Divergent Pathways of Gene Expression Are Activated by the RAGE Ligands S100b and AGE-BSA

Bovine albumin (Fraction V, sterile filtered, endotoxin tested), d(−) ribose, sodium phosphate monobasic, sodium phosphate dibasic, sodium hydroxide, recombinant human epidermal growth factor, hydrocortisone, gelatin, recombinant human TNF-α, and lipopolysaccharide from E. coli 0111:B4 were obtained from Sigma (St. Louis, MO). PBS (10×) was purchased from Roche Diagnostics Corporation (Mannheim, Germany). Endotoxin-free distilled water, sterile PBS without calcium and magnesium, MCDB 131 media, heat-inactivated FBS, l-glutamine, antibiotic/antimicotic, 0.05% trypsin/0.53 mmol/l EDTA, penicillin and streptomycin, TRIzol reagent, and Superscript II Choice System were purchased from Gibco BRL/Life Technologies (Gaithersburg, MD). Sterilization filters (Express filter; 0.22 μm; 250 ml) were obtained from Millipore (Bedford, MA). Bicinchoninic acid (BCA) protein assay kit was purchased from Pierce (Rockford, IL). T-175 Falcon flasks were purchased from Fisher Scientific (Pittsburgh, PA). RNeasy kits were obtained from Qiagen (Valencia, CA). A BioArray High Yield DNA Transcript kit was purchased from ENZO Diagnostics (Farmingdale, NY).

Ribose-derived model AGEs (Rib BSA) were prepared with 500 mmol/l ribose (6-week incubation) and characterized as described previously (21). Endotoxin levels were measured by Associates of Cape Cod (Falmouth, MA) using the gel-clot method and were found to be <0.2 ng/mg AGE-BSA. Control BSA (Ctrl BSA) used in these experiments was the same endotoxin-tested BSA used as starting material for AGE-BSA preparations; however, the BSA was kept frozen until needed. As needed, the BSA was thawed and diluted using dialysis buffer to the same concentration as the stock Rib BSA (48.9 mg/ml). After dialysis, the final protein concentration was determined using the BCA assay. As reported previously, the half-maximal inhibition concentration (IC₅₀) for Rib BSA in a cell-free human soluble RAGE (hsRAGE) binding assay was 0.11 μmol/l (21). In contrast, Ctrl BSA showed no detectable binding affinity for hsRAGE (21).

The Hans Kocher lab (Novartis Pharmaceuticals, Basel, Switzerland) generously provided recombinant human S100b. Endotoxin levels were determined to be 2.5 ng/mg protein. Using the cell-free hsRAGE assay reported previously, the IC₅₀ for S100b was 0.24 μmol/l (21).

Human microvascular ECs (HMEC-4) were obtained from Dr. Edwin Ades (Centers for Disease Control and Prevention, Atlanta, GA). HMEC-4 cells were derived from human foreskin and immortalized by constitutive expression of the T-antigen of SV40 virus (23). Monolayers were propagated in growth medium (MCDB131, supplemented with 10% heat-inactivated FBS, 2 mmol/l l-glutamine, 10 ng/ml epidermal growth factor, 1 μg/ml hydrocortisone [HC], and 1% antibiotic-antimicotic in 5% CO₂ at 37°C). The cells were grown to confluence in T-175 flasks (5 × 10⁶ cells per flask in 20 ml medium). Cells were passaged once a week following mild trypsinization with 0.05% Trypsin-EDTA at 37°C for 5 min. HMEC-4 cells were used at passage 22. When ∼80% confluent, cells were treated for 18 h at 37°C with 0.5 mg/ml Ctrl BSA, 0.5 mg/ml Rib BSA, or 0.2 mg/ml S100b diluted in growth medium, except the FBS, which was used to 5% (three flasks per treatment group).

