Protein Adducts of Iso[4]levuglandin E2, a Product of the Isoprostane Pathway, in Oxidized Low Density Lipoprotein*

Levuglandin (LG) E2, a cytotoxic seco prostanoic acid co-generated with prostaglandins by nonenzymatic rearrangements of the cyclooxygenase-derived endoperoxide, prostaglandin H2, avidly binds to proteins. That LGE2-protein adducts can also be generated nonenzymatically is demonstrated by their production during free radical-induced oxidation of low density lipoprotein (LDL). Like oxidized LDL, LGE2-LDL, but not native LDL, undergoes receptor-mediated uptake and impaired processing by macrophage cells. Since radical-induced lipid oxidation produces isomers of prostaglandins, isoprostanes (isoPs), via endoperoxide intermediates, we postulated previously that a similar family of LG isomers, isoLGs, is cogenerated with isoPs. Now iso[4]LGE2-protein epitopes produced by radical-induced oxidation of arachidonic acid in the presence of protein were detected with an enzyme-linked immunosorbent assay. Iso[4]LGE2-protein epitopes are also generated during free radical-induced oxidation of LDL. All of the LGE2isomers generated upon oxidation of LDL are efficiently sequestered by covalent adduction with LDL-based amino groups. The potent electrophilic reactivity of iso-LGs can be anticipated to have biological consequences beyond their obvious potential as markers for specific arachidonate-derived protein modifications that may be of value for the quantitative assessment of oxidative injury.

Oxidative modification of low density lipoprotein (LDL) 1 is considered a key step in the etiology of atherosclerosis (1,2).
Free radical-induced oxidation of LDL consumes polyunsaturated fatty esters and concomitanty generates lipid-derived electrophiles which modify LDL by covalent adduction with protein-based nucleophiles (3)(4)(5). Receptor recognition of the resulting protein modifications leads to uptake of the oxidized (ox) LDL by macrophages (6 -8). Because uptake is unregulated and processing of oxLDL is inefficient, the macrophages become lipid-laden foam cells, progenitors of atherosclerotic plaques (9). To acquire a fundamental molecular level understanding of atherogenesis and other biological sequelae of oxidative injury, we are identifying the chemical structures of lipid oxidation products that bind with proteins.
Previously, we discovered derivatives of levulinaldehyde with prostaglandin side chains appended at the carbons ␣ and ␤ to the aldehyde group. Named levuglandins (10), e.g. LGE 2 ( Fig. 1), these seco prostanoic acids are cogenerated with prostaglandins (PGs) (11)(12)(13) by rearrangements of the endoperoxide PGH 2 which occur readily (t1 ⁄2 ϭ 5 min at 37°C) under the conditions of its cyclooxygenase (COX)-promoted biosynthesis from arachidonic acid (AA). LGE 2 binds avidly with proteins (14) forming a protein-bound pyrrole, LGE 2 -pyrrole (15), as well as protein-protein (16,17) and DNA-protein (18) crosslinks. We recently reported mass spectral characterization of several lysine-based modifications that are generated by covalent adduction of LGE 2 with proteins (19). 2 Levels of LGE 2protein adducts are markedly elevated in the blood of atherosclerosis and end stage renal disease patients versus healthy controls (21). Furthermore, LGE 2 -modified LDL is recognized by macrophages, taken up, and inefficiently processed in close analogy and competition with oxLDL (8). In effect, LGE 2 -modified LDL may function as a Trojan horse, fostering uptake but then compromising the ability of macrophage proteases to hydrolyze oxidatively damaged LDL protein.
Because COX only converts free AA into PGH 2 , this pathway is regulated by enzymatic release of AA from AA-PC (22,23). In contrast, a free radical pathway oxidizes AA-PC directly to produce phospholipid endoperoxides (24 -26). We previously showed that LGE 2 -protein adducts are also produced during free radical-induced oxidation of LDL (27). While the enzymatic pathway generates a single stereoisomer with trans disposed side chains, peroxy radical cyclization generates an isomeric mixture in which stereoisomers with cis disposed side chains predominate (28) as, for example, in the 2-lysophosphatidylcholine (PC) ester 8-epi-PGH 2 -PC (Fig. 1). Rearrangement of 8-epi-PGH 2 -PC would deliver 8-epi-LGE 2 -PC. However, because the stereocenters at positions 8 and 9 are lost during Paal-Knorr condensation (29) of this LG-phospholipid with ly-* This work was supported by the National Institute of General Medical Sciences, National Institutes of the Health, Grants GM21249 (to R. G. S.), GM42056 and GM15431 (to L. J. R.), and HL52012 (to H. F. H). 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 U.S.C. Section 1734 solely to indicate this fact.
