Molecular mechanism of recombinant liver fatty acid binding protein's antioxidant activity.

Hepatocytes expressing liver fatty acid binding protein (L-FABP) are known to be more resistant to oxidative stress than those devoid of this protein. The mechanism for the observed antioxidant activity is not known. We examined the antioxidant mechanism of a recombinant rat L-FABP in the presence of a hydrophilic (AAPH) or lipophilic (AMVN) free radical generator. Recombinant L-FABP amino acid sequence and its amino acid oxidative products following oxidation were identified by MALDI quadrupole time-of-flight MS after being digested by endoproteinase Glu-C. L-FABP was observed to have better antioxidative activity when free radicals were generated by the hydrophilic generator than by the lipophilic generator. Oxidative modification of L-FABP included up to five methionine oxidative peptide products with a total of ∼80 Da mass shift compared with native L-FABP. Protection against lipid peroxidation of L-FABP after binding with palmitate or α-bromo-palmitate by the AAPH or AMVN free radical generators indicated that ligand binding can partially block antioxidant activity. We conclude that the mechanism of L-FABP's antioxidant activity is through inactivation of the free radicals by L-FABP's methionine and cysteine amino acids. Moreover, exposure of the L-FABP binding site further promotes its antioxidant activity. In this manner, L-FABP serves as a hepatocellular antioxidant.

6p-2. The fusion protein GST/L-FABP was fi rst associated with the glutathione-Sepharose 4B beads, and then the GST/L-FABP was eluted by addition of elution buffer or L-FABP digested from GST by addition of 10 units PreScission Protease and cleavage buffer. Both GST/L-FABP and L-FABP were analyzed by SDS-PAGE and MALDI-quadrupole time-of-fl ight (QqTOF) MS, respectively.

Characterization of the recombinant protein
SDS-PAGE and Western blot analysis. The resolved proteins were either stained with Coomassie blue for SDS-PAGE or transferred to Nitroplus-2000 membranes for Western immunoblotting as previously outlined by our group ( 23 ). For the Western blot, a 15% SDS-polyacrylamide gel was used. Once proteins were separated, bands were transferred onto Nitroplus-2000 membrane (Micron Separations, Westborough, MA). Nonspecifi c antibody binding was blocked by preincubation of membranes in 1× TBS containing 5% skim milk for 1 h at room temperature. Membranes were then incubated overnight at 4°C with a polyclonal antibody raised against rat L-FABP as described previously ( 22 ) in 1× TBS containing 5% skim milk. After being washed, membranes were incubated with donkey anti-rabbit IgG (Amersham) at 1:500 dilution for 1 h (room temperature). Bands were visualized using an enhanced chemiluminescence kit (ECL system; GE Healthcare Bio-Sciences. MALDI-QqTOF MS. L-FABP proteins ( ‫ف‬ 5 ug) were digested with 50 ng sequencing grade trypsin or endoprotease Glu-C (Roche Diagnostic) in 25 mM ammonium bicarbonate and the solution incubated at 37°C for 6 h. Analyses of the proteolytic peptides were performed on Applied Biosystems/MDS Sciex QStar XL QqTOF mass spectrometer by MALDI at positive ionization mode. The instrument was equipped with a MALDI II source and a UV nitrogen laser operating at 337 nm. Samples were prepared at the ratio of 1:1 (v/v) of the peptide digest to matrix (i.e., 2,5-dihydroxybenzoic acid) in 50% acetonitrile/ water and subsequently dried on a stainless steel MALDI plate at room temperature. After MALDI MS mapping, the individual peptide sequences were identifi ed by MS/MS measurements using argon as the collision gas. The intact mass of L-FABP was determined by MALDI and also confi rmed by nanospray ESI TOF mass spectrometry using the same instrument. In the case of ESI, the protein sample was dissolved in 50% methanol/0.1% formic acid solution. Peptide fi ngerprinting masses were searched by MS-Fit program against the National Center for Biotechnology Information database using ProteinProspector at the UCSF web site (http:// prospector.ucsf.edu), whereas the MS/MS ions search on each tandem mass spectrum was performed through the Mascot search engine (MatrixScience; http://www.matrixscience.com). These searches take account of up to three missed enzyme cleavage sites and the modifi cations of methionine oxidation, asparagine and glutamine deamidation to aspartic acid and glutamic acid, and N-terminal pyroglutamation. Mass tolerance between calculated and observed masses used for database search was considered at the range of ±100 ppm for MS peaks and ±0.2 Da for MS/MS fragment ions. If no result was retrieved by the automated database search, then manual data interpretation was conducted on the spectrum based on the L-FABP predicted sequence.
A unique property of L-FABP that has escaped detection is its antioxidant effect. L-FABP is known to bind polyunsaturated fatty acids ( 19 ) and long-chain fatty acid peroxidation products ( 11 ). By binding polyunsaturated fatty acids, L-FABP modulates the availability of these fatty acids to intracellular oxidative pathways ( 19 ) and in this manner controls the amount of reactive oxygen species (ROS) released within the cell from this pathway. In addition to these well-known functions, recent studies have shown that L-FABP plays a further role in the cellular antioxidant defense mechanism (20)(21)(22). Using an L-FABP cDNA transfection model Wang et al. ( 20 ) reported that hepatocytes containing L-FABP were associated with much lower levels of ROS compared with hepatocytes devoid of L-FABP. Using a bile-duct ligated model of cholestasis, this group further showed that clofi brate increased L-FABP levels were associated with improved hepatic function and reduced lipid peroxidation products ( 21 ). The antioxidative function of L-FABP is thought to be due to its amino acid composition. L-FABP contains one cysteine and several methionine groups that are known to take part in cellular redox cycling. Thus, although the actual mechanism for the L-FABP antioxidant property is not known, it is likely to involve these amino acids. Also unknown is whether L-FABP preferentially inactivates ROS released in lipophilic environments, such as membranes, or in the hydrophilic cytosolic environment. In this study, we investigated the mechanism and effi ciency of the recombinant L-FABP antioxidant property when free radicals were released into the aqueous or lipid environments.

