INTRODUCTION

Adenocarcinoma of the pancreas affects approximately 29,000 individuals each year in the United States, and nearly all patients die within months of diagnosis (1). Only a minority of cases is amenable to surgery, and regrettably, surgical resection of localized pancreatic cancer is seldom curative. Thus, prevention and early detection of pancreatic adenocarcinoma remain the best hope for a cure, because once established, this disease is nearly always fatal (2).

A multistep model has recently been proposed for pancreatic adenocarcinomas, in which noninvasive precursor lesions in the pancreatic ducts undergo histologic and genetic progression toward invasive cancer (3, 4). These morphologically distinct, noninvasive lesions have been classified under a uniform nomenclature scheme termed pancreatic intraepithelial neoplasia, or PanIN. We and others have demonstrated that PanINs share many of the genetic aberrations associated with invasive adenocarcinomas, underscoring their classification as “neoplasms” rather than as a reactive/hyperplastic process (5, 6, 7, 8, 9). Some of the genetic alterations are nearly ubiquitous (e.g., telomere shortening) (10), suggesting that these are early events in the ductal epithelium, whereas others, such as loss of function of the tumor suppressor gene BRCA2, occur only in the most advanced PanIN lesions that precede invasive cancer (11). Understanding the molecular mechanisms that facilitate PanIN progression toward invasive adenocarcinomas is critical, because these noninvasive neoplasms represent one of the best targets available for early detection and chemoprevention strategies for pancreatic cancer (12).

We have recently explored the global expression profiles of invasive pancreatic adenocarcinomas with three different gene expression platforms (serial analysis of gene expression, oligonucleotide microarrays, and cDNA microarrays) (13, 14, 15). These studies have identified a multitude of genes that are overexpressed at the transcript level in pancreatic cancer compared with in normal ductal epithelium. We have subsequently validated the “cancer-specific” expression of a subset of these markers at the protein level by immunohistochemistry (16). However, the expression of these tumor markers in the noninvasive precursor ductal lesions remains largely unexplored. The advent of TMAs for high-throughput expression profiling in archival tissues has greatly facilitated the identification of novel diagnostic and prognostic markers in multiple cancer types (17, 18, 19, 20, 21, 22). We have constructed a TMA made up entirely of noninvasive ductal lesions (PanIN TMA) for multicomponent profiling of the pancreatic adenocarcinoma progression model. Using a TMA-based approach, we herein validate molecular abnormalities previously reported in routine histologic sections of PanINs, as well as examine the expression of novel cellular markers that have been discovered by global gene expression profiling of pancreatic adenocarcinomas.

MATERIALS AND METHODS

Tissue samples were obtained from the surgical pathology archives of the Department of Pathology at the Johns Hopkins University School of Medicine. Formalin-fixed, paraffin-embedded blocks were retrieved from 44 patients who underwent pancreaticoduodenectomy (Whipple resection) for pancreatic ductal adenocarcinoma. PanIN lesions were selected by two authors (AM and RHH) and classified into PanIN-1A, PanIN-1B, PanIN-2, and PanIN-3 using previously described criteria (3). For TMA construction, representative areas containing morphologically defined PanINs were circled on the glass slides and used as a template. The PanIN TMA was constructed using a manual Tissue Puncher/Arrayer (Beecher Instruments, Silver Spring, MD) as previously described (10). For each PanIN, a 1.4-mm core was punched from the donor block to ensure that the entire duct lesion and adequate surrounding tissue could be incorporated into the spot. A total of 99 cores (61 PanINs, 11 intra- and interlobular pancreatic ducts, and 27 control tissue cores from various extrapancreatic organs) were arrayed on the recipient block. Six PanIN lesions were not suitable for evaluation because of either loss of tissue cores or exhaustion of the duct lesion in deeper sections, and these were excluded from analysis. The remaining 55 PanIN lesions comprised 16 PanIN 1A, 18 PanIN 1B, 14 PanIN 2, and 7 PanIN 3 lesions).

