Isoprostanes: Formation, Analysis and Use As Indices of Lipid Peroxidation in Vivo *

The formation of prostaglandin-like structures as a product of arachidonic acid (AA) peroxidation in vitro was first reported by Mihelich and others (1–4). However, it was Morrow, Roberts, and co-workers (5, 6) who characterized interfering peaks observed in a GC/MS assay for 9a,11b-PGF2a, a metabolite of PGD2 in urine, as isomers of PGF2a. These compounds, termed F2-isoprostanes (F2iPs), possess a 1,3-dihydroxycyclopentane ring (PGF ring) with hydroxyls mainly in the syn configuration and are formed from arachidonic acid by a free radical mechanism (5). Depending upon which of the labile hydrogen atoms is first abstracted by free radical attack, up to 64 isomers in four structural classes can be generated (6). Compounds analogous to the F2-iPs may be formed from other fatty acid substrates (7, 8). Similarly, free radicalderived isomers of other prostaglandins, leukotrienes, and epoxyeicosatrienoic acids have been reported (9–12).

Group III and Group IV, with identical lower side chains to Group III and Group IV derived from AA. The differences will again be in the upper side chains, which will have one less cis double bond and will be two carbons shorter.
Isoprostanes may be formed by either of two routes of peroxidation (14,15), an endoperoxide mechanism (Fig. 2) or a dioxetane/ endoperoxide mechanism (Fig. 3). In the former, the first oxygen molecule is incorporated into the endoperoxide ring to form the two hydroxyl groups on the PGF ring. In the latter, by contrast, it is the second oxygen molecule that is incorporated into the PGF ring. Also 5-and 15-hydroperoxy radicals can only form Groups VI and III by the dioxetane/endoperoxide mechanism. The radical at position 10 of arachidonic acid, by contrast, can yield iPs by both mechanisms. Thus, hydroperoxy radicals formed at 8 and 12 have the option to proceed to form a dioxetane ring (Fig. 3) or a dioxypentane ring (Fig. 2) on a competitive basis, although it is not yet clear which is favored. Recent attention (see below) has focused upon Group VI iPs. These compounds may be derived from a 9-hydroperoxy radical by the endoperoxide mechanism or from a 5-hydroperoxy radical by the dioxetane/endoperoxide mechanism. However, both are derived from an initial hydrogen atom abstraction at position 7 of arachidonic acid. Abstraction at carbon 13 can give rise to 11-and 15-hydroperoxy radicals, yielding only one series (Group III) of iPs. A radical at position 10 of arachidonic acid gives a radical at 8 and 12, which yields groups V and IV, respectively. If the dioxetane mechanism is operative, the same 8-and 12-hydroperoxy radicals will yield Groups IV and V, respectively.

Isoprostane Analysis
Isoprostanes are formed in a free radical-dependent manner and are chemically stable. They are generated initially in cell membranes at the site of free radical attack from which they are cleaved, presumably by phospholipases, circulate, and are excreted in urine. They have also been reported in body fluids besides blood and urine, such as pericardial fluid (16), bile (17,18), lung condensates (19), and cerebrospinal fluid (20,21).
