Pancreatic cancer is the fourth leading cause of cancer deaths in the United States. The mortality rate is the highest among cancer types, with an overall 5-year survival rate of less than 5% [1]. Because of its asymptomatic nature in early stages and a lack of sensitive and specific diagnostic tools, pancreatic cancer is usually detected after metastasis has occurred and when curative therapy is no longer possible. There are currently no biomarkers available for early detection of the disease, and even the most aggressive monitoring of high-risk patients is inadequate. Furthermore, pancreatic cancer is highly resistant to both chemotherapy and radiation therapy. The identification of overexpressed genes in pancreatic cancer may lead to the development of new biomarkers and screening techniques as well as new immunotherapeutic, molecular targeting, and gene therapy strategies.

Techniques Used in the Identification of Gene Overexpression

The progression of carcinogenesis is thought to occur through the accumulation of multiple genetic alterations, and high-throughput gene expression technologies (“global gene expression profiling”) have been recently used to simultaneously determine the expression patterns of thousands of genes.

These technologies are based either on serial, partial sequencing of cDNAs derived from sample RNA (as in serial analysis of gene expression [SAGE]), or on hybridization of labeled sample-derived cDNAs to specific probes immobilized on a grid (as in cDNA microarrays) [2, 3]. The SAGE technique has the advantage of being able to quantify gene expression without prior sequence information.

Recent global gene expression profiling studies of pancreatic cancer have employed SAGE, cDNA microarray, and oligonucleotide microarray technology to quantify gene expression in cancer cell lines as well as in human tissues. Several overexpressed genes have been identified by these techniques, and overexpression of many of these genes has been confirmed by individual studies of mRNA and/or protein expression. RNA studies have included Northern blots, in situ hybridization, and reverse transcriptase polymerase chain reaction (RT-PCR). Protein studies have included immunohistochemical labeling and Western blots. Gene expression studies which incorporate both RNA and protein analyses are ideal, as posttranslational regulation of gene expression frequently accounts for a disparity between mRNA abundance and protein expression.

Analysis of protein overexpression is particularly important for the development of biomarkers of cancer and for the identification of new therapeutic targets. Differential display proteomics, a biochemical technique that compares protein levels between tissue or fluid samples, allows for the simultaneous differential analysis of a large number of proteins and holds promise for improved diagnosis and therapy in pancreatic cancer [4].

Immunohistochemical staining and in situ hybridization are particularly useful as complements to global gene expression profiling studies performed on bulk tumor, as they are able to localize the site of expression to a particular area of the tumor or to a particular cell type. Newly developed tissue microarray technology allows for large-scale immunohistochemical comparison of a multitude of tissue samples and has recently been used to verify differential protein expression between pancreatic cancer tissue samples, other tumor samples, and normal tissue [5].

Problems Encountered in Global Gene Expression Profiling

One problem with gene profiling performed by comparing pancreatic cancer tissue to normal pancreas is that the desmoplastic reaction of the tumor’s stromal component contributes significantly to measured differences in overall gene expression. Interestingly, the desmoplastic response in chronic pancreatitis mimics that of pancreatic adenocarcinoma and accounts for fewer differentially expressed genes. A recent study by Logsdon et al. [6] incorporated a comparison between adenocarcinoma and chronic pancreatitis into its gene profiling strategy in order to more specifically identify genes differentially expressed by the neoplastic epithelium. The study’s immunohistochemical localization of four of these genes, (14-3-3 sigma/stratifin, beta 4 integrin, S100A6, and S100P) to the neoplastic epithelium validated this approach.

Additional strategies to eliminate the influence of the desmoplastic response on differential gene expression measurements have included fine-needle aspiration of the tumor [7], laser capture microdissection [8], and comparison of pancreatic cancer cell lines (which lack a stromal component) to normal pancreas [9]. Laser capture microdissection, however, has proven to be time-consuming and labor-intensive, and the amount of genetic material obtained is minimal and usually requires amplification [7]. However, studies that profile cancer cell lines in vitro may not accurately reflect differential gene expression, as the cell lines lack in vivo influences on gene expression, including those of the surrounding stroma and the immune system [6].

Another problem facing gene expression profiling is that tumor gene expression is largely ductal, whereas the ductal cell contribution to gene expression in normal pancreas is diluted by the presence of islet and acinar cells. Therefore, genes appearing to be overexpressed in global gene profiling studies of tumor tissue may simply be attributable to the smaller ductal component of the normal sample rather than to neoplastic differential gene expression. To account for this problem, normal ductal cell lines have been used as a control in gene expression analysis, and differentially expressed genes found in studies involving normal tissue have been referenced in SAGE libraries of normal ductal cell lines, and excluded accordingly.

