsaturated fatty acids have side chains that are cis in relation to the prostane ring. Indeed, we have recently confirmed that endoperoxides with cis-side chains predominate over trans-side chain compounds when arachidonate is oxidized in vitro (5). Thus, it would be predicted that the IsoP endoperoxide with a structure identical to PGH 2 generated from the peroxidation of arachidonate in vitro and in vivo would compose a trivial fraction of the total endoperoxides that are formed. Therefore, the amounts of rac-PGE 2 and rac-PGD 2 that are subsequently generated from this endoperoxide intermediate would be present at no more than a few nanograms/ 1000 ng of total E 2 /D 2 -IsoPs (4). Employing the HPLC protocols utilized for the studies described above, we quantified rac-PGE 2 , rac-PGD 2 , and rac-15-E 2t -IsoP in vitro and in vivo and also assessed the relative formation of each enantiomer in the racemic mixture. Losses of endogenous material during the chromatographic procedures were accounted for by determining the percent loss of the respective radiolabeled PG added to the samples prior to purification. As noted in Table II, the amounts of rac-PGE 2 and rac-PGD 2 far exceeded those predicted based upon the observations of O'Connor et al. (4), and the quantities of rac-PGE 2 were at least as great as, if not greater than, those of rac-15-E 2t -IsoP both in vitro and in vivo. In summary, these quantitative data provide support that PGE 2 and PGD 2 and their respective enantiomers are generated in significant amounts via the IsoP pathway.
Excretion of Unesterified rac-PGE 2 and rac-PGD 2 in Rat and Human Urine in Vivo-In addition to detecting the in vivo formation of rac-PGE 2 and rac-PGD 2 esterified in rat liver tissue, we also sought to determine whether these compounds are present unesterified in human and rodent urine at base line and whether they increase in association with oxidant stress. Table III shows the results of studies performed to determine the relative amounts of these eicosanoids under these conditions. As shown, at baseline, the relative levels of both ent-PGE 2 and ent-PGD 2 in humans and rats were low and composed no more than 10% of the total rac-PGE 2 and rac-PGD 2 generated. In addition, in several of the urine samples obtained from normal humans and rats at baseline, the levels of rac-PGE 2 were below the limits of assay detection. On the other hand, after treatment of rats with CCl 4 , the levels of both ent-PGE 2 and ent-PGD 2 increased significantly. This was particularly the case for ent-PGD 2 . If one assumes that an amount of PGD 2 equivalent to that of ent-PGD 2 is generated via the IsoP pathway after administration of CCl 4 , then ϳ15% of PGD 2 present in rat urine under these conditions is formed by a mechanism independent of COX. Analogously, ϳ30% of rac-PGD 2 (total of PGD 2 and ent-PGD 2 ) would thus be predicted to be generated by this mechanism. Fig. 12 (A and B) shows the results from chiral analysis of rat urine for PGD 2 and ent-PGD 2 at base line and after treatment with CCl 4 . As shown, at base line, the chromatographic peak comprising PGD 2 (*) in Fig.  12A greatly exceeded the enantiomer (ϩ), suggesting that COX contributes to the vast majority of PGD 2 production at base line. After CCl 4 administration (Fig. 12B), the levels of ent-PGD 2 (ϩ) increased dramatically in relation to PGD 2 (*), supporting the contention that CCl 4 has induced PG formation via the IsoP pathway.
To provide further evidence that PGs can be generated via the IsoP pathway, we pretreated rats with indomethacin (10 mg/kg intraperitoneally for 24, 12, and 1 h) prior to CCl 4 administration and collected urine for 24 h after the oxidant was given (6,21). Fig. 12C shows the results from the chiral analysis of rac-PGD 2 . As shown, the chromatographic peaks representing PGD 2 and ent-PGD 2 are very similar. The levels of PGD 2 and ent-PGD 2 were ϳ30% of those present in CCl 4treated rats not given indomethacin (Table III) and support the contention that significant amounts of unesterified PGD 2 (and to a lesser extent, PGE 2 ) can be formed by a mechanism independent of COX in settings of oxidant stress. FIG. 11. Chiral HPLC separation of rac-PGD 2 . The putative rac-PGD 2 purified as noted in Table I was subjected to chiral column chromatography using the solvent system 95:5 (v/v) hexane/isopropyl alcohol. Aliquots of fractions that eluted from the HPLC column were then analyzed for PGD 2 by GC/MS. The peak indicated by the asterisk co-chromatographed in the HPLC solvent system with chemically pure PGD 2 , whereas the peak denoted by the plus sign represents ent-PGD 2 . b Total E 2 /D 2 -IsoPs were 561 Ϯ 232 ng/g of liver tissue (n ϭ 6).

