Conceived and designed the experiments: PB AL MO. Performed the experiments: AL MO YH. Analyzed the data: PB AL MO. Contributed reagents/materials/analysis tools: Mv XC AL SL KL. Wrote the paper: PB AL MO.
The authors have declared that no competing interests exist.
Cancers of the pancreas originate from both the endocrine and exocrine elements of the organ, and represent a major cause of cancer-related death. This study provides a comprehensive assessment of gene expression for pancreatic tumors, the normal pancreas, and nonneoplastic pancreatic disease.
DNA microarrays were used to assess the gene expression for surgically derived pancreatic adenocarcinomas, islet cell tumors, and mesenchymal tumors. The addition of normal pancreata, isolated islets, isolated pancreatic ducts, and pancreatic adenocarcinoma cell lines enhanced subsequent analysis by increasing the diversity in gene expression profiles obtained. Exocrine, endocrine, and mesenchymal tumors displayed unique gene expression profiles. Similarities in gene expression support the pancreatic duct as the origin of adenocarcinomas. In addition, genes highly expressed in other cancers and associated with specific signal transduction pathways were also found in pancreatic tumors.
The scope of the present work was enhanced by the inclusion of publicly available datasets that encompass a wide spectrum of human tissues and enabled the identification of candidate genes that may serve diagnostic and therapeutic goals.
The aim of this study is to provide a global assessment of gene expression patterns for the major neoplastic diseases of the pancreas. We demonstrate that the inclusion of a diverse source of diseased and normal pancreatic tissues for analysis using DNA microarrays enhances the identification of gene expression patterns that can be attributed to specific cell types and pancreatic diseases.
Pancreatic adenocarcinoma represents a significant health burden in industrialized countries and represents one of the most lethal human cancers. It is the fourth-leading cause of cancer-related death in the United States with a five-year survival rate of 4%
Islet cell tumors represent a second major pancreatic tumor whose pathogenesis is poorly understood and for which effective therapies are lacking. In contrast to previously published studies that compared gene expression between islet tumors and normal islets, the current study also examines gene expression in islet tumors in the context of the whole pancreas and other pancreatic diseases
The expanding collection of publicly accessible gene expression data also provides an opportunity to compare and enhance the dataset provided by this study. For example, the gene expression profiles of a wide spectrum of normal human tissues
All participating patients provided consent prior to surgery. The resected tissue was immediately frozen in liquid nitrogen, and stored at −80°C. A portion of the sample was processed for histopathology. Normal pancreatic islets and pancreatic ducts were obtained from organ donors. Samples were obtained with approved human subjects protocols from Stanford University, Samsung Medical Center (Seoul, South Korea), University of Washington, and the Cooperative Human Tissue Network. Data derived from 11 adenocarcinoma samples and 5 non-tumor pancreatic samples included in this report were previously published in collaboration with Johns Hopkins University
Pancreatic cancer cell lines were obtained from the American Type Culture Collection (Manassas, VA) and included AsPC-1 (#CRL-1682), BxPC-3 (#CRL-1687), Capan-1 (#HTB-79), Capan-2 (#HTB-80), CFPAC-1 (#CRL-1918), Hs 766T (#HTB-134), MIA PaCa-2 (#CRL-1420), PANC-1 (#CRL-1469), SU.86.86 (#CRL-1837). The cells were grown to 80–90% confluence in media specified for each cell type. Islets were isolated from pancreata as previously described
Total RNA for all samples was extracted using Trizol (Invitrogen, Carlsbad, CA) followed by mRNA isolation using FastTrack® mRNA (Invitrogen, Carlsbad, CA). RNA amplification was performed for the isolated pancreatic duct cells and islets, 3 adenocarcinomas, and the reference mRNA standards using the procedure described by Wang et al.
