Int. J.

Cancer: 112, 100 –112 (2004) Publication of the International Union Against Cancer

© 2004 Wiley-Liss, Inc.

ANALYSIS OF GENE EXPRESSION IN CANCER CELL LINES IDENTIFIES

CANDIDATE MARKERS FOR PANCREATIC TUMORIGENESIS AND
METASTASIS
Edoardo MISSIAGLIA1,2, Ekaterini BLAVERI1, Benoit TERRIS1,3, Yao-He WANG1, Eithne COSTELLO4, John P. NEOPTOLEMOS4,
Tatjana CRNOGORAC-JURCEVIC1 and Nicholas R. LEMOINE1*
1
Cancer Research U.K., Molecular Oncology Unit, Imperial College School of Medicine at Hammersmith Campus,
London, United Kingdom
2
Department of Pathology, Università di Verona, Verona, Italy
3
Hôpital Cochin, Service d’Anatomie Pathologique, Paris, France
4
Department of Surgery, University of Liverpool, Liverpool, United Kingdom

Pancreatic cancer is a highly aggressive type of malignancy
and the prognosis for disease presenting typically at a late
stage is extremely poor. A comprehensive understanding of
its molecular genetics is required in order to develop new
approaches to clinical management. To date, serial analysis
of gene expression and more recently oligo/cDNA microar-
ray technologies have been employed in order to identify
genes involved in pancreatic neoplasia that can be developed
as diagnostic markers and drug targets for this dismal dis-
ease. This study describes the expression profile obtained
from 20 pancreatic cell lines using cDNA microarrays con-
taining 9,932 human gene elements. Numerous genes were
identified as being differentially expressed, some of which
have been previously implicated in pancreatic adenocarci-
noma (S100P, S100A4, prostate stem cell antigen, lipocalin 2,
claudins 3 and 4, trefoil factors 1 and 2) as well as several novel
genes. The differentially expressed genes identified are in-
volved in a variety of cellular functions, including control of
transcription, regulation of the cell cycle, proteolysis, cell
adhesion and signaling. Validation of our array results was
performed by exploring the SAGEmap database and by im-
munohistochemistry for a selection of 4 genes that have not
previously been studied in pancreatic cancer: anterior gradi-
ent 2 homologue (Xenopus laevis), insulin-like growth factor
binding protein 3 and 4 and Forkhead box J1. Immunostaining
was performed using pancreas-specific tissue microarrays
containing core biopsies from 305 clinical specimens. In ad-
dition, using statistical group comparison and hierarchical
clustering, a selection of genes was identified that may be
linked to the site of metastasis from which these cell lines
were isolated.
© 2004 Wiley-Liss, Inc.
Key words: pancreatic cancer cell lines; microarrays; metastasis
Pancreatic cancer has the worst prognosis of all major cancers
and remains the fifth most common cause of cancer death in the
industrialized world. This is mainly due to late diagnosis (the
disease is often asymptomatic in its early stages and there is
presently a lack of sensitive diagnostic tools) and the aggressive-
ness of this type of tumor and its resistance to existing therapeutic
regimes.1,2
Due to the difficulties in obtaining large series of tissue speci-
mens from this type of tumor, human pancreatic carcinoma cell
lines may provide a useful additional model for the analysis of the
epithelial component of pancreatic neoplasia. Recent studies have
demonstrated that the prominent features in the gene expression
patterns still reflect the molecular signature of the tissue from
which the cells originated.3
In this study, we have therefore assessed expression profiles
across 20 pancreatic cell lines using cDNA microarray technology
in order to identify genes that are involved in the development of
pancreatic neoplasia.
To confirm expression of those genes at the protein level, tissue
microarray (TMA) technology was used in 290 primary pancreatic
adenocarcinomas and 15 specimens from normal pancreatic tis-
sues. Thus, 2 high-throughput array technologies were exploited in
order to identify rapidly potentially important biomarkers involved
in pancreatic carcinogenesis. Furthermore, genes that could be
predictive of the potential of pancreatic cancer cells to metastasize
to lymph nodes were identified by means of statistical group
comparison and hierarchical clustering.
MATERIAL AND METHODS
Cell culture and RNA isolation
Nineteen pancreatic cancer cell lines were analyzed in this
study: A818-4,4 AsPC-1,5 Capan-2,6 CFPAC-1,7 Colo-357,8 FA6,9
