Oxidation of methionine residues to methionine sulfoxides does not decrease potential antiatherogenic properties of apolipoprotein A-I.

The initial stage of oxidation of high density lipoproteins (HDL) is accompanied by the lipid hydroperoxide-dependent, selective oxidation of two of the three Met residues of apolipoprotein A-I (apoA-I) to Met sulfoxides (Met(O)). Formation of such selectively oxidized apoA-I (i.e. apoA-I(+32)) may affect the antiatherogenic properties of HDL, because it has been suggested that Met(86) and Met(112) are important for cholesterol efflux and Met(148) is involved in the activation of lecithin:cholesterol acyl transferase (LCAT). We therefore determined which Met residues were oxidized in apoA-I(+32) and how such oxidation of apoA-I affects its secondary structure, the affinity for lipids, and its ability to remove lipids from human macrophages. We also assessed the capacity of discoidal reconstituted HDL containing apoA-I(+32) to act as substrate for LCAT, and the dissociation of apoA-I and apoA-I(+32) from reconstituted HDL. Met(86) and Met(112) were present as Met(O), as determined by amino acid sequencing and mass spectrometry of isolated peptides derived from apoA-I(+32). Selective oxidation did not alter the alpha-helicity of lipid-free and lipid-associated apoA-I as assessed by circular dichroism, and the affinity for LCAT was comparable for reconstituted HDL containing apoA-I or apoA-I(+32). Cholesteryl ester transfer protein mediated the dissociation of apoA-I more readily from reconstituted HDL containing apoA-I(+32) than unoxidized apoA-I. Also, compared with native apoA-I, apoA-I(+32) had a 2- to 3-fold greater affinity for lipid (as determined by the rate of clearance of multilamellar phospholipid vesicles) and its ability to cause efflux of [(3)H]cholesterol, [(3)H]phospholipid, and [(14)C]alpha-tocopherol from lipid-laden human monocyte-derived macrophages was significantly enhanced. By contrast, no difference was observed for cholesterol and alpha-tocopherol efflux to lipid-associated apolipoproteins. Together, these results suggest that selective oxidation of Met residues enhances rather than diminishes known antiatherogenic activities of apoA-I, consistent with the overall hypothesis that detoxification of lipid hydroperoxides by HDL is potentially antiatherogenic.

High density lipoproteins (HDL) 1 are generally regarded as antiatherogenic, an activity commonly attributed to the removal of extrahepatic cholesterol by HDL particles (1,2) and apolipoproteins, mainly apolipoprotein A-I (apoA-I), that dissociate from HDL (3,4). In addition to promoting cholesterol efflux, HDL has also been proposed to be antiatherogenic by aiding the removal and detoxification of proatherogenic oxidized lipids (5)(6)(7). Thus, cholesteryl ester transfer protein (CETP) transfers oxidized lipids from low density lipoproteins (LDL) to HDL (7), and HDL carries the majority of cholesteryl ester hydroperoxides (CE-OOH, the first and major products formed during lipoprotein oxidation) in human plasma (5). In addition, CE-OOH and cholesteryl ester hydroxides (CE-OH) in HDL, but not LDL, are removed rapidly via selective uptake by liver parenchymal cells in vitro (6) and by perfused liver in situ (8). This uptake is associated with cellular detoxification of CE-OOH (6) and is more rapid than that of the corresponding nonoxidized cholesteryl esters (CE) (6,8). Furthermore, CE-OH are also rapidly removed from HDL via hepatic clearance in vivo (9), and this is associated with biliary secretion of the CE-OH-derived cholesterol (9), indicating that oxidation of the fatty acid moiety of CE may aid the elimination of cholesterol from the body.
Once associated with HDL, CE-OOH are reduced to the corresponding CE-OH (10) that no longer contribute to or enhance the oxidation of lipoproteins. This "antioxidant" activity is expressed by reconstituted HDL particles (rHDL) containing apoA-I only and lipid-free apoA-I (11,12) and extends to phospholipid hydroperoxides (11)(12)(13). The reduction of lipid hydroperoxides, added to or formed in HDL exposed to radical oxidants, results in the formation of selectively oxidized apoA-I (i.e. apoA-I ϩ32 ) that contains two Met sulfoxide [Met(O)] residues as the sole modification (11,12). However, it is not known which Met residues of apoA-I become oxidized and how formation of Met(O) affects the ability of apoA-I/HDL to mediate 1 The abbreviations used are: HDL, high density lipoproteins; AAPH, 2,2Ј-azobis(2-amidinopropane) hydrochloride; apoA-I, apolipoprotein A-I; apoA-I ϩ32 , selectively oxidized apoA-I (32 mass units heavier than native apoA-I and containing two Met sulfoxide residues); CD, circular dichroism; CE, cholesteryl esters; CE-OH, cholesteryl ester hydroxides; CE-OOH, cholesteryl ester hydroperoxides; CETP, cholesteryl ester transfer protein; DMPC, dimyristoylphosphatidylcholine; drHDL, discoidal reconstituted HDL; srHDL, spherical reconstituted HDL; LCAT, lecithin:cholesterol acyltransferase; LDL, low density lipoproteins; acLDL, acetylated low density lipoproteins; hMDM, human monocytederived macrophages; Met(O), methionine sulfoxide; PBS, phosphatebuffered saline; ␣-TOH, ␣-tocopherol; RP-HPLC, reverse-phase high pressure liquid chromatography; BSA, bovine serum albumin; ESI-MS, electrospray ionization mass spectrometry. efflux of lipids from macrophages and to act as substrate for lecithin:cholesterol acyltransferase (LCAT). These questions are relevant, because in human atherosclerotic lesions HDL is oxidized to an extent comparable to that of LDL (14) and a decreased and enhanced ability to promote cellular cholesterol efflux has been reported for in vitro oxidized HDL (see e.g. Refs. [15][16][17]. In the present study, we therefore examined the physicochemical characteristics and biological activities of apoA-I ϩ32 .