Finally, many of the genes and gene products identified as being differentially expressed in global gene expression profiling studies have several names or abbreviations and are not yet fully described in terms of location and function [10]. Further progress in these studies will be limited by the amount of information available for each individual gene.

Review of Differentially Expressed Genes

Global gene expression profiling studies, including SAGE, cDNA microarray, and oligonucleotide microarray analyses, were identified in the literature by a MEDLINE search. Genes for which overexpression was confirmed by Northern blot, in situ hybridization, RT-PCR, or immunohistochemistry following initial identification by global gene expression profiling are listed in Table 1. In general, the fact that these genes were chosen for confirmatory testing implies that they showed significant differential expression in global gene expression profiling, or that they have been implicated in carcinogenesis or are otherwise associated with cancer.

Table 1 Overexpressed genes in pancreatic adenocarcinoma: methods of confirmatory testing (as identified by previous studies which confirmed global gene expression findings with individual gene expression analysis).
Full size table

Each gene listed in Table 1 was then referenced in the Cancer Genome Anatomy Project (CGAP) Gene Finder and the SAGE Anatomic Viewer (http://cgap.nci.nih.gov/Genes/) in order to compare the number of tags identified in normal pancreatic DNA libraries with the number found in cancerous pancreatic DNA libraries. The widest discrepancy in data was chosen from among the three data options (tissue only, cell lines only, tissue and cell lines) and is shown in Table 2. Additional tissue types in which greater than 63 tags per 200,000 were observed in cancer are also included in Table 2.

Table 2 Overexpressed genes in pancreatic adenocarcinoma: CGAP data on gene expression (as identified by previous studies which confirmed global gene expression findings with individual gene expression analysis).
Full size table

Diagnostic Implications of Gene Overexpression

The identification of genes that are overexpressed in neoplastic pancreatic tissue versus non-neoplastic pancreatic tissue holds promise for the development of histologic diagnostic markers detectable in small biopsy or cytopathology samples. Furthermore, the identification of secretory or cell-surface gene products that are overexpressed in neoplastic pancreatic tissue versus other tumors or other normal tissues holds promise for the development of biomarkers detectable in the serum and/or pancreatic juice. It should be noted that the overall sensitivity and/or specificity of new diagnostic strategies and screening techniques may be optimized by incorporating a panel of markers that, individually, may lack sensitivity and/or specificity. Therefore, genes that are not exclusively overexpressed in neoplastic pancreatic tissue still warrant consideration as future diagnostic markers.

Identifying genes that are overexpressed in the precursor lesions of pancreatic cancer is particularly important for the development of sensitive and specific screening techniques for high-risk patients and for determining the genetic alterations that produce malignancy. Current methods of monitoring patients with predisposing genetic syndromes or a family history of pancreatic cancer include serum CA 19-9 measurements and endoscopic ultrasound (EUS), both of which are of limited value in detecting small pancreatic lesions. For screening purposes, pancreatic juice may provide an ideal specimen, as it potentially has a higher concentration of cancer-related DNA and protein and lacks many normal components of the serum [11].

Recently, a standard system of nomenclature was established to facilitate comparative studies of precursor lesions of pancreatic cancer, now called PanIN (pancreatic intraepithelial neoplasia). Similar to the progression model of colorectal neoplasia, PanINs are thought to progress from flat and papillary mucinous lesions without dysplasia (PanIN-1A and PanIN-1B) to flat or papillary lesions with mild-to-moderate dysplasia (PanIN-2), to carcinoma in situ (PanIN-3), to infiltrating carcinoma [12]. The current molecular progression model of pancreatic cancer associates this histologic progression with the gradual accumulation of genetic alterations in oncogenes, tumor suppressor genes, and mismatch repair genes. Studies of the validity of this model have identified alterations in several genes including K-ras, HER-2/neu, p16, p53, DPC4, and BRCA2. Point mutations in the K-ras oncogene and overexpression of HER-2/neu are the earliest genetic alterations seen in PanIN-1A and PanIN-1B [13, 14].

Additional support for the progression model of pancreatic cancer has been provided recently by progressively increased immunolabeling of Ki-67 in PanINs [15]. Ki-67 is a nuclear protein that correlates well with cellular proliferation and has been used extensively as a proliferation marker for other types of neoplasia. Further studies of genetic alterations in the context of this model may lead to more effective screening and disease-monitoring strategies and may provide clues to the genetic alterations that are responsible for malignancy.