GC/MS Assays-These have been based on the sensitive and specific capillary GC/negative ion electron capture chemical ionization MS technique. Derivatizing the carboxylic acid, common to all eicosanoids, to the pentafluorobenzyl ester is the key to its sensitivity. Bombardment of a moderating gas by an electron beam produces low energy thermal electrons, which are captured by the electrophilic pentafluorobenzyl moiety. This then cleaves, leaving the carboxylate anion which remains to a large degree intact, yielding a spectrum that is dominated by a single ion and is therefore well suited to selected ion monitoring. The closer the homology between analyte(s) and the stable isotope-labeled internal standard, the more reliable the assay. The original assay of Morrow et al. (22) (23) and is still the most widely used reference technique. It requires two solid phase extraction (SPE) steps, two thin layer chromatography (TLC) steps, and two derivatization steps. A large number of overlapping peaks were observed, and one of them (shown to be resolved from the enzymatic isomers known to be present in urine and itself clearly composed of multiple isomers) was chosen for integration and comparison with the internal standard. The only synthetic F 2 -iP available at the time, iPF 2␣ -III (also known as 8-iso-PGF 2␣ ), eluted as one component of the chosen peak. However, because the target compound and the internal standard were heterologous, differences in recovery in any of the four purification steps, especially the TLC, could cause differential recovery of one or more analytes, relative to the internal standard. The identity, number, or TLC retention characteristics of the analytes were not known. A simplified version eliminated the TLC steps (24). This minimized loss of isomers due to differential recovery during purification. However, this method also cannot resolve iPF 2␣ -III from the other isomers. Indeed, application of this and the earlier assay led to the conclusion that iPF 2␣ -III was a major F 2 -iP isomer (25), which turned out not to be the case. Such minimal purification may also confound the sensitivity and reliability of the GC/MS. Our approach has been to synthesize homologous standards (26,27) and to develop assay conditions that permit the quantitation of a single isomer. Given reports that iPF 2␣ -III was a prominent F 2 -iP and that it had bioactivity in vitro and in vivo (28,29), we focused initially on this compound. We developed an assay that seemed to measure a single isomer by synthesizing [ 18 O 2 ]iPF 2␣ -III and improving the GC/MS characteristics by using the tert-butyldimethylsilyl ether, instead of the trimethylsilyl ether (30). In reality, the assay (one SPE step, two TLC steps, and two derivatizations) was technically demanding. However, we found that iPF 2␣ -III, unlike other F 2 -iPs, could be formed by either COX-1 or COX-2 (30, 31), potentially undermining its value as an index of lipid peroxidation in vitro. Therefore, we focused on iPF 2␣ -VI (formerly known as IPF 2␣ -I) (32), which had promise as a target analyte because it could be easily converted to a cyclic lactone, enabling facile separation from F 2 -iPs of classes III, IV, and V. It is present in urine at concentrations higher than iPF 2␣ -III and is not subject to COX-dependent formation (33).
LC/MS/MS Assays-The application of this technique to iP analysis has been pioneered by Murphy and colleagues (11,23,34,35). Using this approach, we have shown 8,12-iso-iPF 2␣ -VI and 5-epi-8,12-iso-iPF 2␣ -VI to be the most abundant F 2 -iPs in human urine (36). Recent advances in electrospray ionization have rendered LC/MS a practical alternative to GC/MS. Because all F 2 -iPs are isomeric and LC is unable to resolve completely all the isomers, the result is unsatisfactory. However, tandem MS adds another level of selectivity (34 -36), permitting virtual separation of the four classes of F 2 -iPs ( Fig. 4A) as well as an increased signal-tonoise ratio. LC/MS is still 2 or 3 orders of magnitude less sensitive than GC/MS. However, sample preparation can consist of a single SPE step, with no derivatization required, so analyte recovery may be an order of magnitude higher. The method is presently sensitive enough to quantitate a single F 2 -iP isomer from 1 ml of urine (Fig.  4B). As tandem MS instrumentation becomes more common, more sensitive, and less expensive, it will likely become an important method for F 2 -iP analysis.
GC/MS/MS Assays-GC/MS/MS has not contributed significantly to F 2 -iP analysis, with only one such assay being published (37). Although in principle the method can have specificity for the four F 2 -iP classes (34,35) using electron impact ionization, the sensitivity of the negative ion electron capture MS technique is sacrificed. When the more sensitive negative ion method is used, the initial ionization yields the carboxylate anion (m/z 353), and the collision-induced dissociation ions monitored in the second quadrupole originate from the loss of (CH 3 ) 3 SiOH groups, common to all isomers and therefore devoid of any structural information. No distinct advantage in F 2 -iP quantitation is obvious. However, any spurious contribution to the F 2 -iP peak by non-iP impurities would be minimized. Specific quantitation of iPF 2␣ -III requires off-line high pressure liquid chromatography purification (37), incompatible with routine sample preparation.
