Highly expressed genes in pancreatic ductal adenocarcinomas: a comprehensive characterization and comparison of the transcription profiles obtained from three major technologies.

When using gene expression profiling to understand human tumors, one is often confronted with long lists of genes that need to be further categorized into meaningful data. We performed a comprehensive evaluation and comparison of gene expression profiles obtained from pancreatic cancers to determine those genes most differentially expressed and thus with the most promise for translation into clinically useful targets. cDNA was prepared from 50 samples of normal pancreas or duodenal mucosal tissues, 7 samples of chronic pancreatitis, and 39 samples of pancreas cancer tissues or cancer cell lines and hybridized to the complete Affymetrix Human Genome U133 GeneChip set (arrays U133A and U133B) for simultaneous analysis of 45,000 fragments corresponding to 33,000 known genes and 6,000 expressed sequence tags. Genes expressed at levels at least 3-fold greater in the pancreatic cancers as compared with nonneoplastic tissues were identified. Three hundred seventy-seven Affymetrix fragments were identified as having > or = 3-fold expression levels in pancreas cancer specimens as compared with nonneoplastic tissues, corresponding to 234 known genes. Serial analysis of gene expression libraries (http://www.ncbi.nlm.nih.gov/SAGE/) of two normal pancreatic ductal cell cultures (HX and H126) were used to exclude 17 genes with high expression levels in the normal duct epithelium (more than five tags/library). Of the remaining 217 known genes, 75 have been previously reported as highly expressed in pancreatic cancers, while the remaining 142 genes are novel. We used principal components analysis (PCA) to identify the genes among these 217 identified as the most differentially expressed and specific to pancreatic cancer tissues or cell lines. Among the most differentially expressed genes identified by PCA were Mesothelin, Muc4, Muc5A/C, Kallikrein 10, Transglutaminase 2, Fascin, TMPRSS3 and stratifin. The differential expression identified by PCA for these genes indicates they are among the more attractive targets for novel therapeutic targets, tumor markers, or as a means of screening pancreatic cancer samples for information regarding tumor classification or potential therapeutic responses. Our findings were also compared in detail to the previously reported findings of highly expressed genes in other studies of global gene expression in pancreatic cancers. We found that robust changes in gene expression were most often identified by more than one gene expression platform. Forty genes were identified by more than one method (U133 oligonucleotide arrays, cDNA arrays or serial analysis of gene expression), and 6 of these genes were identified by all three methods. Our findings identify a novel set of genes as highly expressed in pancreatic cancer, validate the differential expression of previously reported genes, and provide additional support for those genes most differentially expressed to be translated into clinically useful targets.