Proteomic and Transcriptomic Analyses of Macrophages with an Increased Resistance to Oxidized Low Density Lipoprotein (oxLDL)-induced Cytotoxicity Generated by Chronic Exposure to oxLDL*

The uptake of oxidized low density lipoprotein (oxLDL) by macrophages leads to foam cell formation and fatty streaks, which represent early sites of potential atheroma development. We developed a cell culture model of chronic oxLDL exposure to determine whether hallmark parameters of oxLDL uptake and cytotoxicity are altered during foam cell formation and to determine changes in protein and mRNA expression that distinguish acute and chronic oxLDL exposure. Although the extent of oxLDL uptake did not change, a resistance to oxLDL-induced cytotoxicity was observed in the chronically exposed cells. Macrophages that have been chronically exposed to oxLDL required a 40% higher concentration of oxLDL to achieve 50% survival in a 48-h treatment relative to macrophages subjected to a single oxLDL exposure. A main feature of the differentially expressed proteome was a series of significantly overexpressed antioxidant and antioxidant-related proteins in the oxLDL-exposed cells. A large proportion of these proteins (45%) was overexpressed in the chronically exposed cells prior to the oxLDL treatment, indicative of the unique phenotype produced by the chronic treatment. Analysis of the transcriptome also revealed a broad increase in the expression of antioxidant and antioxidant-related proteins. In addition, the transcriptome experiments found an increased inflammatory response under conditions of both acute and chronic oxLDL exposure. Overall the combined functional, proteomic, and transcriptomic experiments show that macrophages respond to oxLDL by developing an oxidative stress resistance that increases and stabilizes with chronic exposure. Furthermore this protective response and the increased foam cell survival that it supports amplifies their proatherogenic role by promoting a continued inflammatory state.

The uptake of oxidized low density lipoprotein (oxLDL) by macrophages leads to foam cell formation and fatty streaks, which represent early sites of potential atheroma development. We developed a cell culture model of chronic oxLDL exposure to determine whether hallmark parameters of oxLDL uptake and cytotoxicity are altered during foam cell formation and to determine changes in protein and mRNA expression that distinguish acute and chronic oxLDL exposure. Although the extent of oxLDL uptake did not change, a resistance to oxLDL-induced cytotoxicity was observed in the chronically exposed cells. Macrophages that have been chronically exposed to oxLDL required a 40% higher concentration of oxLDL to achieve 50% survival in a 48-h treatment relative to macrophages subjected to a single oxLDL exposure. A main feature of the differentially expressed proteome was a series of significantly overexpressed antioxidant and antioxidant-related proteins in the oxLDL-exposed cells. A large proportion of these proteins (45%) was overexpressed in the chronically exposed cells prior to the oxLDL treatment, indicative of the unique phenotype produced by the chronic treatment. Analysis of the transcriptome also revealed a broad increase in the expression of antioxidant and antioxidant-related proteins. In addition, the transcriptome experiments found an increased inflammatory response under conditions of both acute and chronic oxLDL exposure. Overall the combined functional, proteomic, and transcriptomic experiments show that macrophages respond to oxLDL by developing an oxidative stress resistance that increases and stabilizes with chronic exposure. Furthermore this protective response and the increased foam cell survival that it supports amplifies their proatherogenic role by promoting a continued inflammatory state.

Molecular & Cellular Proteomics 4:1522-1540, 2005.
The aggregation of macrophage-derived foam cells into fatty streaks is a critical event in early atherosclerosis. These foam cells are formed from macrophages residing in the subendothelial space of the vasculature that have internalized large quantities of low density lipoprotein (LDL) 1 (1,2). Although not all fatty streaks lead to advanced atherosclerosis, they represent potential sights for atheroma development (3,4). Macrophage-derived foam cells are also a major component of advanced atheromas (5). In fact, foam cells are present in all atherosclerotic lesions, from the earliest fatty streak formations to the most advanced atheromas (6), meriting a continued interest in macrophage and foam cell biology as therapeutic targets for the prevention and treatment of atherosclerosis (7,8).
Two processes are key contributors to foam cell formation: the modification of LDL and the unregulated uptake of that modified LDL by macrophages. Brown et al. (9) discovered that acetylated and malondialdehyde-modified LDL, but not native LDL, was bound at surface receptor sites of macrophages and internalized in a manner that is not regulated by cellular cholesterol content. Steinberg and co-workers demonstrated that this modification was oxidative and could be achieved by either copper-mediated reactions or in cell-mediated processes produced by incubating LDL with endothelial cells, smooth muscle cells, or macrophages (10,11). Concurrently Chisolm and co-workers demonstrated that oxidized LDL (oxLDL) was cytotoxic to cells in culture (12) and that this toxicity was dependent upon oxidative modification to LDL (13). Thus the "oxidative modification hypothesis of atherogenesis" has evolved to describe the cell-mediated formation of cytotoxic, oxidatively modified LDL that is internalized by macrophages at the earliest stages of atheroma development (14).
Although many factors contribute to the progression of atherosclerosis, foam cells appear to have a uniquely proatherogenic phenotype that goes beyond the lipid deposition that is produced by the unregulated LDL uptake. Foam cells contribute to an inflammatory response through enhanced expression of intercellular adhesion molecule-1 and vascular endothelial growth factor (15). Additionally the inflammatory cytokines interleukin-1 (IL-1), IL-6, IL-8, and tumor necrosis factor (TNF) are expressed by foam cells in human atherosclerotic lesions (16,17). Foam cells participate in the immune response by expressing the lipid presentation molecule CD1 that co-localizes in the arterial wall in T cell-rich areas, implying an interaction between these major immune response cells (17). Foam cells also express and activate matrix metalloproteases that contribute to plaque instability, leading to heart attack and stroke (6,18). Overall the complexity of the foam cell response to oxLDL loading makes this system an excellent subject for system biology approaches such as proteomic and transcriptomic experiments.