Unfortunately, PanIN lesions are often not clinically detectable and are often inaccessible to biopsy. Intraductal papillary mucinous tumors (IPMTs) of the pancreas are uncommon cystic exocrine pancreatic neoplasms that, in general, have a significantly less aggressive course than ductal adenocarcinomas. The IPMTs have been used recently as a clinically detectable model for studying the early genetic alterations of pancreatic carcinogenesis [16]. Like PanIN, IPMTs can progress from hyperplasia to dysplasia to malignancy. Gene expression profiling of this tumor type at different stages may identify those genes that are differentially expressed in precursor lesions of pancreatic adenocarcinoma [17].

Therapeutic Implications of Gene Overexpression

Knowledge of gene overexpression may be used to develop immunotherapeutic strategies, including the creation of a cell-mediated vaccine or the development of antibody-based immunotherapy. The identification of tumor-specific antigens may allow for the development of a vaccine. Recently, serological analysis of recombinant cDNA libraries (SEREX) has been used to identify overexpressed genes that are immunogenic [18, 19]. The identification of genes overexpressed on the cell surface is particularly useful for the development of antibody-based immunotherapy.

As a therapeutic strategy, molecular targeting may involve inhibition of either the expression or function of the gene product itself or disruption of the signaling transduction pathway in which the gene is involved, with consequent anti-mitogenic action. Examples of the first targeting strategy were recently proposed by a study [20] that identified overexpression of HMG-I (Y) and laminin receptor in pancreatic cancer cell lines derived from metastatic lesions of an orthotopic pancreatic tumor in SCID mice. Inhibition of HMG-I (Y) expression may be accomplished by the introduction of antisense RNA [21], and the laminin receptor may be blocked with monoclonal antibodies to inhibit metastasis [22].

Finally, information on overexpression may be used in the development of gene-therapy strategies. In particular, the promoter regions of overexpressed genes may be incorporated into vectors that deliver “suicide genes” to the tumor cells. These genes, when expressed by the neoplastic cells, act to convert systemically administered prodrugs into active chemotherapeutic agents, thereby localizing the cytotoxic effect to the neoplasm [23].

Clinical Implications of Select Overexpressed Genes

Based on our review of the literature, S100 genes, mesothelin, prostate stem cell antigen, and 14-3-3 sigma/stratifin are candidate genes with promising clinical utility in pancreatic cancer. Confirmatory testing with in-situ hybridization, RT-PCR, and/or immunohistochemistry has consistently shown these proteins to be upregulated.

S100 Genes

Overexpression of several members of the S100 family of calcium-binding proteins has been identified in global gene expression studies and confirmatory RNA and/or protein analyses, including S100A4, S100A6, S100A10, and S100P (Table 1). S100 proteins have been implicated in a variety of intracellular interactions resulting in enzymatic regulation, signal transduction and transcription, calcium homeostasis, and cytoskeletal dysfunction. This final action may confer invasive and metastatic properties. Also, many S100 proteins are released into the extracellular space and may be clinically detectable in the serum and/or pancreatic juice [24]. Additional studies of specific interactions and of signaling pathways involving these genes may lead to the development of anti-metastatic molecular-targeting strategies.

Several S100 proteins, including S100A4 and S100A6, exhibit expression in a variety of other tumors [5, 25] but may still be useful in a serum biomarker panel. Evidence that S100A4 overexpression may be attributable to the stromal response rather than to the neoplastic epithelium has been presented in a recent global gene expression profiling study comparing pancreatic adenocarcinoma and chronic pancreatitis [6]. However, S100A4 overexpression has also been shown in RT-PCR studies of pancreatic cancer cell lines [26, 27] and in immunohistochemical localization performed on pancreatic adenocarcinoma [27], suggesting that overexpression may in fact be associated with the neoplastic epithelium. Additional investigation of S100A4 as a potential aid in diagnosis is warranted. Furthermore, evidence of hypomethylation has been found in S100A4 overexpression [26, 27], and this information may be used to develop therapeutic molecular-targeting strategies.

S100P has recently been identified as a potential biomarker for three reasons. First, like most S100 proteins, it may be secreted into the extracellular space and may, therefore, be detectable in the serum or pancreatic juice. Second, overexpression of the gene has been seen in IPMTs, suggesting that the gene may be expressed at an early stage of carcinogenesis and may therefore be useful for early diagnosis [16]. Third, there is recent evidence obtained by cDNA microarray, SAGE library scanning, and tissue microarray immunohistochemistry that differential expression of S100P is exclusive to neoplastic pancreatic tissue, increasing its specificity as a diagnostic marker [5].