Mouse peritoneal macrophages internalize and degrade LGE 2 -LDL if the molar ratio of LGE 2 to LDL protein (apoB) exceeds a threshold somewhere between 10:1 and 38:1 by a receptor mediated uptake that is completely inhibited by ox-LDL (8). Furthermore, uptake of oxLDL is inhibited by LGE 2 -LDL, supporting the conclusion that both LGE 2 -LDL and ox-LDL are recognized by the same receptor. However, the ratio of LGE 2 to apoB in oxLDL does not exceed 2:1. Nevertheless, the total modification of apoB by all of the isomeric levulinaldehyde derivatives produced by oxidation of AA might suffice to account for receptor recognition, uptake, and inefficient processing of oxLDL. Thus, because hydrogen atom abstraction readily occurs nonregioselectively at any doubly allylic methylene, we postulated that the free radical pathway not only can produce a stereoisomeric mixture of levulinaldehyde derivatives with PG side chains, i.e. iso-LGs, but also structurally isomeric levulinaldehyde derivatives with nonprostanoid side chains, i.e. iso [n]LGs. For example, hydrogen atom abstraction from the 10-position of AA-PC followed by cyclization of an intermediate 8-peroxyeicosatetraenoyl radical could lead to iso[4]PGH 2 -PC and then iso [4]LGE 2 -PC ( Fig. 1), where the number in brackets signifies the length of the carboxylic side chain appended to a common 2,3-dioxabicyclo[2.2.1]heptane or levulinaldehyde core. The generation of phospholipid endoperoxides that are structural isomers of PGH 2 by free radical-induced oxidation was postulated previously to account for the formation of isoprostanes (24,25). Thus, iso[4]PGH 2 -PC (12-H 2 -IsoP) is also the putative precursor of isoprostanes that have been designated 12-F 2 -IsoP, 12-E 2 -IsoP, and 12-D 2 -IsoP (33). In analogy with the chemistry of LGE 2 , we expected that iso [4]LGE 2 -PC would form iso [4]LGE 2 -pyrrole by covalent adduction to proteins and concomitant phospholipolysis (Fig.  1). We now report confirmation of this hypothesis. Thus, the generation of iso [4]LGE 2 -protein epitopes during in vitro nonenzy-matic free radical-induced oxidation of LDL was detected with an immunoassay using antibodies raised against an iso [4]LGE 2 -protein adduct. Since iso [4]LGE 2 is formed by the isoprostane pathway but not by the COX pathway, the new antibody allows unambiguous assessment of the formation of iso-LGs from the isoprostane pathway. In a companion paper (19), we report mass spectral characterization of the covalent iso-LG-derived protein modifications that are generated during free-radical induced oxidation of LDL.

General Methods
Centrifugation was done on a Sorvall centrifuge at 5°C and 2000 rpm. Absorbance values of enzyme-linked immunosorbent assays (ELISAs) were measured on a Bio-Rad Microplate Reader using dual wavelength (405 nm to read the plate and 650 nm as a reference).

-Protein Adducts
An Ehrlich pyrrole assay (40,41) was performed to determine the concentration of protein-bound iso [4]LGE 2 -derived pyrroles as described previously for determining the concentration of LGE 2 -derived pyrroles (15,27). Tritium-labeled LGE 2 -HSA and LGE 2 -BSA were used as a standards for the assay. The amount of LGE 2 bound to HSA or BSA was determined by quantitative radiochemical analysis. The data for the standards (Fig. 2) fits the equation: [pyrrole (mol)] ϭ 2.22 (absorbance at 586 nm). The absorption maximum ( max ) for LGE 2 -HSA-derived and iso [4]LGE 2 -BSA-derived Ehrlich pyrrole assay chromophores were 586 and 584 nm, respectively. The concentration of pyrrole was presumed to be equal to the concentration of LGE 2 that is bound to BSA or HSA. This assumes a quantitative yield of protein-bound LGE 2derived pyrrole. Thus, the concentration of pyrrole estimated by Ehrlich assay is an upper limit for protein-bound iso [4]LGE 2 -derived pyrrole. The pyrrole concentration in the iso [4]LGE 2 -KLH solution is 0.52 mM KLH-bound iso [4]LGE 2 .

Iso[4]LGE 2 -BSA Coating Agent
A PBS solution containing 4.54 mM iso [4]LGE 2 (1.6 mg, 4.54 mol) and 0.2 mM BSA (13.2 mg, 11.7 mol of lysyl residues) was incubated at 37°C for 12 h. The solution was then dialyzed against PBS (4 ϫ 500 ml) over 48 h at room temperature. After dialysis, the final volume of the solution was adjusted to 6 ml. The final protein concentration, determined using the Pierce BCA protein assay as described above, was 1.85 mg/ml. The pyrrole concentration, 0.59 mM BSA-bound iso [4]LGE 2 , was determined by an Ehrlich assay using LGE 2 -BSA and LGE 2 -HSA as standards.