Other investigations have focused on characterizing the transcriptome of oxLDL-treated macrophages (19) as well as identifying proteins found in culture media following foam cell formation (20). A proteomic approach has not been applied, however, to the cellular proteins of foam cells. Importantly those previous experiments also used only a single oxLDL exposure, so the effect of chronic oxLDL exposure on the foam cell phenotype is unknown. Because atherosclerosis is a disease seen primarily later in life, it is important to model this long term oxLDL exposure component of the disease progression. In fact, other models of chronic exposure to agents such as ethanol (21) and hydrogen peroxide (22,23) have identified important differences in protein expression, relative to the acute exposure conditions, that contributed to distinct cellular phenotypes.
In the present study, differentially regulated proteins and genes were identified in foam cells derived from J774 murine macrophages treated with oxLDL. A chronic exposure model was developed that used repeated exposure of the J774 cells to oxLDL over a period of several months to generate a derivative line referred to as J774-CE cells. A system biology approach has been used to characterize this chronic exposure model and test the hypothesis that changes in foam cell phenotype are produced by long term exposure to oxLDL relative to both untreated macrophages and foam cells formed by a single oxLDL exposure.

EXPERIMENTAL PROCEDURES
Lipoprotein Preparation-Fresh human plasma was obtained from the Blood Bank of the Cleveland Clinic Foundation, and LDL (1.019 Ͻ d Ͻ 1.063 g/ml) was isolated by differential ultracentrifugation (24). LDL in NaBr solution containing 0.02% EDTA was dialyzed against 0.9% NaCl, 0.02% NaN 3 , 0.02% EDTA (pH 7.4) and stored in the dark at 4°C (25). oxLDL was achieved by dialyzing 500 g/ml LDL in TBS (20 mM Tris, pH 7.8, 150 mM NaCl) for 24 h in Spectra/Por dialysis tubing (molecular mass cutoff, 6 -8 kDa) (Spectrum Laboratories, Rancho Domingues, CA) followed by dialysis in TBS and 5 M CuSO 4 for 24 h. Oxidation was stopped by a final dialysis in TBS and 0.3 mM EDTA for 24 h. The extent of oxidation was monitored by using the trinitrobenzene sulfonic acid assay to estimate the number of free amines. Previous investigators have shown that the apoB-100 protein of oxidatively modified LDL contains fewer oxidizable histidines and lysines in comparison to unmodified LDL (26). In our system, a typical Cu 2ϩ oxidation procedure resulted in 40% modified amines compared with 3% modified amines when LDL was dialyzed in TBS without 5 M CuSO 4 .
Cell Culture Conditions and Treatment-J774A.1 murine macrophages were obtained from American Type Culture Collection (ATCC TIB 67) and maintained in culture media (Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 100 IU/ml penicillin, and 100 g/ml streptomycin) at 37°C in 5% CO 2 . Two J774 sublines were developed through chronic exposure to oxLDL. J774-CE macrophages were treated with 50 g/ml oxLDL for 24 h before each passage for a period of ϳ3 months. J774-CE(Ϫ) macrophages were derived from the J774-CE subline by stopping the chronic oxLDL treatment. All cells were passed twice each week in 75-cm 2 tissue culture flasks. For all experiments, the J774-CE macrophages that were to be used were grown in a separate flask and kept free of oxLDL for 4 -5 days prior to treatment to ensure that the cells were no longer lipid-loaded. Foam cells were generated by adding 50 g/ml oxLDL to culture media of J774 macrophages followed by a 48-h incubation.
Cytotoxicity Assay-Approximately 10 4 cells were seeded into 24well tissue culture plates and incubated overnight. Cells were treated with increasing concentrations of oxLDL (0 -200 g/ml) for 48 h. Cell survival was assayed with the CellTiter 96 Aqueous One Solution cell proliferation assay (Promega, Madison, WI), a colorimetric assay that produces a soluble formazan product that is directly proportional to the number of metabolically active cells. 200 l of media/sample were transferred to a 96-well plate, and the absorbance at 490 nm was read in a microplate reader. Background absorbance of wells without cells was subtracted, and the resulting values were normalized to corresponding values for untreated cell and expressed as percent cell survival.
Lipoprotein Labeling and oxLDL Internalization Assay-Native LDL and oxLDL were labeled with the lipophilic tracer 1,1Ј-dioctadecyl-3,3,3Ј,3Ј-tetramethylindocarbocyanine perchlorate (DiI; Molecular Probes, Eugene, OR). DiI was maintained in the dark at room temperature as a 10 mg/ml stock in DMSO. Lipoprotein labeling was performed by adding 20 l of DiI stock/1 ml of lipoprotein (500 g/ml) or TBS vehicle (150 mM NaCl, 20 mM Tris-HCl, pH 7.8). The DiI mixtures were vortexed briefly, incubated in the dark at 37°C overnight, and centrifuged at 10,000 ϫ g for 5 min to pellet unbound DiI. The supernatants were sterile filtered with a Millex-GP filter unit (0.2-m pore size polyethersulfone; Millipore, Billerica, MA) and kept in the dark until needed. The DiI vehicle was clarified following filtering, verifying the removal of unbound DiI.
Extent of foam cell formation was assayed by measuring DiIlabeled oxLDL internalization. DiI-labeled LDL and TBS vehicle were used as negative controls. Samples were incubated with 50 g/ml DiI-oxLDL, -LDL, or vehicle for fluorescent microscopy or 0 -100 g/ml DiI-oxLDL for dose-response experiments. Following 48 h of incubation at 37°C in 5% CO 2 , DiI internalization was assayed by fluorescence microscopy or in a fluorescence plate reader (excitation ϭ 540 nm, emission ϭ 565 nm, cutoff ϭ 550 nm). Background fluorescence was measured in DiI-oxLDL-incubated (0 -100 g/ml) wells with no cells and subtracted from corresponding raw data values. Results were normalized by total number of cells, assayed by 4Ј,6-diamidino-2-phenylindole, dihydrochloride (DAPI) nuclei staining (excitation ϭ 358 nm, emission ϭ 461 nm, cutoff ϭ 420 nm) or the CellTiter 96 cell survival assay described above.
