Invasive pancreatic cancer is thought to develop through a series of noninvasive duct lesions known as pancreatic intraepithelial neoplasia (PanIN). We used cDNA microarrays interrogating 15,000 transcripts to identify 49 genes that were differentially expressed in microdissected early PanIN lesions (PanIN-1B/2) compared with microdissected normal duct epithelium. In this analysis, a cluster of extrapancreatic foregut markers, including pepsinogen C, MUC6, KLF4, and TFF1, was found to be up-regulated in PanIN. Up-regulation of these genes was further validated using combinations of real-time reverse transcription-PCR, in situ hybridization, and immunohistochemistry in a total of 150 early PanIN lesions from 81 patients. Identification of these gastrointestinal transcripts in human PanIN prompted assessment of other foregut markers by both semiquantitative and real-time reverse transcription-PCR, revealing similar up-regulation of Sox-2, Gastrin, HoxA5, GATA4/5/6, Villin and Forkhead 6 (Foxl1). In contrast to frequent expression of multiple gastric epithelial markers, the intestinal markers intestinal fatty acid binding protein, CDX1 and CDX2 were rarely expressed either in PanIN lesions or in invasive pancreatic cancer. Hedgehog pathway activation induced by transfection of immortalized human pancreatic ductal epithelial cells with Gli1 resulted in up-regulation of the majority of foregut markers seen in early PanIN lesions. These data show frequent up-regulation of foregut markers in early PanIN lesions and suggest that PanIN development may involve Hedgehog-mediated conversion to a gastric epithelial differentiation program.
- pancreatic cancer
- epithelial differentiation
There is now compelling histopathologic and molecular evidence to support the evolution of invasive pancreatic cancer through a series of noninvasive duct lesions called pancreatic intraepithelial neoplasia (PanIN; refs. 1, 2 ). Early duct lesions designated PanIN-1A and PanIN-1B show minimal cytologic and architectural atypia and are associated with activating K-ras mutations (3), shortened telomeres (4), and overexpression of p21WAF1/CIP1 (5). PanIN-2 lesions exhibit mild to moderate cytologic and architectural atypia and are associated with loss of p16INK4A expression (6) and cyclin D1 overexpression (5). PanIN-3 exhibits severe cytologic and architectural atypia and manifests p53 mutations (7) and loss of DPC4/Smad4 expression (8). These molecular aberrations, known to play a role in carcinogenesis, increase in frequency with advancing PanIN lesions and are also observed in invasive cancer. In addition, recent evidence has implicated reactivation of Notch (9) and Hedgehog signaling (10), both of which regulate foregut development, in the evolution of PanIN.
Despite a better understanding of molecular aberrations associated with PanIN lesions, a comprehensive analysis of the PanIN phenotype has not yet been accomplished. Here we assess global gene expression in microdissected PanIN lesions using cDNA microarrays, real-time reverse transcription-PCR (RT-PCR), immunohistochemistry, and in situ hybridization. In addition to providing novel insights regarding the PanIN transcriptome, this analysis revealed up-regulated expression of a panel of extrapancreatic foregut markers in early PanIN lesions (PanIN-1B/2) compared with normal pancreatic ducts. In addition, we show that a similar panel of foregut markers can be up-regulated in nontransformed human pancreatic ductal epithelial cells by overexpression of Gli1, a downstream mediator of Hedgehog signaling. These data imply that the development of early PanIN is associated with characteristic changes in epithelial differentiation, including the ectopic appearance of a gastric epithelial phenotype, and that these changes in gene expression may represent a consequence of aberrant Hedgehog signaling.