Iso[4]LGE 2 -HSA Standard
A PBS solution containing 3.55 mM iso [4]LGE 2 (2.5 mg, 7.10 mol) and 0.16 mM HSA (20.6 mg, 18.30 mol of lysyl residues) was incubated at 37°C for 16 h. The solution was then dialyzed against PBS (3 ϫ 1 liters) for 48 h at room temperature. The final volume of the solution was adjusted to 10 ml. The final protein concentration, determined using Pierce BCA protein assay as described above, was 1.81 mg/ml. The pyrrole concentration, 0.52 mM HSA-bound iso [4]LGE 2 , was determined by Ehrlich assay using LGE 2 -BSA and LGE 2 -HSA as standards.

Immunization
The immunogen, iso [4]LGE 2 -KLH (5.0 mg) containing 0.75 mol of iso [4]LGE 2 per mg of KLH, was diluted to 5 ml with pH 7.4 PBS. An aliquot (500 l) was emulsified in Freund's complete adjuvant (500 l). Each of two Pasturella free, New Zealand White rabbits (Hazelton) were inoculated intradermally into several sites on the back (200 l) and rear leg (200 l). Booster injections of iso [4]LGE 2 -KLH with Freund's incomplete adjuvant were given every 21 days. Antibody titer was monitored 10 days after each inoculation by ELISA as described below.

Antibody Purification
The iso [4]LGE 2 -KLH immune serum from the 73 day bleeding of rabbit 1, containing 34.4 mg/ml protein, as determined by absorbance at 280 nm (A 280 ϭ 1.35 for 1.0 mg/ml), was purified using a protein G column as described previously (21). The resulting antibody solution (8.75 ml) contained 1.47 mg/ml purified IgGs. This corresponded to 13.3% of the total protein in the immune serum.

ELISA
For all ELISAs, unless otherwise noted, duplicates of each sample were run on the same plate.
Antibody Titers-For determination of antibody levels in rabbit blood serum, iso [4]LGE 2 -BSA containing 10 mol of pyrrole/mol of protein, was used as coating agent. The iso [4]LGE 2 -BSA conjugate (100 l of a solution containing 4.4 mg/ml in pH 7.4 PBS) was added to each well of a sterilized Baxter ELISA plate. The plate was then incubated at 37°C for 1 h in a moist chamber. After discarding the coating solution, each well was washed with PBS (3 ϫ 300 l), then filled with 1.0% CEO in PBS (300 l), and incubated at 37°C for 1 h to block remaining active sites on the plastic phase. Each well was washed with 0.1% CEO in PBS (300 L) and then 100 l of rabbit serum from each bleeding diluted 1:10,000 with 0.2% CEO in PBS, or 0.2% CEO in PBS without serum for a blank, was dispensed into the sample wells. Normal rabbit, i.e. prior to inoculation with antigen, serum diluted as above was employed as a negative response control. The ELISA was completed as described previously (21). The antibody titer rose abruptly after 3 weeks, reaching a plateau within about 30 days (Fig. 3).
Competitive Antibody Binding Inhibition Studies-For antibody binding inhibition studies to measure cross-reactivities, an iso [4]LGE 2 -BSA adduct was used as coating agent and iso [4]LGE 2 -HSA was used as a standard. On each ELISA plate, a blank, a positive control containing no inhibitor, and up to 10 serial dilutions of each inhibitor and the iso [4]LGE 2 -HSA standard were run. The standard was prepared by diluting a 1.04 mM HSA-bound iso [4]LGE 2 solution in PBS to 104 M with pH 7.4 PBS. A serial dilution of factor 0.2 was used. Each well of the plate was coated with iso [4]LGE 2 -BSA solution (100 l), prepared by diluting a solution containing 1.08 mM BSA-bound iso [4]LGE 2 in PBS to 432 nM with pH 7.4 PBS. The plate was covered with a plastic lid and placed in incubator at 37°C for 1 h, and then allowed to come to room temperature. After discarding the supernatant, each well was washed with pH 7.4 PBS (3 ϫ 300 l) and then blocked by incubating 1 h at 37°C with 300 l of 1% CEO in pH 7.4 PBS. After coming to room temperature, the supernatant was discarded and the wells rinsed with 0.1% CEO in pH 7.4 PBS (300 l). For each sample and the iso [4]LGE 2 -HSA standard, the undiluted sample solution (150 l) and aliquots (150 l) of up to nine 1:10 serial dilutions with 5 mM pH 7.4 PBS were incubated in test tubes at 37°C for 1 h with antibody solution (150 l) that was prepared by adding the required amount of protein G column purified antibody (0.294 g/ml) in pH 7.4 PBS to 0.2% CEO in pH 7.4 PBS (2.8 l/14 ml of 2% CEO). The remaining ELISA procedure and similar antibody binding inhibition studies with LGE 2 -KLH antibody, LGE 2 -BSA adduct as coating agent, and LGE 2 -HSA as standard were performed as described previously (21).