Immuno/GC/MS Assays-An assay for iPF 2␣ -III has been reported (38). The analyte is selectively extracted on an immunoaffinity column, derivatized, and analyzed by GC/MS. The immunoaffinity column effectively replaces the SPE and TLC steps, with significantly more specificity. The columns must be reused, so sample carryover must be monitored, and the columns have a finite lifespan, requiring a constant supply of antibody.
Immunoassays-Adaptation of EIA and RIA from prostaglandin to iP analysis is complex. Traditionally, the antibodies used for eicosanoid analysis have been tested for cross-reactivity with the other major eicosanoids. The degree of cross-reactivity has usually been low because of the major differences in distinct antigenicity. When extrapolating to F 2 -iPs, however, all of the 64 possible isomers share the same basic ring structure, 1,3-syn-hydroxycyclopentane. It is believed that PG antigenicity is largely directed toward the ring, so the possibility of cross-reactivity among F 2 -iPs may be significant. Given that metabolites may also be present in quantities perhaps greater than the parent compounds, the situation becomes even more complex. This does not negate the use of EIA/ RIA in F 2 -iP analysis, but it does introduce some caveats that to date have not been adequately addressed. Many articles have presented "8-epi-PGF 2␣ " (iPF 2␣ -III) levels in various milieu; none have tested the antibody for cross-reactivity with all other F 2 -iPs. They are thus semiquantitative indices of "8-epi-PGF 2␣ -like immunoactivity" unless stringently proven to be measuring iPF 2␣ -III (39). Further, because the degree of antibody cross-reactivity can vary from batch to batch, quantitative comparisons of data from different antibodies should be made with caution. Uncontrolled losses because of sample preparation before quantitation have also often been ignored.
Despite the attraction of iP analysis in general, there remain some caveats. For example, the putative endoperoxide precursor of iPs, analogous to PGG 2 (22), can spontaneously rearrange to PGD 2 / E 2 -iPs or be reduced to F 2-iPs (9). Thus, measurement of F 2 -iPs reflects not only the peroxidation of arachidonic acid but also the redox status of the microenvironment in which peroxidation occurs. Two individuals with identical rates of peroxidation might differ in F 2 -iP generation. An assay which coincidentally measured an F 2 -iP and a D 2 /E 2 -iP might allow correction for such differences in redox status if they occurred. A more immediate consideration is that iPF 2␣ -III can be formed by either COX isoform, in vitro and ex vivo (30,31,37,40), and COX activation and oxidant stress often coincide (41,42). The COX-dependent pathway does not appear to contribute to urinary iPF 2␣ -III, even in syndromes of COX activation (43,44). Finally, little is known about the metabolism of iPs. Infusion of labeled iPF 2␣ -III into non-human primates and a volunteer identified 2,3-dinor-5,6-dihydro-iPF 2␣ -III as a major urinary metabolite (45). Basu (46) has found that ␣-tetranor-15-keto-13,14-dihydro-iPF 2␣ -III is the major urinary metabolite in the rabbit, whereas both the dinor-dihydro and dinor metabolites of iPF 2␣ -III are present in human urine (47).

Isoprostanes as Indices of Oxidant Stress
Altered generation of iPs has been reported in a variety of syndromes putatively associated with oxidant stress. These include coronary ischemia-reperfusion syndromes (48,49), Alzheimer's disease (20,21), adult respiratory distress syndrome, and chronic obstructive pulmonary disease (44,50). Recently, Davi et al. (51) have reported that immunoreactive iPF 2␣ -III in diabetes was depressed, not only by vitamin E administration but also by control of hyperglycemia (52). There is some evidence that iP generation may increase with age (39). Both cigarette smoking (33,38,43,53) and alcohol have been shown to increase iP generation (54,55).