Materials and Methods
Microdissection and Microarray Hybridization. Consent for tissue collection and experimentation was obtained from the Johns Hopkins Joint Committee for Clinical Investigation. PanIN lesions and normal ducts were identified in nonmalignant regions of fresh frozen pancreatic tissue from patients who underwent pancreatectomy for pancreatic cancer at Johns Hopkins Hospital between 1999 and 2003. Microarray hybridization was done on cDNA microarrays which have been described elsewhere (11, 12) . These microarrays use over 15,000 cDNA clones obtained from sequence verified IMAGE clones (Invitrogen Co., Inc., Carlsbad, CA), 75% of which represent genes with functional annotations, with the remaining 25% representing expressed sequence tags. Multiple 8-μm frozen sections from each specimen were stained with the Histogene kit (Arcturus Engineering, Mountain View, CA). Both PanIN lesions and normal ductal epithelium were microdissected using the Pixcell system (Arcturus Engineering), and processed as paired samples from individual patients. Total RNA was extracted from 500 to 1,000 microdissected cells using the Picopure RNA Isolation kit (Arcturus Engineering), and amplified through two rounds of T7 polymerase-based linear RNA amplification using the RiboAmp kit (Arcturus Engineering). Yield of total RNA from microdissected samples was 15 to 20 ng on average per microdissected PanIN lesion or normal duct (representative lesion shown in Fig. 1A-C ). Each sample required four to eight slides, each of which contained an average of three lesions to acquire a starting total RNA amount of 200 to 400 ng. Labeled cDNA was reverse transcribed from aRNA with a dN6 random primer (Invitrogen) in the presence of Cy3 (for normal duct) or Cy5 (for PanIN) dUTP (Amersham Pharmacia). Alternate labeling (Cy5 for normal duct and Cy3 for PanIN) was also done for duplicate experiments. The labeled targets were purified using Qiaquick PCR purification columns (Qiagen, Valencia, CA). cDNA microarrays interrogating 15,000 genes were cohybridized with paired samples for six patients in 40 μL of hybridization solution (4× SSC, 0.3 μg/μL COT1, 0.2% SDS, and 2× Denhardt's) at 65°C for 16 hours under >60% humidity. After hybridization, the slides were washed at room temperature in 0.5× SSC, 0.1% SDS for 2 minutes; 0.5× SSC, 0.01% SDS for 2 minutes; and 0.06× SSC solution for 2 minutes. The arrays were scanned (Agilent Technologies, Foster City, CA), and relative intensities of Cy3 and Cy5 fluorescent signals were normalized and filtered through quality control variables, including comparison of reverse labeled samples, and used to calculate gene expression ratios between each paired sample using IPLab software (Scanalytics, Fairfax, VA). Relative intensity data was imported into a custom made FileMaker Pro 5.5 database (FileMaker, Inc., San Francisco, CA) for further analysis.
Real-time RT-PCR, In situ Hybridization, and Immunohistochemistry. Quantitative real-time RT-PCR was conducted with the QuantiTect SYBR Green RT-PCR system (Qiagen) on an ABI7700 instrument (Applied Biosystems, Foster City, CA) according to the manufacturers' recommendations using the relative quantitation method. After optimization of each of the primer pairs, samples were assayed in a 50-L reaction mixture containing 10 L of sample RNA and optimal concentration of each of the primers using the SYBR-Green Quantitech reagent kit (Qiagen). The thermal profile for one-step SYBR Green-based RT-PCR consisted of a 30-minute RT step at 50°C and 15 minutes of Taq polymerase activation at 95°C, followed by 40 cycles of PCR at 95°C for 20 seconds (denaturation), 55°C for 30 seconds (annealing), and 72°C for 30 seconds (extension). Each experiment was done in duplicate.
In situ hybridization was done on 8-μm frozen sections of 11 pancreata resected for pancreatic cancer. Sections from 11 patients containing cross-sectional representation of 37 normal ducts, 26 PanIN-1A/1B, and nine PanIN-2 lesions were fixed in 4% PFA. Digoxigen-labeled sense and antisense RNA probes were denatured at 80°C for 8 minutes, then applied to tissue sections, covered with a glass coverslip, and sealed. The slides were hybridized overnight at 56°C. After stringent wash conditions at 56°C, the digoxigen-labeled RNA was detected using rabbit anti-digoxigen-alkaline phosphatase (Roche Diagnostics Co., Indianapolis, IN). Sections were then developed using nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate.