Cross-reactivity of LGE 2 -Lysine Lactam and Hydroxylactam-A mixture of LGE 2 -lysine lactam and hydroxylactam adducts was obtained by incubating LGE 2 with [ 3 H]lysine (27,00 cpm/g) under argon overnight at 37°C. The mixture was applied to a C 18 SepPak cartridge (Waters) that had been preconditioned with methanol (5 ml) and water (10 ml). The SepPak was washed with heptane (10 ml) and heptane/ethyl acetate (1:1, v/v, 10 ml) before elution with methanol/ethyl acetate (2:3, v/v, 10 ml). The eluate was dried, resuspended in 10% aqueous methanol, and subjected to HPLC (4.6 ϫ 250 mm Macrosphere 300 C 18 column from Alltech; 10 min in 0.1% aqueous acetic acid, then 30% acetonitrile in 0.1% aqueous acetic acid; 1 ml/min). Fractions (1 ml) were collected and aliquots subjected to scintillation counting. Aliquots of fractions exhibiting UV absorbance at 205 nm (Fig. 4A) and containing radioactivity, which indicates the incorporation of lysine, were assessed by LC/MS (Fig. 4, B and C) by direct infusion using a Finnigan TSQ7000 spectrometer with the sheath gas held at 70 p.s.i., auxiliary gas at 10 p.s.i., and with 25 volts on the capillary, a capillary temperature of 220°C, and the tube lens voltage at 90 V. Fractions deemed to contain only compounds with the molecular ion of the lactam (m/z 479.4) or hydroxylactam (m/z 495.4) adducts and were combined. None of these fractions contained detectable amounts of the LGE 2 The resulting mixture of lactam and hydroxylactam adducts was again analyzed by LC/MS, revealing no contamination by compounds with the molecular ion of the pyrrole adduct (m/z 463). The putative lactam and hydroxylactam adducts were also subjected to collisioninduced dissociation, which resulted in fragmentation patterns consistent with the structures assigned to the compounds (19). A sample containing 50 ng each of LGE 2 -lysine lactam and hydroxylactam was dissolved in PBS/EtOH (75 l, 4:1, v/v) to give 140 pmol/well as final concentration. LGE 2 -BSA was used as coating agent and LGE 2 -HSA was used as standard to measure cross-reactivity. The ELISA was done as under "Competitive Antibody Binding Inhibition Studies." Autoxidation of Polyunsaturated Fatty Acids (PUFAs) in the Presence of HSA-Fatty acid (2 mg) and HSA (30 mg, 0.45 mol) were dissolved in 0.1 M PBS (10 ml). Autoxidation was started by addition of 20 mM sodium ascorbate (510 l) and 0.8 mM FeSO 4 ⅐7H 2 O (510 l) (42) . The solutions were incubated at 37°C for 24 h under air. After incubation the reaction was quenched by adding 1 mM EDTA (200 l) to each solution which then was dialyzed against pH 7.4 PBS (2 ϫ 2 liters) for 40 h at room temperature. Samples of PUFAs that had been oxidized in the presence of HSA were analyzed for LGE 2 -, iso [4]LGE 2 -, and HNE-derived epitopes by ELISAs using LGE 2 -KLH (27), iso [4]LGE 2 -KLH, or ON-KLH (36) antibodies, respectively, in pH 7.4 PBS containing 0.001% Tween TM 20 and 0.2% CEO.
LGE 2 -and Iso [4]LGE 2 -Protein Immunoreactivity in OxLDL-ELISA of oxLDL was performed the same as the inhibition assays, except a dilution factor of 0.3 was employed. The starting concentration was the undiluted samples. The time dependence of appearance of proteinbound LGE 2 -and iso [4]LGE 2 -derived epitopes during oxidation of LDL was determined as described in our previous study of LGE 2 -pyrrole generation during oxidation of LDL (27).
Trapping ELISA Detection of Free LGE 2 -To detect any free LGE 2 that may be released upon oxidation of LDL, a trapping ELISA was done on the ultrafiltrate from oxLDL. Thus, LDL (0.5 mg/ml) was dialyzed at 5°C for 5 h against pH 7.4 PBS (4 liters), and then for 12 h against fresh buffer (4 liters). The LDL was then incubated at 37°C with 10 M CuSO 4 . The reaction product mixture was then filtered using an Ultrafree-CL filter unit (NMWL: 10,000) for 3 h in a Beckmann centrifuge at 5°C and 4,000 rpm. Each well of a microtiter plate was coated with 100 l of BSA (1 mg/ml) in pH 7.4 PBS and was incubated at 37°C for 1 h. Following washing once with PBS, samples for a standard curve containing LGE 2 (0 -35 pmol/well), or the filtrate from oxLDL were added to the wells (100 l/well). After incubation for 3 h at 37°C followed by washing once with PBS, each well was filled with 300 l of 1% CEO for 1 h at 37°C. After washing once with 0.1% CEO, 100 l of KLH-LGE 2 antibody was added to each well and the plate was gently shaken for 1 h at room temperature. After three washes with 0.1% CEO, 100 l/well of goat anti-rabbit IgG-alkaline phosphatase (1:1,000) was added and the mixture was incubated for 1 h at room temperature. After washing three times with 0.1% CEO, 100 l of disodium p-nitrophenyl phosphate (10 mg) in water (11 ml, pH adjusted to 9.6 using NaOH) containing glycine (50 mM) and MgCl 2 (1 mM) were added and the resulting mixture was incubated for about 20 min at room temperature. The reaction was terminated by adding 3 M NaOH (50 l) to each well, and the absorbance was read at 405 nm on a micro-ELISA plate reader. A standard curve, constructed from absorbance data for solutions containing 0 -35 pmol/well of LGE 2 (Fig. 5) showed a linear increase in absorbance with LGE 2 concentration in the standard solutions. No absorbance was observed for any of the wells treated with ultrafiltrate from oxLDL. [4]LGE 2 -protein Adducts-Iso [4]LGE 2 is a chemically sensitive vinylogous ␤-hydroxy aldehyde that was freshly prepared for reaction with proteins (BSA, HSA, and KLH) to afford iso [4]LGE 2 -protein adducts. We previously showed that for high LGE 2 /protein ratios, Paal-Knorr condensation of LGE 2 with ⑀-amino groups of lysyl residues of proteins gives mainly LGE 2 -derived protein-bound pyrrole (21). Earlier studies also demonstrated that quantitative analysis of LGE 2derived protein-bound pyrroles can be accomplished using the Ehrlich assay that measures the absorbance of a blue-green chromophore generated by the condensation of LGE 2 -pyrrole with DMAB (15).

