Gene expression profiling identifies genes associated with invasive intraductal papillary mucinous neoplasms of the pancreas

Abstract

The molecular pathology of intraductal papillary mucinous neoplasms (IPMNs) of the pancreas has not been well characterized, and there are no reliable markers to predict the presence of an associated invasive carcinoma in IPMNs. Using oligonucleotide microarrays, we performed a large-scale gene expression profiling of 12 IPMNs with or without an associated invasive carcinoma. A subset of genes identified was validated for the gene expression patterns in a large panel of IPMNs by reverse-transcription polymerase chain reaction and/or immunohistochemistry. A total of 673 transcripts were identified as expressed at significantly higher levels (P < 0.05 and at fivefold or greater) in IPMNs relative to normal pancreatic ductal epithelial samples. Of interest, many of the genes identified as overexpressed in IPMNs have also been previously reported to be highly expressed in infiltrating ductal adenocarcinoma of the pancreas. By analyzing genes overexpressed selectively in IPMNs with an associated invasive carcinoma (n = 7), we also identified a panel of genes potentially associated with the invasive phenotype of the neoplasms. Immunohistochemical validation revealed that claudin 4, CXCR4, S100A4, and mesothelin were expressed at significantly high frequency in invasive IPMNs than in noninvasive IPMNs. Notably, the expression of at least two of the four proteins was observed in 73% of 22 invasive IPMNs but in none of 16 noninvasive IPMNs (P < 0.0001). Our findings suggest that preoperative assessment of gene expression profiles may be able to differentiate invasive from noninvasive IPMNs.

Figure 1

An example of the intraductal component of IPMN frozen sections dissected for microarray analysis showing a high neoplastic cellularity (stained with hematoxylin).

Figure 2

Hierarchical cluster analysis of 12 IPMNs with or without an associated invasive carcinoma and 2 normal pancreatic ductal epithelial samples.

Figure 3

A: Microarray expression levels of 11 genes selected for RT-PCR validation in the nonneoplastic ductal cell line (HPDE), microdissected normal ductal epithelium (Avg-NPD), and IPMNs (Avg-IPMNs). B: RT-PCR analysis of 11 selected genes in HPDE and 12 IPMNs. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) serves as a RNA control.

Figure 4

Immunohistochemical validation of five overexpressed genes on IPMN tissue microarrays. Shown are representative tissue cores stained positive for S100A4 (A), CD55 (B), TIMP1 (C), PSCA (D), and mesothelin (E).

Figure 5

Immunohistochemical validation of two selected genes (CXCR4 and claudin 4) associated with the invasive phenotype of IPMNs. CXCR4 is not expressed in normal ductal epithelium (A). Strong immunolabeling of CXCR4 is detected in two cases of invasive IPMNs (B and C), but not in an IPMN associated with colloid carcinoma (D). Claudin 4 is not expressed in normal ductal epithelium (E). Diffuse labeling for claudin 4 is shown in the neoplastic epithelium of invasive IPMNs (F and G), in contrast to negative staining in noninvasive IPMN (H). Note the discrete membrane immunoreactivity of claudin 4 in the neoplastic cells demonstrated by a high-power view (G).

Figure 6

Examples for histological classification of IPMNs (left, H&E staining) and expression profiles of a panel of 38 IPMNs on TMAs determined by immunohistochemical analyses of all of the seven genes tested (right). Filled box and open box indicate positive and negative immunolabeling, respectively. ND, not determined.