Construction, expression, and purifi cation of pGEX-6P-2 with L-FABP
The expression vector pGEX-6P-2 was digested with Bam HI and Xho I for the insertion of a full-length rat L-FABP cDNA, which was cloned by a pair of gene-specifi c PCR primers designed by the Oligo 5.1 computer software according to the GenBank sequence (BC086947). The primers contained Bam HI and Xho I, respectively, and were synthesized by Invitrogen. PCR cloning of rat L-FABP was performed according to our previous publication ( 20 ). The plasmid pGEX-6P-2 with L-FABP was constructed according to the pGEX-6P-2 instruction manual, and successful ligation of L-FABP with pGEX-6P-2 was confi rmed by restriction enzyme digestion and DNA sequencing.
Rat L-FABP was expressed by transformation into bacteria DH5 ␣ and amplifi ed by addition of 0.1 nM isopropyl-␤ -D -thiogalactoside according to standard protocol. After expression in DH5 ␣ , L-FABP was purifi ed according to the manual of pGEX-tosolic proteins. Optimal expression time was evaluated by determining the extent of protein accumulation and degradation using SDS-PAGE analysis of cellular lysates over 1-8 h. Following 6 h of induction, the GST/L-FABP tag reached a peak intracellular concentration of ‫ف‬ 80% of total E. coli intracellular protein, and 0.1 mM isopropyl-␤ -D -thio-galactoside was suffi cient to induce a maximum level of GST/L-FABP expression.
GST/L-FABP was purifi ed to homogeneity by three successive elution steps using GST 4B beads, which precipitated 85% of the GST/L-FABP. This purifi cation method had the highest protein recovery and optimal elution characteristics for GST/L-FABP separation. Following elution, ‫ف‬ 50-60 mg of GST/L-FABP was obtained from a 500 ml culture. Impurities were not present as observed by SDS-PAGE on 15% polyacrylamide gels. Thus, contaminants could be effectively removed from the GST/L-FABP complex by GST 4B beads.
Purifi ed recombinant L-FABP was successfully isolated after incubating the GST/L-FABP complex with PreScission Protease and cleavage buffer. SDS-PAGE results showed that our recombinant L-FABP had the same molecular weight ( ‫ف‬ 14 kDa) as L-FABP isolated from hepatoma cells.

