Free Radical-induced Generation of Isoprostanes in Vivo EVIDENCE FOR THE FORMATION OF D-RING AND E-RING ISOPROSTANES*

We recently reported the discovery that a series of novel prostaglandin (PG)F2-like compounds (F2-isopros- tanes) are produced in vivo independent of the cyclooxygenase as products of free radical-catalyzed lipid per- oxidation. F2-isoprostanes are initially formed in situ from arachidonic acid esterified to phospholipids and then released preformed. We have now investigated whether PGDfiz-like isoprostanes are also produced by rearrangement of the PGG2-like intermediates involved in isoprostane formation. Using a variety of approaches utilizing mass spectrometry, compelling evidence was obtained for the presence of Dfi2-isoprostane-contain-ing phosphospholipids in the liver (85 33 ng/g of liver) and free Dfiz-isoprostanes in the circulation (215 f 90 pg/ml) of rata treated with CCl, to induce lipid peroxidation. In untreated rats, levels of Dfiz-isoprostanes esterified in liver phospholipids were much lower (0.90 * 0.10 ng/g), and free compounds could not be detected in the circulation (c5 pg/d). Intergroup multiple comparisons were performed with one-way analy- sis of variances followed by Dunnett's multiple comparison procedure, while, for nonmultiple comparison, Student's unpaired t test was done. Differences were considered significant at p < 0.05.

mechanism of formation is in contradistinction to the formation of cyclooxygenase-derived prostanoids in which arachidonic acid esterified to phospholipids must first be released prior to oxygenation.
Because the formation of isoprostanes proceeds through endoperoxide intermediates, we investigated the possibility that, in addition to F-ring isoprostanes, compounds with a prostane D-ring and E-ring (Dz/Ez-isoprostanes) may also be formed i n vivo by rearrangement of the endoperoxide intermediates. We present evidence that Dz/Ez-isoprostanes are, in fact, produced i n vivo, that they are present both esterified to phospholipids a n d in free form, and that they are capable of exerting potent biological activity.
Analysis of D21E2-isoprostane~-D~2-isoprostanes were analyzed by gas chromatography (GCYnegative ion chemical ionization (NIC1)mass spectrometry (MS) using a modification of methods described previously for the analysis of PGDz and PGE, (10). Briefly, 1.5 ng of [2H,IPGE, internal standard is initially added to a biological fluid and adjusted to pH 3 with 1 M HCI. The sample is then applied to a C-18 Sep-Pak cartridge that has been prewashed with 5 ml of methanol and 4317 5 ml of H20 (pH 3). The cartridge is then washed with 10 ml of HzO (pH 3) followed by 10 ml of heptane, and compounds are then eluted with 10 ml of ethyl acetate and evaporated to dryness under nitrogen. Compounds are subsequently methoximated by treatment with 250 pl of a 2% solution of aqueous methoxyamine.HC1 for 30 min at room temperature. Compounds are then extracted with 1 ml of ethyl acetate and the organic layer evaporated under nitrogen. Compounds are converted to a pentafluorobenzyl (PFB) ester by the addition of 40 pl of a 10% solution of pentafluorobenzyl bromide in acetonitrile and 20 pl of a solution of 10% diisopropylethanolamine in acetonitrile and allowed to incubate for 30 min at 37 "C. Reagents are dried under nitrogen and the residue reconstituted in 30 pl of chloroform and 20 pl of methanol and chromatographed on a silica TLC plate to 13 cm in a solvent system of ethyl acetate:methanol (98:2, v/v). The methyl ester of PGFzQ and the 0-methyloxime, PFB ester derivative of PGD, (approximately 5 pg each) are chromatographed on a separate lane and visualized with 10% phosphomolybdic acid in ethanol by heating. The RF of PGFZQ methyl ester in this solvent system is -0.25, and the RF of the 0-methyloxime, PFB ester derivative of PGD, is -0.60. Compounds migrating in the region 1 cm above the PGFzQ standard to 0.5 cm below the PGD, standard are scraped from the TLC plate, extracted with 1 ml of ethyl acetate, and dried under nitrogen. silyl (TMS) ether derivatives by the addition of 20 pl ofN,O-bidtrimeth-Following TLC purification, compounds are converted to trimethyl-ylsily1)trifluoroacetamide and 10 pl of dimethylformamide. The sample is incubated at 37 "C for 10 min and then dried under nitrogen. The residue is redissolved for GCMS analysis in 10 pl of undecane which has been stored over a bed of calcium hydride. GCNICI-MS is carried out on a Nermag R10-1OC mass spectrometer interfaced with a DEC-PDP computer. GC is performed using a 15-m, 0.25-pn film thickness, DB-1701 fused silica capillary column (J and W Scientific, Folsom CA). The column temperature is programmed from 190 to 300 "C at 20 "C/ min. The major ion generated in the NICI mass spectra of the PFB ester, 0-methyloxime, TMS ether derivative of PGDz and PGE,, which would be the same ion generated by D&,-isoprostanes, is the m/z 524 carboxylate anion [M-181 (M-'CH,C,F,)]. The corresponding ion generated by the [2H41-PGE, internal standard is m/z 528. Levels of endogenous D#,-isoprostanes in a biological sample are calculated from the ratio of intensities of the ions m/z 524 to 528. In some experiments, compounds were reacted with 1-butaneboronic acid or subjected to catalytic hydrogenation following TLC purification as described (11). D&zisoprostanes were also analyzed by GC/electron ionization-MS as methyl ester, 0-methyloxime, TMS ether derivatives. Purification and derivatization of compounds for analysis by GC/electron ionization-MS was as noted above except the methyl ester derivatives were formed by treatment of compounds with excess ethereal diazomethane.