DNA microarrays containing 41,125 clones, representing approximately 24,473 unique genes were printed on microscope slides and processed for hybridization as previously described (
Features containing artifacts or blemishes were flagged for exclusion from further analysis using GenePix Pro 5.0 (Axon Instruments). The raw data was deposited in the Stanford Microarray Database (
The entire primary dataset is available at the Stanford Microarray Database (
Immunohistochemistry was performed using Dako Envision Plus (Glostrup, Denmark) following the manufacturer's instruction. Rabbit anti-axin2 serum was from Zymed (Invitrogen, Carlsbad, CA). Tissue microarrays constructed to evaluate axin2 expression contained 13 normal pancreata and 26 pancreatic adenocarcinomas. Rabbit antibodies against LGALS4 were produced against peptide sequence NH2-PPYPGPGHCHQQLNS, and for MUC13 peptide sequence NH2-DPEEKHSMAYQDLHSEC (AnaSpec, Inc., San Jose, CA).
Gene expression profiles were obtained using cDNA microarrays for a total of 80 surgically removed pancreatic tissue samples (
|
| |
| 6 | |
|
| |
osteoclast giant cell tumor (1 patient) | 2 | |
liposarcoma | 1 | |
lymphoma | 1 | |
|
| |
chronic pancreatitis | 3 | |
pseudocyst | 1 | |
mucinous cyst | 1 | |
|
| |
|
| |
|
| |
|
|
3 adenocarciinomas were used solely as amplified samples.
Additional samples included primary cultures of human pancreatic duct cells, purified human islets from normal donors, and nine commercially available pancreatic cancer cell lines. The human pancreatic duct cells and human islets were harvested from 5 and 9 different donors, respectively, and were processed separately and not pooled. RNA derived from the pancreatic duct cells and purified islets required antisense RNA amplification for the microarray analysis because of lower amounts of starting material.
We first used an unsupervised hierarchical clustering approach to organize and explore the data and highlight groups of genes and groups of samples with similar gene expression patterns
Unsupervised hierarchical clustering of the patterns of variation in expression for 9,259 genes (represented by 16,959 cDNA) in 80 pancreatic tissue specimens (A). The image uses a color code to represent relative expression levels. Red represents expression levels greater than the mean for a given gene across all samples. Green represents expression levels less than the mean across samples. A color bar (
Islet cell tumors were derived from four patients. Three samples from one patient representing separate metastatic lesions are included. Although immunohistochemistry revealed different secretory products for 3 patients, the tumors shared sufficient similarity in gene expression profiles to be clustered together.
Other pancreatic tumors that did not cluster with the adenocarcinomas or islet cell tumors included a pancreatic sarcoma, lymphoma, and osteoclast-like giant cell tumor. Although only single cases were represented for these rare tumors, they clustered together with the major common feature consisting of genes characteristic of the desmoplastic response, and relatively high expression of genes associated with proliferation.
Non-tumor and normal whole pancreatic tissues consistently clustered separately from the diseased tissues. The analysis did not distinguish between uninvolved non-tumor tissue from patients with cancer and pancreatic tissue from normal donors without any history of pancreatic disease.