Effect of recombinant L-FABP on DCF fl uorescence in vitro
The DCF fl uorescence assay was used as a convenient screening method for assessing the extent of the antioxidative potential of our recombinant L-FABP. Since intracellular L-FABP has been reported to play a major role in suppressing oxidative stress ( 20,22 ), a reduction in ROS by L-FABP would be observed in an in vitro DCF oxidation assay. As shown in Fig. 1 , our recombinant L-FABP inhibited H 2 O 2 -induced free radical levels as observed by a decrease in DCF fl uorescence intensity. Moreover, inhibition of the solution in ethanol with 0.5 ml of 0.01 N NaOH for 30 min at room temperature in the dark. The mixture was neutralized with 2.5 ml of 20 mM sodium phosphate buffer (pH 7.0) to give a fi nal concentration of 60 M of the activated DCFH dye stock solution. The deesterifi ed DCFH-DA could then be oxidized by free radicals to a highly fl uorescent dichlorofl uorescein (DCF) whose absorbance could be quantitated at 504 nm spectrophotometrically. Oxidation reactions were carried out in 96-well CoStar plates using 10-30 M dye stock solution with 200 M hydrogen peroxide (H 2 O 2 ) and different concentrations of L-FABP.

Isolation of lipoproteins
Plasma fractions of <1.21 densities were separated from fresh human plasma by ultracentrifugation in the presence of 1 mM EDTA. After dialysis, plasma lipoproteins were applied on a lysine-Sepharose 4B affi nity chromatography column. Unbound lipoproteins were used to prepare Lp(a)-free LDL (density 1.019-1.063) using ultracentrifugation. Lipoproteins were stored in sealed tubes fi lled with nitrogen and kept in the dark at 4°C to prevent oxidization during storage. Protein concentrations of LDL were measured by a BCA protein kit (Fisher Scientifi c).

Lipid peroxidation induced by AAPH or AMVN
Lipoprotein peroxidation was induced by two oxygen-derived free radical generators: a hydrophilic radical generator, AAPH, and a lipophilic radical generator, AMVN. All the free radical generators were prepared fresh and preincubated with LDL (1 mg cholesterol/ml) for 90 min at 37°C. AAPH was dissolved in deionized water, whereas AMVN was dissolved in 95% ethanol. The lipid peroxidation product (malonaldehyde, MDA) was determined by the thiobarbituric acid-reactive substance (TBARS) method ( 24 ). Briefl y, the TBARS assay was used to demonstrate reactive aldehydes from lipid peroxidation, which has been widely accepted as a general marker of free radical production. After incubation for 90 min, the reaction was terminated by addition of 1 ml TBA reagent (0.67%, w/v, TBA in a 15%, w/v, trichloroacetic acid solution and 0.25 N HCl). The reaction mixture was heated to 100°C for 15 min and then cooled on ice. Samples were then centrifuged, resulting in the development of a pink chromogen whose absorbance was measured at 535 nm in a spectrophotometer. Freshly diluted malondialdehyde bis(dimethyl acetal 1,1,3,3-tetramethoxypropane) was used as a reference standard. Thiobarbituric acid reactive substances were expressed as MDA equivalents.

Statistical analysis
Results are expressed as mean ± SD. Appropriate statistical analysis included Student's t -test (unpaired) where two groups were compared. Two-way ANOVA was used for multiple comparisons. Statistical signifi cance was considered at P < 0.05. The n value refers to the number of experimental assays in each study.

Expression and purifi cation of recombinant glutathione S -transferase/L-FABP in Escherichia coli
The cDNA fragment encoding the complete rat L-FABP sequence was cloned into the pGEX-6P-2 plasmid downstream of the hybrid glutathione S -transferase (GST) tag promoter to allow for the inducible and effi cient intracellular expression of rat GST/L-FABP in E. coli . Following sonication and removal of the bacterial cell pellet, the soluble GST/L-FABP complex was isolated from bacterial cy- able to reduce MDA production by 60% ( P < 0.001) ( Fig.  3A ), whereas no signifi cant protective effect against lipid peroxidation was observed in the 0-40 mM concentration range in the AMVN assay. At a concentration of 5 M, ␣ -tocopherol reduced >70% of MDA formation ( Fig. 3A ). The same concentration of ␣ -tocopherol inhibited MDA production by <50% in the AMVN-induced lipid peroxidation assay ( Fig. 3B ), showing that ␣ -tocopherol has a weaker effect in the inhibition of LDL oxidation by AMVN.