EXPERIMENTAL PROCEDURES
Isolation and Selective Oxidation of HDL-Isolated HDL was prepared from fresh plasma obtained from normolipidemic donors (18), and its protein content was determined by the bicinchoninic acid method (Sigma) using bovine serum albumin (BSA) (Sigma) as standard. HDL was oxidized in phosphate-buffered saline (PBS), pH 7.4, containing 1 mM EDTA by aerobic incubation at 37°C for 19 h in the presence of 2,2Ј-azobis(2-amidinopropane) hydrochloride (AAPH, 80 mol/liter and mg HDL protein), a generator of aqueous peroxyl radicals. After oxidation, AAPH was removed by gel filtration (11).
Isolation of Native and Selectively Oxidized ApoA-I-ApoA-I and apoA-I ϩ32 were isolated by semipreparative reverse-phase high pressure liquid chromatography (RP-HPLC), as described in detail (11,19). Peaks corresponding to apoA-I and apoA-I ϩ32 were collected, placed immediately on dry ice, freeze-dried, redissolved in reconstitution buffer (3 M guanidine HCl, 10 mM Tris, 0.01% EDTA, pH 8.2), and then dialyzed extensively against PBS (two buffer changes) and 0.5 mM Tris, pH 7.4 (three changes). Protein concentrations were determined by A 280 nm or by the bicinchoninic acid method described as above, and the phospholipid content was determined using an enzymatic kit (Roche Molecular Biochemicals).
Preparation and Characterization of Discoidal Reconstituted HDL-Discoidal reconstituted HDL (drHDL) containing either apoA-I (drH-DL A-I ) or apoA-I ϩ32 (drHDL A-Iϩ32 ) were prepared by cholate dialysis (20) using the respective purified apolipoproteins. Compositional analyses were performed on a Cobas autoanalyzer (Roche Diagnostics). The Stokes' diameter and surface charge of the particles were determined by nondenaturing polyacrylamide gradient and agarose gel electrophoreses, respectively (20). The composition and properties of drHDL A-I and drHDL A-Iϩ32 were comparable with regards to the molar ratio of phospholipid/cholesterol/apoA-I (i.e. 91/12/1 and 90/11/1 for apoA-I and apoA-I ϩ32 , respectively, with a stoichiometric variation of Ͻ5% between preparations), electrophoretic mobility, and particle size. Because a change in electrophoretic migration of rHDL is indicative of alteration of the secondary and/or tertiary structure of apoA-I (21,22), the results indicate that introduction of two Met(O) does not cause a major conformational change of lipid-associated apoA-I.
Apolipoprotein Dissociation Studies-Spherical reconstituted HDL A-I (srHDL A-I ) and srHDL A-Iϩ32 were prepared from drHDL A-I and drHDL A-Iϩ32 , respectively, and purified by ultracentrifugation (23). The phospholipid/cholesterol/CE/apoA-I molar ratios were 24/25/4/1 and 24/ 25/3/1 for srHDL A-I and srHDL A-Iϩ32 , respectively. Dissociation of apoA-I and apoA-I ϩ32 from srHDL was assessed by incubating the particles with CETP and the phospholipid/triglyceride emulsion Intralipid (20% triglyceride, Kabivitrum AB). Details of the incubations are in the legend to Fig. 1. Changes in srHDL size were determined by nondenaturing gradient gel electrophoresis (23). Dissociation was determined by immunoblotting with anti-apoA-I polyclonal antibodies and visualization by enhanced chemiluminescence (Amersham Pharmacia Biotech). A Sharp JX-610 high resolution scanner was used to scan Coomassie-stained gradient gels and immunoblots.
Apolipoprotein Self-association Studies-Size-exclusion HPLC of lipid-free apoA-I and apoA-I ϩ32 was performed using a Phenomex Biosep SEC-S2000 column (300 ϫ 7.8 mm, Bio-Rad), using 0.05 M phosphate, 0.5 M NaCl, pH 7.5, as the mobile phase at 0.5 ml/min and UV detection at 214 nm. Protein size was calculated using a gel filtration standard (Bio-Rad) for comparison. For cross-linking experiments, apoA-I or apoA-I ϩ32 (25 g/ml PBS) was preincubated for 30 min at 37°C, followed by further incubation for 30 min at room temperature in the presence of a 100-fold molar excess of bis-sulfosuccinimidyl suberate (Pierce) and separation of the proteins on 8% Tris-glycine SDS-polyacrylamide gel electrophoresis. Bands were visualized by silver staining.