H&E staining and immunohistochemistry were done on 4-μm serial sections of paraffin-embedded, formalin-fixed tissue. Three tissue microarrays and six additional nonarrayed specimens containing pancreatic cancer and PanIN were obtained from material resected at the Johns Hopkins Hospital, Baltimore, MD, between the years 1996 and 2001. Overall, paraffin-embedded samples consisted of (a) PanIN lesions and normal ducts from pancreata resected for either chronic pancreatitis or pancreatic cancer (n = 53), in which chronic pancreatitis specimens contained normal ducts and PanIN-1A/1B lesions, whereas pancreata resected for pancreatic cancer contained examples of all grades of PanIN as well as normal ducts; and (b) pancreatic cancer without adjacent PanIN (n = 19). On the tissue microarrays, each specimen was represented by up to four 1.4-mm cores to obtain adequate representation (13). PanINs were defined using criteria originally established at the National Cancer Institute think tank in Park City, UT (14) and updated over time (1, 2) . Sections were deparaffinized in xylene and rehydrated through a series of alcohols. Antigen retrieval was achieved by microwave heating in citrate buffer at pH 6.0. Endogenous peroxidase activity was quenched in 3% hydrogen peroxide in methanol and nonspecific binding of secondary antibody blocked by incubation with normal horse serum. Individual sections were incubated with antibodies to either mucin 6 (MUC 6, 1:100, mouse monoclonal clone ON459 U.S. Biological, Swampscott, MA) or Trefoil factor 1 (TFF 1, 1:400, mouse monoclonal clone BC04, Zymed Laboratories, Inc., San Francisco, CA) overnight at 4°C. For each marker, conditions omitting primary antibody were used as negative controls. Because the caudal-type homeodomain transcription factors CDX1 and CDX 2 are thought to be important in regulating intestinal differentiation during development, immunohistochemistry for CDX 1 (1:200, rabbit polyclonal, generous gift from Dr. Debra Silberg, Department of Medicine, University of Pennsylvania, Philadelphia, PA) and CDX 2 (1:400, mouse monoclonal, Biogenex, San Ramon, CA) was also done. A streptavidin-biotin peroxidase detection system was used in accordance with the manufacturer's instructions (Vectastain Elite Kit; Vector Laboratories, Inc., Burlingame, CA) and then developed using 3,3′-diaminobenzidine as substrate. Sections were counterstained with hematoxylin. Samples of human and mouse (adult and embryonic) gastric epithelium, as well as human small intestine, were used as positive controls.
For semiquantitative RT-PCR, 2 μg of RNA per sample of microdissected PanIN lesions and normal ducts were subjected to cDNA synthesis using the Ready-To-Go first strand synthesis kit (Pharmacia, Uppsala, Sweden). PCR was subsequently done to analyze expression of foregut markers in the linear phase of amplification for each gene, with expression levels categorized into one of three groups: low or none, medium, and high. For all samples, glyceraldehyde-3-phosphate dehydrogenase was used as a loading control.
Transfection of Human Pancreatic Ductal Epithelial Cells with Gli1. Immortalized human pancreatic duct epithelial (HPDE) cells (15) were cultured in keratinocyte serum-free medium supplemented with bovine pituitary extract and human epidermal growth factor (Life Technologies, Grand Island, NY). Transfection was done using the FuGENE-6 transfection reagent (Roche Diagnostics) according to the manufacturer's protocol. Briefly, cells (1-2 × 106) were seeded on 10-cm Petri dishes. The following day, transfection was done using 20 μg of either Gli1-zfd-pSR-α DNA (gift of Dr. Philip Beachy, Department of Molecular Biology and Genetics, Johns Hopkins University, Baltimore, MD) or empty vector DNA. The cells were maintained in antibiotic free growth medium (K-SFM) for 72 hours. Total RNA from cells harvested at 48 and 72 hours post-transfection for both Gli1 and empty vector–transfected cells were used for real-time RT-PCR.
Microarray Data. Analysis of high-quality hybridization data for a total of 8,458 transcripts representing ∼6,500 unique genes revealed 32 genes that were up-regulated ≥3-fold in PanIN-1B/2 lesions compared with normal ducts and 17 genes that were down-regulated by a factor of ≥3 in at least four of six individual-paired experiments ( Table 1 ). Poor-quality hybridization data, as assessed by intensity analysis, was excluded. Up-regulated genes included those previously reported to be up-regulated in pancreatic cancer [e.g., Heat shock protein 70 (16), S100 calcium-binding protein P (17), and jagged 1 (9); Table 1] as well as a number of genes not previously associated with pancreatic neoplasia [e.g., aquaporin 5, FAT tumor suppressor (Drosophila homologue)]. One striking pattern emerging from these data involved a panel of genes, including pepsinogen C, TFF1, Kruppel-like factor 4 (KLF4), and MUC6, which are normally expressed in adjacent foregut epithelium, but not classically known to be expressed at high levels in normal human pancreas.