2D Gel Electrophoresis-Approximately 10 6 cells were plated in 100-mm tissue culture plates and incubated overnight. The following day, cells were treated with 50 g/ml oxLDL or an equivalent volume of vehicle (20 mM Tris-buffered saline, pH 7.8). Cells were incubated for 48 h, then harvested by scraping in Dulbecco's modified Eagle's medium, and washed once in sterile PBS. Cell pellets were suspended in lysis buffer (25 mM Tris, pH 7.5, 2.5 mM MgCl 2 , 0.5% SDS) and boiled for 5 min with occasional vortexing. Samples were cooled to room temperature and treated with 100 g/ml DNase I and 100 g/ml RNase A for 15 min. Protein concentration was determined using an SDS-compatible Lowry-based method (DC protein assay kit, Bio-Rad), and sample volumes were adjusted with lysis buffer to equal concentrations. For 2D gel analysis (27), 1 mg of total protein (typically 200 -400 l of total cell lysate) was precipitated in acetone (80%, v/v) overnight. Protein precipitate was resuspended in a rehydration buffer consisting of 7 M urea, 2 M thiourea, 1% dithiothreitol, 1% CHAPS, 1% Triton X-100, and 1% ampholytes (Bio-Rad Bio-Lyte 3/10). A Bio-Rad Protean IEF Cell was used to rehydrate 11-cm IPG strips with 200 l of samples (1 mg of total protein) at 50 V for 16 h. An isoelectric focusing program (8000 V maximum) was applied that accumulated 35,000 V-h over a 5-h period, separating proteins in one dimension by net charge. The focused IPG strips were equilibrated in SDS under reducing and alkylating conditions with 6 M urea, 2% SDS, 20% glycerol containing first 100 mM dithiothreitol (reduction)and then 100 mM iodoacetamide (alkylation) for 10 min. Excess amounts of each reagent (2 ml/strip) were used, and the DTT solution was removed prior to addition of the iodoacetamide. The second dimension of separation was achieved by subjecting the focused IPG strip to SDS-PAGE in a 12.5% Tris-HCl gel. Gels were fixed in 50% ethanol, 10% acetic acid, washed in distilled water 3 ϫ 15 min, and stained overnight in Bio-Rad Gel Code Blue. Stained gels were washed in distilled water and stored in heat-sealable bags (Kapak, Minneapolis, MN) with 2 ml of 5% acetic acid, 10% glycerol.
Analysis of Differential Protein Expression-Differentially expressed protein bands were determined by comparing 2D SDS-PAGE gels representing oxLDL-treated versus untreated cell cultures. The four samples of interest (untreated and oxLDL-treated J774, untreated and oxLDL-treated J774-CE) were each represented by three replicate gels originating from independent cell culture experiments. Each gel was scanned with a Bio-Rad GS-700 densitometer and imported in Bio-Rad PDQuest 2D analysis software. Replicate gels were combined into groups and normalized to the total density of detected bands, and protein bands were matched across the set of 12 gels. Resulting band intensities were an average across each replicate group, and standard deviations were calculated based on the band intensities of the individual gels. Over-and underexpressed protein bands were determined using a Student's t test within PDQuest with significance level set to 95% (p Ͻ 0.05). The resulting band sets were visually inspected to verify band quality and the integrity of the statistical significance.
Protein Identification by Tandem Mass Spectrometry-Protein bands from 2D SDS-PAGE gels were excised, destained in 50% ethanol, 5% acetic acid, and dehydrated in acetonitrile. Gel pieces were dried in a SpeedVac, rehydrated with 5 l of trypsin solution (20 g/ml in 50 mM ammonium bicarbonate), and incubated overnight at room temperature. Peptides were extracted with 30 l of 50% acetonitrile, 10% formic acid, and these extracts were concentrated in a SpeedVac for 30 min to ϳ5 l. Sample volume was adjusted to 30 l with 1% acetic acid.
Protein digests were analyzed with a Finnigan LTQ linear ion trap mass spectrometer system incorporating a 6.5-cm ϫ 75-m-inner diameter Phenomenex Jupiter C 18 reversed-phase capillary chromatography column (Phenomenex, Torrance, CA). Peptide samples were autoinjected and eluted from the column by an acetonitrile, 0.05 M acetic acid gradient at a flow rate of 1 l/min. Digests were analyzed in a data-dependent mode during which the instrument acquired a full mass scan spectrum to determine peptide molecular weights followed by product ion spectra to determine amino acid sequence in successive scans.
The CID spectra produced by each tryptic digest were searched against the National Center for Biotechnology Information (NCBI) non-redundant database using Mascot software (Matrix Science, Boston, MA). Peptide tolerance was set to Ϯ2 Da allowing 1 missed cleavage, and peptide charges of ϩ2 and ϩ3 were fragmented. MS/MS tolerance was set to Ϯ1 Da. All identified proteins gave Mascot total scores greater than 200, which was calculated from the expression Ϫ10 ϫ log(p) where p is the probability that the match is a random event (p Ͻ 10 Ϫ20 ). Individual peptide scores of 40 or greater represented extensive homology, and at least two of these CID spectra from each analysis were visually inspected to verify the peptide sequence. Typical sequence coverage was at least 20%.
Gene Chip Analysis-Total RNA was isolated from cell samples according to the manufacturer's instructions using the Qiagen RNAeasy kit (Qiagen Inc., Valencia, CA). RNA samples were submitted to the Gene Expression Array Core Facility at Case Western Reserve University for cRNA preparation and gene chip analysis. Instrumentation included the Affymetrix GeneChip Instrument System with a GeneChip Scanner 3000 and autoloader. Message RNA levels were analyzed using the GeneChip® Mouse Expression Set 430 chip set, which consists of 45,037 probe sets representing 34,323 genes. Each probe set consists of 11 unique probe pairs that are 25 nucleotides in length. Each probe pair consists of a perfect match (PM) oligonucleotide and a single base mismatch (MM) oligonucleotide. Thus, a probe set representing a specific gene hybridizes mRNA to 11 different sequence segments with a negative control for each sequence to determine nonspecific hybridization.