Animal Model of Free Radical-induced Lipid Peroxidation-Free radical-catalyzed lipid peroxidation was induced in rats by intragastric administration of CC14 as described previously (12). In some cases, animals were pretreated with indomethacin with a dosing regimen shown previously to inhibit the cyclooxygenase >90% (5 mg subcutaneously at 24, 12, and 2 h prior to the administration of cc14) (13). At various time intervals, animals were sacrificed and the livers removed, snap frozen in liquid N,, and either processed immediately or stored at -70 "C. In some experiments, blood was also removed from the animals and plasma processed and analyzed for isoprostanes as described above.
Extraction, Purification, and Hydrolysis of Phospholipids-Lipids from the livers of CCI,-treated rats were extracted as described (14). Depending on the experiment, 0.005% butylated hydroxytoluene was added to the lipid extracts during the extraction procedure. The lipid extracts (containing approximately 1 pmol of phospholipid) were then hydrolyzed by reaction with A. mellifera venom phospholipase Az (approximately 200 pg) as described (15) and subsequently analyzed for free Dflz-isoprostanes. As a positive control for phospholipase A Z activity, phosphatidylcholine containing [3Hlarachidonate in the sn-2 POsition was added to the incubation mixture, and the percent of radiolabeled arachidonate released was determined as described (7). In all experiments > 95% of esterified [3H]arachidonate was released. In some experiments, phospholipid extracts were first reduced with NaBH4 and then hydrolyzed using methanolic KOH (7,11).
Analysis oflipid Extracts fiom Livers of CCl,-treated Rats by Normal Phase HPLC-Normal phase HPLC analysis of lipid extracts was performed on a 25-cm x 4.6-mm Econosil SI column (Alltech, Deerfield IL) with 5-pm particles using an isocratic solvent system of hexane: isopropyl alcoho1:water (4:6:1, v/v/v) at a flow rate of 1 mumin. UV absorbance was monitored continuously at 205 nm. Aliquots of fractions eluted were subjected to hydrolysis using bee venom phospholipase Az and then analyzed for free Daz-isoprostanes as described above.
Analysis ofPhospholipids by Negative Liquid Secondary Ion Tandem Mass Spectrometry-Negative liquid secondary ion MS and tandem MS (MS/MS) were carried out on a Kratos Concept I1 HH four-sector tandem mass spectrometer. The instrument was operated with an accelerating voltage of 8 kV. It was equipped with a cesium ion gun that was operated at 5 kV. Samples were dissolved in a matrix of 20% triethanolamine in dimethyl sulfoxide. Collision-induced dissociation was performed in the collision cell situated at the focal point between the two mass spectrometers. Argon was used as the collision gas at a pressure of 5 x torr with the collision cell floated at 4 kV. Precursor and product ion spectra were obtained using the Sun Mach I11 data system; four to six scans were averaged for each spectrum.
Biological Effects of 8-Epi-PGE2 in the Kidney-Renal clearance studies were performed using 200-250 male Sprague-Dawley rats as described elsewhere (16). In the first group, time control rats (n = 31, base-line measurements were performed and repeated during a 30-min infusion of vehicle (0.1 N phosphate-buffered saline). In the second group of rats (n = 9), whole kidney measurements were performed during base-line conditions and during the infusion of 8-epi-PGEz in the following doses; 1.0 pg/kg/min (n = 3), 2.0 pgkg (n = 3), and 4.0 pgkg (n = 3). In a third group of rats (n = 6), 8-epi-PGEz was administered at a rate of 2.0 pgkglmin (n = 3) or 4.0 pgkg/min (n = 3) in the presence of a thromboxane A, receptor antagonist, SQ29548, administered by a continuous intravenous infusion of 3 mgkgh. Treatment with 9629548 was started 30 min before the initial base-line measurements. Glomerular filtration rate (GFR), renal plasma flow (RPF), and arterial blood pressure values were expressed as mean .c S.E. of three determinations. Intergroup multiple comparisons were performed with one-way analysis of variances followed by Dunnett's multiple comparison procedure, while, for nonmultiple comparison, Student's unpaired t test was done. Differences were considered significant at p < 0.05.

