Gene expression databases and data mining.

The DNA microarray technology has arguably caught the attention of the worldwide life science community and is now systematically supporting major discoveries in many fields of study. The majority of the initial technical challenges of conducting experiments are being resolved, only to be replaced with new informatics hurdles, including statistical analysis, data visualization, interpretation, and storage. Two systems of databases, one containing expression data and one containing annotation data are quickly becoming essential knowledge repositories of the research community. This present paper surveys several databases, which are considered "pillars" of research and important nodes in the network. This paper focuses on a generalized workflow scheme typical for microarray experiments using two examples related to cancer research. The workflow is used to reference appropriate databases and tools for each step in the process of array experimentation. Additionally, benefits and drawbacks of current array databases are addressed, and suggestions are made for their improvement.

Classification of human astrocytic gliomas on the basis of gene expression: a correlated group of genes with angiogenic activity emerges as a strong predictor of subtypes.

The development of targeted treatment strategies adapted to individual patients requires identification of the different tumor classes according to their biology and prognosis. We focus here on the molecular aspects underlying these differences, in terms of sets of genes that control pathogenesis of the different subtypes of astrocytic glioma. By performing cDNA-array analysis of 53 patient biopsies, comprising low-grade astrocytoma, secondary glioblastoma (respective recurrent high-grade tumors), and newly diagnosed primary glioblastoma, we demonstrate that human gliomas can be differentiated according to their gene expression. We found that low-grade astrocytoma have the most specific and similar expression profiles, whereas primary glioblastoma exhibit much larger variation between tumors. Secondary glioblastoma display features of both other groups. We identified several sets of genes with relatively highly correlated expression within groups that: (a). can be associated with specific biological functions; and (b). effectively differentiate tumor class. One prominent gene cluster discriminating primary versus nonprimary glioblastoma comprises mostly genes involved in angiogenesis, including VEGF fms-related tyrosine kinase 1 but also IGFBP2, that has not yet been directly linked to angiogenesis. In situ hybridization demonstrating coexpression of IGFBP2 and VEGF in pseudopalisading cells surrounding tumor necrosis provided further evidence for a possible involvement of IGFBP2 in angiogenesis. The separating groups of genes were found by the unsupervised coupled two-way clustering method, and their classification power was validated by a supervised construction of a nearly perfect glioma classifier.

Gene expression profiling in breast cancer: understanding the molecular basis of histologic grade to improve prognosis.

BACKGROUND: Histologic grade in breast cancer provides clinically important prognostic information. However, 30%-60% of tumors are classified as histologic grade 2. This grade is associated with an intermediate risk of recurrence and is thus not informative for clinical decision making. We examined whether histologic grade was associated with gene expression profiles of breast cancers and whether such profiles could be used to improve histologic grading. METHODS: We analyzed microarray data from 189 invasive breast carcinomas and from three published gene expression datasets from breast carcinomas. We identified differentially expressed genes in a training set of 64 estrogen receptor (ER)-positive tumor samples by comparing expression profiles between histologic grade 3 tumors and histologic grade 1 tumors and used the expression of these genes to define the gene expression grade index. Data from 597 independent tumors were used to evaluate the association between relapse-free survival and the gene expression grade index in a Kaplan-Meier analysis. All statistical tests were two-sided. RESULTS: We identified 97 genes in our training set that were associated with histologic grade; most of these genes were involved in cell cycle regulation and proliferation. In validation datasets, the gene expression grade index was strongly associated with histologic grade 1 and 3 status; however, among histologic grade 2 tumors, the index spanned the values for histologic grade 1-3 tumors. Among patients with histologic grade 2 tumors, a high gene expression grade index was associated with a higher risk of recurrence than a low gene expression grade index (hazard ratio = 3.61, 95% confidence interval = 2.25 to 5.78; P < .001, log-rank test). CONCLUSIONS: Gene expression grade index appeared to reclassify patients with histologic grade 2 tumors into two groups with high versus low risks of recurrence. This approach may improve the accuracy of tumor grading and thus its prognostic value.

Gene expression profiling of isolated tumour cells from anaplastic large cell lymphomas: insights into its cellular origin, pathogenesis and relation to Hodgkin lymphoma.