DISCUSSION
This study describes the formation of PGE 2 and PGD 2 independent of COX and involving the free radical-catalyzed peroxidation of arachidonate. We have reported that significant amounts of rac-PGE 2 and rac-PGD 2 are generated in vitro and in vivo in settings of oxidant stress. Unlike PGs formed via COX, generation of eicosanoids by this mechanism results in the formation of compounds as racemic mixtures because the oxygenation of arachidonic acid does not occur stereospecifically (2,3,6). Our initial interest in determining whether PGs are generated via the IsoP pathway emerged from the observation that compounds with the same molecular weights and GC retention times as PGE 2 and PGD 2 are present when analyzed by GC/MS in mixtures of arachidonate oxidized in vitro and in rat liver hydrolysates. Utilizing a variety of high resolving chromatographic, chemical, and mass spectrometric approaches, we have found that substantial quantities of these racemic PGs can be generated. Analysis of putative rac-PGE 2 and rac-PGD 2 by chiral HPLC revealed that each compound is composed of two enantiomers generated in equal amounts in vitro and in liver tissue from rats exposed to oxidant stress. rac-PGE 2 and rac-PGD 2 were also present in the unesterified form in significant amounts in urine from rats treated with CCl 4 , and their formation was unaffected by COX inhibition. That COX is not involved in the formation of these compounds is also supported by the findings that these PGs could be generated in vitro without COX and were present in vivo esterified in phospholipids. COX is not active on arachidonate esterified in phospholipids (1). Finally, compounds with retention times and molecular weights identical to those of PGE 2 and PGD 2 were present when liver tissue from COX-1 Ϫ/Ϫ /COX-2 Ϫ/Ϫ mice was analyzed for E 2 /D 2 -IsoPs.
We propose that the formation of PGs independent of COX involves the generation of two IsoP endoperoxide intermediates (rac-15-H 2t -IsoP and rac-15-H 2c -IsoP) that isomerize to rac-15-E 2t -IsoP and rac-15-D 2c -IsoP, respectively. These eicosanoids subsequently undergo rapid epimerization to compounds identical in all respects to racemic PGE 2 and PGD 2 , respectively (Fig. 2). A number of lines of evidence that we and others have obtained support this proposed mechanism of formation. As noted, we have previously shown that IsoPs contain E/D-, F-, and thromboxane-type prostane rings (7). However, an important distinction between IsoPs and PGs is that IsoP bicycloendoperoxide intermediates contain side chains that are predominantly (Ͼ90%) oriented cis in relation to the prostane ring (4). Indeed, we have recently confirmed that endoperoxides with cis-side chains predominate over trans-side chain compounds when arachidonate is oxidized (5). One IsoP that is formed in abundance in vivo is 15-E 2t -IsoP, which is generated from the endoperoxide intermediate 15-H 2t -IsoP (10). It would also be predicted that the endoperoxide 15-H 2c -IsoP can rearrange to form the analogous D-ring IsoP 15-D 2c -IsoP. In contrast to other types of prostanoids, E 2 /D 2 -IsoPs are ␤-hydroxyketonecontaining compounds that can undergo reversible keto-enol tautomerization under both acidic and basic conditions, allowing rearrangement of the side chains that are initially cis to the more stable trans-configuration. That the trans-configuration is highly favored has been demonstrated by the finding that, when PGE 2 is subjected to conditions that induce keto-enol tautomerism, Ͻ10% of the compound rearranges to the cis-side chain isomer 15-E 2t -IsoP (11). Also, attempts to synthesize 15-D 2c -IsoP have been unsuccessful because epimerization at C-12 readily occurs during synthesis to yield PGD 2 (12).
In this study, we have shown that chemically synthesized 15-E 2t -IsoP is unstable and rapidly epimerizes nonenzymatically to PGE 2 in phosphate buffer at physiological pH. It is likely that the isomerization is further enhanced in the presence of protein-containing biological solutions, which have been shown to facilitate epimerization and dehydration of other eicosanoids (22). Whether the isomerization can be catalyzed enzymatically is unknown. That epimerization of IsoP endoperoxides occurred in the in vitro and in vivo studies reported herein is strongly supported by the fact that comparable amounts of rac-PGE 2 and rac-15-E 2t -IsoP were generated from the peroxidation of arachidonate. In addition, the abundance of rac-PGD 2 lends credence to the hypothesis that epimerization occurs readily. As noted by our findings, the formation of rac-PGD 2 predominates over that of rac-PGE 2 both at base line and after oxidant stress, perhaps because this compound would be predicted to form more readily from the epimerization of rac-15-D 2c -IsoP compared with PGE 2 from 15-E 2t -IsoP. In this study, the lack of a chemically synthesized 15-D 2c -IsoP standard precludes our detection of this compound in vitro and in vivo, although it would be predicted that it would not be present in significant amounts.
Our results also suggest that epimerization of 15-E 2t -IsoP and 15-D 2c -IsoP to PGE 2 and PGD 2 , respectively, occurs to a significant extent while these compounds are esterified in phospholipids based on two lines of evidence. First, a pattern of peaks virtually identical to that shown in the chromatograms in Fig. 5 was obtained after hydrolysis of rat liver phospholipids that had been treated with methyloxime HCl prior to hydrolysis. Second, in control experiments, the conversion of exogenously added 15-E 2t -IsoP to PGE 2 occurred to a negligible extent during sample workup.