C57BL/6J mice were obtained from The Jackson Laboratory at 4–6 weeks of age and were allowed to acclimate for at least 1 week before use. The mice were 6–10 weeks of age (∼25 g) at the time of the experiment. Model AGEs were injected intravenously in a volume of 10 ml/kg or intraperitoneally at a volume of 50 ml/kg. Mice were administered with a single intravenous injection of 400 mg/kg Ctrl BSA or Rib BSA (∼10 mg/mouse) or of 4 mg/kg (∼0.1 mg/mouse) of LPS (three mice/group). After 24 h, mice were killed by CO₂ asphyxiation and the liver was removed for RNA isolation. In a separate experiment, mice were injected with 400 mg/kg i.p. of Ctrl BSA or Rib BSA daily for 4 days (five mice/group). On day 5, mice were killed and the liver was removed for RNA isolation. The use and care of laboratory animals at the Novartis Institutes for Biomedical Research through institutional policy complies with or exceeds all requirements mandated by the Animal Welfare Act and state and local laws governing the use of animals in research.

Total RNA was isolated from cultured cells and murine liver tissue using TRIzol reagent according to the manufacturer’s instructions. The total RNA was further purified using the clean-up protocol in the Qiagen RNeasy kit according to the manufacturer’s instructions. Final RNA concentrations were determined spectrophotometrically at 260 nm. Quality of the total RNA (300 ng/lane) was determined by subjecting the samples to 1% agarose gel electrophoresis. RNA integrity was confirmed by ribosomal 18S and 28S RNA ethidium bromide staining.

Purified total RNA was used to synthesize double-stranded cDNA using Superscript Choice System. The cDNA was then transcribed in vitro using Enzo BioArray high-yield transcript labeling kit to form biotin-labeled cRNA. The labeled cRNA was fragmented and hybridized to the microarray for 16 h at 45°C. The array was washed and stained using the GeneChip Fluidics station. For HMEC-4 cells, cRNA was hybridized to Affymetrix hg U133A chips. For the murine tissue–derived cRNA, the Affymetrix MG U74Av2 chips were utilized. The array was scanned and the data were captured using the Affymetrix GeneChip Laboratory Information Management System (LIMS). The Affymetrix GeneChip MAS4.0 software was used to generate the average difference calls (AvgDiff).

For each experiment, pairwise comparison of replicates showed that there were no outliers and that the twofold difference could be considered significant. Therefore, data were filtered using the following criteria: fold change twofold or greater with (Student’s t test P < 0.05) and mean AvgDiff values ≤200. Note: some of the probes recognize multiple genes within a family; therefore, the gene sequence recognized by the probe is either identical to the sequence provided under the listed gene accession number or similar to that gene sequence.

Hierarchical clustering to generate an experimental tree was performed using GeneSpring software and the default settings (measure similarity by standard correlation with a separation ratio of 0.5 and a minimum distance of 0.001). Experiment trees were generated using two different lists of genes. The first list identified genes that differed in expression between mice treated with a single bolus of either Rib BSA (10 mg/mouse) or Ctrl BSA (10 mg/mouse). Selection criteria included a mean average difference of at least 200 (a twofold difference between the two treatment groups; P < 0.05 Welsh T-test, unequal variance, no additional Bonferroni corrections). The second list identified genes that differed in expression between mice treated with LPS versus Ctrl BSA using the same selection criteria described above. Of note, another group of mice was injected with a lower dose of Rib BSA (0.3 mg/mouse) and no significant changes in gene expression were observed compared with Ctrl BSA treatment (data not shown).

Statistical analysis was performed in Excel (Microsoft, Redmond, WA). Triplicate experiments were analyzed unless otherwise noted. Experiment tree graphs were created in GeneSpring (Silicon Genetics, Redwood City, CA).