RESULTS
Evidence for the Formation of DzlE2-isoprostanes in Vitro during Storage of Plasma-Previously, we had shown that plasma arachidonic acid undergoes autoxidation during storage of plasma at -20 "C for several months, resulting in the formation of large quantities of Fz-isoprostanes (11). Concentrations of Fz-isoprostanes in fresh plasma that has not been stored are approximately 20 pg/ml, and storage of plasma for several months increases levels consistently to greater than 1,000 pg/ml. Therefore, we initially explored whether Dfizisoprostanes are also formed in vitro by analyzing plasma that had been subjected to storage at -20 "C for approximately 6 months. The selected ion current chromatograms obtained from this analysis monitoring rnlz 524 for Dfi,-isoprostanes and rnlz 528 for the [2H4]PGEz internal standard are shown in peak with that of the starred (*) peak in the rnlz 528 chromatogram. The concentration of the compounds in the rnlz 524 chromatogram was calculated to be 2,500 pglml. Four additional plasma samples that had been stored for various periods of time were then analyzed, and the concentrations found ranged from between 90 and 21,000 pg/ml. In contrast, these compounds could not be detected (lower limits of detection = 5 pg/ml) when fresh plasma from normal volunteers was analyzed, suggesting that the compounds detected in stored plasma were formed during storage.
The finding that large quantities of a series of compounds were formed during storage of plasma that had TLC and GUMS properties similar to those of PGEz would be consistent have identical molecular weights and thus generate the same major "181 ion when analyzed by NICI-MS, it is not possible to differentiate whether the putative isoprostane compounds detected in the rnlz 524 ion current chromatogram in Fig. 1 have a D-type or E-type prostane ring. However, since the endoperoxide PGH, derived from the cyclooxygenase rearranges in aqueous solutions to form both PGDz and PGE2 (171, one would expect that the isoprostane endoperoxide intermediates would also rearrange to form both D-ring and E-ring isoprostanes.
Additional experimental approaches were then employed in an attempt to obtain further evidence that the compounds detected in stored plasma were Dz/Ez-isoprostanes. First, no peaks were present when rnlz 523 was monitored, indicating that the rnlz 524 peaks were not natural isotope peaks of compounds generating a n ion less than 524 Da. When the compounds were analyzed as ['Hg]TMS ether derivatives, the rnlz 524 peaks all shifted upwards 18 Da, indicating the compounds have 2 hydroxyl groups. When the compounds were analyzed as a [,H31 0-methyloxime derivative, the rnlz 524 peaks all shifted upward 3 Da, indicating that they contain one carbonyl group. These data thus indicate that these compounds have the same functional groups as PGD, and PGE, and are consistent with their being Dz/En-isoprostanes.
Analysis for the Presence of DzlE2-isoprostanes Esterified to Phospholipids in Vivo-Since the above results suggested that Dz/E,-isoprostanes could be formed in vitro, we investigated whether these compounds may also be formed in vivo. Previously, we had shown that Fz-isoprostanes are initially formed in situ from arachidonic acid esterified in tissue phospholipids and subsequently released preformed (7). Therefore, we examined whether D2Ez-isoprostanes might be esterified to phospholipids in livers of rats that had been treated with CC14 to induce lipid peroxidation. To investigate this, lipids were extracted from the livers, subjected to hydrolysis using A. mellif-

Retention Time (min)
era venom phospholipase A,, and subsequently analyzed for free compounds. Enzymatic hydrolysis of phospholipids was employed to circumvent the problem of dehydration of the prostane DE-rings during chemical hydrolysis with strong base.
