OxLDL and mmLDL used in the experiments were prepared from the same native LDL each time as in Miller et al. 18. Oxidized LDL were analyzed by measuring thiobarbituric reactive substances (TBARS, 30–40 μmol/g protein) and EO6-reactive phospholipid oxidation products. mmLDL contained early lipid peroxidation products, but it did not contain any measurable TBARS or EO6-reactive substances above that of native LDL 1319. The mmLDL modification appeared to be very reproducible, and a successful modification was documented in a biological assay in which mmLDL induced spreading of J774 macrophages in cell culture 1319. All LDLs were tested for endotoxin level with a LAL kit (BioWittaker, Walkersville, MD). The level of endotoxin contamination was below 2.5 pg/ml. All other chemicals, if not further specified, were from Sigma (St. Louis, MO).

Experiments were performed with cells isolated from healthy human subjects who were volunteers recruited by announcements. There were no restrictions regarding sex, age, and ethnicity. All participants signed a consent form approved by the Human Investigation Committees at the University of Virginia, Charlottesville, VA and at Cleveland Clinic, Cleveland, OH. Whole blood (100 ml) was drawn from the antecubital vein of donors. Dextran 500 was added to the blood for sedimentation of red blood cells. After incubation at room temperature, the buffy coat was removed and placed on Histopaque 1.077 (Sigma Diagnostics, Inc., St. Louis, MO). As an alternative to whole blood, we used buffy coats delivered by Virginia Blood Services (Richmond, VA). These buffy coats were used for Affymetrix GeneChip^® U133A arrays. Following centrifugation, the mononuclear layer was removed and washed with PBS containing 0.02% EDTA.

Monocytes were isolated using adherence to plastic as in Bey et al. 20. Briefly, the mononuclear cell pellet was resuspended in 1X H-lyse Buffer (R&D Systems Inc., Minneapolis, MI), and then washed with 1X Wash Buffer. Cells were resuspended in RPMI 1640 +10% FCS (Invitrogen Corporation, Carlsbad, CA) and plated on 60 mm or 100 mm dishes. After one hour incubation at 37°C, the dishes were washed twice with PBS and all non-adherent cells were removed.

For Affymetrix GeneChip^® experiment, isolated adherent monocytes were resuspended in Macrophage Serum-Free Medium (MSFM, Gibco, Invitrogen Corporation, Carlsbad, CA) in the presence of 1% media supplement Nutridoma-HU (Roche Molecular Biochemicals, Indianapolis, IN). The concentration of insulin (500 nM) and glucose (17.5 mM) in this medium allowed us to mimic conditions of impaired glucose tolerance or type 2 diabetes (hyperinsulinemia, hyperglycemia). This insulin concentration is greater than in diabetic patients 21. However, this is a regular insulin concentration for the prolonged cultivation of the cells under serum-free conditions.

For additional experiments, we used an insulin- and glucose-deficient medium (glucose-free RPMI 1640 (Invitrogen Corporation, Carlsbad, CA)+10% dialyzed FCS (HyClone, Logan, UT)) supplemented with varying concentrations of both glucose and insulin. Cells incubated in MSFM and in insulin- and glucose-deficient medium in the presence of 500 nM insulin and 17.5 mM glucose showed similar pattern of gene and protein expression as assessed by RT-PCR and flow cytometry (data not shown). Differentiation of monocytes was induced by macrophage colony-stimulating factor (M-CSF) or platelet factor 4 (PF₄) (each at concentration 100 ng/ml). We used these stimulators since they have been identified in human atherosclerotic lesions 2223 and play a pathophysiological role in atherogenesis.

After 6–8 days of incubation with M-CSF or PF₄ in a CO₂-incubator (95% O₂/5% CO₂), the cell medium was changed for a similar one without stimulator, but with native LDL, OxLDL or mmLDL (100 μg of protein/ml) and/or different concentrations of insulin and glucose. Directly before use, all LDLs were filtered through a 0.45 μ Acrodisc syringe filter to remove aggregates. As a negative control, cultured monocytes/macrophages were exposed to a medium without additives under the same experimental conditions. Cells were further incubated in a CO₂ incubator for 48 h. Some dishes with differentiated macrophages were also incubated either with monocyte chemoattractant protein-1 (MCP-1 or CCL2) or growth-regulated oncogene alpha (GROα or CXCL1) for 5 h to investigate an acute effect of CC or CXC chemokines. CCL2 and CXCL1/CXCL8 are among the chemokines that have been implicated most strongly in atherogenesis 24. Data on CCL2 and CXCL1 effects were used only as input data (in combination with other data generated from MDM) with the intent to increase reliability and decrease standard deviations of Hierarchical Clustering and Self Organizing Map analysis. Cell concentration and viability after incubation were assessed using a Guava Personal Cytometer PCA (Guava Technologies, Inc., Hayward, CA).