When the analysis was extended to a threshold of at least 1.8-fold up-regulation or down-regulation in at least four of six paired samples, an additional 102 genes were found to be up-regulated in PanIN-1B/2 versus normal, whereas 16 additional down-regulated genes were identified (Supplementary Table 1). Among up-regulated genes in this expanded list were the S100 calcium binding proteins A6 and A11, previously shown to be up-regulated in invasive pancreatic cancer (18), as well as a cluster of ribosomal proteins (RPL3, RPL12, RPL15, RPL18a, RPL35a, RPS7, and RPS14), some of which have recently been identified as haploinsufficient tumor suppressors in zebrafish (19).
Validation of Microarray Results. A subset of differentially expressed genes, including identified foregut markers, were validated using semiquantitative and real-time RT-PCR, in situ hybridization, and immunohistochemistry and are shown in bold type in Table 1. Table 2 shows data confirming differential expression of genes based on real-time RT-PCR using RNA extracted from microdissected PanIN-1B/2 lesions versus normal ducts in paired samples from six individual patients numbered 1 to 6. TFF 1 and S100 calcium-binding protein were up-regulated ≥2-fold in five of six samples; pepsinogen C, MUC6, and rhomboid were up-regulated in four of six samples; KLF4 in three of six. Retinol binding protein was down-regulated ≥2-fold in three of six samples.
Up-regulated expression of KLF4 and pepsinogen C in early PanIN lesions compared with normal ducts was further confirmed by in situ hybridization. Frozen sections of pancreata resected for pancreatic cancer showed KLF 4 expression in PanIN lesions from 5 of 11 cases (45%), with characteristic marked expression in PanIN lesions (>90% of cells staining), as compared with normal appearing ducts (<5% of cells; Fig. 2A ). Transcripts for pepsinogen C were seen in <5% of cells of normal ducts of all 11 patients but were highly and uniformly expressed in >90% of cells in the PanIN lesions from 6 of 11 patients (55%; Fig. 2B). As assessed by immunohistochemistry, TFF 1 and MUC6 were expressed in PanIN-1A/1B lesions, but not in normal ducts ( Fig. 2C and D). TFF1 expression was seen in three patterns, either homogeneously expressed in >90% of cells, focally expressed in contiguous areas that constitute >25%, but <90% of cells, or <5% of cells. TFF1 was highly expressed (>90% of cells staining positive) in 24 of 29 PanIN-1A (94%), an additional three showed focal expression, 13 of 19 PanIN-1B (68%, 5 focal), but decreased in frequency in more advanced PanIN lesions, with staining observed in one of six PanIN-2 (17%, two focal) and zero of three PanIN-3 lesions. No TFF1 expression was detected in any of the normal ducts identified in a total of 22 resected specimens. In contrast to early PanIN lesions, none of the invasive carcinomas of 19 patients expressed TFF1. MUC6 expression (>50% of cells positive compared with <5%) was seen in 26 of 29 PanIN-1A (90%), 17 of 19 PanIN-1B (89%), four of six PanIN-2 (67%), and one of three PanIN-3 lesions (33%), but not in the normal ducts of 22 patients. MUC6 was expressed in 12 of 19 cases (63%) of pancreatic cancer. No difference in staining for either TFF1 or MUC6 was observed between PanINs identified in specimens resected for PC or chronic pancreatitis.
Up-regulated Expression of a Broad Panel of Foregut Markers in PanIN-1B/2. Confirmed up-regulation of foregut markers identified on microarray analysis prompted examination of additional foregut markers not represented on the microarrays using both semiquantitative and real-time RT-PCR ( Fig. 3A and B ). This analysis revealed that a number of markers expressed at high levels in adult and embryonic foregut, but not in normal adult pancreas, were also frequently up-regulated in early PanIN lesions compared with normal ductal epithelium. These frequently up-regulated genes included gastrin, GATA4, GATA5, GATA6 (20), villin 1, villin 2 (21), Sox2 (22), HoxA5 (23), and Forkhead6 (Fkh 6, Foxl1; ref. 24). In contrast, the gastric parietal cell transcripts H+/K+ ATPase and intrinsic factor were not expressed in PanIN. IFABP, a marker of intestinal epithelium, was also not up-regulated. The homeobox gene HlxB9, which when ectopically expressed is capable of converting developing endoderm from a pancreatic to an intestinal-like fate (25), was expressed in both normal ductal epithelium and in early PanIN lesions. Quantitative real-time RT-PCR analysis was done using RNA from pooled PanIN-1B/2 lesions and pooled normal ductal epithelium from additional four patients. In these samples, HOXA5, GATA 5, Gastrin, Villin 1, and Villin 2, as well as pepsinogen C, TFF1, KLF4, and MUC4 were all up-regulated by at least 2-fold in PanIN epithelium compared with normal ( Fig. 3C).