The results of this analysis are shown in Fig. 2. A series of rnlz 524 peaks were present in a pattern very similar to that obtained from analysis of stored plasma, although the relative abundances of the various peaks differ somewhat (cf: Fig. 1). Table I compares the amounts of the putative Dz/Ez-isoprostanes with the amounts of F,-isoprostanes measured following hydrolysis of lipids from the same livers of both untreated and CC14-treated rats. The quantities of Dz/Ez-isoprostanes measured following hydrolysis of lipids from livers of CCl.,-treated rats were approximately 100-fold higher than those in untreated rats. Levels of free Dfi2-isoprostanes measured in lipid extracts that were not subjected to hydrolysis were < 1% of the levels measured following hydrolysis (n = 41, suggesting that the compounds detected following treatment with phospholipase A, were released from an acyl linkage on phospholipids. Pretreatment of animals with indomethacin prior to CC14 administration with a dosage regimen shown previously to inhibit cyclooxygenase activity by >90% (13) did not affect levels of the compounds measured ( p > 0.8, Student's t test, n = 41, indicating that the cyclooxygenase enzyme is not involved in their formation. Previously we had shown that butylated hydroxytoluene markedly suppresses the formation of F2-isoprostanes by autoxidation in vitro (11). The presence of butylated hydroxytoluene (0.005%) in the extraction solution, however, did not affect levels of D,/E,-isoprostanes measured ( p > 0.8, n = 41, arguing that these compounds are not formed ex vivo by autoxidation during sample processing. Experiments were then carried out to obtain further evidence for the identity of the compounds represented by the rnlz 524 peaks in Fig. 2 . 3). No peaks were detected at rnlz 526 or 530. This indicated that all of the compounds contained two double bonds. Collectively, these results indicated that the compounds represented by the rnlz 524 peaks contain the same functional groups and number of double bonds as would be expected for DE-ring isoprostanes. An additional experimental approach was employed to obtain further evidence for the presence of DE-ring isoprostanes esterified to phospholipids. In this experiment, phospholipids extracted from livers of CC1,-treated rats were first reduced by treatment with NaBH, and then subjected to hydrolysis with methanolic KOH. The hydrolysate was then analyzed for the presence of Fz-isoprostanes with trans-cyclopentane ring hydroxyls. The rationale of this experimental approach is based on the fact that we had shown previously that the cyclopentane ring hydroxyls of Fz-isoprostanes that are produced in vivo are exclusively oriented cis (5, 7). Chemical reduction of the carbonyl on DE-ring compounds, however, would res_ult in a mixture of F-ring compounds with cis-and trans-cyclopentane ring hydroxyls. Thus, a finding of Fz-isoprostanes with cyclopentane ring hydroxyls oriented trans in the hydrolysate of phospholipids that had first been reduced with N&H4 would provide compelling indirect evidence that DIE-ring isoprostanes were present esterified to the phospholipids. Orientation of the hydroxyls on F-ring prostanoids can be assessed by determining The results of the analyses for Fz-isoprostanes hydrolyzed from phospholipids that had been reduced with NaBH4 are shown in Fig. 4. At the top of Fig. 4A is shown the rnlz 569 selected ion current chromatogram obtained from the analysis for Fz-isoprostanes hydrolyzed from phospholipids that had not been reduced with NaBH,. In this analysis, compounds were not treated with 1-butaneboronic acid. In the lower three chromatograms are the results obtained following treatment of these compounds with n-butylboronic acid. The major ion generated by Fz-isoprostanes as a PFB ester, TMS ether, cyclic butylboronate derivative is the "181 ion, rnlz 491 (11). As is evident, the mlz 569 peaks disappear when reacted with l-butaneboronic acid, and there is a coincident appearance of intense peaks in the rnlz 491 chromatogram shown at the bottom of Fig. 4A. These results thus indicate that essentially all of the Fz-isoprostanes formed a cyclic boronate derivative. Fig. 4B shows the results of similar analyses for Fz-isoprostanes hydrolyzed from phospholipids that had first been reduced with N&H4. The rnlz 569 chromatogram at the top of Fig. 4B was obtained when the compounds were not treated with l-butaneboronic acid. The pattern of the rnlz 569 peaks is similar to that obtained from the analysis of the hydrolysate of lipids that had not been reduced (Fig. 4A, top chromatogram). However, the level of compounds measured is approximately 1.6-fold higher in 4B than in 4 A . In the lower set of three chromatograms are the results obtained from the analysis of these compounds following treatment with 1-butaneboronic acid. Similar to the results in Fig. 4A, intense rnlz 491 peaks appear (Fig. 4 B ,  bottom chromatogram), indicative of compounds that formed a cyclic boronate derivative. In contrast to the results shown in Fig. 4 A , however, peaks also remain in the rnlz 569 chromatogram after treatment with 1-butaneboronic acid. This suggested the presence of F-ring compounds with trans cyclopentane hydroxyls. Of note, the quantity of the compounds that did A. 6. not form a cyclic boronate derivative was calculated to be approximately 215 nglg of liver. Since chemical reduction of the carbonyl on DE-ring compounds produces a mixture of compounds with both cis-and trans-cyclopentane hydroxyls, it can be estimated that the total quantity DE-ring compounds that were present would be approximately 215 x 2 = 450 nglg of liver. This figure is almost exactly the difference found in the levels of F-ring compounds in hydrolysates of lipids that had been reduced with NaBH4 (1230 ng/g of liver; Fig. 4B, top chromatogram) and the levels measured in the hydrolysate of lipids that had not been reduced by NaBH, (750 nglg of liver, Fig. 4A, top chromatogram). This suggests that the quantity of DE-ring isoprostanes that are formed esterified to phospholipids in vivo is approximately 40% of the quantity of esterified Fz-isoprostanes that are formed. However, this may be somewhat of a n underestimate. For our analyses of DE-ring isoprostanes, we have arbitrarily scraped the region of the TLC plate which lies between the location where the methyl ester of PGFz, migrates up to 0.5 cm below the location where the 0-methyloxime, PFB ester derivative of PGDz migrates. However, we have also analyzed the region on the TLC plate 0.5 cm below to 1 cm above where the 0-methyloxime, PFB ester of PGDz migrates for Dflz-isoprostanes and have detected additional rnlz 524 peaks that also elute from the GC with a retention time similar to that of PGEz. Although these compounds may likely represent additional DE-ring isoprostanes, they have not been scrutinized carefully, e.g. as to the number and type of functional groups that they contain, etc. Therefore, we cannot conclude with confidence that these compounds represent additional Dflz-isoprostanes. We have noted, however, that the levels of these compounds increase in parallel in animal models of oxidant injury with the rnlz 524 peaks routinely quantified.