Anaplastic large cell lymphoma (ALCL) is a main type of T-cell lymphomas and comprises three distinct entities: systemic anaplastic lymphoma kinase (ALK) positive, systemic ALK(-) and cutaneous ALK(-) ALCL (cALCL). Little is known about their pathogenesis and their cellular origin, and morphological and immunophenotypical overlap exists between ALK(-) ALCL and classical Hodgkin lymphoma (cHL). We conducted gene expression profiling of microdissected lymphoma cells of five ALK(+) and four ALK(-) systemic ALCL, seven cALCL and sixteen cHL, and of eight subsets of normal T and NK cells. The analysis supports a derivation of ALCL from activated T cells, but the lymphoma cells acquired a gene expression pattern hampering an assignment to a CD4(+), CD8(+) or CD30(+) T-cell origin. Indeed, ALCL display a down-modulation of many T-cell characteristic molecules. All ALCL types show significant expression of NFkappaB target genes and upregulation of genes involved in oncogenesis (e.g. EZH2). Surprisingly, few genes are differentially expressed between systemic and cALCL despite their different clinical behaviour, and between ALK(-) ALCL and cHL despite their different cellular origin. ALK(+) ALCL are characterized by expression of genes regulated by pathways constitutively activated by ALK. This study provides multiple novel insights into the molecular biology and pathogenesis of ALCL.

Test of four colon cancer risk-scores in formalin fixed paraffin embedded microarray gene expression data.

BACKGROUND: Prognosis prediction for resected primary colon cancer is based on the T-stage Node Metastasis (TNM) staging system. We investigated if four well-documented gene expression risk scores can improve patient stratification. METHODS: Microarray-based versions of risk-scores were applied to a large independent cohort of 688 stage II/III tumors from the PETACC-3 trial. Prognostic value for relapse-free survival (RFS), survival after relapse (SAR), and overall survival (OS) was assessed by regression analysis. To assess improvement over a reference, prognostic model was assessed with the area under curve (AUC) of receiver operating characteristic (ROC) curves. All statistical tests were two-sided, except the AUC increase. RESULTS: All four risk scores (RSs) showed a statistically significant association (single-test, P < .0167) with OS or RFS in univariate models, but with HRs below 1.38 per interquartile range. Three scores were predictors of shorter RFS, one of shorter SAR. Each RS could only marginally improve an RFS or OS model with the known factors T-stage, N-stage, and microsatellite instability (MSI) status (AUC gains < 0.025 units). The pairwise interscore discordance was never high (maximal Spearman correlation = 0.563) A combined score showed a trend to higher prognostic value and higher AUC increase for OS (HR = 1.74, 95% confidence interval [CI] = 1.44 to 2.10, P < .001, AUC from 0.6918 to 0.7321) and RFS (HR = 1.56, 95% CI = 1.33 to 1.84, P < .001, AUC from 0.6723 to 0.6945) than any single score. CONCLUSIONS: The four tested gene expression-based risk scores provide prognostic information but contribute only marginally to improving models based on established risk factors. A combination of the risk scores might provide more robust information. Predictors of RFS and SAR might need to be different.

ET-1 and NOS III gene expression regulation by plaque-free and plaque-prone hemodynamic conditions