To identify genes regulated in the endothelium after exposure to AGE-BSAs with high RAGE binding affinity, HMEC-4 cells were treated for 18 h at 37°C with 0.5 mg/ml Rib BSA or Ctrl BSA. Total RNA was isolated from the treated cells and used for microarray analysis. As shown in Tables 1 and 2, analysis identified only five genes that were significantly upregulated and only four genes were significantly downregulated. The responses were very modest, with no gene increasing by more than threefold. Overall, the genes found to be regulated did not comprise the expected proinflammatory mRNA phenotype. In fact, the genes found to be upregulated showed no obvious expression pattern. However, several genes that have been suggested to be involved in the control cell proliferation were downregulated after treatment with Rib BSA wk6, including inhibitors of DNA binding-1, -2, and -3 (Table 2).

When HMEC-4 cells were treated for 18 h at 37°C with S100b, 44 genes were significantly upregulated and 10 were significantly downregulated as assessed by microarray analysis (Tables 3 and 4). Many of the genes upregulated were indicative of an activated endothelium, including genes encoding a number of chemokines and adhesion molecules and genes encoding proteins involved in antigen presentation, including expression of a variety of major histocompatibility complex (MHC) class I and II alleles and subunits of the proteasome (Table 3).

For comparison, HMEC-4 cells were also treated with two known inflammatory triggers—TNF-α (20 ng/ml) or LPS (200 ng/ml) for 4 h at 37°C. As expected, numerous proinflammatory genes were differentially regulated (84 genes after TNF-α treatment; 165 genes after LPS treatment). Table 5 shows the top 35 genes that were upregulated in both TNF-α and LPS treatments. HMEC-4 cells treated with TNF-α or LPS resulted in a strong inflammatory cellular response, which included an increase in the expression of several cytokines/chemokines, adhesion molecules, transcription factors/regulators, and proteins involved in apoptosis to name a few.

While the effects observed in cultured cells are often indicative of what happens in vivo, the cultured cell model system is limited because it is unable to account for interactions between different cell types that occur within an animal. Therefore, the effects of acute administration of exogenous AGEs to healthy mice were evaluated. Healthy C57BL/6 mice were intravenously injected with either a single bolus of Rib BSA or Ctrl BSA (10 mg/mouse). After 24 h, the livers were removed and total RNA was isolated. Another group of healthy mice was intraperitoneally injected for 4 days with either Rib BSA or Ctrl BSA (10 mg · mouse⁻¹ · day⁻¹). On the 5th day, the livers were removed and total RNA was isolated. The preparations used in this study were essentially endotoxin free (≤0.2 ng/mg AGE-BSA) according to the gel-clot method. However, to be sure the genes differentially regulated in animals injected with Rib BSA were not due to trace amounts of endotoxin, the expression pattern in AGE-treated animals was compared with animals injected with endotoxin. Using GeneSpring software, cluster analysis illustrated the samples from LPS-treated mice did not cluster with the model AGE–treated mice (data not shown). These data suggest that the biological responses induced by model AGEs are not similar to the responses elicited by LPS; therefore, the biological activity of the model AGEs used in this study is not likely a result of endotoxin contamination.

Table 6 lists selected genes upregulated in the liver from mice treated with a single bolus of model AGE compared with mice treated with a single bolus of Ctrl BSA. The list further demonstrates that although some genes upregulated by Rib BSA are also upregulated by LPS, the overall expression patterns differed. For example, of the 43 genes that changed at least 10-fold after LPS treatment, only 4 of those genes were also upregulated by Rib BSA (serum amyloid A1, serum amyloid A3, monocyte chemotactic protein-1, and MARCO). In contrast, both M and P lysozyme were upregulated by Rib BSA, but not by LPS. No genes were found significantly downregulated in the liver.

In mice treated with model AGEs for 4 days (10 mg/mouse i.p.), 26 genes were upregulated in the liver at least twofold, and no genes were significantly downregulated. Genes with at least a 2.5-fold upregulation are listed in Table 7. LPS was not included as a control in this experiment. A number of genes that had been identified after a single injection of model AGEs were also elevated after four injections of model AGEs, including M and P lysozyme and the macrophage scavenger receptors MARCO and CD5L.