Mass Spectrometry-Mass spectra were acquired using a single quadrupole mass spectrometer equipped with an electrospray ionization source (Platform, VG-Fisons Instruments). Samples (ϳ50 pmol, 10 l) were injected into a moving solvent (10 l/min; acetonitrile/water (1:1, v/v), containing 0.1% trifluoroacetic acid) coupled directly to the ionization source via polyether ether ketone tubing (127 m x 40 cm). The source temperature was 50°C and N 2 was used as nebulizer and drying gas. Sample droplets were ionized at approximately 3 kV and transferred to the mass analyzer with a cone voltage of 50 V. The peak width at half height was 1 Da. Spectra of proteins were acquired in multichannel acquisition mode over the mass range of 700 -1800 Da in 5 s then calibrated with horse heart myoglobin (Sigma). Spectra of peptides were also acquired in multichannel acquisition mode over the mass range 250 -2000 Da in 10 s. Some spectra were recorded using an LC/MSD 1100 mass spectrometer (Hewlett Packard, Palo Alto, CA) employing similar conditions. Automated N-terminal Sequencing-Peptides (typically 50 -200 pmol) were N-terminally sequenced using an automated protein sequencer (model 470A, Applied Biosystems).
Biophysical Characterization-The secondary structure was analyzed by circular dichroism (CD). Spectra of apoA-I or apoA-I ϩ32 (0.1 mg/ml 0.5 mM Tris, pH 7.4) and drHDL A-I or drHDL A-Iϩ32 (0.16 mg of protein/ml of 1 mM PBS, pH 7.4) were recorded on a JASCO 720 CD spectropolarimeter at 25°C using 1-and 0.1-mm path length cells, respectively. Acquisition parameters were: range ϭ 184 -260 nm; resolution ϭ 0.5 nm; bandwidth ϭ 1.0 nm; response time ϭ 1 s; scan speed ϭ 20 nm/min; and number of scans ϭ 16. The mean residue ellipticity was calculated by MRW ϭ obs ϫ MRW/10 ϫ c ϫ l, where obs is the observed raw data in millidegrees, MRW is the mean residue weight (i.e. 115.0 for apoA-I), c is the concentration, and l is the path length of the cell. The secondary structure was predicted by analyzing the CD spectra using the program "variable selection" (VARSELEC), where fitting of a data base of 33 proteins of known secondary structures is performed (24).
The lipid affinity of apoA-I and apoA-I ϩ32 was determined using the dimyristoylphosphatidylcholine (DMPC, Sigma) clearance assay (25). Briefly, the decrease in turbidity of multilamellar DMPC vesicles (0.5 mg/ml) upon addition of apoA-I and apoA-I ϩ32 (0.2 mg/ml) was measured over time at 325 nm and 24 Ϯ 0.2°C using a UV-visible Lambda 40 spectrophotometer (Perkin-Elmer).
Kinetics of the LCAT Reaction-LCAT, prepared as described (23) and used for the experiments, esterified 340 nmol of CE/ml of LCAT/h under assay conditions. drHDL was labeled by placing [ 3 H]cholesterol (48 Ci/mmol, Amersham Pharmacia Biotech) in a test tube, and the organic solvent was removed under N 2 . Ethanol (20 l) and drHDL containing cholesterol were then added, and the mixture was incubated at 37°C for 30 min before dilution with buffer and supplementation with ␤-mercaptoethanol (4 mM, final concentration) and fatty acid-free BSA (4 mg/ml). The final volume of the incubation mixtures was 135 l, and the cholesterol concentration was varied from 0.1 to 5.0 M. Following incubation at 37°C for 30 min under N 2 , the LCAT reaction was initiated by adding 5 l of a 1:8 (v/v) dilution of the LCAT preparation or buffer (control), and the mixture was incubated under N 2 at 37°C for 1 h. Following the reaction, radiolabeled CE were quantified by the digitonin precipitation method, exactly as described by Piran and Morin (26).
Cell Culture-Monocytes were isolated by counter-flow centrifugal elutriation (27) from buffy coats prepared from blood from normolipidemic volunteers. Cells were plated in 12-well tissue culture plates (Falcon) at a density of 1.5 ϫ 10 6 cells per ml of RPMI medium containing 10% heat-inactivated human serum and left to adhere and differentiate into macrophages (hMDM) over the next 10 days. Subsequently, cells were washed and incubated in RPMI medium containing 10% lipoprotein-depleted serum in the presence of 100 g/ml acetylated LDL (acLDL) for 48 h to generate lipid-laden "foam cells" as described (28). For metabolic labeling of cellular cholesterol pools, acLDL was first labeled with [ 3 H]cholesterol (51 Ci/mmol, Amersham Pharmacia Biotech) for 6 h (28), before [ 3 H]cholesterol-acLDL was diluted in RPMI (100 g/ml acLDL and 2 Ci/ml [ 3 H]cholesterol), and then incubated with cells. The "labeling" medium was then replaced with RPMI medium without serum for 18 h to equilibrate labeled cholesterol among cellular pools. The same procedure was used for metabolic labeling of cellular ␣-tocopherol (␣-TOH) pools, using 0.4 Ci/ml all-rac-[ 14 C]␣tocopherol (a generous gift of Eisai Co. Ltd.) instead of [ 3 H]cholesterol. To label phospholipids, hMDM were incubated with acLDL as described above and labeled during the 18-h equilibration period with 5 Ci/ml methyl-[ 3 H]choline chloride (Amersham Pharmacia Biotech) in the presence of 0.1% BSA (29). Subsequently, cells were washed twice with RPMI, incubated with fresh medium for 1 h, and washed twice with RPMI containing 0.1% BSA and then twice with RPMI before efflux incubations.