Mesothelin

Mesothelin is a glycosylphosphatidyl inositol-anchored, membrane-bound glycoprotein thought to function in cell adhesion. It is a component of a precursor protein that is proteolytically cleaved to mesothelin and soluble MPFs (mesothelin-related proteins) [28]. Normally, it is expressed in mesothelial cells, with relatively little expression in other normal tissues, including the pancreas. Mesothelin overexpression has been detected in several cancer types, including squamous cell carcinoma of the lung, ovarian cancer, and malignant mesothelioma. Its location in the cell membrane makes it a potential target for antibody-mediated immunotherapy or the development of a cell-mediated vaccine. Immunotherapeutic targeting of mesothelin in ovarian cancer, malignant mesothelioma, and lung cancer has yielded success in preliminary studies [29, 30].

Mesothelin overexpression has been identified in global gene expression profiling studies of pancreatic cancer, and subsequent in-situ hybridization and immunohistochemical labeling has confirmed that it is expressed in the neoplastic epithelium rather than in the surrounding stroma or nonneoplastic epithelium [31].

Immunohistochemical localization of mesothelin at luminal borders increases the likelihood of its detection, or the detection of MPF, in pancreatic juice or serum [31]. However, its expression in other tumor types decreases its specificity as a serum biomarker, and recent weak immunohistochemical labeling found in a small amount of non-neoplastic pancreatic tissue suggests that it may be most useful alongside other diagnostic studies or in a biomarker panel for pancreatic cancer [31].

Furthermore, mesothelin was not overexpressed in a recent global gene expression profiling study of IPMTs [16], and so it may be of limited use in the detection of precursor lesions. As in S100A4, PSCA (see below), and 14-3-3 sigma, hypomethylation has been identified as a frequent epigenetic alteration of this gene and is highly correlated with overexpression [26].

Prostate Stem Cell Antigen (PSCA)

Prostate stem cell antigen, like mesothelin, is a glycosylphosphatidyl inositol-anchored, membrane-bound glycoprotein and is normally expressed in the prostate, with relatively little expression in other normal tissues, including the pancreas. Overexpression is seen in prostate cancer and is positively correlated with grade, stage, and metastasis in this cancer type [32]. Like mesothelin, it holds promise for new immunotherapeutic strategies.

Immunohistochemical confirmatory tests of global gene expression profiling studies have shown PSCA to be expressed in the neoplastic epithelium rather than in the surrounding stroma or non-neoplastic epithelium. Like mesothelin, its membranous location and expression at luminal borders may allow detection in pancreatic juice and/or serum and may lead to new diagnostic imaging techniques. Also, as in mesothelin, evidence suggests that it may be most effective as a marker when used alongside other diagnostic studies or in a biomarker panel for pancreatic cancer [33]. It was not overexpressed in a recent global gene expression profiling study of IPMTs [16], and so it may be of limited use in the detection of precursor lesions. As in mesothelin, S100A4, and 14-3-3 sigma, hypomethylation has been identified as a frequent epigenetic alteration of this gene and is highly correlated with overexpression [26].

14-3-3 sigma (Stratifin)

14-3-3 sigma/stratifin has been found to act as a tumor-suppressor gene that inhibits cell cycle progression through the G2 phase. A loss of 14-3-3 sigma expression is seen in breast, lung, head and neck, bladder, and hepatocellular carcinomas. However, global gene expression profiling has identified overexpression of this gene in pancreatic adenocarcinoma. The reverse transcriptase PCR and immunohistochemistry have both been used to confirm overexpression in both cancer tissue and cancer cell lines (Table 1). Recent immunohistochemical labeling of 14-3-3 sigma showed expression primarily in the cytoplasm, with membranous accentuation [34], and the gene has potential as a histologic biomarker.

Findings of overexpression suggest a role in pancreatic carcinogenesis, in contrast to its role as a tumor suppressor in other tissues. An anti-apoptotic role has been previously described for this gene [35]. In addition, the gene may play a role in resistance to chemotherapy, as it was found to be highly expressed in pancreatic cancer cell lines selected for chemotherapeutic resistance [36].

As in mesothelin, S100A4, and PSCA, hypomethylation has been identified as a frequent epigenetic alteration of this gene and is highly correlated with overexpression [26, 34].

Conclusions

Global gene expression profiling is a useful tool for the initial identification of biomarker candidates and potential therapeutic targets. However, these large-scale techniques require subsequent confirmatory testing of mRNA and protein overexpression in order to sufficiently evaluate the proposed gene’s sensitivity and specificity for pancreatic cancer. Several genes have been identified in the recent literature as holding promise for the development of new diagnostic and therapeutic strategies, including the S100 family of calcium-binding proteins, mesothelin, prostate stem cell antigen, and 14-3-3 sigma/stratifin.