Statistical significance of each mRNA identification was confirmed as follows. A discrimination score (R) was calculated as the ratio of target-specific hybridization to overall hybridization where R ϭ 1 is a perfect match with no nonspecific binding: R ϭ (PM Ϫ MM)/(PM ϩ MM). Detection p values were calculated using the one-sided Wilcoxon's signed rank test, which ranks the probe pairs based on the separation of R from a threshold value Tau (Tau ϭ 0.015 for this analysis). mRNA transcripts were denoted as "present" if the result of this statistical test was p Ͻ 0.04. For the purposes of our analysis, mRNA transcripts were considered only if the transcript was present in the untreated sample of an underexpressed pair or in the oxLDLtreated sample of an overexpressed pair. Signal intensity also incorporated the values from each probe pair in a probe set and was calculated using the one-step Tukey's biweight estimate.
The statistical significance of the differential expression for each transcript was determined by the Wilcoxon signed rank test. Each probe set in the differentially expressed sample was compared with its counterpart, and a p value was calculated to determine the degree of change. A value of p Ͻ 0.002 denoted an increase in transcript, and a value of p Ͼ 0.998 denoted a decrease in transcript. All of the differentially expressed transcripts reported here were determined to be increased or decreased in the oxLDL-treated sample of at least one sample pair (J774 untreated/oxLDL-treated or J774-CE untreated/oxLDL-treated) according to this statistical test.
Statistical analysis of the gene chip data was carried out in Affymetrix GeneChip Operating Software. Further data analysis was preformed in Microsoft Access and Excel, and functional categorizations were assigned using Lucidyx Searcher (Lucidyx LLC, Cleveland, OH).

Development of J774-CE Macrophages-
The effects of chronic oxLDL exposure on J774 macrophages were studied by exposing cells to 50 g/ml oxLDL for 24 h twice per week for 3 months, the equivalent of 25 oxLDL exposures. This concentration of oxLDL is consistent with the amount of ox-LDL used by other laboratories to generate foam cells in culture. The twice weekly exposure allowed significant uptake of lipoprotein by the cells with no toxicity. Each week, the cells were harvested by scraping and passaged for continued exposure. This chronic oxLDL exposure subline to the J774 parent line was designated as J774-CE. No differences were observed in the morphology, growth rate, or cell cycle progression of the J774-CE cells relative to the parent J774 cells. The intent of these experiments was to investigate the effects of chronic oxLDL exposure on the two hallmarks of the oxidative hypothesis of atherosclerosis: oxLDL internalization and cytotoxicity.
Lipid Uptake in J774 and J774-CE Macrophages-The extent of foam cell formation was determined by monitoring the uptake of lipid by the cells using two different methods: oil red O staining of fixed cells or DiI labeling of the LDL and oxLDL. Results of the oil red O experiments are shown in Fig. 1, and results of the DiI labeling are shown in Fig. 2.
In the oil red O experiments ( Fig. 1), foam cell formation is indicated by red-stained lipid droplets within the cell cytoplasm. Fig. 1c shows the extent of foam cell formation following incubation of cultured macrophages with 50 g/ml oxLDL for 48 h. In contrast, untreated (Fig. 1a) and LDL-treated ( Fig.  1b) cells exhibit minimal staining. The small number of lipid droplets in LDL-treated macrophages are likely a product of cell-mediated oxidation of LDL over the course of the incubation period. Oil red O, however, is a general neutral lipid stain that can bind to other lipids such as triglycerides. This lack of specificity is particularly evident, for example, when cultured primary human monocyte-macrophages internalize significant quantities of triglyceride during incubation in human serum-containing media and produce extensive oil red O staining (28). Therefore, lipoprotein uptake was confirmed by monitoring the uptake of LDL and oxLDL labeled with the fluorescent lipophilic tracer DiI. Macrophages incubated with DiI-labeled oxLDL are positive for DiI internalization (Fig. 2c) in contrast to macrophages treated with vehicle ( Fig. 2a) and DiI-labeled LDL (Fig. 2b).
The cell nuclei were also counterstained with the fluorescent nucleic acid stain DAPI. A quantitative assessment of DiI internalization was made by determining the ratio of DiI fluorescence at 565 nm relative to DAPI fluorescence at 460 nm ( Fig. 2d). This experiment demonstrated negligible DiI uptake in vehicle-treated cells, whereas DiI-oxLDL-treated cells internalized 5-fold more lipid than DiI-LDL-treated cells. Fig. 3 shows a dose-response analysis of DiI-labeled oxLDL internalization over a range of 0 -100 g/ml oxLDL in J774 and J774-CE macrophages. No statistically significant differences were present in oxLDL internalization when comparing the two cell lines. The general pattern of oxLDL internalization was consistent with other experiments where internalization begins to plateau at ϳ50 g/ml oxLDL, a common concentration used to generate foam cells in culture.
OxLDL-induced Cytotoxicity in J774 and J774-CE Macrophages-A second component of the oxidative modification hypothesis of atherosclerosis is the cytotoxicity of oxLDL. Therefore, oxLDL-induced toxicity was investigated in both the J774 and J774-CE macrophages. Fig. 4 shows the results of a typical cytotoxicity assay in which J774 and J774-CE macrophages were treated with increasing concentrations of oxLDL (ranging from 0 to 200 g/ml) for 48 h. Previous investigators have determined that low doses of oxLDL induce a proliferative response in macrophages (29,30). That observation is verified by these experiments in which 10 and 25 g/ml oxLDL induced proliferation in the both the J774 and J774-CE cell lines. At toxic concentrations of oxLDL, the J774 macrophages required 98 g/ml oxLDL to reduce the cell survival to 50%, whereas the J774-CE macrophages required 140 g/ml to reach 50% cell survival, a 43% increase in oxLDL dose required to produce 50% cytotoxicity. Therefore, the chronic exposure of the J774 macrophages produces a significant resistance to oxLDL-mediated cytotoxicity.