Lipid Efflux Studies-After appropriate labeling, cells were incubated at 37°C with 2 ml of RPMI per dish and, unless indicated otherwise, 25 g/ml either lipid-free or lipid-associated apoA-I or apoA-I ϩ32 . Previous studies showed that 25 g of apoA-I is sufficient for saturating cholesterol efflux from primary macrophages (30). To investigate initial kinetics, efflux experiments were performed as described previously (31). At the indicated times, aliquots of efflux media were collected and centrifuged to remove cell debris, and the radioactivity was determined by ␤-counting. Where indicated, the label remaining within the cells was also determined. For this, cells were washed twice with PBS containing 0.1% BSA, washed twice with PBS, and lysed in 0.2 N NaOH for 10 min on ice, and aliquots of the lysates then used to count radioactivity and to determine protein using the bicinchoninic acid assay. After equilibration and before efflux (i.e. time 0), cells in separate dishes were lysed, and the lipids were extracted with methanol/hexane (1:5; v:v) and analyzed by HPLC with UV 210 nm detection (for cholesterol and CE), with electrochemical or fluorescent detection (for ␣-TOH), and with on-line radiometric detection (for [ 14 (7,32). In some experiments, lipids in the efflux media were also extracted and analyzed as above. For experiments with methyl-[ 3 H]choline-labeled cells, media and cells were stored (Ϫ80°C) until protein determination and extraction of phospholipids with 2 ϫ 1 ml of hexane/isopropanol (3:2; v:v) (18 and 1 h for the first and second extractions, respectively). The combined organic supernatants were then dried under nitrogen and extracted further using the method of Bligh and Dyer (33). The resulting CH 3 Cl fraction (1 volume) was back-washed four times with 1 volume of methanol:H 2 O (1:1) to remove any remaining radiolabeled choline, and the radioactivity in the washed CH 3 Cl fraction was counted. The cellular phospholipid mass at time 0 was determined using an enzymatic kit (Roche Molecular Biochemicals).

RESULTS
Characterization of ApoA-I and ApoA-I ϩ32 -To assess HDL protein oxidation and to isolate pure native apoA-I and apoA-I ϩ32 , we adapted the HPLC method previously described (11,12,19). In native HDL, apoA-I and apoA-II were the major proteins and oxidized forms of apoA-I or apoA-II were not detected. Oxidation of the same HDL with AAPH (as described under "Experimental Procedures") decreased the content of apoA-I and apoA-II with the concomitant formation of new peaks. The two apoA-I-derived peaks have been shown previously to correspond to modified apoA-I containing one (referred to as apoA-I ϩ16 ) and two (apoA-I ϩ32 ) Met(O) instead of Met (11,12). In the present study, the mass of collected apoA-I ϩ32 was 28,112.5 Ϯ 2 Da (mean Ϯ S.D., n ϭ 10), 33 Ϯ 2 mass units greater than that obtained for apoA-I (28,079.1 Ϯ 1 Da, n ϭ 10; the value predicted from the amino acid composition of apoA-I is 28,078.7 Da) ( Table I). This is consistent with the mass difference of 32 units determined previously (11,12).
Isolated apoA-I and apoA-I ϩ32 were essentially devoid of phospholipids, and at concentrations between 0.1 and 1 mg/ml, they were present predominantly as dimers (Table I). When tested at 25 g/ml, i.e. the concentration used for most cellular studies, apoA-I or apoA-I ϩ32 dimers were also the major bands and there was no apparent difference between native and selectively oxidized apoA-I. This was revealed by apolipoprotein samples chemically cross-linked (see "Experimental Procedures") and subsequently subjected to gel electrophoresis (not shown). Together, this suggests that apoA-I and apoA-I ϩ32 behave equally in their ability to undergo self-association.
To assess differences in the dissociation of apo-AI and apoA-I ϩ32 from HDL, srHDL A-I and srHDL A-Iϩ32 were incubated with CETP and Intralipid for 1, 3, 6, 12, and 24 h. During such incubation, CE and triglycerides are transferred between srHDL and Intralipid, resulting in an initial increase in srHDL size (as a consequence of triglyceride enrichment). However, with continuing depletion of core lipids, srHDLs decrease in size and apoA-I dissociates (34). The respective diameters of srHDL A-I and srHDL A-Iϩ32 were 9.2 and 9.3 nm, respectively, at the beginning and remained unaltered during incubation for 24 h in the absence of CETP (Fig. 1). The time courses in size change of srHDL A-I and srHDL A-Iϩ32 in the presence of CETP and Intralipid are shown in Figs. 1a and 1b, respectively. As can be seen, after 24 h all srHDL A-Iϩ32 were converted to small particles (8.0 nm), whereas at the same time point some of the original srHDL A-I were still present. This, together with the appearance of more large, triglyceride-rich particles at the 6-, 12-, and 24-h time points for srHDL A-I than those for srHDL A-Iϩ32, indicates that core lipid depletion was more rapid in srHDL A-Iϩ32 than in srHDL A-I . The immunoblots show dissociation of apoA-I from srHDL A-Iϩ32 (Fig. 1d), but not srHDL A-I (Fig. 1c), after 24 h of incubation with CETP and Intralipid (lanes 7). This indicates that CETP mediates the dissociation of apoA-I from srHDL A-Iϩ32 more readily than from srHDL A-I .
To identify the Met residues oxidized in apoA-I ϩ32 , AspN digests of purified apoA-I and apoA-I ϩ32 were subjected to RP-HPLC (Fig. 2) and the mass of individual collected peptides was determined by ESI-MS. Among the peptides isolated from apoA-I, peptides 73-88, 128 -149, and 102-127 (corresponding to peaks 4, 6, and 7, respectively, in Fig. 2A) contained Met as verified by N-terminal sequencing (Table II). The corresponding peptides 73-88 and 102-127 derived from apoA-I ϩ32 eluted with slightly shorter retention times (peaks 4Ј and 7Ј, respectively, in Fig. 2B), and their masses were exactly 16 mass units heavier than their counterparts derived from apoA-I (Table II). By contrast, peptide 128 -149 eluted at 16.8 min (peak 6 in Fig.  2), and its mass was 2579 Da (Table II), independent of whether it was derived from apoA-I or apoA-I ϩ32 . This demonstrates that Met 86 and Met 112 (but not Met 148 ) are oxidized in apoA-I ϩ32 .