Unsupervised hierarchical clustering also highlighted clusters of genes specifically associated with adenocarcinomas, islet cell tumors, and normal pancreata. The diversity of tumors and cell types in this sample set gave rise to a diversity of gene expression patterns, many of which appear to represent specific cell types. Defining the source of the distinctive gene expression patterns was facilitated by projecting the expression data for clonal pancreatic cancer cell lines, primary pancreatic duct cultures, and isolated pancreatic islets alongside the data derived from the previous clustering analysis (
Of particular interest in this study was the segregation of genes associated with adenocarcinomas cells themselves from those involved in the desmoplastic response. Two gene clusters, labeled “Adeno 1” (297 genes) and “Adeno 2” (88 genes) (
We used significance analysis of microarrays (SAM), a supervised approach of analysis that incorporates an assessment of statistical significance, to identify genes that are differentially expressed in adenocarcinomas compared to the normal pancreas and other tumors in the dataset
Islet tumors were identified by a cluster of 311 genes with elevated expression (correlation ≥0.75). Consistent with their endocrine origin, these tumors were characterized by elevated levels of transcripts encoding chromogranins A and B (CHGA, CHGB), secretogranin II (SCG2), synaptophysin (SYP), peptidylglycine α-amidating monooxygenase (PAM), dopa decarboxylase (DDC), and somatostatin receptor 1 (SSTR1) (
A group of genes was expressed at higher levels in non-tumor and normal pancreatic tissues than in either islet cell tumors or adenocarcinomas. Acinar cells represent 90% of the cellular composition of the pancreas and are expected to dominate the expression profile of the normal pancreas. Genes encoding digestive enzymes produced by acinar cells such as amylase (AMY2A), elastase (ELA3A, ELA3B), carboxypeptidase A (CPA2), pancreatic lipase (PNLIP), and trypsin (PRSS2) were enriched in this cluster (
Genes known to be associated with the desmoplastic response in pancreatic cancer were clustered separately from the adenocarcinoma clusters. These genes included osteonectin (SPARC), fibronectin (FN1), versican (CSPG2), and collagen type 1A2 (COL1A2), tissue inhibitor of metalloproteinase 1 (TIMP1), biglycan (BGN), matrix metalloproteinase 2 (MMP2), and stromelysin-1 (MMP3)
Additional analysis was performed to identify genes potentially useful for diagnostic and therapeutic purposes. As secretory or membrane proteins may be especially useful as diagnostic markers or therapeutic targets, each gene enriched in pancreatic cancer or islet cell tumors was further analyzed using algorithms to identify signal peptide or transmembrane domains. For pancreatic cancer, only genes within the “Adeno” clusters were analyzed because they are enriched in adenocarcinoma cells and do not include the stromal genes common to other pancreatic diseases. Of the 259 genes in the “Adeno” clusters, 126 were classified as secretory or membrane proteins by amino acid sequence analysis. An additional 71 genes, for a total of 197, were identified using a recently defined dataset by Diehn et al. in which a genome-wide screen for secretory proteins was empirically determined using similar DNA microarrays
Gene Symbol | Name | Clone ID |
| ||
LAMC2 | Transcribed locus, homologous to laminin, gamma 2 | IMAGE:460403 |
CTSE | Cathepsin E | IMAGE:243202 |
GPX2 | Glutathione peroxidase 2 (gastrointestinal) | IMAGE:587847 |
LGALS4 | Lectin, galactoside-binding, soluble, 4 (galectin 4) | IMAGE:511068 |
GPRC5A | G protein-coupled receptor, family C, group 5, member A | IMAGE:595037 |
MMP14 | Matrix metallopeptidase 14 (membrane-inserted) | IMAGE:270505 |
ITGA2 | Integrin, alpha 2 (CD49B, alpha 2 subunit of VLA-2 receptor | IMAGE:525246 |
AGR2 | Anterior gradient 2 homolog (Xenopus laevis) | IMAGE:2321113 |
COL17A1 | Collagen, type XVII, alpha 1 | IMAGE:252259 |
TSPAN8 | Tetraspanin 8 | IMAGE:509731 |
GPRC5A | G protein-coupled receptor, family C, group 5, member A | IMAGE:1457276 |
CEACAM7 | Carcinoembryonic antigen-related cell adhesion molecule 7 | IMAGE:509688 |
| ||
SCG2 | Secretogranin II (chromogranin C) | IMAGE:174627 |
RASSF7 | Ras association (RalGDS/AF-6) domain family 7 | IMAGE:1573778 |
TTR | Transthyretin (prealbumin, amyloidosis type I) | IMAGE:1868551 |
SGNE1 | Neuroendocrine protein 1 (putative) | IMAGE:878836 |
INSM1 | Insulinoma-associated 1 | IMAGE:22895 |
PCSK2 | Proprotein convertase subtilisin/kexin type 2 | IMAGE:24254 |