Résumé: Les environnements hémodynamiques, favorisant ou protégeant contre la formation de la plaque, induisent tout deux une augmentation de la production d'anion superoxide dans les cellules endothéliales (ECs). Par ailleurs, une régulation différente de l'expression des gènes a été décrite dans les cellules exposées à ces différentes conditions. Dans le but d'investiguer le rôle de l'augmentation du stress oxydatif dans l'expression des gènes régulée par le flux, nous avons d'abord exposé les EC à un flux unidirectionnel, non pulsé. Dans ces conditions, l'état oxydatif des cellules endothéliales est augmenté de façon transitoire. L'expression du gène de l'endothéline 1 (ET-1) est aussi induite de façon transitoire par un tel flux, alors que l'expression du gène de la nitiric oxyde synthase endothéliale (NOS III) est stimulé de façon durable. Au contraire, un flux unidirectionnel pulsé, qui induit une augmentation durable de la production d'anion superoxide, augmente aussi de façon durable l'expression des gènes de ET-1 comme de NOS III. Un flux oscillatoire (favorisant la plaque), qui lui aussi ,a des effets à long terme sur la production d'anion superoxide, a uniquement augmenté l'expression de ET-1. De plus, l'utilisation d'un antioxydant, a seulement partiellement inhibé la stimulation de l'expression du gène NOS III par le flux unidirectionnel pulsé, alors qu'il a complètement abrogé la stimulation de l'expression du gène ET-1 par le flux unidirectionnel pulsé et oscillatoire. Ceci suggère que les forces mécaniques régulent l'expression des gènes dans les EC par un double mécanisme dépendant et indépendant du stress oxidatif des cellules. Par ailleurs, ces résultats supportent ultérieurement l'hypothèse que la balance entre la réponse oxidative et anti-oxidante dans les cellules endothéliales exposées à un environnement hémodynamique est une des clés de la prédisposition à un dysfonctionnement endothélial observé dans des régions exposées à des flux perturbés. Abstract: Both plaque-free and plaque-prone hemodynamic environments induce an increase in the oxidative state of endothelial cells (ECs), whereas differential gene expression regulation was described in cells exposed to these conditions. In order to investigate the role of the increased oxidative state in flow-regulation of gene expression, we first exposed EC to non-pulsed unidirectional shear stress. These conditions only slightly increases ECs oxidative state and endothelin-1 (ET-1) mRNA expression, whereas endothelial nitric oxide synthase (NOS III) mRNA level were significantly up-regulated. On the contrary, both ET-1 and NOS III gene expression were significantly induced in EC exposed to pulsed-unidirectional flow (plaque-free). Only ET-1 gene expression was up-regulated by oscillatory flow (plaque-prone). Moreover, use of an antioxidant only partially inhibited NOS III gene up-regulation by unidirectional flow, whereas it completely abrogated ET-1 gene up-regulation by unidirectional and oscillatory flows. Thus suggesting that mechanical forces regulate gene expression in ECs both via oxidative stress-dependent and -independent mechanisms.

eVOC: a controlled vocabulary for unifying gene expression data.

Expression data contribute significantly to the biological value of the sequenced human genome, providing extensive information about gene structure and the pattern of gene expression. ESTs, together with SAGE libraries and microarray experiment information, provide a broad and rich view of the transcriptome. However, it is difficult to perform large-scale expression mining of the data generated by these diverse experimental approaches. Not only is the data stored in disparate locations, but there is frequent ambiguity in the meaning of terms used to describe the source of the material used in the experiment. Untangling semantic differences between the data provided by different resources is therefore largely reliant on the domain knowledge of a human expert. We present here eVOC, a system which associates labelled target cDNAs for microarray experiments, or cDNA libraries and their associated transcripts with controlled terms in a set of hierarchical vocabularies. eVOC consists of four orthogonal controlled vocabularies suitable for describing the domains of human gene expression data including Anatomical System, Cell Type, Pathology and Developmental Stage. We have curated and annotated 7016 cDNA libraries represented in dbEST, as well as 104 SAGE libraries,with expression information,and provide this as an integrated, public resource that allows the linking of transcripts and libraries with expression terms. Both the vocabularies and the vocabulary-annotated libraries can be retrieved from http://www.sanbi.ac.za/evoc/. Several groups are involved in developing this resource with the aim of unifying transcript expression information.

Meta-analysis of gene expression profiles in breast cancer: toward a unified understanding of breast cancer subtyping and prognosis signatures.

INTRODUCTION: Breast cancer subtyping and prognosis have been studied extensively by gene expression profiling, resulting in disparate signatures with little overlap in their constituent genes. Although a previous study demonstrated a prognostic concordance among gene expression signatures, it was limited to only one dataset and did not fully elucidate how the different genes were related to one another nor did it examine the contribution of well-known biological processes of breast cancer tumorigenesis to their prognostic performance. METHOD: To address the above issues and to further validate these initial findings, we performed the largest meta-analysis of publicly available breast cancer gene expression and clinical data, which are comprised of 2,833 breast tumors. Gene coexpression modules of three key biological processes in breast cancer (namely, proliferation, estrogen receptor [ER], and HER2 signaling) were used to dissect the role of constituent genes of nine prognostic signatures. RESULTS: Using a meta-analytical approach, we consolidated the signatures associated with ER signaling, ERBB2 amplification, and proliferation. Previously published expression-based nomenclature of breast cancer 'intrinsic' subtypes can be mapped to the three modules, namely, the ER-/HER2- (basal-like), the HER2+ (HER2-like), and the low- and high-proliferation ER+/HER2- subtypes (luminal A and B). We showed that all nine prognostic signatures exhibited a similar prognostic performance in the entire dataset. Their prognostic abilities are due mostly to the detection of proliferation activity. Although ER- status (basal-like) and ERBB2+ expression status correspond to bad outcome, they seem to act through elevated expression of proliferation genes and thus contain only indirect information about prognosis. Clinical variables measuring the extent of tumor progression, such as tumor size and nodal status, still add independent prognostic information to proliferation genes. CONCLUSION: This meta-analysis unifies various results of previous gene expression studies in breast cancer. It reveals connections between traditional prognostic factors, expression-based subtyping, and prognostic signatures, highlighting the important role of proliferation in breast cancer prognosis.