Synthesis of Iso
For the present study, iso [4]LGE 2 -protein adducts, rich in iso [4]LGE 2 -pyrrole, were prepared by exposing various proteins to an excess of iso [4]LGE 2 . The levels of protein-bound iso [4]LGE 2 -derived pyrrole in these adducts were determined by Ehrlich assays (40, 41) using LGE 2 -protein adducts as standards since the availability of radiolabeled LGE 2 allowed an accurate independent assessment of LGE 2 content in these standard samples. As expected, the chromophore generated by the condensation of iso [4]LGE 2 -pyrrole with DMAB is very similar to that from LGE 2 -pyrrole. Thus, the absorption maxima ( max ) observed for the LGE 2 -HSA-DMAB and iso [4]LGE 2 -BSA-DMAB chromophores are 586 and 584 nm, respectively. It is reasonable to presume that the structurally similar Ehrlich chromophore derived from an iso [4]LGE 2 -pyrrole has the same extinction coefficient as that derived from an LGE 2 -pyrrole.
A linear correlation was obtained for a plot of pyrrole concentration verses absorbance at 586 nm for the DMAB chromophore of LGE 2 -derived protein-bound pyrroles in LGE 2 -BSA and LGE 2 -HSA (see Fig. 2). The concentration of LGE 2 -derived protein-bound pyrrole in LGE 2 -HSA was taken to be equal to the total amount of protein-bound LGE 2 (0 -500 nmol/sample) as determined by quantitative radiochemical analysis. This assumes a quantitative yield for pyrrole formation. Therefore, the use of LGE 2 -protein-derived pyrrole as a standard for the Ehrlich assay provides an upper limit for the concentration of iso [4]LGE 2 -protein-derived pyrrole. The final protein concentrations in iso [4]LGE 2 -protein adducts were determined by BCA protein assay (39) and the ratios of iso [4]LGE 2 -pyrrole to protein were calculated (Table I).
Lactam, and Hydroxylactam Epitopes in LGE 2 -Protein Adducts-Studies detailed elsewhere (19), employing mass spectral detection of lipid-modified lysine to characterize epitopes generated by covalent adduction of LGE 2 with proteins, uncovered oxidative modifications that append one or two atoms of oxygen to protein-bound LGE 2 -derived pyrroles. Thus, while LGE 2 -lysine adduct containing the expected lysine-LGE 2 -pyrrole could be prepared if oxygen is rigorously excluded, expo-sure to air or enzymatic proteolysis of LGE 2 -protein adduct produced only mono-and dioxygenated lysine-LGE 2 -pyrrole. These oxidized pyrroles almost certainly are lactams and hydroxylactams (Fig. 6) generated by well known free radicalinitiated reactions of molecular oxygen with electron-rich pyrroles (45,46). These oxidized derivatives of LGE 2 -pyrrole cross-react strongly with LGE 2 -KLH antibodies. Thus, a sample containing a mixture of lysine-LGE 2 -lactam and lysine-LGE 2 -hydroxylactam was isolated by HPLC from a Paal-Knorr condensation of LGE 2 with radiolabeled lysine and subsequent oxidation by adventitous oxygen. Quantitative radiochemical analysis in conjunction with an ELISA comparison of binding with LGE 2 -KLH antibodies, showed 256% cross-reactivity for the hydroxylactam-lactam mixture relative to the LGE 2 -HSA standard.