Additional experiments were carried out to determine the stability of this cytotoxicity-resistant phenotype. The repetitive oxLDL treatment was stopped for a population of the J774-CE cells for a period of 2 months, and both cytotoxicity and lipid uptake were assayed. These cells were designated J774-CE(Ϫ). As seen in Fig. 4, no difference was observed in the dose-response course of the oxLDL-mediated cytotoxic- ity indicating that the J774-CE(Ϫ) cells retained the resistant phenotype of the J774-CE cells. Similarly the pattern of DiIlabeled oxLDL uptake was also unchanged in the J774-CE(Ϫ) cells (data not shown). Therefore, these results indicate that chronic exposure of macrophages to oxLDL does not affect lipid uptake characteristics of foam cell formation but does produce a significant degree of resistance to oxLDL-induced cytotoxicity that is stable for at least 2 months after the oxLDL challenges are stopped. An investigation into the differential proteome and transcriptome of the J774-CE cells, in comparison with parental J774 cells, was used to begin the investigation of the proteins and genes that may contribute to this resistant phenotype.
Proteomic Analyses-The differential proteomes of the J774 and J774-CE cell lines were analyzed to determine the effects of foam cell formation on protein expression. A series of 2D SDS-PAGE gels representing total cell lysates were run in replicate experiments covering the pI range from 3 to 8 in overlapping sections of pI 3-6 and pI 5-8. The images were processed to produce a master image of protein bands (Fig.  5a), and the relative amounts of the detected proteins were determined and compared. Statistically significant (p Ͻ 0.05) over-and underexpressed bands were determined in oxLDLtreated J774 or J774-CE macrophages relative to respective untreated controls, and these data are summarized in Table I. Examples of oxLDL-induced overexpression are shown for aldose reductase-related protein 2 (Fig. 5b) and superoxide dismutase (Cu,Zn) (Fig. 5c). The results in Table I are presented as the -fold change in protein quantity for four different comparisons designed to detect differences in protein expression in response to oxLDL treatment. The primary comparison was the effect of oxLDL treatment on protein expres-sion in either the J774 or J774-CE cells (columns A and B, respectively). A total of 45 bands were overexpressed with statistical significance following oxLDL treatment. These bands were identified by tandem mass spectrometry and found to represent 28 unique proteins. Similarly a total of seven bands, representing six unique proteins, were underexpressed with statistical significance following oxLDL treatment. Subsequent evaluations of the protein expression data determined differences in the expression of these protein between the J774 and J774-CE cells before the oxLDL treatment (column C) and the difference in the magnitude of the change in oxLDL-induced protein expression between the J774 and J774-CE cells (column D). Also included in Table I are the gene identification (ID) number, an NCBI reference sequence (RefSeq) number, and a summary of the cellular process of the respective identified proteins.
Of particular interest are the differentially regulated proteins in J774-CE macrophages that have significantly altered ox-LDL-induced responses relative to the response seen in the oxLDL-treated J774 cells. For example, approximately half (25 of 52) of the differentially expressed proteins shown in Table I were differentially expressed (p Ͻ 0.05) in the J774-CE cells, relative to the J774 cells, prior to any treatment. Following oxLDL treatment, 19 of these proteins had statistically different responses in the J774-CE cells compared with the J774 parental cells. The expression patterns for this subset of proteins are shown in Fig. 6. The differences in regulation between the J774 and J774-CE cells provide direct evidence of novel effects of chronic oxLDL treatment that are not seen in single exposure experiments. These differences in protein expression demonstrate the unique phenotype developed in the J774-CE cell by the chronic oxLDL exposure regimen. A 2D gel image generated by Bio-Rad PDQuest software represents the entire set of analyzed gels (four unique samples, three gels per sample) (A). Numbered bands (arrows) represent statistically significant (p Ͻ 0.05) differentially regulated bands identified and tabulated in Table I. Examples of oxLDL-induced overexpression are shown for aldose reductase-related protein 2 (B) and superoxide dismutase (Cu,Zn) (C). Gene Chip Analysis of J774 and J774-CE Macrophages and Foam Cells-Differential mRNA expression was analyzed in an Affymetrix GeneChip experiment using GeneChip mouse expression set 430 microarrays, which contain over 45,000 unique mRNA transcripts. J774 macrophages and the J774-CE subline were analyzed following oxLDL-induced foam cell formation for 48 h and compared with untreated controls. A representative set of differentially expressed transcripts was assembled from mRNA that was Ϯ4-fold differentially expressed in oxLDL-treated J774 or J774-CE macrophages. Under these criteria, a total of 41 transcripts were overexpressed (Table II), and 52 transcripts were underexpressed (Table III). The identified transcripts were grouped according to functional class defined by GeneOntology entries in the NCBI database as provided by Mouse Genome Informatics. Genes that were not assigned a function under GeneOntology were grouped as "undefined." GeneOntology process classifications are listed for each gene when assigned or are otherwise classified as undefined.
Further focus was placed on regulation of genes involved in antioxidant defense by adjusting the selection criteria to transcripts that had greater than Ϯ2-fold differential expression in oxLDL-treated J774 or J774-CE macrophages. Table IV lists the 13 overexpressed genes related to antioxidant responses. It is interesting to note that no antioxidant-related genes were underexpressed by more than 2-fold. The genes listed in Table IV lead directly to antioxidant production or contribute an essential component to antioxidant activity. Glutamate-cysteine ligase (GCL) is an example of indirect antioxidant activity because it catalyzes the rate-limiting step in the synthesis of the antioxidant glutathione. Overexpression of the GCL modifier subunit, which comprises one half of the GCL heterodimer, potentially contributes to an increase in antioxidant response. The Alpha 3 isozyme of GST is overexpressed ϳ10and 20-fold in oxLDL-treated J774 and J774-CE macrophages, respectively. Excluding Alpha 3 GST, the remaining antioxidant defense genes are up-regulated ϳ2-fold in ox-LDL-treated J774 and J774-CE macrophages.