Given the abundant information that oxidation of LDL confers properties thought relevant to atherogenesis (56), there has been interest in the potential of iP measurements both to identify patient populations for interventional studies and for dose selection for antioxidants employed in such clinical trials (57). When LDL is oxidized in vitro, either by exposure to copper or by coincubation with endothelial cells, iPs are formed in a time-dependent manner, along with more conventional indices of lipid peroxidation (24,33,58). However, it is unknown how oxidizability of LDL, either in vitro or ex vivo, relates to actual LDL oxidation in vivo. Isoprostanes are present in human atherosclerotic plaque (59,60), as are isoeicosatrienoic acids, which appear to be even more abundant (12). Isoprostanes circulate in increased amounts esterified in the LDL of patients with hypercholesterolemia (61). Furthermore, urinary iPs are also increased in patients with hypercholesterolemia and appear to correlate with the levels esterified in LDL (61).
To address the hypothesis that urinary iPs might be used for antioxidant dose selection, we studied the hypercholesterolemic, apoE-deficient mouse that develops atherosclerosis-type lesions on a chow diet (62). Dosage of vitamin E was selected such that elevated urinary iPF 2␣ -VI in the apoE knock-outs was suppressed to levels observed in wild type mice. This intervention also suppressed the elevated levels of iPF 2␣ -VI esterified in LDL and in vascular tissue and retarded the development of atherosclerosis, despite persistent hypercholesterolemia (63). The confused picture that has emerged from prospective clinical trials of antioxidants to date may reflect the selection of inappropriate doses and/or inclusion of patients who were not rational targets for antioxidant therapy.

Isoprostanes as Mediators of Oxidant Stress
Several isoprostanes are known to have biological effects in vitro via membrane receptors for prostaglandins. Thus, iPF 2␣ -III is a potent smooth muscle cell constrictor and a mitogen and modulates platelet as well as other cell functions in vitro (64 -67). These effects are prevented by TP antagonists. Although there is some evidence suggesting a distinct receptor for this molecule, it is inconclusive. Although iPs may act as incidental ligands for prostanoid receptors, their effects may differ from the cognate ligand. For example, 8,12-iso-iPF 2␣ -III, which ligates the prostaglandin F 2␣ receptor and causes a hypertrophic response in cardiomyocytes, activates distinct as well as overlapping downstream signaling pathways when compared with PGF 2␣ (68). Given the inflexibility of their structures, iPs may also theoretically contribute to alterations in membrane biophysics under conditions of oxidant stress.
Despite these observations, it is difficult to relate the concentrations of iPs used to evoke biological effects in vitro to what might pertain in vivo. First, a myriad of products is formed under condi- tions of oxidant stress, and yet, to date, studies have concentrated on single isomers. Second, these compounds may be subject to rapid reesterification after release from membrane phospholipid. Finally, the mechanisms and regulation of their release are poorly understood. Nonetheless, there are hints that they might have some relevance in vivo. A TP antagonist is more effective at preventing platelet-dependent coronary occlusion in dogs after thrombolysis than aspirin, despite complete inhibition of Tx formation by the latter (69). This observation suggests activation of the TP by a ligand distinct from TxA 2 and iPF 2␣ -III is known to increase during coronary reperfusion in this model (48). Finally, vitamin E suppresses not only iPF 2␣ -III in patients with diabetes but also the elevated levels of a Tx metabolite. Perhaps, as suggested by Davi et al. (51), platelet-active iPs, such as iPF 2␣ -III, contribute to the enhanced platelet activation in these patients.

Conclusion
Analysis of F 2 -iPs has emerged as a credible approach to the study of lipid peroxidation in vivo. Given the complexity of the iP family, it seems judicious to focus upon sensitive assays of specific isomers and their metabolites. The significance of iPs as biological mediators remains less certain, although the possibility that they might represent a family of primitive autacoids and perhaps, signaling molecules, remains intriguing.