Protection against lipid peroxidation of recombinant L-FABP in AAPH and AMVN
We were interested in knowing whether L-FABP is more effective as a hydrophilic or lipophilic free radical scavenger. Since L-FABP is water soluble, it is thought that most of the antioxidant properties may be directed in the cytosol. However, as a protein it also imparts some lipophilic properties, and as such some of the antioxidant property may be directed at the lipophilic or cell membrane environment. Thus, we investigated the LDL oxidation levels mediated by 40 mM AAPH or 10 mM AMVN in the absence or presence of different L-FABP concentrations using an in vitro system. Figure 2B shows a dose-dependent inhibition in oxidized LDL by 1-20 µM L-FABP on AAPHand AMVN-induced MDA production. Using AAPH to produce free radicals, a concentration of 1 M L-FABP was able to reduce MDA formation by 40% ( P < 0.001), and a 90% ( P < 0.001) reduction in MDA formation was obtained when the concentration of L-FABP was increased to 20 M. Figure 3A shows that 10 M L-FABP inhibited a similar amount of MDA production as 10 mM ascorbic acid but inhibited more MDA production than 10 M ␣ -tocopherol ( P < 0.01). At a concentration of 20 M L-FABP, L-FABP protected against free radical damage much more than either 20 M ␣ -tocopherol ( P < 0.001) or 20 mM ascorbic acid ( P < 0.001).
L-FABP was less potent against MDA production when free radicals were induced by AMVN. Figure 3B shows that released free radicals was increased with increased L-FABP concentration. Suppression of free radical release was greatest at 200 M L-FABP, which was observed to be 66 ± 1% of control values ( P < 0.001). To determine the possibility that L-FABP quenched the DCF fl uorescence by binding DCF or by some other unknown mechanism, Rajaraman et al. ( 22 ) investigated the quenching potential of L-FABP. They reported that fl uorescence was not affected by L-FABP concentration. Thus, results in this study support the notion that the associated decrease in DCF fl uorescence was in fact due to L-FABP inactivation of free radicals.

Effect of recombinant L-FABP on lipid peroxidation in the presence of AAPH and AMVN in vitro
To compare the effectiveness of antioxidants in both lipophilic and hydrophilic free radical generating systems, it was necessary to examine the effects of AAPH and AMVN free radical generators in a LDL solution. Levels of the lipid peroxidation product (MDA) were fi rst examined as a function of varying concentrations of AMVN or AAPH ( Fig. 2A ). The lipid-soluble peroxyl radical generator (AMVN) induced greater MDA production in LDL than the water-soluble generator (AAPH). Thus, for subsequent comparison, it was necessary to use AAPH and AMVN concentrations that induced similar amounts of MDA production. The same amount of MDA was produced by 10 mM AMVN and 40 mM AAPH. Therefore, those concentrations were used in all in vitro studies.

Protection by ascorbic acid and ␣ -tocopherol against lipid peroxidation induced by AAPH or AMVN
To compare the difference of antioxidant activity between L-FABP and ascorbic acid or ␣ -tocopherol, LDL was incubated with 40 mM AAPH or 10 mM AMVN in the absence or the presence of different concentrations of ascorbic acid or ␣ -tocopherol ( Fig. 3A , B ). These two antioxidants inhibited the oxidation of LDL in a dose-dependent manner. Using the AAPH generator, 1 mM ascorbic acid was  Table 1 . 20 M L-FABP provided greater antioxidant effect against lipid peroxidation induced by AAPH (90%) than that of AMVN (55%). Figure 3B also shows that 10 M L-FABP inhibited 50% of MDA production and had a similar effect to that of 10 M ␣ -tocopherol in the AMVN-induced lipophilic free radical generating system. Thus, the antioxidant activity of L-FABP was comparable to ␣ -tocopherol and much greater than ascorbic acid at the same molar concentrations.

Antioxidant effect of L-FABP upon binding of palmitate or ␣ -bromo-palmitate
To understand whether long-chain fatty acid binding to L-FABP infl uences its antioxidant activity, LDL was incubated with 40 mM AAPH or 10 mM AMVN in the absence or presence of 10 M L-FABP, which was preincubated with 30 M ␣ -bromo-palmitate ( ␣ -bromo-palmitate is not metabolized) or 30 M palmitate (reversible binding). In the AAPH-induced lipid peroxidation system, 10 µM L-FABP was able to signifi cantly reduce MDA production by 70 ± 2% ( P < 0.001), while ␣ -bromo-palmitate and palmitate partially blocked the L-FABP antioxidative activity by 29 ± 2% and 19 ± 1%, respectively (see Fig. 4 ; compared with L-FABP no binding; P < 0.001). Figure 4 also shows that in the AMVN-induced lipid peroxidation system, 10 µM L-FABP was able to reduce MDA production by 50 ± 1%, while ␣ -bromo-palmitate and palmitate blocked L-FABP antioxidative activity by 43 ± 1% and 11 ± 1%, respectively (compared with L-FABP no binding; P < 0.001). These results indicated that by blocking the L-FABP binding cavity some of the L-FABP antioxidant activity was eliminated.