To gain a better understanding of the role RAGE ligands play in cellular activation, we evaluated the effects of model AGEs or recombinant S100b on EC gene expression using microarray technology. AGE treatment of HMEC-4 cells did not induce an inflammatory mRNA phenotype as predicted by the literature (see below). However, treatment with S100b induced an mRNA phenotype of activated endothelium (Tables 3 and 4, Fig. 1). In addition, treatment of ECs with known inflammatory triggers TNF-α or LPS induced a strong inflammatory immune response defined by an increase in the expression of genes, including cytokines, cytokine receptors, chemokines, MMP-1, and various cell adhesion molecules (Table 5, Fig. 1). These data suggest that RAGE binding AGEs may not be general drivers of inflammation. In contrast, this work confirmed S100b as mediator of inflammation either by activating RAGE or through other pathways yet to be elucidated. Future work will be required to determine whether the various reported RAGE ligands activate different cellular responses.

Numerous studies have been published showing cells exposed to AGEs resulted in significant alterations in the expression of many genes, including IL-1β (24), TNF-α (12), IL-6 (13), platelet-derived growth factor (25), insulin-like growth factor-1 (IGF-1) (26), thrombomodulin, VCAM-1 (11), and TF (14,27,28). Of these genes, all were present on the chip; however, the majority were considered below detection level. The only exception was platelet-derived growth factor, which displayed a low level of expression that did not differ between Ctrl BSA and Rib BSA treatments. Treating cells with AGE-BSAs for multiple time periods may provide a more complete picture. For example, longer/chronic exposure of cells might result in changes in the above-mentioned genes. However, treatment of HMEC-4 cells up to 72 h failed to induce VCAM-1 as assessed by enzyme-linked immunosorbent assay (22). In addition, total RNA extracted from HMEC-4 cells treated for 4 h with endotoxin-free AGE-BSA also did not show an increase in proinflammatory gene expression (data not shown).

Gene expression profiling of S100b-treated endothelial cells confirmed S100b as the mediator of inflammation as previously reported (7) and, therefore, also confirmed the validity of our in vitro model system. In our studies, S100b triggered an immune response defined mainly by expression of chemokines, adhesion molecules, and genes involved in antigen presentation, which included MHC class I and II alleles and several proteasome subunits (summarized in Fig. 1). In addition, increased expression of those gene classes is dependent on activation of the NF-κB pathway (29), confirming previous reports that NF-κB is a central transcription factor in the cellular response to S100b (7). Taken together, the changes in gene expression observed after S100b treatment described an activated endothelium.

Although S100b treatment induced some similar gene expression changes compared with TNF-α or LPS, the overall pattern of gene expression varies greatly (compare Tables 3 and 5). Thus, the effects on gene expression observed when HMEC-4 cells were treated with S100b are unlikely due to contaminating LPS. In addition, the gene expression changes observed after treatment of HMEC-4 cells with TNF-α or LPS further validated our in vitro model system, showing that the HMEC-4 cells are responsive to proinflammatory triggers.

Exposure of healthy animals to high doses of model AGEs triggered a modest immune response in liver tissue defined mainly as a macrophage-based clearance/detoxification response. Overall, mice injected with model AGEs failed to display gene expression changes indicative of a strong induction of the NF-κB pathway. At least six of the nine genes that increased in expression after a single intravenous administration of model AGEs are associated with macrophage activation and differentiation (monocyte chemotactic protein-1, M and P lysozymes, MARCO, CD5L, and thymosin) (30) (Table 6, Fig. 1). These results were confirmed by a second experiment measuring gene expression changes in livers from mice treated for 4 days with high doses of model AGEs. These mice also showed an increase in a large number of the genes associated with macrophage activation or differentiation, including lysozyme P and M, MARCO, CD5L, TYRO, CD68, properdin factor, and complement C1qB (30). This suggests that the majority of the cellular responses that followed exposure to exogenous AGEs were derived from the macrophage-derived Kupffer cells. The upregulation of lysozyme is interesting, because lysozyme has been reported to bind AGEs and improve renal excretion of AGEs (31). In addition, a 2.8-fold increase in VCAM-1 was observed in liver tissue from mice treated for 4 days with model AGEs (total amount of intraperitoneally injected AGE: 40 mg/mouse). In these same mice, a two- to fourfold increase in soluble VCAM-1 was measured by enzyme-linked immunosorbent assay (data not shown). Stern and Schmidt (11) reported that healthy mice injected with a single bolus 0.50 mg/mouse of model AGE showed a two- to threefold increase in VCAM-1 expression in the lung according to immunohistological staining. We did not see an increase in liver VCAM-1 mRNA after a single injection of 10 mg/mouse (a 20 times larger dose).