CD measurements were performed to examine the secondary structure of lipid-free and lipid-associated apoA-I and apoA-I ϩ32 , using the fitting procedure given in Table III. In agreement with a previous report (35), ␣-helix was the predominant structural type of apoA-I, accounting for 52.0 Ϯ 2.5% of the protein. The corresponding value for apoA-I ϩ32 was the same (Table III). As for the lipid-free proteins, almost identical ␣-helical contents were obtained for rHDL A-I and rHDL A-Iϩ32 , although the values were higher (i.e. 69% and 68%, respectively), again in agreement with literature data (35). In contrast, significant albeit small differences were observed for the lipid-free I Characterization of lipid-free apoA-I and apoA-I ϩ32 Molecular mass, phospholipid content, and protein size of "lipid-free" apoA-I and apoA-I ϩ32 were determined by ESI-MS, enzymatic assay and by size exclusion HPLC, respectively, as described under "Experimental Procedures." HPLC was performed using three separate preparations of apoA-I and apoA-I ϩ32 at 0.1, 0. The protein size is given as the ratio of the molecular weight measured (M r(meas) ) to the molecular weight of the monomeric apolipoprotein (M r(mono) ).
apolipoprotein when comparing the parallel ␤-sheet (4.3 Ϯ 1.1 and 1.7 Ϯ 1.5 for apoA-I and apoA-I ϩ32 , respectively) with the ␤-turns (16.0 Ϯ 1.7 versus 20.7 Ϯ 1.5); there was no significant difference in the content of antiparallel ␤-sheet. However, when associated with phospholipids, these differences were no longer apparent (Table III).
Biochemical Characterization-ApoA-I acts as a cofactor for LCAT, and repeat 6 (residues 143-164) is thought to regulate this process (36,37), although residues 99 -120 (containing Met 112 ) may also participate (38). As can be seen in Table IV, the ability of drHDL A-I and drHDL A-Iϩ32 to activate LCAT was comparable. Changes in the apparent kinetic parameters of LCAT of Ն3-fold are needed for differences to be physiologically meaningful (39). Our findings are consistent with Met 148 not being oxidized in apoA-I ϩ32 (Table II).
Because the ability to associate with lipids is a feature determining the ability of apoA-I to promote cellular lipid efflux, we next examined the rate of clearance of multilamellar DMPC liposomes by lipid-free apoA-I and apoA-I ϩ32 . As can be seen in Fig. 3, the kinetics of liposome clearance were similar, although the time required for the initial relative turbidity [(A 0 Ϫ A ϱ )/A 0 ] to decrease to half (i.e. t 1/2 ) was significantly shorter for apoA-I ϩ32 than apoA-I. Thus, the rate constant, k 1/2 (k 1/2 ϭ 1/t 1/2 ) (25), was 0.7 Ϯ 0.4 and 1.6 Ϯ 0.8 min Ϫ1 for apoA-I and apoA-I ϩ32 , respectively (mean Ϯ S.D., n ϭ 9; p ϭ 0.0003). The resulting ratio of k 1/2 A-Iϩ32 /k 1/2 A-I was 2.4 Ϯ 0.5, indicating that apoA-I ϩ32 converted multilamellar liposomes to small unilamellar vesicles two to three times faster than did apoA-I. This suggests that the introduction of the sulfoxide moieties increased the ability of lipid-free apoA-I to interact with phospholipids.
Cellular Studies-ApoA-I promotes the efflux of cholesterol from peripheral cells (3), and ␣-helices containing Met 86 and Met 112 are thought to be important for this process (40,41). To test whether selective oxidation of apoA-I altered cholesterol efflux, we preincubated hMDM for 48 h with medium containing acLDL and [ 3 H]cholesterol and then overnight with medium without supplements (see "Experimental Procedures"). During such lipid loading and subsequent equilibration, cells acquired 6.6% of the radioactivity added (i.e. 1.1 Ϯ 0.1 ϫ 10 6 cpm/mg of cell protein; mean Ϯ S.D.; n ϭ 3). Cellular [ 3 H]cholesterol and cholesterol mass were distributed equally between unesterified and esterified cholesterol pools as determined by HPLC with UV 210 nm and on-line radiometric detection (not shown). Three independent experiments showed that 51-73% of cholesterol label and mass were present as CE. Lipid-free apoA-I and apoA-I ϩ32 promoted a time-dependent efflux of [ 3 H]cholesterol (Fig. 4). During short term incubations (Fig. 4,  inset) apoA-I ϩ32 achieved significantly greater efflux than The final volume of the incubation mixtures was 0.23 ml. Aliquots of the unprocessed incubation mixtures were subjected to electrophoresis on 3-40% nondenaturing polyacrylamide gradient gels and immunoblotted for apoA-I. The remainder of the incubation mixtures were subjected to ultracentrifugation, the isolated srHDL (d Ͻ 1.25 g/ml) electrophoresed on a 3-30% nondenaturing polyacrylamide gradient gel, and the gels were stained with Coomassie Blue, destained, and then scanned as described under "Experimental Procedures." Such scans for srHDL A-I and srHDL A-Iϩ32 are shown in a and b, respectively. Scans of the corresponding immunoblots are shown in c and d. Lanes 1 and 2 show, respectively, srHDL maintained at 4°C and incubated at 37°C, in the absence of CETP. Lanes 3-7 show srHDL incubated with Intralipid and CETP for 1, 3, 6, 12, or 24 h, respectively. Lipid-free apoA-I is shown in lane 8.