An additional analysis of the transcriptome data was carried out to focus on genes involved in immune response (Table V). For purposes of this classification, "proinflammatory" genes were defined as being involved in leukocyte recruitment or cell proliferation at a site of inflammation. This definition includes cytokines that stimulate leukocyte production and promote leukocyte attachment, cytokines that stimulate smooth muscle cell proliferation or extracellular matrix production, and chemokines that promote leukocyte chemotaxis. Four transcripts were assigned to the inflammatory response process (TNF-␣ and the macrophage inflammatory proteins 1␣, 1␣ receptor, and 2) by GeneOntology designations. The overexpressed TNF receptor superfamily transcript was included in this subset of proinflammatory genes for the purposes of this analysis. Additionally the overexpressed transcripts IL-6 (31-33), oncostatin M (34 -36), and prostaglandin-endoperoxide synthase 2 (37-39) were designated as proinflammatory in this analysis considering their links to the inflammatory re- sponse. This set of eight proinflammatory transcripts were overexpressed an average of 2.5-and 2.4-fold in oxLDLtreated J774 and J774-CE macrophages, respectively, compared with untreated macrophages. However, these transcripts were 2.5-fold underexpressed in the untreated J774-CE subline compared with untreated J774 parent macrophages (steady-state comparison). Similarly proinflammatory transcripts were 2.9-fold underexpressed in oxLDLtreated J774-CE subline macrophages compared with ox-LDL-treated J774 parent macrophages (stimulated state comparison). Therefore, although oxLDL treatment up-regulates inflammatory response in both J774 and J774-CE macrophages, the proinflammatory transcripts were decreased as a whole in the J774-CE subline.

DISCUSSION
Foam cell formation is initiated by the exposure of macrophages to oxidatively modified LDL. These cells then bind the oxLDL to scavenger receptors and take up the oxLDL in an unregulated manner that internalizes large quantities of lipid. Despite the toxic nature of oxLDL, this lipid loading is not immediately toxic to the foam cell. Rather foam cells aggregate at so-called fatty streaks, which become potential sites for the development of advanced atherosclerotic lesions. An important aspect of foam cell formation is the reversible nature of the lipid-loading process. Cholesterol efflux in macrophages is mediated by lipid acceptors such as apolipoprotein A-I in high density lipoprotein, and this process is still functional in the early stages foam cell formation.
A major difference that exists between the in vivo events and in vitro experiments that investigate foam cell formation is the fact that the macrophages commonly used in atherosclerosis research, such as the J774 murine macrophage cell line used here, mouse peritoneal macrophages, and elutriated human monocyte-macrophages, have not been previously exposed to oxLDL. In the experiments described here, a chronic exposure subline (the J774-CE cells) was generated from parent J774 macrophage cells by repeated exposure to oxLDL. The goal was to model the long time frame over which the disease develops that is an intrinsic component of atherosclerosis. Pathological examinations have shown the lipid accumulation in the vasculature begins in the first decade of life and proceeds steadily as one ages (40,41). The morbidity and mortality of the disease, however, typically does not become apparent until after 40 -50 years. The development of the J774-CE subline was intended to model in vitro any effects of this repetitive, long term exposure.
The availability of the J774-CE subline has a secondary effect of providing an additional test of important trends in the data. Of particular interest in these experiments were the proteins and genes for which expression was not altered by the single exposure but was altered by the chronic exposure. Not only do these proteins and transcripts represent a system biology confirmation of the importance of modeling the chronic exposure, but they are also novel observations that could not be made in the standard acute exposure experiments.
As the cells were exposed to oxLDL, foam cell formation was validated in oxLDL-treated J774 murine macrophages by lipid-staining experiments that were carried out either after the lipid had accumulated in the cell (oil red O) or before the lipid was placed on the cells (DiI). Positive staining with both stains verified formation of foam cells. It is known that LDL in culture media with macrophages can be oxidized to some degree by cell-mediated processes, and this form of oxLDL can be taken up by macrophages. Therefore, over the course of a 48-h incubation period, some degree of lipid uptake is expected in LDL-treated macrophages. Although oil red O-stained droplets are apparent in the LDL-treated cells, the occurrence is minimal in comparison to the oxLDL-treated macrophages. The use of DiI labeling not only verified the oil red O staining results but also allowed quantitative analysis of the extent of oxLDL uptake.
The chronic exposure phenotype was tested at two levels: the ability of macrophages to internalize oxLDL and the cytotoxicity of oxLDL. These tests were based on the key events of unregulated oxLDL accumulation through scavenger receptors and the cytotoxicity of oxLDL described in the oxida-tive modification hypothesis of atherosclerosis (14). A doseresponse analysis of DiI-labeled oxLDL internalization revealed that there is no difference in lipid loading between the parental J774 macrophages and the J774-CE and J774-CE(Ϫ) sublines. Therefore, chronic exposure to oxLDL did not change the lipid loading characteristics of foam cell formation. Cell survival experiments, however, showed that chronic exposure to oxLDL did produce a significant resistance in the cells to oxLDL-mediated cytotoxicity. Furthermore this difference in survival was also seen in the J774-CE(Ϫ) subline, indicating that the cytotoxicity-resistant phenotype is stable for at least 2 months.
The initial experiments to characterize the foam cells used a series of proteomic experiments to detect and identify differentially expressed proteins. Previous investigations of foam cell formation have been primarily carried out on the single protein level. A system biology approach has the ability to provide important new information about functional changes in foam cells by determining coordinated changes in groups of proteins. All of the proteomic experiments were carried out in triplicate so that traditional statistical tests could be used to detect differentially expressed proteins.