Generation of LGE 2 -HSA and Iso [4]LGE 2 -HSA Immunoreactivity by Fe 2ϩ -catalyzed Oxidation of AA but Not Linoleic Acid (LA) or Docosahexaenoic Acid (DHA)-In vitro free radical oxidations of a variety of PUFAs with iron and ascorbate were performed in the presence of HSA. Immunoreactive proteinbound epitopes were detected by ELISAs with LGE 2 -KLH ( Fig.  9) and iso [4]LGE 2 -KLH (Fig. 10) antibodies in the reaction product mixture from AA but not in the reaction product mixture from LA. Similar experiments with ␥-linolenic (␥-LA), DHA, ETA, and EPA acids revealed the generation of protein epitopes that cross-react with LGE 2 -KLH and iso [4]LGE 2 -KLH antibodies from ␥-LA, ETA, and EPA, but not DHA (see below).
LGE 2 -Protein and Iso [4]LGE 2 -Protein Adduct Immunoreactivity in OxLDL-LDL was oxidized by dialyzing an aqueous solution of LDL in air against a buffer containing Cu 2ϩ , an in vitro model (2) for physiological oxidation of LDL. Oxidation was halted after various time periods by sequestration of Cu 2ϩ with Na 2 EDTA added to an aliquot of the reaction mixture. After an induction period, during which the endogenous antioxidants presumably were consumed, immunoreactivity toward both LGE 2 -KLH (Fig. 11) and iso [4]LGE 2 -KLH (Fig. 12) antibodies increased rapidly, reaching a plateau after several hours. The immunoreactivity detected for LGE 2 -protein and iso [4]LGE 2 -protein epitopes in the oxLDL corresponded to a final ratio of 1:4, respectively.
Free LGE 2 Is Not Present in OxLDL-A trapping ELISA was used detect any free LGE 2 that might be present in the reaction product mixture generated by in vitro oxidation of LDL in the presence of Cu 2ϩ . Free LGE 2 can be trapped by the protein coating agent (BSA) to give immunoreactive LGE 2 -pyrrole epitope. Thus, a linear increase in absorbance was found for increasing concentrations of free LGE 2 (see Fig. 5). However, the wells treated with ultrafiltrate from oxLDL showed no absorbance, indicating that they contained no free LGE 2 .

DISCUSSION
IsoLGs-The chemistry of LDL oxidation is quite complex. A plethora of lipid oxidation products is generated, and some of these covalently modify LDL protein, apolipoprotein (apo) B (3,4). Two aldehydic fragmentation products, MDA and HNE, have been studied extensively because they form adducts with apoB, and because the MDA-LDL (47) and HNE-LDL (48) adducts could be atherogenic, in contrast with native LDL. Besides protein-bound HNE, free HNE is detectable in oxLDL. In a recent study, free HNE was quantitatively analyzed by an "HNE-trapping ELISA" based on the detection of epitopes generated when HNE is trapped by a protein that has been coated   onto an immunoplate (49). This study demonstrated that a considerable amount of free HNE is released from human plasma LDL upon Cu 2ϩ -promoted oxidation. In contrast, employing an analogous LGE 2 -trapping-ELISA, we now find no evidence for the presence of free LGE 2 in LDL that has undergone Cu 2ϩ -promoted oxidation. This is expected because, as we have noted elsewhere (8,19), LGE 2 binds with proteins far more avidly than HNE. There is a physiological steady-state concentration of free HNE in human venous blood plasma (50,51). In contrast, the generation of LGs and iso [n]LGs in vivo must be inferred from detection of protein-bound derivatives. Studies employing the new iso [4]LGE 2 -KLH antibodies to detect iso [4]LGE 2 -derived protein epitopes in vivo are in progress. Preliminary results show that these protein modifications are present in human blood plasma, confirming the hypothesis that a family of levulinaldehyde derivatives is generated in vivo by a free radical-induced oxidation of AA-PC (Fig. 13). Thus, nonregioselective hydrogen atom abstraction from the 7, 10, and 13 positions of an arachidonyl ester produces three regioisomeric pentadienyl radicals. These then react with molecular oxygen to afford four regioisomeric peroxyeicosatetraenoyl radicals that undergo peroxy radical cyclization (28,52) to deliver four structurally isomeric endoperoxides. Besides the geometrically enforced cis relationship of the endoperoxide oxygens and a preference for peroxy radical cyclization to produce stereoisomers with cis disposed side chains (28), each structurally isomeric endoperoxide is expected to be generated as a mixture of 16 stereoisomers that are referred to collectively as isoPGH 2 or iso[n]PGH 2 where [n] specifies the number of carbon atoms in the carboxyl side chain of the non-prostanoid structural isomers. Each endoperoxide rearranges to form two structurally isomeric isoLGs or iso [n]LGs, designated as E series if the acetyl substituent is nearer than the formyl substituent to the carboxyl group or as D series if the formyl is nearer than the acetyl to the carboxyl. Paal-Knorr condensation of the eight structurally isomeric  isoLGs and iso [n]LGs with protein primary amino groups produces eight different pyrrole epitopes. We previously reported chemical evidence for the initial formation of pyrroles that incorporate the ⑀-amino group of protein lysyl residues (15). Our recent studies employing mass spectral detection of lipidmodified lysine uncovered the facile oxidation of LGE 2 -derived pyrroles leading to lactam and hydroxylactam derivatives, and confirmed that isoLG-derived lysyl group modifications are present in oxLDL (19). Those studies also demonstrated the formation of LGD 2 epimers in the free radical-induced oxidation of AA. Since LGD 2 -protein and LGE 2 -protein adducts can be produced by the enzymatic COX pathway, only detection of iso [n]LG-protein adducts, i.e. with nonprostanoid side chains, can provide unambiguous evidence for the operation in vivo of the free radical-promoted oxidative pathway summarized in Fig. 13. We now have two orthogonal polyclonal rabbit antibodies, i.e. that recognize and strongly discriminate between, LGE 2 -protein and iso [4]LGE 2 -protein adducts.