FIG. 2. HPLC traces of proteolytic peptides of apoA-I (A) and
apoA-I ؉32 (B) obtained after endoproteolytic digests. 50 g of apolipoproteins were digested using endoprotease AspN and applied to a C18 RP column to elute peptides as described under "Experimental Procedures." Peaks 1-10 represent or contain peptides of apoA-I or apoA-I ϩ32 with identified mass and sequence (see Table II apoA-I at all time points (p Ͻ 0.001). After 6 h, 11.3 Ϯ 1.4% and 7.0 Ϯ 1.2% of the cellular [ 3 H]cholesterol was released into the medium for apoA-I ϩ32 and apoA-I, respectively (Fig. 4), whereas only 2.6 Ϯ 0.4% of the cellular radioactivity was released in controls (medium alone) (p Ͻ 0.05). All of the effluxed [ 3 H]radioactivity was associated with unesterified cholesterol, as determined by HPLC with UV 210 nm and on-line radiometric detection (not shown). After 24-h incubation 2.3-and 2.8-fold more [ 3 H]cholesterol effluxed in the presence of apoA-I and apoA-I ϩ32 , respectively, compared with control incubations; cholesterol efflux remained significantly greater for apoA-I ϩ32 than apoA-I but only by a factor of 1.2 (Fig. 4). Thus, the enhanced removal of cellular cholesterol by lipid-free apoA-I ϩ32 is observed primarily during early stages of incubation of the apolipoprotein with the cells, consistent with the enhanced lipid affinity of apoA-I ϩ32 versus apoA-I (Fig. 3).
We next examined the concentration dependence of [ 3 H]cholesterol efflux from lipid-laden hMDM by lipid-free apoA-I and apoA-I ϩ32 . [ 3 H]Cholesterol efflux increased dose dependently and was more pronounced for apoA-I ϩ32 than for apoA-I, independent of the incubation time used (Fig. 5). The enhanced efflux by apoA-I ϩ32 seemed more apparent at low protein concentration, suggesting that dissociation of even a small proportion of apoA-I ϩ32 from oxidized HDL may enhance cholesterol efflux compared with apoA-I dissociated from unoxidized HDL.
The association of phospholipids with apoA-I yields distinct lipoprotein particles that are better acceptors of cellular cholesterol than is lipid-free apoA-I (3). Consistent with this, more [ 3 H]cholesterol was released from hMDM to drHDL A-I and drHDL A-Iϩ32 than the corresponding lipid-free apolipoproteins in short and long term incubations (compare Figs. 4 and 6). After 6 h of incubation, 16.1 Ϯ 2.1% (for drHDL A-I ) and 16.7 Ϯ 1.6% (for drHDL A-Iϩ32 ) of the cellular [ 3 H]cholesterol was present in the medium. However, drHDL A-I and drHDL A-Iϩ32 were equally effective (Fig. 6), in contrast to the lipid-free apolipoproteins. Thus, once associated with substantial amounts of phospholipids and cholesterol, apoA-I and apoA-I ϩ32 have similar cholesterol efflux activity.

TABLE IV
Lipid-associated apoA-I and apoA-I ϩ32 have a comparable ability to activate LCAT LCAT activation was examined using drHDL containing apoA-I or apoA-I ϩ32 as described under "Experimental Procedures." Michaelis-Menten kinetic parameters were determined in three separate experiments, each using a different preparation of drHDL. V max is given in nanomoles of CE/ml of LCAT/h, and k cat represents the efficiency of the enzyme-substrate reaction (k cat ϭ V max /K m ). The correlation coefficient for the respective Lineweaver-Burk plots is represented by r.

TABLE II
ESI-MS and amino acid sequences of isolated peptides produced by endoproteolytic digests of apoA-I and apoA-I ϩ32 Apolipoproteins isolated from RP-HPLC were digested with endoprotease AspN. The resulting peptides were then separated by RP-HPLC (see Fig. 2), collected, lyophilized, and subjected to ESI-MS as described under "Experimental Procedures." Peptides shown are numbered according to the peak labels shown in Fig. 2 TABLE III Secondary structure prediction of lipid-free and lipid-associated apoA-I or apoA-I ϩ32 CD spectra were recorded on a Jasco spectrometer between 184 and 260 nm as described under "Experimental Procedures." Analysis of the secondary structure was performed using the "variable selection" procedure described under "Experimental Procedures." Cholesterol loading enhances the apolipoprotein-mediated efflux of cholesterol and phospholipids, and apolipoproteinmediated cholesterol efflux appears to be critically dependent on the initial removal or microsolubilization of membrane phospholipids (42). Fig. 7 shows that, similar to cholesterol, the efflux of [ 3 H]phospholipids from lipid-laden hMDM was greater for lipid-free apoA-I ϩ32 than for apoA-I.