Identifi cation of L-FABP amino acids involved in the antioxidant effect by MALDI QqTOF MS
Identifi cation of the modifi ed recombinant L-FABP. Recombinant L-FABP was analyzed following separation using chromatography by MALDI QqTOF MS. The amino acid sequence of our recombinant rat L-FABP ( Table 1 ) was identical with that of L-FABP isolated from rat liver ( 25 ). MALDI-TOF analysis showed presence of a major mass peak at m/z of 15,275.9 ( Fig. 5A ). Total measured mass (15,274.9 Da) of L-FABP was greater than that predicted based on the amino acid sequence (14,272.450 Da). The increased mass was attributed to additional amino acids on the N-terminal end and S -glutathionylation of the cysteine group, which was observed following peptide mass fi nger- Fig. 3. Comparison of the effect of ascorbic acid, ␣ -tocopherol, and L-FABP on AAPH-and AMVNinduced MDA production on LDL. A: LDL (1 mg cholesterol/ml) was incubated with 40 mM AAPH ± different concentrations of ascorbic acid, ␣ -tocopherol, and L-FABP for 90 min at 37°C. B: LDL (1 mg cholesterol/ml) was incubated with 10 mM AMVN and different concentrations of ␣ -tocopherol and L-FABP for 90 min at 37°C. Values are presented as means ± SD, n = 6, *** P < 0.001.    ) containing fragments, each provide evidence for oxidant modifi cation of methionine by AAPH-induced free radicals. In the MALDI MS spectrum of AMVN-modifi ed L-FABP ( Fig. 6B ), the ob-and function of these cellular components and are involved in the pathogenesis of various degenerative dis eases, including heart disease ( 28 ), diabetes ( 29 ), infl am mation ( 30 ), and neurodegenerative disorders ( 31 ). Although cells contain specifi c endogenous antioxidant enzymes, such as superoxide dismutase, catalase, glutathione peroxidase, and compounds such as ␤ -carotene and tocopherols, levels of these antioxidants may be too low to combat the high concentrations of released free radicals especially during periods of oxidative stress. Moreover, many of these antioxidants direct their activity toward the aqueous milieu. However, ␣ -tocopherol is a fat-soluble vitamin that may have some of its activity directed toward lipophilic structures ( 32 ).
Recent work has shown that in addition to being the intracellular counterpart to extracellular albumin, L-FABP served product 104-127+OX ( m/z 2758.468) (OX = oxidized) corresponded to one oxidized methionine (Met 113 ) containing fragment, which provided evidence for oxidant modifi cation of methionine by AMVN-induced free radicals.

