To comprehensively catalog CNAs in pancreatobiliary cancers, we carried out array CGH-based genomic profiling of a set of 37 cancers (31 exocrine pancreatic cancers and 6 distal bile duct cancers) expanded as xenografts to enrich for tumor cells, using cDNA microarrays representing ~22,000 genes with a median interprobe spacing of ~15 Kb. We identified numerous CNAs, among which 17 focal high-level DNA amplifications (i.e. fluorescence ratios ≥3, corresponding to at least 5-fold amplification 
) and 7 presumptive homozygous deletions (i.e. fluorescence ratios ≤0.25) were particularly informative in pinpointing known or novel candidate cancer genes (). By profiling gene expression in parallel, we also defined the subset of amplified genes exhibiting elevated expression (), a characteristic of oncogenes. For a subset of presumptive homozygous deletions, we validated homozygous loss by polymerase chain reaction (PCR) using human gene-specific primers ( and Figure S1
High-level amplifications and homozygous deletions.
Among the focal amplifications, we identified gain at 18q11.2 in 19% of pancreatobiliary cancers (5 of 31 pancreatic, and 2 of 6 bile duct). Notably, we found gains spanning 18q11.2 to be less common in other tumor types we had profiled on the same array platform, including cancers of the breast (3 gains in 89 (3%) tumors, 1 in 49 (2%) cell lines), prostate (0 in 64 (0%) tumors), lung (4 in 76 (5%) tumors, 4 in 52 (8%) cell lines) and colon (1 in 29 (3%) cell lines) 
(and unpublished data), and in these other tumor types the gains when present were not focal, suggesting the putative driver oncogene within this locus may be specific to pancreatobiliary cancer. Strikingly, the smallest shared region of amplification among the xenograft specimens spanned just two annotated genes, GATA6
(GATA binding protein 6) and CTAGE1
(cutaneous T-cell lymphoma (CTCL)-associated antigen 1) (). GATA6 
belongs to the GATA factor family of transcriptional regulators, whose members are expressed in distinct developmental and tissue-specific patterns and regulate cell-restricted programs of gene expression 
. Because GATA6
was known to regulate normal pancreas development 
, we sought to explore a possible functional connection of GATA6
gene amplification and pancreatobiliary cancer.
GATA6 is focally amplified in pancreatobiliary cancer.
A single bile duct cancer xenograft specimen (B291) with focal high-level DNA amplification was particularly informative in defining amplicon boundaries. Using a custom Agilent ultra-high definition CGH array with probes tiling 18q11.2 with an average 343nt spacing, we first confirmed the amplicon boundaries in B291, finding the amplicon peak indeed spanned just GATA6
(). We also validated GATA6
amplification in B291 by quantitative (Q)-PCR (), and by fluorescence in situ
hybridization (FISH) in the parent tumor from which the B291 xenograft was derived (, left
panel), the latter excluding the possibility of amplification arising during xenograft growth. Focal 18q11.2 gain was also present in 3 of 18 (17%) pancreatic cancer cell lines (AsPC1, Panc3.27 and Capan1) we had previously profiled by array CGH (
) (, and data not shown), a finding we confirmed by FISH ().
Consistent with an oncogenic role, GATA6
exhibited increased mRNA expression by microarray in B291 () and among the group of xenograft specimens with 18q11.2 gain (; P
0.003, Mann-Whitney U-Test), a finding also confirmed by Q-reverse transcription (RT)-PCR (). In contrast, expression of the neighboring gene CTAGE1
was not detectable by Q-RT-PCR (data not shown). We also observed increased GATA6 protein levels by Western blot in pancreatic cancer cell lines with 18q11.2 gain, compared to pancreatic cancer cell lines without gain or to the nontumorigenic human pancreatic ductal epithelial line HPDE (), and by immunohistochemistry (IHC) in the parent tumor from which the B291 xenograft was derived (). To assess the frequency with which GATA6 exhibited elevated expression in primary pancreatic cancer, we performed IHC on a tissue microarray (TMA) that included cases of normal pancreas, pancreatitis and pancreatic ductal adenocarcinoma. We observed moderate and strong GATA6 nuclear staining respectively in 15 (28%) and 25 (46%) of 54 primary pancreatic cancers compared to just 3 (9%) and 0 (0%) of 33 normal pancreas specimens surveyed (P
test) (). GATA6 expression was also elevated in pancreatitis (). There was no significant relation between GATA6 staining and tumor grade (P
GATA6 is overexpressed when amplified.
GATA6 is overexpressed in primary pancreatic tumors.
Since GATA6 is a transcriptional regulator, we sought to identify co-expressed genes, which might include its downstream transcriptional targets and suggest functional involvements. Using Significance Analysis of Microarrays (SAM) 
, we identified 86 genes whose expression was significantly (False discovery rate, FDR, <1%) increased (73 genes) or decreased (13 genes) in xenografts with elevated GATA6 expression (). The SAM-identified gene set spanned diverse biological processes, and included known cancer genes like FGF1
. Gene Set Enrichment Analysis (GSEA) 
confirmed an enrichment of putative upstream GATA factor binding sites among the genes whose expression correlated with elevated GATA6 levels (P
0.004) (). Interestingly, by GSEA the top functional gene sets associated with elevated GATA6 expression all related to mitochondrial activities connected to oxidative phosphorylation ().
To directly assess the functional significance of GATA6
amplification and overexpression in pancreatic cancer, we used RNA interference (RNAi) to target GATA6 knockdown in two pancreatic cancer cell lines, AsPC1 and Panc3.27, with GATA6
gain and overexpression. Transfection of two independent On-TARGETplus short interfering RNAs (siRNAs) targeting GATA6
, designed and chemically modified to minimize off-target effects 
, led to decreased GATA6 protein levels (), and to decreased cell proliferation compared to a negative control siRNA pool (). While the reduction in cell proliferation was relatively modest, it was statistically significant and reproducible in multiple independent experiments (not shown). In contrast, siRNA transfection of a pancreatic cancer cell line, PL45, without GATA6
amplification and overexpression () did not diminish cell proliferation (, right
panel), supporting the specificity of GATA6
targeting. We examined in more detail the effect of GATA6 knockdown in AsPC1 cells, where the reduced cell proliferation was attributable to decreased cell-cycle progression (as evidenced by decreased S-phase fraction; ) but not increased apoptosis (). GATA6 knockdown in AsPC1 cells also led to reduced colony formation in liquid culture ().
GATA6 amplification/overexpression contributes to cell proliferation.
In complementary experiments, we attempted to overexpress GATA6 by retroviral transduction in nontumorigenic human pancreatic ductal epithelial HPDE cells, and in the pancreatic cancer cell line PL45 harboring activated KRAS but no 18q11.2 gain. Though GATA6 expression was initially detected in infected cells by Western blot (data not shown), expression was lost upon expansion of cell pools under selection, suggesting GATA6 conferred negative fitness in these cell contexts.