In addition to cholesterol and phospholipids, ␣-TOH is another important constituent of all cell membranes. We therefore assessed whether apoA-I ϩ32 is also more effective than apoA-I in removing this lipid from [ 14 C]␣-TOH-labeled hMDM. Incubation of such cells with lipid-free apolipoprotein A-I resulted in a time-dependent efflux of [ 14 C]␣-TOH. The extent of this efflux was greater for apoA-I ϩ32 than apoA-I (Fig. 8). As ␣-TOH and cholesterol efflux experiments were performed under comparable conditions (see "Experimental Procedures"), the relative extents of release of the two lipids were compared. Cells acquired 2.25 Ϯ 0.02 ϫ 10 5 cpm/mg of cell protein (mean Ϯ S.D., n ϭ 3) or 4.9% of the [ 14 C]␣-TOH added. 4.4 Ϯ 0.5% and 5.6 Ϯ 0.6% of cell [ 14 C]␣-TOH was released after 3 h, compared with 4.5 Ϯ 0.5% and 6.1 Ϯ 0.8% of cellular cholesterol for apoA-I and apoA-I ϩ32 , respectively. Thus, apolipoprotein A-I removes these two lipids with comparable efficacy from hMDM, although apoA-I ϩ32 is superior in doing this compared with apoA-I. Similar to the efflux of [ 3 H]cholesterol, lipid-associated apoA-I ϩ32 and apoA-I had comparable efflux-promoting activity for [ 14 C]␣-tocopherol (not shown).

DISCUSSION
Oxidized LDL is now generally thought to contribute to atherogenesis, because it can cause lipid accumulation in macrophages and has a number of other potentially proatherogenic properties (43). There is convincing evidence for the presence of oxidized LDL in human atherosclerotic lesions (see e.g. Ref. 44), and this has recently been extended to HDL (14). Therefore, studying the properties of oxidized HDL may be relevant to atherogenesis. The present study characterized and compared known properties of in vitro selectively oxidized (i.e. apoA-I ϩ32 ) with native, nonoxidized apoA-I. Compared with native apoA-I, lipid-free apoA-I ϩ32 had a greater affinity for lipids and showed an enhanced ability to promote lipid efflux from lipid-laden human macrophages. This increased activity was no longer seen with lipid-associated apoA-I ϩ32 (drHDL A-Iϩ32). Selective oxidation was not associated with a decrease in the ability of lipid-associated apolipoprotein to activate LCAT, and compared with apoA-I, apoA-I ϩ32 dissociated more readily from srHDL in the presence of CETP. Together our findings suggest that selective and limited oxidation does not decrease, but rather increases, the potential antiatherogenic properties of lipid-free apoA-I.
HPLC-purified apoA-I ϩ32 was used throughout the present study. Compared with native apoA-I, apoA-I ϩ32 contains two Met(O) instead of Met residues (11,19). The presence of Met(O) was confirmed indirectly by coelution from the HPLC column of apoA-I ϩ32 used in the present study with modified forms of apoA-I produced during AAPH-induced oxidation of HDL (not shown) and the increased mass (Table I). Therefore, the enhanced ability of apoA-I ϩ32 to associate with phospholipids ( Fig. 3)  moieties were located at Met 112 and Met 148 (19). In contrast, the two sulfoxide moieties in apoA-I ϩ32 formed during the oxidation of HDL with aqueous peroxyl radicals (this study) are located at Met 86 and Met 112 , as demonstrated unambiguously by amino acid sequencing and mass determination of the proteolytic fragments of apoAI ϩ32 (Table II). This suggests that different Met residues may become oxidized depending on the prevailing oxidizing conditions. In preliminary experiments, we observed that H 2 O 2 also caused the conversion of Met 86 and Met 112 to the corresponding Met(O), independent of whether lipid-free or lipid-associated apoA-I is used. 2 Possible reasons for the apparent discrepancies between these findings and those reported earlier by others (19) are presently being investigated. In any case, the results obtained with reagent H 2 O 2 contrasts those obtained with peroxyl radicals and other conditions giving rise to one electron oxidants, where lipid association, i.e. the formation of lipid hydroperoxides, is required for the selective oxidation of apoA-I, as we have demonstrated previously (11,12). Our observations obtained with apoA-I ϩ32 isolated from AAPH-oxidized, lipid hydroperoxide-containing HDL may be of general relevance, because radical oxidants generally induce the formation of CE-OOH (11,12). Also, as CE-OOH are detected in HDL of human plasma (6) and atherosclerotic lesions (14), our findings may be relevant in vivo.
The mechanism responsible for the enhanced ability of isolated apoA-I ϩ32 to associate with liposomal and cellular lipids is not known at present. The fact that enhanced efflux is seen with three different classes of lipids (phospholipids, cholesterol, and ␣-TOH) suggests that it relates to a general characteristic of the modified apolipoprotein. Amphipathic helices of apoA-I are responsible for its association with lipids. Because there was no significant difference in the ␣-helical content of apoA-I ϩ32 and apoA-I (Table III), an oxidation-induced change in this secondary structure per se is not likely to explain the present findings. The hydrophobic residues Met 86 and Met 112 are each located in the nonpolar face adjacent to the polar face at the surface of amphipathic helices 2 and 4 of apoA-I, respectively ( Fig. 9) (45). Jonas et al. (45) suggested that introduction of the polar sulfoxide moiety at the boundary between the polar and nonpolar faces changes the ratio of charged to neutral surface area. Such change decreases the hydrophobic moment and hence could alter the lipid affinity of the corresponding helices of the apolipoprotein (Fig. 9) (45). Met 112 represents a hydrophobic residue in one of the apoA-I canonical "modified" heptad repeats, and this modified heptad repeat is particularly rich in aromatic residues and has a high hydrophobic moment (45,46). The interruption of a canonical heptad repeat has been proposed to rotate the orientation of the hydrophobic residues that follow (46). Therefore, introduction of a polar sulfoxide could reorient the direction of the hydrophobic face following Met 112 , and this could conceivably alter the lipid affinity. In any case, the fact that apoA-I ϩ32 (11) and its oxidized peptides (Fig. 2) elute before apoA-I and the respective nonoxidized peptides, respectively, on RP-HPLC, clearly indicates a substantial change in overall polarity of repeats 2 and 4 as the sole result of conversion of the Met residues into Met(O).