Immunohistochemistry was performed as previously described (16). Briefly, unstained 5-μm sections were cut from the paraffin block selected and deparaffinized by routine techniques. The TMAs were placed in 200-mL of Target Retrieval Solution, pH 6.0 (Envision Plus Detection Kit, DAKO) for 20 minutes at 100° C. After cooling for 20 minutes, slides were quenched with 3% H2O2 for 5 minutes before incubating with primary antibody using the Dako Autostainer. The list of primary antibodies used, with respective dilutions and source of antibodies, is given in Table 1. Labeling was detected with the Dako Envision system according to the manufacturer's protocol. Negative controls (primary antibody replaced by serum from appropriate species) were used for each antibody in each run. Internal positive controls were available from the plethora of pancreatic and extrapancreatic control cores on the PanIN TMA itself (for example, basal cells in the prostate express prostate stem cell antigen; normal pancreatic ductal epithelium expresses Dpc4, etc.); additionally, external controls were examined with each run (e.g., breast carcinoma for MUC1, colonic adenocarcinoma for MUC2, gastric tissue for MUC5, etc.). For all immunohistochemical stains, the pattern of staining in the normal pancreatic ductal epithelium was considered to be the baseline for comparison with progressive PanIN lesions.

Table 1 Immunohistochemical Antibodies
Full size table

The following scoring criteria were used for assessment of immunolabeling: for the 11 proteins that are not expressed or minimally expressed in the normal pancreatic ductal epithelium (p53, cyclin D1, Ki-67, topoisomerase II α [topo II], MUC1, MUC2, MUC5, mesothelin, fascin, prostate stem cell antigen [PSCA], and 14-3-3σ), immunolabeling in <5% of cells within a PanIN lesion was considered negative; immunolabeling in 5–25% was considered focal, and >25% labeling was considered diffuse. Only appropriate subcellular localization of detectable signal was considered for assessment of percentage staining (for example, only nuclear staining was counted for p53, cyclin D1, Ki-67, and topo II, and cytoplasmic or membrane labeling was counted for the rest). Complete absence of cytoplasmic and nuclear expression of Dpc4, or complete absence of nuclear expression of p16, was considered negative (abnormal) for the respective stain, as previously described (8, 23, 24). Finally, the expression of β-catenin, which is normally present only in a membranous distribution, was graded based on percentage of cells with strong nuclear and cytoplasmic labeling, as previously described (25).

Statistical analyses for significant differences in immunohistochemical abnormalities between high-grade (PanIN 2 and 3) and low-grade (PanIN 1A and 1B) lesions were performed using two-tailed Fisher's exact test, with a significant P value set as < .05. All statistical comparisons were performed using the Analyze-it add-in for Microsoft Excel.

RESULTS AND DISCUSSION

A summary of immunohistochemical abnormalities in the 55 PanIN lesions in this study is tabulated in Table 2 and is pictorially illustrated using Eisen's TREEVIEW software (http://rana.lbl.gov/EisenSoftware.htm) in Figure 1. The results for the 14 individual proteins are described in the ensuing discussion, in the context of each protein's functional class and relevant literature.

Table 2 Summary of Immunohistochemical Abnormalities in PanIN Lesions
Full size table
FIGURE 1
figure 1

Pictorial representation of progressive molecular abnormalities in PanIN lesions depicted using Eisen's TREEVIEW software (http://rana.lbl.gov/EisenSoftware.htm). Individual PanIN lesions are in rows, and the 14 proteins examined are in columns. PanIN 1A lesions are designated as A1–A16; PanIN 1B lesions, as B1–B18; PanIN 2 lesions, as C1–C14, and PanIN 3 lesions, as D1–D7. The original TREVIEW color scheme for relative gene expression is retained for designating relative protein expression in this figure. Thus, green implies no expression, red designates diffuse expression, and brown designates focal expression. Depending on the protein of interest, therefore, red color could imply retained normal expression (e.g., Dpc4/Smad4), or aberrant overexpression (MUC5, PSCA, etc.).