Analysis for the Presence of Free Concentrations of Dz/Ezisoprostanes in the Circulation in CCZ,-treated Rats-We have
demonstrated previously that Fz-isoprostanes are initially formed esterifed to tissue phospholipids in CCL-treated rats and subsequently released into the circulation preformed (7). Since the above results suggested that Dz/Ez-isoprostanes are also formed esterified to phospholipids, we examined whether increased concentrations of these isoprostanes could also be detected free in the circulation of rats 4 h following administration of CC14 to induce lipid peroxidation. As was found in fresh human plasma, compounds presumably representing DE-ring isoprostanes could not be detected in plasma from normal rats that had not been treated with CC14 (lower limit of detection = 5 pglml, n = 4). However, following treatment of rats with CC14, putative Dz/Ez-isoprostanes were detected in plasma at concentrations of 215 3 90 pg/ml ( n = 41, representing a n increase of greater than 43-fold. The pattern of peaks was essentially identical to that shown in Fig. 2. Concentrations of Fz-isoprostanes measured in the same plasma samples were 640 2 70 pg/ml ( n = 4). The increased levels of DE-ring isoprostanes in the circulation of CC14-treated rats were not affected by pretreatment of animals with indomethacin (266 2 72 pg/ml (n. = 3), p > 0.5, Student's t test). As before, analysis of these compounds in the circulation as deuterated TMS ether and deuterated 0-methyloxime derivatives also confirmed that they had, as expected, two hydroxyl groups and one carbonyl group.
Although the above results were consistent with the identity of the compounds detected in the circulation as DzEz-isoprostanes, we also considered the possibility that they might represent metabolites of Fz-isoprostanes in which the side chain hydroxyl had been metabolized by a dehydrogenase to a keto group. First, analogous to cyclooxygenase-derived PGFz, in which metabolism to 15-keto compounds represents a major pathway of metabolism (19), a dehydrogenase may also be involved in metabolism of Fz-isoprostanes. In fact, we reported recently the identification of urinary metabolites of Fz-isopros- tanes which contained a keto group (20). Second, metabolites of Fz-isoprostanes with a keto group would have the same functional groups as Da,-isoprostanes, i.e. two hydroxyl groups and one keto group. These metabolites, therefore, would generate the same "181 ion, mlz 524, as DaZ-isoprostanes when analyzed as a PFB ester, 0-methyloxime, TMS ether derivative by GCLNICI-MS. However, it should be possible to differentiate metabolites of F-ring isoprostanes with a keto group on a side chain from Dfiz-isoprostanes by determining whether the compounds form a cyclic boronate derivative. Cyclic boronate derivatives will only form bridging vicinal hydroxyls or hydroxyls separated by no more than a single carbon atom. Thus, Fa-isoprostane metabolites will form a cyclic boronate derivative, whereas Daz-isoprostanes will not. When we treated the compounds detected in the circulation with 1-butaneboronic acid, they did not form a cyclic boronate derivative (not shown). This result suggested that these compounds were not keto metabolites of Fz-isoprostanes but were DE-ring isoprostanes. Analysis of Lipid Extracts by HPLC for Phospholipids Containing D21E2-isoprostanes-The results obtained up to this point provided evidence suggesting strongly that DzE,-isoprostanes are formed in vivo esterified to phospholipids. We had

Retention Time (min)
demonstrated previously that Fz-isoprostane-containing phospholipids exhibited much more polar characteristics on normal phase HPLC than nonoxidized phospholipids (7). One would expect, therefore, that DE-ring isoprostane containing phospholipids would also exhibit polar characteristics on HPLC. Therefore, to substantiate further the existence of DE-ring isoprostane-containing phospholipids, lipid extracts from livers of CC14-treated rats were subjected to normal phase HPLC analysis using a a solvent system that separates phosphatidylcholine from other phospholipids and neutral lipids (21)(22)(23). To detect lipids containing esterified Dfi,-isoprostanes, aliquots of fractions collected were hydrolyzed with bee venom phospholipase Az and free Dz/Ez-isoprostanes quantified by GCMS. Consistent with the presence of DE-ring isoprostane-containing species of phosphatidylcholine, fractions in which the majority (-70%) of free DzE2-isoprostanes were detected after hydrolysis eluted at a much more polar retention volume (25-41 ml) than nonoxidized phosphatidylcholine, which eluted between 14 and 22 ml (Fig. 5). In addition, fractions also eluted between 2 and 8 ml in which free D2E2-isoprostanes were detected after hydrolysis. Although not analyzed further, these were thought likely to represent phospholipid species other In chromatogram E is shown the quantities of free D&,-isoprostanes detected by G C N S after bee venom hydrolysis of aliquots of fractions that were collected and pooled as indicated by the widths of the burs. As noted, the majority of Da2-isoprostane containing lipids eluted at a more polar retention time (25-41 m l ) than unoxidized phosphatidylcholine.