A striking feature of the present study is that the observed increased ability of apoA-I ϩ32 to cause lipid efflux from cells was lost upon reconstitution of the apolipoprotein into discoidal particles. It is well established that the conformation of apoA-I is altered upon association with phospholipids (46). These polar 2 M. Raftery, U. Panzenböck, and R. Stocker, unpublished. lipids may override any polarity difference between apoA-I ϩ32 and apoA-I.
Lipid-free apoA-I binds to incompletely understood cell surface sites that may include the ABC1 transporter (47). In combination with physicochemical factors, specific binding may contribute to the removal of cellular lipids such as cholesterol (3). In the current study, we have not addressed the relative importance of these two pathways for the observed lipid efflux induced. However, a novel observation is that the extent of removal of cellular cholesterol and ␣-TOH was comparable for both apoA-I and apoA-I ϩ32 . This suggests that apoA-I is an important acceptor of cellular ␣-TOH and that ATP-binding cassette transporter 1 is involved in the transport of the vitamin across the cell membranes. Interestingly, ␣-TOH can promote the formation of apoA-I ϩ32 (11), and this could conceivably contribute to subsequent enhanced efflux of cholesterol from lipid-laden cells.
Previous studies on the effect of oxidation on HDL's ability to promote cholesterol efflux have yielded contradicting results. Where a decreased ability was reported (15,17), harsh oxidizing conditions were used and the oxidative modification inflicted on apoA-I was not characterized in detail. The physiological relevance of the oxidizing conditions used in these studies is not known. Heinecke and coworkers (48) exposed HDL to tyrosyl radicals, which may be formed in human atherosclerotic lesions. "Tyrosylated" HDL had an enhanced ability to decrease cholesterol mass in lipid-laden mouse macrophages and was more potent than native HDL at inhibiting cholesterol esterification by acyl-coenzyme A:cholesterol acyltransferase (16). Tyrosyl radicals can cross-link proteins (49), including apoA-I and apoA-II in HDL (50), and rHDL, containing the apoA-I/apoA-II heterodimer, has an enhanced ability to promote cholesterol efflux from fibroblasts and to inhibit accumulation of LDL-derived cholesterol mass by fibroblasts (50). Our studies and those of Francis et al. (50) thus establish a precedent that selective oxidation of HDL's apolipoproteins can promote rather than inhibit cellular lipid efflux. Interestingly, however, although the association with lipids increased the lipid efflux promoting activity of the apoA-I/apoA-II het-erodimer (50), the difference in this activity between apoA-I ϩ32 and apoA-I observed here was lost upon reconstitution of the protein with lipid. This distinguishes our observations from those of Francis and coworkers, although both report a promotion of lipid efflux by oxidized apoA-I/HDL.
The physiological relevance of the present findings is presently unclear. At least two parameters need to be considered, namely, the occurrence of HDL oxidation leading to the formation of human apoA-I ϩ32 and the presence of lipid-free apoA-I/ apoA-I ϩ32 in vivo. As indicated earlier, the contents of CE-OOH and CE-OH associated with human lesion HDL are comparable to those in LDL (14), and CE-OOH themselves or conditions that lead to their formation, including tyrosyl radicals, convert apoA-I to apoA-I ϩ32 (11,12). Thus, it is conceivable that formation of Met(O) in apoA-I takes place in lesions. Indeed, preliminary results indicate that a substantial proportion of apoA-I present in HDL isolated from human lesions coelutes on HPLC with apoA-I ϩ32 . 3 There is also evidence that apoA-I dissociates from HDL (4). The existence of lipid-free apoA-I in vivo has been indicated both in plasma and in interstitial fluid (51). Most apoA-I secreted by cells and pre-␤-1 HDL, thought to represent a major cholesterol acceptor in vivo (52), is associated with substantial amounts of phospholipids. This is not surprising, given the high affinity of apoA-I for phospholipids and the ready availability of phospholipids in vivo. However, this does not exclude the possibility that lipid-free apoA-I exists in vivo. K m values for the reaction of lipid-free apoA-I with cells have been reported to be on the order of 0.2-1 ϫ 10 Ϫ7 M, i.e. only around 1/500 of its concentration in plasma (51). Therefore, only one of several hundred apoA-I molecules is necessary to be free to carry out cellular cholesterol and phospholipid efflux at V max (51). Indeed, upon removal of core lipids, lipid-free apoA-I may dissociate from the surface of HDL (3) and be transferred to cell surface sites to generate pre-␤-1 HDL with cellular lipid at a high rate (51). This reaction has been suggested to be relevant to physiological conditions (51). Our finding, i.e. apoA-I ϩ32 dissociates more readily from reconstituted HDL than apoA-I in the presence of CETP (Fig. 1), supports the notion that this process is relevant for apoA-I ϩ32 present on oxidized HDL in vivo. If a similar process were to occur in the intima of a developing lesion where HDL oxidation takes place (14), lipidfree apoA-I ϩ32 may be present and contribute to cholesterol efflux. Future studies are needed to investigate this possibility.