Different Effects of BORIS/CTCFL on Stemness Gene Expression, Sphere Formation and Cell Survival in Epithelial Cancer Stem Cells.

Cancer stem cells are cancer cells characterized by stem cell properties and represent a small population of tumor cells that drives tumor development, progression, metastasis and drug resistance. To date, the molecular mechanisms that generate and regulate cancer stem cells are not well defined. BORIS (Brother of Regulator of Imprinted Sites) or CTCFL (CTCF-like) is a DNA-binding protein that is expressed in normal tissues only in germ cells and is re-activated in tumors. Recent evidences have highlighted the correlation of BORIS/CTCFL expression with poor overall survival of different cancer patients. We have previously shown an association of BORIS-expressing cells with stemness gene expression in embryonic cancer cells. Here, we studied the role of BORIS in epithelial tumor cells. Using BORIS-molecular beacon that was already validated, we were able to show the presence of BORIS mRNA in cancer stem cell-enriched populations (side population and spheres) of cervical, colon and breast tumor cells. BORIS silencing studies showed a decrease of sphere formation capacity in breast and colon tumor cells. Importantly, BORIS-silencing led to down-regulation of hTERT, stem cell (NANOG, OCT4, SOX2 and BMI1) and cancer stem cell markers (ABCG2, CD44 and ALDH1) genes. Conversely, BORIS-induction led to up-regulation of the same genes. These phenotypes were observed in cervical, colon and invasive breast tumor cells. However, a completely different behavior was observed in the non-invasive breast tumor cells (MCF7). Indeed, these cells acquired an epithelial mesenchymal transition phenotype after BORIS silencing. Our results demonstrate that BORIS is associated with cancer stem cell-enriched populations of several epithelial tumor cells and the different phenotypes depend on the origin of tumor cells.

Up-regulation of macrophage migration inhibitory factor gene and protein expression in glial tumor cells during hypoxic and hypoglycemic stress indicates a critical role for angiogenesis in glioblastoma multiforme.

Glioblastoma multiforme (GBM) is the most malignant variant of human glial tumors. A prominent feature of this tumor is the occurrence of necrosis and vascular proliferation. The regulation of glial neovascularization is still poorly understood and the characterization of factors involved in this process is of major clinical interest. Macrophage migration inhibitory factor (MIF) is a pleiotropic cytokine released by leukocytes and by a variety of cells outside of the immune system. Recent work has shown that MIF may function to regulate cellular differentiation and proliferation in normal and tumor-derived cell lines, and may also contribute to the neovascularization of tumors. Our immunohistological analysis of MIF distribution in GBM tissues revealed the strong MIF protein accumulation in close association with necrotic areas and in tumor cells surrounding blood vessels. In addition, MIF expression was frequently associated with the presence of the tumor-suppressor gene p53. To substantiate the concept that MIF might be involved in the regulation of angiogenesis in GBM, we analyzed the MIF gene and protein expression under hypoxic and hypoglycemic stress conditions in vitro. Northern blot analysis showed a clear increase of MIF mRNA after hypoxia and hypoglycemia. We could also demonstrate that the increase of MIF transcripts on hypoxic stress can be explained by a profound transcriptional activation of the MIF gene. In parallel to the increase of MIF transcripts, we observed a significant rise in extracellular MIF protein on angiogenic stimulation. The data of our preliminary study suggest that the up-regulation of MIF expression during hypoxic and hypoglycemic stress might play a critical role for the neovascularization of glial tumors.