Total RNA was isolated using RnEasy Mini Kit (Qiagen Inc., Valencia, CA), treated by RNase-free DNase (Qiagen Inc.) to avoid possible DNA contamination and was used for cRNA synthesis. cRNA was used for hybridization to a total of eighteen Affymetrix GeneChip^® U133A arrays (nine conditions in duplicates, the conditions are defined in the abscissa of Figure 4 and in Additional figure 1 in Additional file 1). Additional plates were used for confirmation of Affymetrix array data by RT-PCR and for analysis of proteins by Western blot. Aliquots of supernatants were used for detection of 5-HETE and cytokine production by macrophages.

Real time reverse transcriptase-polymerase chain reaction was performed on RNA isolated from MDM and foam cells cultivated under varying concentrations of insulin and glucose. The primers used to analyze mRNA for insulin receptor were: sense-TGCTGGTGTTCATCAGAACAG and antisense-ACAAACACCACCAGGTACATG. Following the reverse transcription step (30 min at 50°C), PCR was performed at 94°C (15 seconds), 60°C (30 seconds), 72°C (30 seconds) with data collection at 79.5°C (15 seconds) for 40 cycles. SYBR Green I (Molecular Probes, Eugene OR) was used as the fluorescent reporter. The specificity of the product was confirmed by analysis of characteristic melting curve and by electrophoresis. Negative controls included RNAse I treatment and omission of the reverse transcription step. Upon treatment with DNAse I, there was no significant change in insulin receptor mRNA expression, indicating an absence of genomic DNA in the samples. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as an internal standard, and all data were normalized to GAPDH expression. Real Time RT-PCR was performed using an Applied Biosystems ABI Prism 7700 Sequence Detection System.

Following incubations, MDM cultures were rinsed three times with PBS without Ca²⁺ and Mg²⁺, scraped into 1 ml of double distilled water and kept frozen at -80°C until chemical analysis was performed. Total and esterified cholesterol was determined enzymatically with the fluorescence method described by Gamble 25. Macrophage protein content was determined by the Lowry method using BSA as a standard. Cytokine analysis was performed using Cytometric Bead Array kit (BD Pharmingen, San Diego, CA) according to standard manufacturer protocols. High performance liquid chromatography (HPLC) was performed under at least two different conditions for clear identification of 5(s) HETE peak (please see Additional file 1 for details).

Data were compared with either 1-way ANOVA followed by Bonferroni Correction Post Hoc test or Student T test to evaluate 2-tailed levels of significance. Determining Pearson correlation coefficients compared continuous variables. Local Pooled Error (LPE) method 26 was used to determine statistical differences between the changes in the gene expression on the chips.

Two algorithms, Hierarchical Clustering analysis 27 and Self Organizing Map (SOM) 28 were used to identify the clusters of genes that are concomitantly regulated by various LDLs. The GENECLUSTER software 28 produced and displayed SOMs of gene expression data. Genes in biologically interesting clusters identified by these approaches were searched across bioinformatic database GenMAPP.

We measured cholesterol content as a direct indicator of foam cell formation (Figure 1). As expected 129, the largest accumulation of cholesterol was detected in OxLDL-treated cells.

Next, we measured inflammatory cytokines in cell supernatants of macrophages incubated with various types of LDL. To identify the direct insulin effect on cytokine production by macrophages, MDM were incubated in insulin-deficient media or with 10 nM insulin (Figure 2, first 2 columns). A moderate dose of insulin (10 nM) did not affect the production of IL-lβ, whereas high insulin (500 nM) in combination with high glucose significantly increased IL-lβ, and the response was further exacerbated by OxLDL (Figure 2A). Similar data was obtained for IL-12 (not shown). IL-6 30 was also increased under high glucose high insulin conditions, but the effect was further exacerbated specifically by mmLDL (Figure 2B).