Analysis of DIE-ring Isoprostanes by Electron Ionization-MS"T0
obtain more direct evidence that the compounds analyzed by selected ion monitoring MS were Dz/Ez-isoprostanes, the compounds hydrolyzed from the polar lipids eluted from the HPLC column shown in Fig. 5 (from about 20 g of liver tissue) were analyzed as a methyl ester, 0-methyloxime, TMS ether derivative by electron ionization MS. This analysis yielded multiple similar mass spectra of compounds eluting over approximately a 40-9 period from the capillary GC column with characteristics of the electron ionization mass spectra of the same derivative of PGDz and PGEz (24). A mass spectrum obtained from the major eluting peak is shown in Fig. 6. The other mass spectra obtained were similar to that shown in Fig.  6, differing primarily in the relative ion abundances of some of the fragment ions. In the mass spectrum shown, there is an intense molecular ion present at the expected rnlz 539. Other characteristic high mass ions present are: rnlz 524 ("15, loss of 'CH3); rnlz 508 ("31, loss of 'OCH3); rnlz 468 ("71, loss of 'CHz(CHz)3CH3); rnlz 449 ("90, loss of Me3SiOH); rnlz 438 ("101, loss of 'CHz(CHz)&OOCH3), rnlz 418 [M-(90+31)1, rnlz 398 ("141, loss of 'CHzCH=CH(CHz)3COOCH3), and rnlz 378 [M-(90+71)1. The intense ions resulting from the loss of 71 Da and the ion at rnlz 398 resulting from the loss of 141 Da suggests that this compound has the same basic structure as cyclooxygenase-derived prostaglandins with a cyclopentane ring at carbons C-8 through C-12. Prominent low mass ions are also present which are also prominent ions in the mass spectra of PGEZ and PGDz such as rnlz 199 representing the lower side chain (+CH=CH-CH(Me3SiOH)(CHz)4)CH3) and rnlz 173 (Me3SiO+=CH(CH2),CH3). The origins of the other low mass ions present remain to be established but presumably arise from coeluting different positional isomeric compounds that are present in this mixture. These data thus provided further evidence for the identity of these compounds as Dz/Ez-isoprostanes.

Analysis of D2 lEz-Isoprostane-containing Phospholipids by Negative Liquid Secondary Zon Tandem Mass Spectrometry-
We sought to obtain further direct evidence for the presence of phospholipids containing esterified Dflz-isoprostanes by analyzing the polar lipids eluted from the HPLC column (see Fig.   5) by negative liquid secondary ion mass spectrometry. The negative liquid secondary ion mass spectrum of the these phospholipid species showed expected high mass intense ions at rnlz 843 and 798, corresponding to "15 [M-CH31-and "60 [M-HN(CH3)3]-, respectively, of phosphatidylcholine containing stearate and a D,/E,-isoprostane (7,25,26). An expected ion at rnlz 772 [M-CHzCHN(CH3)3]-was not sufficiently intense to be observed above background chemical noise. Other structurally significant fragment ions were also present including an ion at rnlz 351, corresponding to the carboxylate anion of a D#,isoprostane, and an ion at rnlz 283, corresponding to the carboxylate anion of stearic acid. Collision-induced dissociation on rnlz 843 resulted in the formation of a series of structurally relevant daughter ions (Fig. 7). In particular, the ions at rnlz 283 and 351 confirmed the presence of stearic acid and a Dz/ Ez-isoprostane, respectively. The prominent ion a t rnlz 208 represents the phosphatidylcholine backbone resulting from the loss of the Dz/E,-isoprostane and stearate. Collision-induced dissociation on rnlz 351 produced daughter ions at rnlz 333 181,168,137,97, and 79. Ions at rnlz 333,315, and 279 are also major fragment ions in the decomposition spectra of authentic PGDz and PGEz (27). Because of the limited amount of material that was analyzed, it difficult to be certain whether all of the other lower mass ions listed were decomposition ions of rnlz 351 or background noise. However, analogous to the previously reported decomposition spectra of rnlz 353, representing the Fz-isoprostane moiety of Fz-isoprostane-containing species of phosphatidylcholine, some of these lower mass ions likely arise from positional isomers of the isoprostane in the mixture (25). These data, therefore, provided further confirmation of the presence of a Dz/Ez-isoprostane esterified to phosphatidylcholine.