Full size image

Tumor Suppressor Genes (p53 and Dpc4)

P53 is commonly inactivated by mutations in invasive pancreatic adenocarcinomas, with a frequency of 50–75% of cases, depending on the series (26, 27, 28). Immunohistochemical detection of nuclear p53, as a surrogate marker for mutation, usually demonstrates a slightly lower frequency of abnormalities (∼30–50%) (29, 30, 31), consistent with the lower sensitivity of this technique. Nuclear overexpression of p53 has also been reported in PanIN lesions in the pancreas. For example, DiGiuseppe et al. (31) found abnormalities in p53 labeling in 2 of 17 (12%) histologically high-grade lesions (carcinomas in situ, equivalent to PanIN 3 in the current classification scheme). In contrast, all histologically lower-grade duct lesions labeled normally, suggesting that p53 gene inactivation is a late event in genetic progression in pancreatic ducts. Similarly, Apple et al. (32) reported p53 accumulation in 20% of “dysplastic” ductal lesions but in no “ductal hyperplasias.” In concert with these findings, we report diffuse nuclear overexpression of p53 (>25% nuclei) in 4 of 7 (57%) PanIN 3 lesions but in no normal ducts, PanIN 1A, PanIN 1B, or PanIN 2 foci. Thus, we confirm that p53 mutation, as assessed by nuclear overexpression of p53 protein, is a “late” event in the progression model of pancreatic adenocarcinoma, occurring only in the most advanced PanIN lesions (Fig. 2).

FIGURE 2
figure 2

Representative photomicrographs illustrating immunohistochemical abnormalities in various histologic categories of PanIN lesions. The protein of interest is designated in rows, and the PanIN histologic category is designated in columns. Row 1, absence of nuclear p53 labeling in PanIN 1A, 1B, and 2, with diffuse nuclear labeling in PanIN 3. Row 2, absence of nuclear MIB-1 labeling in PanIN 1A and 1B, with focal nuclear labeling in PanIN 2 and 3. Row 3, absence of MUC1 labeling in PanIN 1A and 1B; note the apical labeling in normal ducts subjacent to PanIN lesions. In contrast, strong and diffuse apical labeling is seen in PanIN 2 and 3 lesions. Row 4, unlike the case with MUC1, MUC5 labeling is present in low-grade PanIN 1A and 1B lesions, with persistent expression in PanIN 2 and 3 lesions. Row 5, absence of fascin expression in PanIn1A, with focal or diffuse labeling in subsequent PanIN 1B, 2 and 3 lesions; note intense diffuse cytoplasmic labeling in PanIN 3 lesion. Row 6, absence of 14-3-3σ labeling in PanIN 1A and 1B lesions, with expression in PanIN 2 and 3 lesions.

Full size image

The tumor suppressor gene SMAD4 (Small Mothers Against Decapentaplegic 4) or DPC4 (Deleted in Pancreatic Carcinoma 4) located on chromosome 18q, is a member of the transforming growth factor β (TGFβ) family (33). Inactivation of SMAD4/DPC4 by homozygous deletions, or a combination of intragenic mutation and allelic loss of heterozygosity, is present in approximately 55% of invasive pancreatic adenocarcinomas (33). Immunohistochemical labeling for the SMAD4/DPC4 gene product has recently been shown to mirror SMAD4/DPC4 gene status (34). Wilentz et al. (8) studied a large series of routine histologic sections from pancreata using the B-8 anti-Dpc4 antibody. Dpc4 expression was intact in all of the low-grade duct lesions that they examined, whereas 30% of the histologically high-grade lesions had a complete loss of Dpc4 expression. By using the PanIN classification scheme on our PanIN TMA, we detected complete loss of Dpc4 expression in 2 of 7 (28%) PanIN 3 lesions. Dpc4 expression was intact in all of the intra- or interlobular ducts, PanIN 1A, PanIN 1B, and PanIN 2 foci examined. Our findings are virtually identical to those findings of Wilentz et al. (8) and confirm that, similar to the case of the p53 gene, inactivation of the DPC4 gene appears to be a “late” event in the genetic progression of pancreatic cancer.