FIG. 10. Inhibition curves for binding of anti-iso[4]LGE 2 -KLH to iso[4]LGE 2 -BSA by iso[4]LGE 2 -HSA standard (q), iso[4]LGE
As expected, LGE 2 -protein and iso [4]LGE 2 -protein immunoreactivity are produced by free radical oxidation of AA but not LA, the most abundant polyunsaturated fatty acid in LDL. We now have a panel of five antibodies that specifically detect epitopes produced by the adduction of different lipid oxidation products with proteins (Table III). In the reaction product mixture from in vitro oxidation of AA in the presence of HSA, pyrrole epitopes derived from HNE and 5-hydroxy-8-oxo-6-octenoic acid (HOOA) were detected previously. HNE-pyrrole was also detected in the reaction product mixture from in vitro oxidation of LA in the presence of HSA. On the other hand, pyrrole epitopes derived from 9-hydroxy-12-oxo-10-dodecenoic acid (HODA) are a selective marker for LA oxidation in the presence of protein.
HNE-pyrrole epitope (detected with ON-KLH antibody) was generated in the oxidation of -6 but not -3 PUFAs in the presence of HSA. Thus, the -6 acids LA, ␥-LA, AA, and ETA all afforded immunoreactivity detectable with ON-KLH antibodies while the -3 acids DHA and EPA did not (Table IV). Although LA and AA are the major PUFAs in normal human serum phospholipids, oxidative cleavage of ETA in vivo may produce significant amounts of HNE.
Oxidation of ␥-LA, ETA, and EPA in the presence of HSA produces protein epitopes that cross-react with LGE 2 -KLH and iso [4]LGE 2 -KLH antibodies. The levels of ETA and EPA in human LDL vary greatly with diet (Table IV) and, therefore, the LDL from some individuals can contain levels of these PUFAs that may contribute significantly to the generation of LGE 2 -KLH or iso [4]LGE 2 -KLH immunoreactivity. The selective generation of iso [4]LGE 2 -KLH immunoreactivity from ␥-LA, LGE 2 -KLH immunoreactivity from ETA, and both LGE 2 -KLH and iso [4]LGE 2 -KLH immunoreactivity from EPA (Table  IV) is a reasonable consequence of the fact that only a close structural analogue of iso [4]LGE 2 is expected to be generated upon oxidation of ␥-LA, a LGE 2 analogue upon oxidation of ETA, and analogues of both iso [4]LGE 2 and LGE 2 upon oxidation of EPA (Fig. 14).
LGE 2 and Iso [4]LGE 2 Epitope Families-Owing to concerns that LG-derived protein-bound pyrroles would be readily modified by oxidation, our earliest efforts to detect LGE 2 -derived protein epitopes immunologically relied upon cross-reactivity of those epitopes with antibodies raised against a stable pyrazole isostere-derived antigen (53). Quite unexpectedly, the immunoreactivity generated by the reaction of LGE 2 with proteins showed no decrease over several weeks. While this could be the result of some stabilizing influence of the protein matrix on an otherwise readily oxidizable pyrrole hapten, we recognized the possibility of an alternative explanation. Thus, if the molecular fragment responsible for antibody recognition is preserved in secondary products derived from the initially formed pyrroles, e.g. the corresponding lactam or hydroxylactam (Fig.  6), in particular two prostanoid side chains appended to neighboring sp 2 carbons on a five-membered ring, then large changes in antibody binding need not accompany transformations of the LGE 2 -pyrrole into these secondary products. Thus, in contrast with the excellent discrimination for variations in the side chains appended to the pyrrole ring at positions 3 and 4, both the LGE 2 -KLH and iso [4]LGE 2 -KLH antibodies could show a high tolerance for modifications at the 2 and 5 positions of the pyrrole ring. Furthermore, the LGE 2 -pyrrole and iso [4]LGE 2pyrrole antigens most probably were oxidized after administration to rabbits, and therefore, some or all of the LGE 2 -KLH and iso [4]LGE 2 -KLH antibodies in the polyclonal mixtures were raised against lactam or hydroxylactam epitopes. Since the side chains on the pyrrole, lactam, and hydroxylactam epitopes are appended to coplanar sp 2 -hybridized carbons, they are restricted to the same coplanar geometry. This conformational rigidity is probably responsible for the excellent discrimination by LGE 2 -KLH and iso [4]LGE 2 -KLH antibodies for LGE 2 -and iso [4]LGE 2 -derived haptens, respectively. Thus, although the functionality in the side chains of LGE 2 -and iso [4]LGE 2 -protein adducts is the same, the different lengths of the side chains  and restriction of conformational possibilities for their disposition results in strong but geometrically different