DISCUSSION
Attack by free radicals on lipophilic structures, such as cell membranes or intracellular membranes, may result in severe forms of cell damage and/or ultimately cell death. Lipid peroxidation products are especially of importance since they are highly reactive and able to form adducts with many cellular lipophilic macromolecules, including cell membrane constituents, intracellular proteins, and nucleic acids ( 27 ). These adducts modify the structure  stable products or react with molecular oxygen to generate hydrophilic peroxyl radicals ( 39 ). In the presence of LDL, AAPH released free radicals oxidize the aqueous (outer) environment of LDL ( 40 ). AMVN is a synthetic azo compound that dissociates spontaneously to form carboncentered free radicals. Generation of carbon-centered radicals is directed at the lipophilic environment of membranes, where its activity is suffi cient to cause membrane phospholipid peroxidation ( 41 ). Using the free radical generators AAPH and AMVN together with use of the positive controls ascorbic acid (hydrophilic antioxidant) and ␣ -tocopherol (lipophilic antioxidant), we assessed the effectiveness of L-FABP as an antioxidant. Both ␣ -tocopherol and ascorbic acid were effective in protecting LDL from free radicals generated by AAPH. Using AMVN, however, ascorbic acid had no effect on inactivating free radicals, while not surprisingly ␣ -tocopherol was very effective. L-FABP was effective in inactivating free radicals released by both AAPH and AMVN. L-FABP is known to coat the surface of anionic phospholipid membranes ( 42 ) and can prevent propagation of phospholipid oxidation reactions by reacting with the ROS in this environment. Where this becomes of importance is in the protection of cellular membranes during episodes of oxidative stress whereby lipophilic antioxidants have been shown to be effective in suppressing the progression of fatty liver disease ( 43 ), coronary heart disease ( 44 ), and aging ( 45 ).
MALDI-TOF analysis was used to identify the mechanism of L-FABP's antioxidant activity. Amino acid sequence identifi cation showed that our recombinant protein contained one cysteine and seven methionines, which was identical to the sequence reported by others ( 25,46 ). Methionine residues in proteins are known to be oxidized to methionine sulfoxides [Met(O)] by ROS and can be reduced back to the original form by methionine sulfoxide reductase ( 47 ). Sequence identifi cation showed that in the AAPH hydrophilic system, fi ve of the seven methionine amino acids were oxidized. These included Met 19 19 . Met 85 and Met 74 were unavailable for reaction with free radicals. This is likely due to the positioning of these amino acids within the protein itself. Met 74 is located at the L-FABP binding cavity side chain located on the ␤ E ␤ F hairpin turn. This is an area of low solvent accessibility making it diffi cult to be accessible to free radicals ( 6 ). Therefore, it is not surprising to observe that Met 74 was not easily oxidized in either the AAPH or AMVN free radical systems. Binding of free radicals to one amino acid also may cause a slight change in protein structure through ring opening ( 48 ), which could make other amino acids available or unavailable for reaction. Using the lipophilicfree radical generator AMVN, three of the seven methionines in our L-FABP were available for reaction. While Met 74 and Met 85 were unreactive in both systems, Met 91 was unreactive in the lipophilic system. Met 113 was the only methionine residue with the highest degree of availability to free radicals (83%) in the AMVN system. Thus, MALDI-TOF possesses strong antioxidant properties. In fact, when clofi brate was used to increase L-FABP levels in animals subjected to bile-duct ligation, clofi brate-treated animals were shown to be associated with reduced levels of hepatic lipid peroxidation products and improved hepatic function ( 21 ). Clofi brate-treated animals were also associated with lower levels of mortality (G. Wang, H. Shen, G. Rajaraman, M.S. Roberts, Y. Gong, P. Jiang, F.Burczynski, unpublished observations). The authors speculated that L-FABP functions as an antioxidant during periods of cellular oxidative stress. Moreover, in a cell transfection model, hepatocytes containing L-FABP were associated with much lower ROS levels than those devoid of L-FABP ( 20 ). In both models, levels of the endogenous antioxidants were unaffected by the treatment regime. The mechanism whereby L-FABP inactivated ROS, however, was not understood. Understanding the protein's antioxidant mechanism of action required in vitro studies and large amounts of rat L-FABP.
In this study, recombinant L-FABP was successfully expressed in E. coli using a simple purifi cation system. The system made use of a GST affi nity tag that allowed for a more effi cient purifi cation of the recombinant protein in E. coli ( 33 ). The GST tag contained a site whereby ProScission protease could cleave the GST tag allowing for greater sequence specifi city in generating the target protein. A small sequence chain remained on the target protein at the N terminus after the removal of the GST tag. This prevented protein degradation and maintained immunogenicity of the protein ( 26 ). After removal of the GST tag from the soluble GST/L-FABP fusion, the resulting L-FABP covered a full-length amino acid sequence of rat L-FABP ( 25 ).