Cell Cycle Regulatory Genes (p16 and Cyclin D1)

The p16 gene on chromosome 9p encodes for a regulator of the cell cycle, and it is inactivated by homozygous deletions, mutations, or methylation of the p16 promoter in approximately 95% of invasive pancreatic adenocarcinomas (35, 36). Recent evidence suggests that similar mechanisms for p16 gene inactivation are also active in the noninvasive pancreatic ductal lesions, and that there is a progressive accumulation of these genetic abnormalities with histologic progression in the ducts (9, 37). Loss of nuclear p16 protein expression is considered to be a reliable surrogate for complete inactivation of p16 gene function (23, 24). Wilentz et al. (7), using immunohistochemical labeling for the p16 gene product on routine histologic sections, showed that 30% of “flat duct lesions without significant atypia” (equivalent to PanIN 1A), 27% of “papillary duct lesions without significant atypia” (equivalent to PanIN 1B), 55% of “papillary duct lesions with atypia” (equivalent to PanIN 2), and 71% of carcinoma in situ (PanIN 3) lesions had loss of expression of nuclear p16. In our TMA-based study, we found that 5 of 16 (31%) PanIN 1A, 8 of 18 (44%) PanIN 1B, 7 of 14 (50%) PanIN 2, and 6 of 7 (85%) PanIN 3 lesions had loss of nuclear p16. Thus, we were able to confirm that p16 inactivation is a common and “early” event in the multistep model of pancreatic adenocarcinoma, that the frequency of this loss increases with increasing grades of dysplasia, and that it precedes both p53 and Dpc4 inactivation. Nevertheless, loss of one cell cycle checkpoint protein appears to be inadequate to initiate uncontrolled cell division in PanIN lesions, as demonstrated by the concomitantly examined proliferation antigens (see below). It is likely that because of the redundancy of cell cycle checkpoint controls, more than one “hit” is required for initiation of deregulated proliferation.

The cyclin D1 or PRAD1 gene, located on chromosome 11q13, is a positive regulator of the cell cycle. Overexpression of cyclin D1 protein has been documented in several human cancers, such as breast, lung, and colon carcinoma, and in mantle cell lymphoma (38, 39, 40, 41). Between 60% and 85% of invasive pancreatic adenocarcinomas demonstrate nuclear overexpression of cyclin D1 protein by immunohistochemistry (42, 43, 44). Cyclin D1 overexpression in pancreatic cancer is associated with a poor prognosis and decreased survival (42, 45), whereas inhibition of cyclin D1 expression results in increased chemosensitivity and decreased expression of multiple chemoresistance genes (46). In the current study, focal or diffuse nuclear overexpression of cyclin D1 was seen in 0 of 11 normal pancreatic ducts, 0 of 16 PanIN 1A, 0 of 18 PanIN B, 4 of 14 (29%) PanIN 2, and 4 of 7 (57%) PanIN 3 lesions. Based on nuclear cyclin D1 expression in a third of PanIN 2 lesions, cyclin D1 abnormalities would best be classified as an “intermediate” event in the multistage progression of pancreatic adenocarcinoma, preceding p53 mutations and inactivation of Dpc4 in a subset of cases.

Oncogene (β-Catenin)

Beta-catenin is a target of the canonical wnt signaling pathway, implicated in both development and tumorigenesis. Oncogenic mutations of the β-catenin gene, which result in stabilization and nuclear translocation of β-catenin protein, have been reported in a variety of human cancers. β-catenin mutations are uncommon in pancreatic adenocarcinomas (47) but are frequent in nonductal neoplasms of the pancreas (solid pseudopapillary tumors, pancreatoblastomas, and acinar cell carcinomas) (25, 48, 49). All 11 normal and 55 PanIN foci examined in this study demonstrated normal membranous expression of β-catenin, with no convincing cytoplasmic or nuclear expression in any lesion. The complete absence of abnormal β-catenin expression in PanIN lesions reiterates the existence of two distinct, genetically divergent pathways of neoplasia in the pancreas, one resulting in the more common, conventional ductal adenocarcinomas and the other resulting in the less common nonductal neoplasms.