Transcriptional regulation of CD4 gene expression by T cell factor-1/beta-catenin pathway.

By interacting with MHC class II molecules, CD4 facilitates lineage development as well as activation of Th cells. Expression of physiological levels of CD4 requires a proximal CD4 enhancer to stimulate basic CD4 promoter activity. T cell factor (TCF)-1/beta-catenin pathway has previously been shown to regulate thymocyte survival via up-regulating antiapoptotic molecule Bcl-xL. By both loss and gain of function studies, in this study we show additional function of TCF-1/beta-catenin pathway in the regulation of CD4 expression in vivo. Mice deficient in TCF-1 displayed significantly reduced protein and mRNA levels of CD4 in CD4+ CD8+ double-positive (DP) thymocytes. A transgene encoding Bcl-2 restored survival but not CD4 levels of TCF-1(-/-) DP cells. Thus, TCF-1-regulated survival and CD4 expression are two separate events. In contrast, CD4 levels were restored on DP TCF-1(-/-) cells by transgenic expression of a wild-type TCF-1, but not a truncated TCF-1 that lacks a domain required for interacting with beta-catenin. Furthermore, forced expression of a stabilized beta-catenin, a coactivator of TCF-1, resulted in up-regulation of CD4. TCF-1 or stabilized beta-catenin greatly stimulated activity of a CD4 reporter gene driven by a basic CD4 promoter and the CD4 enhancer. However, mutation of a potential TCF binding site located within the enhancer abrogated TCF-1 and beta-catenin-mediated activation of CD4 reporter. Finally, recruitment of TCF-1 to CD4 enhancer was detected in wild-type but not TCF-1 null mice by chromatin-immunoprecipitation analysis. Thus, our results demonstrated that TCF/beta-catenin pathway enhances CD4 expression in vivo by recruiting TCF-1 to stimulate CD4 enhancer activity.

Liver hyperplasia and paradoxical regulation of glycogen metabolism and glucose-sensitive gene expression in GLUT2-null hepatocytes. Further evidence for the existence of a membrane-based glucose release pathway.

We investigated the impact of GLUT2 gene inactivation on the regulation of hepatic glucose metabolism during the fed to fast transition. In control and GLUT2-null mice, fasting was accompanied by a approximately 10-fold increase in plasma glucagon to insulin ratio, a similar activation of liver glycogen phosphorylase and inhibition of glycogen synthase and the same elevation in phosphoenolpyruvate carboxykinase and glucose-6-phosphatase mRNAs. In GLUT2-null mice, mobilization of glycogen stores was, however, strongly impaired. This was correlated with glucose-6-phosphate (G6P) levels, which remained at the fed values, indicating an important allosteric stimulation of glycogen synthase by G6P. These G6P levels were also accompanied by a paradoxical elevation of the mRNAs for L-pyruvate kinase. Re-expression of GLUT2 in liver corrected the abnormal regulation of glycogen and L-pyruvate kinase gene expression. Interestingly, GLUT2-null livers were hyperplasic, as revealed by a 40% increase in liver mass and 30% increase in liver DNA content. Together, these data indicate that in the absence of GLUT2, the G6P levels cannot decrease during a fasting period. This may be due to neosynthesized glucose entering the cytosol, being unable to diffuse into the extracellular space, and being phosphorylated back to G6P. Because hepatic glucose production is nevertheless quantitatively normal, glucose produced in the endoplasmic reticulum may also be exported out of the cell through an alternative, membrane traffic-based pathway, as previously reported (Guillam, M.-T., Burcelin, R., and Thorens, B. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 12317-12321). Therefore, in fasting, GLUT2 is not required for quantitative normal glucose output but is necessary to equilibrate cytosolic glucose with the extracellular space. In the absence of this equilibration, the control of hepatic glucose metabolism by G6P is dominant over that by plasma hormone concentrations.

Transcriptional profiling reveals divergent roles of PPARalpha and PPARbeta/delta in regulation of gene expression in mouse liver.