Notable proteins identified in these experiments with increased expression in oxLDL-exposed cells were a series of antioxidant proteins that react directly with oxidant species, including superoxide dismutase, aldose reductase-related protein, and esterase D. Other proteins that contribute indirectly to oxidant defense through the production of antioxidant molecules or cofactors were also overexpressed with oxLDL treatment. Biliverdin reductase B produces the antioxidant riboflavin and NADPH, a cofactor that represents reducing equivalents for enzymes such as glutathione peroxidase. The overexpression of transaldolase, an enzyme of the nonoxidative pentose phosphate pathway, also suggests an increased need for NADPH. This enzyme is a part of the pentose phosphate pathway that coordinates with glycolysis to produce varying combinations of NADPH, NADH, ATP, or ribose 5-phospate. The precise balance between different components of the pathway depends on the metabolic requirements of the cell. Similarly the overexpression of the three late stage glycolytic enzymes (phosphoglycerate kinase, phosphoglycerate mutase, and enolase) is consistent with an increased demand for ATP. The increased ATP requirement coincides with the overexpression of the ATP synthase ␤ subunit that was also seen. The underexpression of glutamate dehydrogenase that was observed is consistent with this pattern because it would prevent L-glutamate from entering the Krebs cycle and reserve it for use in the synthesis of glutathione, a cofactor used by a number of antioxidant enzymes. Glutathione is also used by glutathione transferase to detoxify reactive aldehydes that are formed by lipid and amino acid oxidation. This pattern of protein expression is consistent with an increased expression of antioxidant and antioxidantrelated proteins in response to oxLDL treatment that would Other changes in protein expression agree with known changes in foam cell function. For example, increased expression of enolase 1 was seen in oxLDL-treated J774 and J774-CE cells. Although enolase plays a role in the final stages of glycolysis, it has also been identified as a cell surface receptor for plasminogen, the zymogen of plasmin (42,43). Therefore, overexpression of enolase could potentially contribute to increase in plasmin activity and fibrinolysis. This change is consistent with findings that macrophages contribute to the degradation of extracellular matrix proteins (44), leading to thrombosis in advanced atherosclerosis. Although an overexpressed band representing full-length ␣-enolase was identified (molecular mass, 50 kDa; pI 7.0), six additional bands were identified as ␣-enolase at various molecular masses and isoelectric points. The bands range from 30 to 50 kDa and pI 5.5 to 7.0, and each band is overexpressed in oxLDL-treated J774 and J774-CE macrophages. Recent studies have identified ␣-enolase as a target of nitration under oxidative stress conditions (45,46), and such a modification may lead to inhibition of activity and targeted degradation. Other proteins that have been described as targets of protein nitration were identified in the current study, including ATP-synthase ␤-chain, farnesyl pyrophosphate synthase, and superoxide dismutase (Cu,Zn). In addition, protein nitration has been implicated in the reversible inhibition of the antioxidant heme oxygenase (47) and glutathione S-transferase (48,49), which were up-regulated at the message level in this study. There is evidence that proteins within foam cells of fatty streaks undergo extensive nitration (50), but as of yet, no studies have addressed the role of protein nitration as a regulatory mechanism in foam cell formation.
Finally a number of the differentially expressed proteins have unclear roles in foam cell formation. For example, Lplastin is a 64-kDa actin-bundling protein involved in integrin activation and leukocyte adhesion (51). Although two bands identifying this protein are overexpressed, they migrate to ϳ15 and 50 kDa on the 2D gel. The molecular masses and peptide sequencing indicate that together these bands represent the full-length protein, suggesting proteolysis as a regulatory mechanism. The overall effect of this proteolysis would lead to an inhibition of L-plastin activity. Perilipin and fatty acid-binding protein 5 (FABP5) bind to lipid droplets in adipocytes and likely play a similar role in lipid-laden foam cells. Perilipin, which is overexpressed in foam cells of ruptured atheromas (52), is overexpressed here in both the J774 and J774-CE foam cells. FABP5 has been detected at the message and protein level in macrophages (53) but as of yet has no link to atherosclerosis. The earliest research linking other fatty acid-binding proteins and atherosclerosis showed that a decrease in arterial FABP coincided with increased arterial cholesterol content and lesion development in rabbits (54,55). In a more recent study, a decrease in adipocyte FABP was linked to decreased lesion development in mice (apoEϪ/Ϫ) (56,57). FABP5 in particular was shown to be overexpressed in the culture media of differentiated THP-1 monocyte-macrophages following oxLDL treatment (20). In this system, FABP5 is intriguing for its overexpression in oxLDL-treated versus untreated J774-CE macrophages but even more for its increased expression in untreated J774-CE cells relative to untreated J774 cells. This increased expression of a lipid-binding protein in a non-stimulated state may provide the J774-CE macrophages with an increased tolerance for oxLDL-induced lipid loading. The novel implication that this protein may be linked to lesion development is an

TABLE III
Down-regulated mRNA in oxLDL-treated J774 or J774-CE macrophages Gene ID and NCBI RefSeq database IDs refer specifically to Entrez Gene and Entrez Nucleotide databases, respectively. An alternative NCBI mRNA ID is listed when no RefSeq entry is available. Functional and process classifications are listed according to GeneOntology designation or are labeled as "undefined" if no GeneOntology classification exists. NC indicates no change in expression where -fold change was less than 2-fold. RANTES, regulated on activation normal T cell expressed and secreted; LPS, lipopolysaccharide.    The transcriptome of the parent J774 and J774-CE cells was also investigated in a gene chip microarray experiment. The goal of these experiments was to complement the proteomic studies by creating a parallel dataset of differentially expressed transcripts that respond to the oxLDL treatment. As expected, a broad set of mRNA transcripts were up-and down-regulated, but particular interest was given to gene expression changes related to antioxidant defense.
The overexpressed genes listed in Table IV are involved directly in oxidant removal or participate in an upstream event that leads to antioxidant expression. There were no genes involved in antioxidant defense that were reduced by oxLDL treatment. Most of the genes listed in Table IV detoxify hydrogen peroxide or the products of lipid peroxidation. Catalase directly reduces H 2 O 2 , and peroxiredoxin 5 reduces H 2 O 2 , lipid hydroperoxides, and peroxynitrites. Thioredoxin reductase participates in the reactivation of oxidizable proteins, such as peroxiredoxin, through its reduction of thioredoxin. Glutamate-cysteine ligase is the rate-limiting step in the synthesis of the antioxidant glutathione, and glutathione S-transferase conjugates glutathione to lipid hydroperoxides in a detoxification process. Heme oxygenase 1 is an inducible enzyme that converts heme to biliverdin. Biliverdin and its by-product bilirubin are potent antioxidants that detoxify peroxyl radical and hydrogen peroxide (58,59).