Assessment of the Biological Activity of the Ez-isoprostane 8-Epi-PGE, in the Kidney-We had shown previously that the Fz-isoprostane, 8-epi-PGFz,, exerts potent biological activity in the kidney. Specifically it has been found to be an extremely potent renal vasoconstrictor (5). Recently we have obtained compelling evidence that 8-epi-PGFZ, is, in fact, one of the more abundant Fz-isoprostanes that are produced in vivo (28). The intermediate endoperoxide that undergoes reduction to yield 8-epi-PGF,, is 8-epi-PGGz. Since D/E-ring isoprostanes arise from rearrangement of the intermediate endoperoxides, 8-epi-PGGz would also be expected to undergo rearrangement to form 8-epi-PGEz. In this regard, we have been able to obtain preliminary evidence that 8-epi-PGEz is, in fact, one of the E-type isoprostanes that is produced in vivo by demonstrating that the major 0-methyloxime isomer of synthetic 8-epi-PGEz cochromatographs perfectly on capillary GC with the major peak (*) in the upper chromatogram in Fig. 2 (data not shown). The minor 0-methyloxime isomer chromatograms in the middle of the unresolved series of peaks eluting approximately 25 s earlier. Since we had shown previously that 8-epi-PGFz, is a potent vasoconstrictor of renal vasculature, it was of interest, therefore, to explore whether 8-epi-PGEz also exerted biological activity in the kidney.
When infused into the renal artery, 8-epi-PGEz decreased both GFR and RPF in a dose-dependent manner (Fig. 8) with no effect on arterial blood pressure. A significant decrease compared with vehicle alone was identified at every dose used in the study and, at 4 pgkglmin, both parameters decreased by 80%. Simultaneous intravenous administration of SQ29548 at 3 mgkg/h completely abolished this effect at the intrarenal arterial infusion rate of both 2 and 4 pg of 8-epi-PGEz/kg/min with no effect on arterial blood pressure.

DISCUSSION
These studies report the discovery that isoprostanes with DE-type prostane rings are formed in vivo. Analogous to the formation of F,-isoprostanes, DzEz-isoprostanes are also formed in situ esterified to phospholipids and released in free form, presumably by a phospholipase(s). The mechanism involved in the formation of DzE2-isoprostanes is outlined in Fig.  9. This mechanism is identical to that outlined previously for the formation of Fa-isoprostanes through the formation of the bicyclic endoperoxide intermediates (11). In the formation of D&,-isoprostanes, however, the endoperoxides undergo rearrangement to form DE-ring compounds rather than reduction to form F-ring compounds. Analogous to the formation of FZisoprostanes, four regioisomers of D&,-isoprostanes are formed, each of which can theoretically be comprised of eight racemic diastereomers. It is also conceivable that the endoperoxides might be stabilized in the hydrophobic environment when esterified to phospholipids and may exist for prolonged periods of time before rearranging to Dz/Ez-isoprostanes. However, the finding of high concentrations of D,Ez-isoprostanes free in the circulation of rats following administration of CCl4 establishes that the endoperoxides do undergo rearrangement in vivo to form D,E,-isoprostanes. I t has been shown previously that autoxidation of unsatur- ated fatty acids can result in the formation of compounds with a cyclopentane ring (29). Interestingly, it was found that the side chains of these compounds are almost exclusively oriented cis which is in contrast to prostaglandins formed enzymatically by the cyclooxygenase in which the side chains are exclusively oriented trans (29). Based on this information, 8-epi prostaglandins would be expected to be formed as isoprostanes. In this regard, recently we obtained evidence that 8-epi-PGF2, is, in fact, one of the F2-isoprostanes that is produced in abundance in vivo (28). Therefore, it would be expected that 8-epi-PGG2, the endoperoxide intermediate that is reduced to form 8-epi-PGF2,, also undergoes rearrangement to form 8-epi-PGE2. Thus, we examined whether this E2-isoprostane exhibited biological activity. Since 8-epi-PGF2, is a very potent renal vasoconstrictor (51, we thought it of interest to investigate the effects of 8-epi-PGE2 on renal hemodynamics. 