QPCT | Glutaminyl-peptide cyclotransferase | IMAGE:711918 |
FGB | **Fibrinogen beta chain | IMAGE:84713 |
PEX7 | Peroxisomal biogenesis factor 7 | IMAGE:2018758 |
PTPRN2 | Protein tyrosine phosphatase, receptor type | IMAGE: 812968 |
SERPINA1 | Serpin peptidase inhibitor | IMAGE:294578 |
ARF3 | ADP-ribosylation factor 3 | IMAGE:291097 |
Mutational activation of the Wnt pathway is one of the most common genetic steps in the pathogenesis of gastrointestinal cancers. Expression of AXIN2, a marker for
DNA microarray analysis of global gene expression patterns in a diverse collection of normal and diseased pancreatic tissues allowed identification of specific gene expression profiles for pancreatic adenocarcinomas, islet cell tumors, and the normal pancreas. Interpretation of the gene expression profiles was facilitated by the parallel analysis of gene expression in primary pancreatic ducts and islets, and pancreatic cancer cell lines. The approach identified genes that are specifically expressed by neoplastic or stromal cells. Supporting the approach is the identification of many genes previously associated with pancreatic adenocarciomas or islet cell tumors. Many other genes not previously associated with these cancers were also newly identified. The availability of publicly available gene expression data provided an opportunity to compare the gene expression patterns of pancreatic adenocarcinomas and islet cell tumors to those of many other normal tissues. Other available datasets enabled the association of differentially expressed genes with areas of genomic DNA gains and losses. Similar gene expression datasets for other adenocarcinomas including hepatic, gastric, and breast tumors provide an opportunity for additional comparisons
The present study also provides a characterization of the gene expression profile for islet cell tumors in the context of the whole pancreas. The gene expression profile for neuroendocrine tumors in patients with the MEN1 syndrome was recently reported
Among pancreatic tumors, the adenocarcinomas are the most common and the most deadly. Distinct and characteristic differences in gene expression patterns have revealed subclasses of common tumors, such as breast cancer, that also differ with respect to their natural history or response to therapy
Our results provide evidence that the Wnt pathway may often be activated in pancreatic adenocarcinoma and islet cell tumors. Evidence for Wnt pathway activation in pancreatic cancer has been inconclusive because nuclear staining for β-catenin, a mediator of Wnt signaling, is observed in less than 5% of cases
The present study provides a comprehensive analysis of gene expression in pancreatic tumors. The analysis of a broad spectrum of pancreatic tissues in conjunction with publicly available datasets and software tools enhanced the identification of genes that may participate in disease pathogenesis, or may serve as preferred targets for diagnostic or therapeutic strategies.
Clinical staging and outcomes for the samples used in
(0.03 MB XLS)
List of genes within the Adeno 1 and 2 clusters. Annotated gene list identified by significance analysis of microarrays versus all other non-adenocarcinoma pancreatic tissues. Only the 31 adenocarcinomas grouped together in
(1.91 MB XLS)
Table of differentially expressed genes identified by SAM in adenocarcinomas whose genomic location showed high DNA copy gains or losses.
(0.03 MB XLS)
List of genes differentially expressed between primary pancreatic duct cells and whole normal pancreatic tissue obtained using SAM as described in the
(2.12 MB XLS)
List of genes within the islet cell tumor cluster. Annotated list of genes identified by SAM that are differentially expressed in islet tumors compared to all other pancreatic tissues in the dataset.
(0.43 MB XLS)
Identification of candidate genes capable of discriminating pancreatic adenocarcinomas or islet cell tumors from a diverse collection of normal human tissues. Putative secretory or membrane proteins enriched in pancreatic adenocarcinomas or islet cell tumors in the current study were examined in the context of a wide variety of normal tissues reported in reference
(0.05 MB XLS)
The authors wish to acknowledge the contributions of Ward Trueblood, Augusto Bastidas, Mark Vierra, Sherry Wren, Nancy Kaduk, George Yang, Janet Mitchell, and Kelli Montgomery from the Departments of Surgery and Pathology at Stanford University, and Suet Y. Leung (University of Hong Kong, Hong Kong, China) for assistance in tissue acquisition; and to Christopher Savard for assistance in isolating the pancreatic ducts and islets.