Proliferation Antigens (Ki-67 and Topo II)

Unchecked proliferation is an intrinsic feature of neoplasia. The proliferation antigen Ki-67, as detected by the monoclonal antibody MIB-1, is expressed in the majority of invasive adenocarcinomas of the pancreas (50, 51), whereas we have demonstrated up-regulation of topo II in invasive pancreatic cancers by using both oligonucleotide and cDNA microarrays (13, 15). The expression of one of these proliferation antigens—Ki-67—has been systematically examined by Klein et al. (52) in routine histologic sections of PanINs. In their study, Klein et al. (52) found increasing Ki-67–labeling indices with increasing grades of dysplasia in PanINs; thus, PanIN 1A had a labeling index of 0.69%, PanIN 1B, of 2.33%, PanIN 2, of 14.08%, and PanIN 3, of 22.01%. There are no studies to date examining topo II expression in the precursor lesions of pancreatic adenocarcinoma. In the current study, no normal pancreatic ducts, PanIN 1A, or PanIN 1B expressed Ki-67 or topo II, whereas only focal nuclear Ki-67 and topo II expression was seen in 1 of 14 (7%) PanIN 2 and 5 of 7 (71%) PanIN 3 lesions; there was complete concordance between Ki-67 and topo II expression in the PanIN lesions. Our percentages for Ki-67 labeling are discordant from those in the study of Klein et al. (52) because we did not use a nuclear labeling index but rather calculated expression based on a two-tier focal (5–25%) versus diffuse (>25%) labeling scheme. In their study, Klein et al. (52) found a mean Ki-67–labeling index of 22% for PanIN 3, which is consistent with the absence of diffuse labeling in the PanIN 3 lesions in our series. Cyclin D1 expression demonstrated a marginally better concordance with proliferation than p53; 5 of 6 (83%) of Ki-67– and topo II–positive lesions also overexpressed nuclear cyclin D1, whereas 4 of 6 (67%) concomitantly overexpressed p53. There was no correlation between p16 loss and Ki-67/topo II overexpression, given the discordance in the timing of their occurrence (early versus late) in the multistep model of pancreatic cancer progression.

Epithelial Mucins (MUC1, MUC2, and MUC5)

The apomucins MUC1, MUC2 and MUC5 are frequently overexpressed in epithelial cancers, particularly those arising in the gastrointestinal tract and pancreas (53, 54, 55, 56). MUC1 is the principal pancreatic mucin, being expressed in normal pancreatic ducts and acini (57). MUC1 expression is commonly observed in invasive pancreatic adenocarcinomas at both the transcript and protein levels (55, 56, 57, 58), approaching 100% of cases in some series. In our study, MUC1 expression was present in 11 of 11 (100%) normal intra- and interlobular pancreatic ducts, 6 of 14 (43%) PanIN 2, and 6 of 7 (85%) PanIN 3 lesions but in only 1 of 16 (6%) PanIN 1A and 1 of 18 (5%) PanIN 1B lesions. Thus, in the multistep progression of pancreatic adenocarcinomas, MUC1 expression within normal intra- and interlobular ducts appears to be decreased in the low-grade PanINs (PanIN 1A and 1B). MUC1 appears to be subsequently re-expressed in the advanced PanIN lesions, and this expression persists into invasive adenocarcinoma. The expression of MUC1 is detected primarily in the apical membranes of normal ducts and PanINs, but only in the latter is occasional cytoplasmic expression also present. The transfection of MUC1-expressing construct into MUC1-negative tumor cell lines results in increased tumorigenicity and invasiveness in vivo (59), whereas functional studies have demonstrated MUC1 being upstream of the oncogenic Grb2-SOS-ras signaling pathway (60, 61). Thus, irrespective of the mechanism, re-expression of MUC1 in advanced PanIN lesions would appear to favor cell growth and invasion.

Unlike MUC1, the expression of the apomucin MUC2 is uncommon in both normal pancreas and in invasive ductal adenocarcinomas (57, 58). In contrast, MUC2 expression is commonly seen in intraductal papillary mucinous neoplasms (IPMNs) and their associated invasive colloid carcinomas (56). Adsay et al. (55) have proposed a dichotomy in the pathways to tumorigenesis within the pancreas, with MUC1-expressing invasive adenocarcinomas arising from PanINs and MUC2-expressing invasive adenocarcinomas arising in the backdrop of IPMNs. Consistent with these previous observations, none of the 66 normal or PanIN lesions in our study expressed MUC2. Thus, MUC2 expression is unlikely to play a significant role in the progression model of ductal adenocarcinoma of the pancreas. The basis for absence of MUC2 expression remains largely unknown, although it may be attributed to epigenetic inactivation by methylation of the MUC2 gene promoter (62).