Our results do not show that AGEs trigger a strong inflammatory response. Previously, animals injected with a single dose of exogenous AGEs have been reported to increase the expression of a variety of inflammatory mediators, including liver IL-6 and heme oxygenase (10,13). However, our experiments showed no evidence of an increase in IL-6 expression. Furthermore, if IL-6 expression was induced in our study, then a significant increase in the expression of acute-phase proteins such as C-reactive protein and fibrinogen would have been observed. Although we feel our data accurately reflect the effects of AGEs in vivo, there are several differences between this study and previously published studies, including 1) strain of mouse (SJL vs. C57BL/6), 2) dose of model AGE (0.5 vs. 10 or 40 mg/mouse), and 3) likely the composition of the AGE preparations (4) time point (6 vs. 24 h or 5 days). In addition, a longer-term study using AGE-modified mouse serum albumin might result in induction of inflammatory mediators. Although many of the known AGE structures that have been shown to form under in vitro conditions have also been found in vivo (32–35), model AGEs may not accurately reflect the chemical composition of AGEs formed in vivo.

The present study is one of the first to look at AGE-induced effects on gene expression using this oligonucleotide array technology. Ideally all of the genes observed to change should be confirmed by additional techniques, such as Northern blot or RT-PCR. In the HMEC-4 cell system, we have confirmed, using a cell-based enzyme-linked immunosorbent assay, that the VCAM-1 gene expression changes elicited by TNF-α, LPS, and S100b described herein reflect a change in protein levels as well (22). In the animal studies, the increase in mRNA expression of P lysozyme in mice injected with model AGEs was confirmed by Northern blot analysis (data not shown).

Exposure of healthy animals to high doses of model AGEs triggered a modest immune response in the liver tissue defined mainly as a macrophage-based clearance/detoxification response. The significance of these changes is unclear. Future work will be required to decipher whether these effects reflect specific AGE-mediated cellular responses or these effects may actually be an artifact resulting from injection of high concentrations of modified proteins. For example, although Ctrl BSA is used for comparison, the ribose-modified protein may be denatured and elicit nonspecific effects. Unlike previous reports, the observed immune response in AGE-treated mice did not entail expression of high levels of inflammatory mediators, although very modest changes in the inflammatory mediators VCAM-1 and monocyte chemotactic protein-1 were noted in mice exposed to model AGEs. Overall, our data do not convincingly demonstrate that model AGEs are signaling molecules, despite previous work showing that the AGEs bind to RAGE with high affinity (21).

Although our work suggests that AGEs do not trigger a significant inflammatory immune response, accumulation of AGEs on macromolecules is known to adversely affect both the functional properties and clearance of these molecules. The resulting biomechanical changes to these molecules have been shown to contribute to the pathology of several disease states, including atherosclerosis and diabetic complications (36–38). Thus, the biomechanical effects of AGEs may prove to be more detrimental in vivo than the proposed cell-surface receptor-mediated pathways.

We thank Marlene Dressman for her contributions in analyzing the gene expression changes observed in mice following a single administration of model AGEs. We thank Shari Caplan for generously providing probes for Northern blot analysis. We would also like to thank Arco Jeng’s lab for their expertise in Northern blot analysis. Helpful discussions from John Rediske are kindly acknowledged.