Purifi ed L-FABP showed high biological activity with respect to its antioxidant activity. Initial studies using H 2 O 2 as a nonspecifi c free radical generator showed that our recombinant protein at 200 µM inactivated ‫ف‬ 66 ± 2% of the released free radicals. Since L-FABP makes up ‫ف‬ 2% of total cellular protein in hepatocytes or ‫ف‬ 200-400 µM ( 34 ), this concentration of L-FABP supplies a signifi cant amount of antioxidant activity within hepatocytes. Following clofibrate treatment, the concentration of L-FABP could rise to 2-to 3-fold ( 35 ), which would have an even greater impact on intracellular free radical levels and further contribute a signifi cant amount of free radical scavenging activity. It is, therefore, not surprising that L-FABP protected the liver during periods of oxidative stress as was associated with cholestatic liver disease ( 21 ), hepatic steatosis ( 10 ), or alcohol-induced liver injury ( 36 ).
Free radicals such as O˙ 2 and OH˙ were initially thought to penetrate lipid bilayers with ease ( 37 ). However, more recent evidence suggested that superoxides do not penetrate deeply into lipid bilayers but rather react at the surface or just below the surface of membranes ( 38 ). AMVN and AAPH are azo initiators that produce a slow and steady source of free radicals by known chemical decomposition mechanisms. AAPH is a water-soluble azo compound. Decomposition produces molecular nitrogen and two carbon radicals. The carbon radicals may combine to produce to occur for heart-FABP and membranes ( 53 ). In this manner, L-FABP could interact with ROS at lipophilic interfaces. The fact that only 10% of the released free radicals were scavenged by L-FABP in the presence of ␣ -bromo-palmitate indicated that Met 113 in L-FABP binding cavity is likely the key amino acid that is mostly responsible for L-FABP antioxidant activity in AMVN-induced lipid peroxidation.
In summary, the GST fusion system provided a simple and convenient way to produce large quantities of recombinant L-FABP. The TBARS assay showed that L-FABP protected LDL from oxidation with a similar potency to ␣ -tocopherol but much greater potency than ascorbic acid. MALDI-TOF MS analysis showed that the methionine groups of L-FABP were responsible for the majority of its antioxidant activity in both lipophilic and hydrophilic systems. Met 113 located in the L-FABP binding cavity was shown to be the most important amino acid for reacting with ROS in the lipophilic free radical generating system. These results demonstrate that rat L-FABP can act as a potent cellular antioxidant.
MS analysis showed that, indeed, several of the methionine groups in L-FABP do react with free radicals. Similarly, the single cysteine group (Cys 69 ) of L-FABP in the presence of free radicals can form a sulfonic acid intermediate (Cys-SOH) that is subsequently reduced back by forming a disulfi de bond through the action of thioredoxin ( 49 ). Reactivity of various amino acids was investigated by Xu and Chance ( 48 ). They showed that cysteine and methionine were the most reactive of all the amino acids. The cysteine group in our recombinant L-FABP was unavailable for reaction with free radicals because it was S -glutathionylated during the GST/L-FABP elution process by reduced glutathione. However, cysteine is well known to be highly effective in scavenging free radicals ( 50 ) and thus also makes it a likely candidate in L-FABP's antioxidant activity.
To understand the contribution of L-FABP's binding pocket to its antioxidant activity in both the hydrophilic and lipophilic free radical generating systems, we incubated L-FABP with palmitate or ␣ -bromo-palmitate. Palmitate binding to L-FABP is a reversible process with distinct association and dissociation rate constants ( 51 ). ␣ -Bromopalmitate does not undergo esterifi cation to form triglycerides; its lack of metabolism allows it to accumulate in the cytoplasm making it relatively more effective at binding to L-FABP compared with palmitate ( 52 ). Binding of palmitate to L-FABP decreased L-FABP's ability to inactivate free radicals, i.e., binding of long-chain fatty acids to L-FABP increased the amount of free radicals within the hydrophilic and lipophilic environments. The amount of ROS generated by AMVN was slightly higher than that generated by AAPH, suggesting that when palmitate is bound to the protein some methionine group(s) may be inaccessible for reaction with free radicals. This is especially evident in the lipophilic environment. Binding of ␣ -bromopalmitate to L-FABP resulted in a much more dramatic increase in ROS levels using the AMVN generator. While both systems resulted in higher ROS levels, these levels were much higher in the lipophilic system. Met 113 is known to be located in the binding cavity of L-FABP ( 6 ). Since ␣ -bromo-palmitate is not metabolized, it would be expected to bind to a larger extent than palmitate and block the accessibility of ROS to this amino acid. Similarly, binding of polyunsaturated long-chain fatty acids possessing binding constants much higher than that of palmitate would be expected to have higher binding occupancy rates. In such cases, Met 113 would not be available for interaction with ROS. The methionine group within the L-FABP binding cavity is therefore of more importance when scavenging free radicals within a lipophilic environment. Met 1 and Met 22 are the only other two methionine groups that have the ability to react with ROS but to a much lesser extent than Met 113 . This is likely the reason why binding of ␣ -bromo-palmitate to L-FABP in the AMVN system did not totally abolish the antioxidant effect of L-FABP. One could speculate that as L-FABP comes in close proximity to the membrane surface; the protein orientates itself such that the ligand binding pocket face is open to the membrane. This could occur by electrostatic interactions between the cell surface and protein. Such interactions have been shown