A number of important physiological and pharmacological issues emerge from the this study. The first relates to the fact that formation of bioactive PGs occurs in vivo to a significant extent via the IsoP pathway in settings of oxidative stress and potentially in other inflammatory situations. Although levels of PGs derived via this mechanism are low at base line in normal humans and animals, they represent up to 15% of PGD 2 present in the urine of rats treated with CCl 4 , and these PGs are formed independent of COX inhibition. IsoPs have been implicated as mediators of oxidant stress (23)(24)(25). Thus, it will be important to investigate the extent to which not only IsoPs, but PGs, contribute to adverse sequelae of oxidative injury.
Although the biological properties of PGE 2 and PGD 2 have been well characterized (1), our studies suggest that equal amounts of the enantiomers of these PGs are also produced. It will thus be of interest to explore the bioactivity of ent-PGE 2 and ent-PGD 2 . In this respect, the former compound was syn- thesized for the studies reported herein, and experiments to determine its biological relevance will likely yield important insights into its role in oxidative injury. The metabolism of PGE 2 and PGD 2 has been extensively studied in animals and humans (1). The metabolism of parent PGs via the formation of C-13,14-dihydro-15-keto derivatives and subsequent ␤or -oxidation generally renders them inactive. However, this is not the case for the one IsoP whose metabolism has been studied in detail, 15-F 2t -IsoP. The major metabolite of this compound is 2,3-dinor-5,6-dihydro-15-F 2t -IsoP, which results from one step of ␤-oxidation and an unusual C-5-C-6 double bond reduction (26). Interestingly, this metabolite displays bioactivity as a vasoconstrictor similar to that of 15-F 2t -IsoP (27). Thus, studying the metabolism of ent-PGs, in addition to their biological activities, may provide important insights into their role as mediators of oxidant stress. In this regard, we have recently found that, unlike PGE 2 , ent-PGE 2 is a poor substrate for 15-hydroxyprostaglandin dehydrogenase, suggesting that the metabolism of this eicosanoid is significantly different from that PGE 2 . 2 The studies reported herein are highly relevant with regard to human pharmacology in that they suggest that a second pathway operates in vivo to generate PGs and is independent of COX. That this pathway contributes to the formation of PGs in settings of oxidant stress has been discussed above. On the other hand, the extent to which it contributes to PG production in other disease states or at base line has not been elucidated. Administration of nonsteroidal anti-inflammatory agents to humans has been shown to significantly decrease production of PGs and PG metabolites, although the degree of suppression varies depending on the eicosanoid measured. For example, administration of high doses of nonsteroids (e.g. 1.5 g of aspirin or more or the equivalent) to normal human volunteers is associated with a 90% reduction in thromboxane formation and a Ͼ80% reduction in PGI 2 (28 -30). In contrast, the same doses of these agents have been reported to be associated with no greater than a 60% decrease in PGE 2 excretion (28). In this regard, we have made similar observations (31). The reasons for this discrepancy are unknown; but in light of our findings that PGs, particularly PGE 2 , are formed via a non-COX mechanism, it is intriguing to postulate that part of the reason that aspirin-like drugs fail to inhibit PGE 2 production compared with other PGs in certain settings is that the former compound can be produced from IsoP intermediates.
In summary, we report that a second pathway exists for the formation of bioactive PGs in vivo that is independent of COX. This finding is likely of physiological and pharmacological importance because it would be predicted that the generation of PGs via this mechanism would not be inhibited by aspirin or other COX inhibitors. The extent to which formation of PGs independent of COX contributes to human physiology and pathophysiology remains to be elucidated. FIG. 12. Chiral analysis of rat urine for rac-PGD 2 . The putative rac-PGD 2 in each case was purified as noted in Table I and under "Results" and subjected to chiral column chromatography using the solvent system 95:5 (v/v) hexane/isopropyl alcohol. Aliquots of fractions that eluted from the HPLC column were then analyzed for PGD 2 by GC/MS. The peaks indicated by the asterisks co-chromatographed in the HPLC solvent system with chemically pure PGD 2 , whereas the peaks denoted by the plus signs represent ent-PGD 2 . A, chiral analysis of rac-PGD 2 in urine from a normal rat. The predominant stereoisomer is PGD 2 . B, chiral analysis of rac-PGD 2 in urine from a rat treated with CCl 4 (2 ml/kg) to induce oxidant stress. As shown, the relative height of the chromatographic peak representing ent-PGD 2 compared with PGD 2 is significantly higher than that in A. C, chiral analysis of rac-PGD 2 in urine from a rat pretreated with indomethacin prior to receiving CCl 4 . Although the relative height of the peak representing PGD 2 decreased significantly compared with that in B, the height of the peak representing ent-PGD 2 was largely unchanged. See Table III for details.