The apomucin MUC5 is a gastric foveolar mucin that is normally expressed in the gastric epithelium but is down-regulated in the course of intestinal metaplasia (54). In our study, MUC5 was focally or diffusely expressed in 0 of 11 normal pancreatic ducts, in 12 of 16 (75%) PanIN 1A, in 13 of 18 (72%) PanIN 1B, in 13 of 14 (93%) PanIN 2, and in 7 of 7 (100%) PanIN 3 lesions. These findings are comparable to the findings of Kim et al. (63), who detected MUC5 expression in 71–90% of PanIN lesions in regular histologic sections. MUC5 is similar to MUC1 in that it is also expressed in the majority of invasive ductal adenocarcinomas (56, 57, 63). In contrast to MUC1, however, MUC5 is not expressed in normal ducts, but its expression is up-regulated even in the earliest PanIN lesions and persists thereafter in the majority of lesions of all histologic grades.

Novel Tumor Markers (PSCA, Mesothelin, Fascin, and 14-3-3σ)

Our recent global gene expression analyses of pancreatic adenocarcinomas have identified a large number of transcripts that are differentially up-regulated in primary pancreatic cancers and in pancreatic cancer cell lines (13, 14, 15, 64). We have subsequently validated a subset of these genes at the protein level in routine tissue sections of invasive pancreatic adenocarcinomas. For example, prostate stem cell antigen (PSCA) is a glycosylphosphatidyl inositol (GPI)–anchored protein with a limited tissue distribution, being most strongly expressed in the basal cells of the prostate (65). Argani et al. (66) have demonstrated up-regulation of PSCA transcripts in pancreatic adenocarcinomas by SAGE analysis (http://www.ncbi.nlm.nih.gov/SAGE) and have demonstrated overexpression of PSCA gene product in 60% of infiltrating adenocarcinomas by immunohistochemistry. In the current study, focal or diffuse PSCA expression was present in 0 of 11 normal ducts, in 7 of 16 (43%) PanIN 1A, in 5 of 18 (28%) PanIN 1B, in 6 of 14 (43%) PanIN 2, and in 4 of 7 (57%) PanIN 3 lesions. The pattern of immunolabeling was similar to that seen in invasive adenocarcinomas, being primarily cytoplasmic with focal luminal accentuation. Thus, up-regulation of PSCA can be classified as an “early” event in the progression model of pancreas cancer.

Along the same lines, mesothelin is a membrane-bound GPI-anchored protein that plays a role in cell adhesion and is normally most abundantly overexpressed by mesothelium (67). Mesothelin is also overexpressed in ovarian carcinomas and in squamous carcinomas of the lung, cervix, and esophagus (68, 69). Argani et al. (70) reported up-regulation of mesothelin transcripts in SAGE libraries of pancreatic adenocarcinomas and confirmed the overexpression of mesothelin protein in 100% (83% diffuse, 17% focal) of primary pancreatic adenocarcinomas (70). In the current study, focal or diffuse mesothelin labeling was seen in 0 of 11 normal ducts, 1 of 16 (6%) PanIN 1A, 0 of 18 PanIN 1B, 2 of 14 (14%) PanIN 2, and 1 of 7 (14%) PanIN 3 lesions. The labeling pattern of mesothelin resembles that of PSCA, being primarily cytoplasmic with luminal accentuation; in addition, luminal contents are frequently positive. In contrast to PSCA, however, mesothelin expression is uncommon in PanINs but is strongly up-regulated in the process of, or after tissue invasion by neoplastic ductal cells. Thus, mesothelin up-regulation would be classified as a “late” event in the progression model of pancreas cancer.

The human homolog of the sea urchin fascin is an actin-binding cytoskeletal protein that has been implicated in cell motility (71, 72). Iacobuzio-Donahue et al. (13) have reported the up-regulation of fascin transcripts in pancreatic adenocarcinoma tissues and cell lines using oligonucleotide microarrays, whereas Maitra et al. (16) have confirmed the overexpression of fascin protein in 95% of primary pancreatic adenocarcinomas. In the current study, focal or diffuse cytoplasmic fascin expression was seen in 0 of 11 normal ducts, 4 of 16 (25%) PanIN 1A, 5 of 18 (28%) PanIN 1B, 8 of 14 (57%) PanIN 2, and 4 of 7 (57%) PanIN 3 lesions. Therefore, fascin up-regulation would be considered an “intermediate” event in pancreatic adenocarcinoma progression, not uncommon in early PanIN lesions, but substantially up-regulated in advanced PanINs and nearly universal in invasive cancers.