8-Epi-PGE2 was also found to be a potent renal vasoconstrictor, albeit somewhat  . " I V less potent than 8-epi-PGF,, (5). This is an important finding because it suggests that the formation of the DD2-isoprostanes is not simply a phenomenon of biochemical curiosity, but one that may also have biological relevance. The finding that 8-epi-PGE, is a vasoconstrictor of renal vasculature is somewhat unexpected since PGE2 derived via the cyclooxygenase is a vasodilator (30). In contrast, both PGFZu derived via the cyclooxygenase and the F2-isoprostane 8-epi-PGF2, are vasoconstrictors (5). Thus, one might anticipate that 8-epi-PGE2 and 8-epi-PGF2, would also have opposing biological effects on vascular smooth muscle. However, this may be an oversimplistic expectation based on the structurdfunction relationships of prostaglandin molecules. We reported previously that the renal vasoconstricting actions of 8-epi-PGF2, could be abrogated by SQ29548, a thromboxane receptor antagonist, suggesting that the vascular actions of 8-epi-PGF,, are mediated via an interaction with thromboxane receptors (16). This was an interest-ing finding in that thromboxane receptors have not been implicated in the vasoconstricting activity of PGFza. More recently, however, we reported that 8-epi-PGFz, appears to interact with a distinct novel receptor on vascular smooth muscle, which is apparently antagonized by SQ29548, which may be similar, but different, from the thromboxane receptor (31). Thus, simple inversion of the stereochemistry of the upper side chain of PGFz, may be a key determinant of receptor interaction. Of interest, we found that the vasoconstricting actions of 8-epi-PGE2 in the kidney could also be abrogated by SQ29548. Future studies investigating whether 8-epi-PGE2 also interacts with the same distinct receptor as 8-epi-PGFz, (owing to the fact that the stereochemistry of the upper side chain is inverted), or interacts with the thromboxane receptor, will be of considerable interest. In this regard, studies investigating further the spectrum of the biological activity of 8-epi-PGE2 and the biological actions of other Dz/E2-isoprostanes will also be of interest and of potential importance since these compounds may participate as mediators in the pathophysiology of oxidant injury.
It should be noted that the quantities of Dz/Ez-isoprostanes that are formed in vivo are only slightly less than the amounts of F2-isoprostanes formed. Since the levels of Fz-isoprostanes in normal human biological fluids are approximately an order of magnitude higher than cyclooxygenase derived prostaglandins ( 5 ) , the amounts of D2/Ez-isoprostanes which are produced in vivo are not trivial. At present, enzymes andlor nonenzymatic substances that may be involved in the reduction of the isoprostane endoperoxide intermediates in vivo to F-ring isoprostanes remain to be identified. Nonetheless, understanding factors regulating the production of DE-ring isoprostanes in relation to F-ring isoprostanes might be important since changes in the relative abundance of these different isoprostanes may have different biological consequences owing to potential variations in their biological actions.
It should also be mentioned that there are potentially important biological ramifications associated with the formation of Dz/Ez-isoprostanes esterified to phospholipids. We reported previously that molecular modeling of phospholipids with F2isoprostanes esterified at the sn-2 position revealed them to be extremely distorted molecules (7). Thus, the formation of isoprostane-containing phospholipids in settings of oxidant stress may have deleterious effects on membrane fluidity and integrity, well recognized sequelae of oxidant injury. Since we have now discovered that Dz/E2-isoprostanes are also formed esterified to phospholipids in only slightly less abundance compared with Fz-isoprostanes, the total quantities of isoprostane-containing phospholipids that may be formed in settings of free radical injury are substantially greater than thought previ-We demonstrated that D2/Ez-isoprostane-containing phospholipids are substrates for bee venom phospholipase in vitro. We also were able to detect free Dz/E2-isoprostanes in the circulation of rats treated with CCll to induce lipid peroxidation. Presumably the free compounds arose primarily from hydrolysis of isoprostane-containing phospholipids. The type(s) of mammalian phospholipase responsible for hydrolysis of Dz/E2isoprostanes from phospholipids has not yet been identified. However, understanding how this process is regulated is of ously.
importance. As discussed above, isoprostane-containing phospholipids are very distorted molecules that should have profound effects on the integrity and fluidity of cellular membranes. However, we have demonstrated that once released in free form, D2/Ez-isoprostanes are capable of exerting biological activity. How these two processes are balanced, therefore, will determine the overall net biological ramifications associated with the formation of isoprostanes in vivo.
In summary, we report the discovery that Dz/E2-isoprostanes capable of exerting potent biological activity are formed in vivo as products of nonenzymatic free radical-catalyzed lipid peroxidation. Analogous to the formation of Fz-isoprostanes, D&,isoprostanes are formed in situ esterified to phospholipids and subsequently released in free form. Further understanding the biological consequences of the formation of these novel compounds and factors that can influence their formation may provide valuable insights into the pathophysiology of oxidant injury.