Little is known about the role of the transcription factor peroxisome proliferator-activated receptor (PPAR) beta/delta in liver. Here we set out to better elucidate the function of PPARbeta/delta in liver by comparing the effect of PPARalpha and PPARbeta/delta deletion using whole genome transcriptional profiling and analysis of plasma and liver metabolites. In fed state, the number of genes altered by PPARalpha and PPARbeta/delta deletion was similar, whereas in fasted state the effect of PPARalpha deletion was much more pronounced, consistent with the pattern of gene expression of PPARalpha and PPARbeta/delta. Minor overlap was found between PPARalpha- and PPARbeta/delta-dependent gene regulation in liver. Pathways upregulated by PPARbeta/delta deletion were connected to innate immunity and inflammation. Pathways downregulated by PPARbeta/delta deletion included lipoprotein metabolism and various pathways related to glucose utilization, which correlated with elevated plasma glucose and triglycerides and reduced plasma cholesterol in PPARbeta/delta-/- mice. Downregulated genes that may underlie these metabolic alterations included Pklr, Fbp1, Apoa4, Vldlr, Lipg, and Pcsk9, which may represent novel PPARbeta/delta target genes. In contrast to PPARalpha-/- mice, no changes in plasma free fatty acid, plasma beta-hydroxybutyrate, liver triglycerides, and liver glycogen were observed in PPARbeta/delta-/- mice. Our data indicate that PPARbeta/delta governs glucose utilization and lipoprotein metabolism and has an important anti-inflammatory role in liver. Overall, our analysis reveals divergent roles of PPARalpha and PPARbeta/delta in regulation of gene expression in mouse liver.

Characterization of the murine high Km glucose transporter GLUT2 gene and its transcriptional regulation by glucose in a differentiated insulin-secreting cell line.

In pancreatic beta-cells, the high Km glucose transporter GLUT2 catalyzes the first step in glucose-induced insulin secretion by glucose uptake. Expression of the transporter has been reported to be modulated by glucose either at the protein or mRNA levels. In this study we used the differentiated insulinoma cell line INS-1 which expresses high levels of GLUT2 and show that the expression of GLUT2 is regulated by glucose at the transcriptional level. By run-on transcription assays we showed that glucose induced GLUT2 gene transcription 3-4-fold in INS-1 cells which was paralleled by a 1.7-2.3-fold increase in cytoplasmic GLUT2 mRNA levels. To determine whether glucose regulatory sequences were present in the promoter region of GLUT2, we cloned and characterized a 1.4-kilobase region of mouse genomic DNA located 5' of the translation initiation site. By RNase protection assays and primer extension, we determined that multiple transcription initiation sites were present at positions -55, -64, and -115 from the first coding ATG and which were identified in liver, intestine, kidney, and beta-cells mRNAs. Plasmids were constructed with the mouse promoter region linked to the reporter gene chloramphenicol acetyltransferase (CAT), and transiently and stably transfected in the INS-1 cells. Glucose induced a concentration-dependent increase in CAT activity which reached a maximum of 3.6-fold at 20 mM glucose. Similar CAT constructs made of the human GLUT2 promoter region and the CAT gene displayed the same glucose-dependent increase in transcriptional activity when transfected into INS-1 cells. Comparison of the mouse and human promoter regions revealed sequence identity restricted to a few stretches of sequences which suggests that the glucose responsive element(s) may be conserved in these common sequences.

Infrequent expression of the MAGE gene family in uveal melanomas.

It has previously been reported that MAGE-1, -2, -3 and -4 genes are expressed in human cancers including cutaneous melanoma. MAGE-1 and MAGE-3 represent targets for specific immunotherapy because they encode peptide antigens which are recognised by cytotoxic T lymphocytes (CTL) when presented by HLA class I molecules, and pilot clinical trials with these peptides are currently in progress. It is likely that other members of the MAGE gene family may also encode antigens recognised by CTL. Uveal melanomas, like cutaneous melanomas, arise from melanocytes that are derived from the neural crest. To determine if uveal melanoma patients would be suitable for MAGE-peptide immunotherapy, the expression of MAGE-1, -2, -3 and -4 genes was assessed by reverse transcription followed by polymerase chain reaction (RT-PCR) amplification and ethidium bromide staining. Expression of MAGE genes was not detected in any of 27 primary tumours. Either MAGE-1 or MAGE-4 was expressed in only 2 of 26 metastatic samples, but expression of MAGE-2 or -3 was not detected. Our data suggest that, unlike cutaneous melanomas, uveal melanomas may not be suitable candidates for MAGE-peptide immunotherapy.