14-3-3σ is a p53-regulated gene that belongs to the family of 14-3-3 proteins, which plays a role in a variety of cellular processes like signal transduction, cell cycle regulation, apoptosis, stress response, and cytoskeletal organization (73, 74). Several lines of evidence reiterate the role of 14-3-3σ as a classic tumor suppressor gene in that loss of expression of 14-3-3σ mRNA or protein has been reported in breast carcinomas (75), squamous cell carcinomas of the head and neck (76), primary bladder cancers (77), and hepatocellular carcinomas (78). Iacobuzio-Donahue et al. (15) demonstrated paradoxical up-regulation of 14-3-3σ transcripts in pancreatic adenocarcinomas by cDNA microarray analysis and confirmed the overexpression of the 14-3-3σ protein in 90% of invasive pancreatic adenocarcinomas. Additionally, they were able to demonstrate cancer-specific hypomethylation of the 14-3-3σ promoter compared with normal pancreatic ductal epithelium, which lacks expression of 14-3-3σ. We have subsequently demonstrated the overexpression of 14-3-3σ in other gastrointestinal adenocarcinomas by immunohistochemistry (unpublished data). In the current study, 0 of 11 normal pancreatic ducts, 2 of 16 (13%) PanIN 1A, 2 of 18 (11%) PanIN 1B, 3 of 14 (21%) PanIN 2, and 6 of 7 (85%) PanIN 3 demonstrated focal or diffuse expression of 14-3-3σ. Thus, up-regulation of 14-3-3σ is another example of a “late” molecular event in the multistep progression of pancreas cancers, possibly occurring in concert with p53 abrogation and/or significant chromosomal instability within the ductal epithelium.

In summary, we have examined 14 cellular proteins encompassing a variety of cellular pathways and functions in the precursor lesions of pancreatic ductal adenocarcinomas using a TMA-based approach. Although the numbers of PanIN lesions in individual histologic categories, particularly PanIN 3, are small, we have demonstrated that frequency of abnormalities for a given marker using TMAs is quite comparable, if not remarkably similar, to the frequency detected using routine histologic sections. This study underscores the overall reliability of a TMA-based approach demonstrated in cancers and extends this concept to the study of precancerous lesions as well. Needless to say, the logistical benefits and cost-effectiveness of studying >50 ductal lesions on a single slide are enormous. In addition, the use of a TMA approach allows one to compare multiple parameters in the same precancerous lesions, a prerequisite for dissecting the molecular pathways to carcinogenesis. Figure 3 demonstrates our current understanding of molecular abnormalities in the multistep progression model of pancreas cancer (“PanINgram”) and underscores the tremendous advances made in our understanding of these precursor lesions since the earliest version of this illustration appeared, merely 3 years ago (79). The separation of early molecular events in PanIN lesions (up-regulation of PSCA) from late changes (up-regulation of mesothelin and 14-3-3σ) demonstrated in this study will hopefully lead to rational early-detection strategies in patients at risk for developing pancreatic cancer (for example, patients belonging to high-risk pancreatic cancer families or patients with one of the known pancreas cancer syndromes). Thus, one could envision that in such patients undergoing screening for pancreas cancer, the presence of elevated mesothelin levels in the pancreatic juice is likely to suggest a more advanced ductal lesion, if not occult carcinoma, as compared with the presence of elevated PSCA alone. Invasive pancreatic adenocarcinoma, once established, is almost always fatal, and therefore, the systematic identification and targeting of molecular abnormalities in the precursor lesions of invasive cancer remains one of our strongest avenues for combating this lethal disease.

FIGURE 3
figure 3

A “PanINgram” illustrating our current understanding of the molecular changes in the multistep progression model of pancreas adenocarcinomas.

Full size image