Because the caudal-type homeodomain transcription factors CDX1 and CDX2 are thought to be integral in promoting intestinal and colonic (but not gastric) epithelial differentiation, their expression was also assessed in PanIN lesions and cancer. Immunohistochemistry for both CDX1 and CDX2 revealed weak staining for CDX2 in only 1 of 42 PanIN-1B lesions (2%) and no staining for CDX 1. Neither CDX1 nor CDX2 were detected in normal ducts, and only 1 of 31 cancers (3%) expressed CDX1, whereas none expressed CDX2 despite robust labeling of duodenal epithelium present on the arrays (data not shown).
Hedgehog Signaling Activates Foregut Markers in HPDE Cells. Recent reports suggest that Hedgehog pathway activation represents a characteristic feature of human PanIN, and it has been suggested by Thayer et al. that aberrant Hedgehog signaling may contribute to the generation of a gastrointestinal phenotype within premalignant pancreatic epithelium (10). To directly test this hypothesis, we induced ectopic Hedgehog pathway activation in nontransformed pancreatic ductal epithelial cells by overexpression of Gli1. Following transient transfection with a Gli1 expression vector, HPDE cells showed significant up-regulation of a panel of foregut epithelial markers very similar to that observed in early PanIN lesions ( Fig. 3D). Gli1-induced up-regulation of several foregut markers was detectable by 48 hours post-transfection and became striking by 72 hours, with a number of markers including GATA6, GATA5, GATA4, and FKH6 demonstrating >5-fold overexpression. HlxB9 expression was not up-regulated at 48 or 72 hours post-transfection, but did show >2-fold up-regulation at 96 hours. In general, the panel of foregut markers induced by Gli1 overexpression in HPDE cells was very similar to that observed to be up-regulated in microdissected human PanIN. However, some differences were also apparent. In addition to most of the markers found to be up-regulated in microdissected human PanIN, Gli1-transfected cells showed additional up-regulation of parietal cell (H+/K+ ATPase, intrinsic factor) and intestinal epithelial (IFABP) transcripts, consistent with heightened plasticity in immortalized HPDE cells compared with in vivo pancreatic epithelium.
The current study represents the first completed analysis of differential gene expression in human PanIN using microarray analysis of paired PanIN and normal ductal epithelial elements generated by microdissection of individual specimens. This analysis identified 49 genes differentially expressed in PanIN versus normal ductal epithelium, including a subset of 13 genes also known to be abnormally expressed in invasive pancreatic cancer. Among the transcripts found to be up-regulated in early PanIN were KLF4, pepsinogen C, TFF1, and MUC6, genes not normally expressed at high levels in pancreatic tissue but known to be expressed in the epithelium of adjacent stomach and/or intestine. Each of these genes were confirmed to be up-regulated in PanIN by both semiquantitative and real-time PCR, as well as in human tissue using either in situ hybridization for KLF4 and pepsinogen C, or immunohistochemistry for TFF1 and MUC6. Further analysis of other foregut markers using semiquantitative and real-time PCR revealed up-regulation of multiple extrapancreatic epithelial transcripts, including GATA4, GATA5, GATA6, gastrin, villin1, villin2, HoxA5, Sox2 and Forkhead6 (Foxl1). Although a significant degree of variability was observed between individual PanIN lesions, these data suggest that the development of early PanIN may occur through a process involving changes in cellular differentiation directed towards a nonpancreatic foregut identity. In addition, transfection of HPDE cells with Gli1, a downstream mediator of hedgehog signaling, resulted in up-regulation of foregut transcripts seen in early PanIN lesions, implying that these changes may be the result of aberrant hedgehog signaling in premalignant pancreatic epithelium.
Although not commonly considered to be a feature of pancreatic neoplasia, changes in epithelial identity are commonly observed in carcinoma of the esophagus (26), stomach (27), and uterine cervix (28), reflecting an apparent reprogramming of tissue-specific epithelial precursors. In stomach, intestinal metaplasia precedes the development of gastric cancer and is thought to represent a direct precursor of gastric adenocarcinoma (27). During the development and progression of intestinal metaplasia of the stomach, there is a shift in mucin expression from gastric-type mucins (MUC5AC and MUC6) to intestinal-type mucin (MUC2; ref. 22). This shift is associated with down-regulation of Sox2, a putative gastric epithelial–specific transcription factor, and up-regulation of CDX1 and CDX2 (22, 27, 29) , factors not expressed in normal stomach but known to be essential for normal differentiation of intestinal and colonic epithelium (22). Further evidence supporting a role for these caudal-related transcription factors in gastric cancer precursors is provided by the finding that misexpression of a CDX2 transgene in murine gastric mucosa results in the initiation of intestinal metaplasia (30).