interactions of the polar functional groups in each side chain with the respective antibodies. Quantitative Analysis of LGE 2 and Iso [4]LGE 2 -Previously, we used quantitative radiochemical analysis to accurately determine the amount of LGE 2 contained in protein adduct standards. Because radiolabeled iso [4]LGE 2 is not presently available, we had to employ a less direct method to determine the amount of iso [4]LGE 2 -derived pyrrole present in the iso [4]LGE 2 -KLH antigen, iso [4]LGE 2 -BSA coating agent, and iso [4]LGE 2 -HSA standard. While the Ehrlich assay is not sensitive enough to detect the low concentrations of iso [4]LGE 2derived pyrroles present in human blood or generated upon oxidation of LDL, it was feasible to use this assay to compare the concentrations of LGE 2 -derived and iso [4]LGE 2 -derived pyrroles in the protein adducts prepared as standards. The iso [4]LGE 2 to protein ratios, i.e. 21, 30 and 1257 mol/mol, calculated for the BSA, HSA, and KLH adducts, are higher than found previously for analogous LGE 2 -protein adducts, i.e. 10.5, 11.9, and 951 mol/mol of BSA, HSA, and KLH (21). Furthermore, in an earlier study, when BSA was exposed to a large excess (125 equivalents) of tritium-labeled LGE 2 , one molecule of BSA was found to bind a maximum of about 16 molecules of LGE 2 (14). It seems reasonable to expect that a similar limit would apply to binding of iso [4]LGE 2 , Especially important is the concentration of iso [4]LGE 2 -pyrrole determined for iso [4]LGE 2 -HSA because this standard was used to calculate the amount of iso [4]LGE 2 -pyrrole in oxLDL samples. The 30:1 ratio determined indirectly by Ehrlich assay for iso [4]LGE 2 -HSA seems to overestimate the actual levels by factor of two. The concentrations of iso [4]LGE 2 -protein adduct indicated in the figures and tables must be interpreted in light of this caveat.
Possible Etiological Importance of LGs and Iso [n]LGs in Artheriosclerosis-With mouse peritoneal macrophages, we previously showed that the covalent adduct of LGE 2 with human LDL (LGE 2 -LDL) is internalized and degraded if the molar ratio of LGE 2 to LDL protein, apoB, exceeds a threshold somewhere between 10:1 and 38:1 (8). OxLDL, but not acetyl-LDL that is recognized by the prototypical scavenger receptor, efficiently competed for receptor binding and uptake of LGE 2 -LDL. This result suggests that LGE 2 -LDL was recognized by a class of scavenger receptor that demonstrated ligand specificity for oxLDL but not for acetyl-LDL. However, our previous study of LDL oxidation found that only 1-2 mol of LGE 2 -protein adduct are generated per mole of apoB (27). Nevertheless, it is reasonable to anticipate that macrophage recognition of iso [n]LG-LDLs will be similar to that of LGE 2 -LDL, and that total levels of LG and iso [n]LG protein adducts in oxLDL are sufficient to account for the recognition and uptake of oxLDL by human monocyte-macrophages in the arterial wall, a key step in the etiology of atherosclerosis. Thus, substantial evidence now suggests that atherosclerotic plaques form when monocytes are recruited into the arterial intima to become macrophages where they grow into bloated, lipid-laden foam cells by accumulating large amounts of oxLDL (1,9,54).
Studies on the localization of immunoreactive LG-protein and iso [4]LG-protein epitopes in human atherosclerotic plaques are in progress in our laboratories. The details of these studies will be reported in due course. Since deficient processing of oxLDL in macrophages leads to foam cell formation, it is especially noteworthy that processing of LGE 2 -LDL exhibits an inefficiency similar to that found for oxLDL and, therefore, that incompletely processed LGE 2 -LDL accumulates in macrophages (8). The resistance to lysozomal degradation of oxLDL which accumulates in macrophages may be a consequence of continued oxidative modification or aggregation of the particles which occurs following uptake (55). In this regard, it is especially pertinent that LGE 2 binds avidly (within minutes) with proteins (14), and the reaction of LGE 2 with proteins generates reactive electrophilic intermediates that are responsible for a slower process, protein-protein cross-linking (14,17). In other words, LG-and iso [n]LG-protein adducts are expected to be "sticky," readily forming protein-protein cross-links by binding to additional protein-based nucleophiles. It is tempting to speculate that such cross-links with proteolytic enzymes interfere with processing of oxLDL.