Carbonyl reductase, which is also listed as an anti-inflammatory mediator, specifically reduces the lipid peroxidation product 4-oxonon-2-enal. However, one product of this reaction is another reactive aldehyde and a component of oxLDL, 4-hydroxynon-2-enal (60). There is currently no research that addresses a role for carbonyl reductase in the macrophage.
Other overexpressed genes provide defense against other types of oxidants. Esterase D (S-formylglutathione hydrolase) contributes to the removal of formaldehyde, a toxic metabolite indicated in endothelial injury and potential atheroma development (61,62). NAD(P)H dehydrogenase (quinone 1) detoxifies quinone, derived from phenolic metabolites of benzene, and is active in macrophages (63,64). Quinone oxidoreductase-like 1 may have similar functionality, but little is know about this gene product other than its high degree of homology to -crystallin (65). However, an NCBI nucleotide megablast search of the mRNA sequence reveals zinc-dependent alcohol dehydrogenase activity, which coincides with the upregulation of two other alcohol dehydrogenase genes, retinol dehydrogenase and the retinol-specific alcohol dehydrogenase 7 (66). These enzymes convert retinol isomers to their corresponding aldehyde retinal isomers and produce the reducing equivalent NADH in the process (67). Retinal isomers are further oxidized by aldehyde dehydrogenases to corresponding isomers retinoic acid, including 9-cis-retinoic, a ligand for the retinoid X receptor (68). Activation of retinoid X receptor in combination with peroxisome proliferator-activated receptor ␥ activation can increase cholesterol efflux and modulate foam cell formation despite triggering an increase in CD36 scavenger receptor expression (69,70). Consequently the up-regulation of retinol and alcohol dehydrogenases may contribute to antioxidant defense through NADH production and contribute to foam cell prevention by stimulating an increase in cholesterol efflux. An additional observation made in the analysis of the microarray data was the change in expression of transcripts linked to the inflammatory response. The link between inflammation and atherosclerosis has gained significant attention in recent years. Several proinflammatory mediators have been investigated for their contribution to atherosclerosis, including interferon ␥ (71), interleukin-18 (72), interleukin-6 and -8 (73), macrophage colony-stimulating factor (74), and others. Additionally anti-inflammatory mediators such as interleukin-10 are a focus of therapeutic applications for atherosclerosis (75). Much of the attention has focused on the effect of inflammation on smooth muscle cells and endothelial cells, but less information is available about the foam cell and its contribution to inflammation and immune response in general. The results of the gene chip analysis in this experiment describe an intriguing balance between up-regulated and downregulated mRNA transcripts involved in immune response.
A large number of overexpressed genes in Table III are proinflammatory, including interleukin-6, macrophage inflammatory protein 1␣ and its receptor, macrophage inflammatory protein 2, oncostatin M, and tumor necrosis factor ␣. Prostaglandin-endoperoxide synthases 1 and 2, more commonly known as COX1 and COX2, both influence a downstream proinflammatory response. COX1, the constitutively expressed enzyme, is down-regulated in both the J774 and J774-CE foam cells, whereas COX2, the corresponding inducible enzyme, is up-regulated in both cell lines. Also contributing to the overall proinflammatory state is the underexpression of transcripts for the anti-inflammatory interleukin-10 and the guanylate-binding proteins 1, 2, and 4. Taken as a group, these changes are consistent with a shift in pro-/antiinflammatory balance to a distinct proinflammatory state.
Despite this up-regulation of proinflammatory response, the response of the foam cell transcriptome to oxLDL treatment showed an intriguing down-regulation in other aspects of macrophage functionality. For example, oxLDL treatment reduced the expression of a number of interferon-induced genes, including 2Ј,5Ј-oligoadenylate synthetase proteins, interferon ␣-inducible protein, interferon ␥-inducible protein, interferon-induced proteins 35 and 44, and interferon-activated gene 203. Interferon-induced proteins often participate in the antipathogenic, antiangiogenic, and cell cycle inhibitory roles of innate immunity. Another critical component of the innate immune response to pathogens are the toll-like receptors, which were also down-regulated following oxLDL treatment. In addition, genes associated with adaptive immune response were also largely down-regulated. Indeed only one transcript, histocompatibility-2 complex class 1-like mRNA, was overexpressed in J774 and J774-CE foam cells. The other genes of this class were down-regulated and include genes encoding complement components, histocompatibility proteins, and lymphocyte antigen complex. The overall effect of this down-regulation is a decrease in the normal antipathogenic response of the macrophage following exposure to oxLDL.
In summary, this investigation used a system biology approach to examine the transition of macrophages into foam cells at the protein and message levels and included a novel focus on macrophages that have been chronically exposed to oxLDL. These chronically exposed cells exhibited an increased capacity for survival under increasing concentrations of oxLDL but showed no difference in lipid loading when compared with naïve macrophages. The functional results are consistent with the changes seen in a group of antioxidant proteins that were up-regulated at either the message or protein level. A subset of these proteins were overexpressed significantly more in macrophages subjected to chronic ox-LDL exposure. The increased resistance to oxLDL-induced cytotoxicity in the J774-CE macrophages is attributed to this increase in antioxidant defense proteins. This elevated protective response is not dependent on the presence of oxLDL because an increase in survival was maintained several weeks after chronic oxLDL treatment was terminated. Our results also detected a group of proinflammatory transcripts that were up-regulated following oxLDL exposure in addition to transcripts of an anti-inflammatory counter-response. However, the chronically oxLDL-exposed macrophages presented an attenuated steady-state and oxLDL-stimulated proinflammatory response when compared with control macrophages. These results support the hypothesis that chronic exposure to oxLDL produces important changes in foam cell function that are not seen in single oxLDL exposure experiments. A large number of changes in protein and mRNA expression are produced by these treatments. As a group, these changes are consistent with both known (for example an enhanced inflammatory response) and novel (for example an increased resistance to oxLDL-mediated cytotoxicity) differences between foam cells and macrophages and between single and multiple oxLDL treatments. Further experimentation is ongoing to begin the process of examining the exact role for these altered proteins in the foam cell phenotype.
land Clinic Foundation, Case Western Reserve University, and National Institutes of Health Grant P30 CA43703. 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.