Next, we examined the effect of LDL treatment on gene expression in macrophages incubated under HIHG conditions using Affymetrix GeneChip^® arrays. Affymetrix arrays generated an enormous amount of data that were analyzed by comparing vehicle-treated cells with the cells treated with modified LDL. Among other genes, we found that many genes directly related to insulin signaling were significantly modulated by modified LDL. They include the insulin receptor, IRS2 (expressed in monocytes at much higher level than IRS-1 9 and our unpublished data), the catalytic subunit δ of phosphatidylinositol 3' kinase (PIK3CD, subunit δ is predominantly expressed in leukocytes 31), Insulin-Induced gene 1 and glucose-6-phosphate dehydrogenase gene. Both OxLDL and mmLDL significantly upregulated the expression of glucose-6-phosphate dehydrogenase (2.3-and 2.0-fold, respectively, p < 0.000001 for both values). OxLDL downregulated the expression of the Insulin-Induced gene 1 and the catalytic subunit δ of phosphatidylinositol 3' kinase 5.4-and 1.9-fold, respectively (p < 0.000001 for both values), but upregulated the expression of the genes encoding the insulin receptor and IRS2 3.9- and 2.4-fold (p < 0.004 and p < 0.000001), respectively. mmLDL downregulated the expression of the Insulin-Induced gene 1 1.4-fold (p < 0.05), and upregulated the expression of the insulin receptor gene 1.7-fold (p < 0.05), but had no effect on the IRS2 and the phosphatidylinositol 3' kinase genes. The changes in expression of these genes by OxLDL and mmLDL are shown in Table 1. The last four genes were significantly correlated across 9 experimental conditions: while insulin receptor and IRS2 genes were positively correlated, they were negatively related to the genes encoding PIK3CD and Insulin-Induced gene 1 (Table 2).

We performed real-time RT-PCR analysis for the insulin receptor (INS-R) (Figure 3) and PIK3CD (data not shown) to confirm the data of the Affymetrix GeneChip^® arrays and to obtain additional data under different experimental conditions, i.e. lower concentrations of insulin and glucose. The results obtained by the Affymetrix GeneChip^® arrays (Table 1) were confirmed for OxLDL (Figure 3) and specificity of the change for HIHG conditions was demonstrated. Since insulin at high concentrations may act through the receptors for IGF-1 32, we evaluated whether IGF-1 has similar effects on the expression of the insulin receptor (INS-R). Indeed IGF-1 in combination with high glucose induced INS-R expression (Figure 3).

To identify the clusters of genes that are concomitantly regulated by various LDLs, Hierarchical Clustering analysis 27 was applied. An example is shown in Figure 4. The genes encoding Glucose-6-phosphate dehydrogenase (G6PD) and Leukotriene A₄ hydrolase (LTA4H) belong to the same cluster (two upper rows, Figure 4) as well as the genes encoding 5-lipoxygenase activating protein (ALOX5AP) and Phosphoinositide-3-kinase (PIK3CD; row 5 and 6). The last observation was confirmed using another statistical method, Self Organizing Maps (please see Additional figure 2 in Additional file 1). Arachidonate is a precursor for the biosynthesis of eicosanoids generated by 5-lipoxygenase/ALOX5AP and by other enzymes. Its release from membrane phospholipids is mediated by the activity of phosholipase A₂ (PLA₂). We found significant modulation of one of the PLA₂ isozymes, namely group IVC cytosolic calcium-independent PLA_2-γ (PLA2G4C). The data on the variations in the genes related to the eicosanoid cascade are presented in Table 3.

Since we found alterations in the expression of a number of genes belonging to insulin cascade, such as PIK3CD, we further explored if the process of foam cell formation under HIHG conditions affects insulin-responsive pathways on protein level. Among several parameters tested, we analyzed the expression of the insulin-regulated aminopeptidase (IRAP) and Akt phosphorylation by Western blot (Figure 5). IRAP represents an effector of insulin action. IRAP expression was stimulated by mmLDL, but slightly inhibited by OxLDL under HIHG conditions. Akt phosphorylation plays a key role in receptor-mediated signaling, including insulin signaling. Both OxLDL and mmLDL significantly increased Akt phosphorylation in macrophages cultured under hyperinsulinemic hyperglycemic conditions (Figure 5A lower and 5B). This effect was probably mediated not only by the insulin receptor, but also by the IGF1-receptor, because the insulin-like growth factor 1 (IGF1) caused a similar response (Figure 5B). Total levels of Akt protein were similar under all conditions examined (not shown).

Finally, since we found that some genes belonging to the 5-lipoxygenase pathway were regulated concomitantly with the genes of the insulin signaling cascade (see Figure 4 and Additional figure 2 in Additional file 1), we also assessed the levels of the 5-lipoxygenase product 5(s) HETE in the cell supernatants by HPLC. The chromatograms were crowded with many large peaks and chromatography under at least two different conditions had to be performed for clear peak identification (Additional figure 3 in Additional file 1). OxLDL enhanced production of 5(s) HETE in HIHG conditions, as measured by chiral HPLC (Figure 6). OxLDL incubated in the same conditions but without cells did not contain any significant amount of 5(s) HETE (not shown).