With respect to the current study of neoplastic precursors in human pancreas, the foregut markers found to be up-regulated in early PanIN are normally expressed in adjacent stomach, either in the adult (pepsinogen C, TFF1, KLF4, MUC6, gastrin, GATA4, GATA5, GATA6, and Sox-2) or during foregut development (HOXA5, Forkhead6, villin, and HlxB9). In addition, a selected number of these markers are also expressed in other regions of the gastrointestinal tract (i.e., intestinal expression of KLF4, HOXA5, and villin). In contrast, the intestinal markers IFABP, CDX1, and CDX2 are not expressed in human PanIN. Similarly, PanIN lesions and pancreatic cancer express gastric-specific mucins MUC5AC and MUC6 (31), but do not express the intestinal mucin MUC2 (13). Together with abnormal activation of GATA5 and Sox2, factors felt to specify gastric epithelial identity, these findings suggest that abnormal epithelial differentiation observed in human PanIN most closely recapitulates a gastric epithelial paradigm, as opposed to an intestinal or colonic phenotype.
During normal development, Hedgehog signaling plays a critical role in foregut patterning and seems to promote gastric and intestinal fates, and repress pancreatic identity in developing foregut (32, 33) . Notochord repression of sonic hedgehog signaling in a discrete region of anterior endoderm allows realization of a pancreatic differentiation program and is required for normal pancreas development (34). Misexpression of sonic hedgehog within foregut endoderm of Pdx1-Shh transgenic mice results in abnormal tubular structures within the pancreas that closely resemble human PanIN lesions. Pancreatic tissue from these mice rapidly develop oncogenic K-ras mutations, another common feature observed in early human PanIN (10). Additional studies have suggested abnormal Hedgehog pathway activation in both human PanIN and in invasive pancreatic cancer (10, 35) . In the current study, transfection of HPDE cells with Gli1 resulted in the up-regulation of foregut epithelial transcripts including those observed in early PanIN lesions, suggesting that up-regulation of nonpancreatic, gastric epithelial genes in human PanIN may indeed represent a consequence of abnormal Hedgehog signaling.
In addition to the foregut markers described above, we have identified a number of other genes differentially expressed in human PanIN, some of which may contribute to the neoplastic phenotype. Among the genes found to be up-regulated in PanIN, S100 calcium-binding protein P is also known to be up-regulated in invasive pancreatic cancer, as are other S100 proteins such as A4 (17, 36) . Drosophila rhomboid is capable of modulating epidermal growth factor receptor signaling (37), a pathway known to play an important role in pancreatic tumorigenesis. Genes down-regulated in the current study included Retinol-binding protein 1, an intracellular carrier molecule involved in the retinol transport, which is also down-regulated in ovarian cancer (38).
In summary, these data show expression of a significant number of extrapancreatic markers in human PanIN, suggesting that PanIN development is associated with activation of an epithelial differentiation program characteristic of adjacent foregut. A similar pattern of gene expression was induced by ectopic Hedgehog activation in nontransformed pancreatic ductal epithelial cells, suggesting a probable mechanism for altered expression of foregut markers observed in human PanIN. Based on up-regulation of the putative gastric epithelial-specific transcription factors GATA5 and Sox2 and the absence of both CDX1 and CDX2, this program seems to most closely recapitulate certain aspects of gastric, rather than intestinal, differentiation. In addition to the reported panel of foregut markers, the current study has identified a number of other genes that may contribute to the development and progression of pancreatic cancer precursors.
Grant support: NIH grant DK-56211, International Hepato-Pancreato-Biliary Association, National Health and Medical Research Council of Australia, family of Margaret Lee, and Paul K. Neumann Professorship in Pancreatic Cancer at Johns Hopkins University (S.D. Leach).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this are available at Cancer Research Online (http://cancerres.aacrjournals.org).
- Received April 27, 2004.
- Revision received November 22, 2004.
- Accepted December 14, 2004.
- ©2005 American Association for Cancer Research.