HPAF-II,10 Hs766T,11 IMIM-PC-2,12 MDAPanc-3,13 MiaPaCa-
2,14 PaCa-3,15 Panc-1,16 PaTu-I, PaTu-II,17 PT45,18 RWP,19 SUIT-
220 and T3M4.21 These cell lines were obtained from Cancer
Research U.K. cell production services (London Research Institute
Clare Hall Laboratories, Potters Bar, U.K.) and cultured in E4
medium supplemented with 10% heat-inactivated fetal calf serum
(Gibco-Life Technologies, Paisley, U.K.) until harvest when they
reached 80% confluence. Twenty-four hours before the cells were
harvested for RNA isolation, the medium was replaced with E4
supplemented with 0.5% fetal calf serum.
A human pancreatic duct epithelial cell line (HPDE) immortal-
ized by transduction with the E6/E7 genes of human papilloma
virus 16 (HPV16) was kindly provided by Dr. Ming-Sound Tsao
(University of Toronto, Ontario, Canada). This cell line was grown
in serum-free keratinocyte medium as described.22 To account for
the contribution of biologic sources of variation, each cell line was
grown in 2 independent cultures and each array experiment was
carried out independently on RNA extracted individually from
each cell culture. All RNA samples were isolated using TRIzol
reagent from Sigma (St. Louis, MO).
Grant sponsor: Cancer Research U.K.; Grant sponsor: E.U. Biomed;
Grant number: QLG1-CT-2002-01196/PACA Targets; Grant sponsor:
Fondazione Giorgio Zanotto, Verona, Italy; Grant sponsor: Associazione
Italiana Ricerca sul Cancro, Milan, Italy.
The first 2 authors contributed equally to this paper.
*Correspondence to: Cancer Research U.K., Molecular Oncology Unit,
Department of Cancer Medicine, Imperial College School of Medicine,
Hammersmith Campus, London W12 0NN, United Kingdom.
Fax: ⫹44-208-383-1708. E-mail: nick.lemoine@cancer.org.uk
Received 16 December 2003; Accepted after revision 6 April 2004
DOI 10.1002/ijc.20376
Published online 2 June 2004 in Wiley InterScience (www.interscience.
wiley.com).
 PANCREATIC TUMORIGENESIS AND METASTASIS
Microarray hybridization
The cDNA microarrays used in this study were 10K Sanger
Human Arrays version 1.2.1 and were obtained through the Cancer
Research U.K.-Ludwig Institute-Wellcome Trust consortium. The
glass arrays were manufactured and quality-controlled at the
Sanger Center (Cambridge, U.K.). Each slide contained 9,932
human gene elements: 8,817 cDNAs derived from the Human
Genome Mapping Project (HGMP) IMAGE clone collection, 647
cDNAs derived from the Research Genetics IMAGE collection
and 468 chromosome 22 gene-specific PCR products. The spotting
patterns and the complete annotated list of these cDNAs are
available at the CRUK Microarray Web site (http://www.sanger.
ac.uk/Projects/Microarrays/informatics/hver1.2.1.shtml).
Labeling of 50 ␮g of total RNA was achieved by direct incor-
poration of Cy5-dCTP or Cy3-dCTP (Amersham, Buckingham-
shire, U.K.) in a reverse transcription reaction using anchored
oligodeoxythymidylate primers (Cancer Research U.K. Oligonu-
cleotide Service, London Research Institute Clare Hall Laborato-
ries) and Superscript II reverse transcriptase (Gibco). The details
of the hybridization and washing protocols are available online
(http://www.cgal.icnet.uk/exprotocols/protocols.html).
The same complementary DNA derived from the HPDE cell
line was used as the control sample in all hybridizations. For each
experiment, the Cy5-dCTP-tagged cDNA from an individual pan-
creatic cancer cell line was mixed with Cy3-dCTP-tagged cDNA
from HPDE cells and subsequently cohybridized to a microarray.
All the experiments were performed in duplicate.
Following hybridization, arrays were scanned using an Af-
fymetrix 428 dual-laser microarray (Affymetrix, High Wycombe,
U.K.) and separate images were acquired for Cy3 and Cy5 fluo-
rescence. All raw data files are available at http://sci.
cancerresearchuk.org/axp/mphh/ijc04/.
Image and data analysis
The signal intensity values of each element were extracted using
the ImaGene v 4.2 software program (BioDiscovery, Los Angeles,
CA). The data were normalized using a curve-fitting transforma-
tion23 (marrayNorm package) and a variance-stabilizing transfor-
mation24 (VSN package) in R.25
The log square ratio of the normalized values from both the Cy5
and Cy3 channels was exported and used for further analysis in
order to determine differentially expressed genes. Only genes with
expression changes of at least 4-fold in more than 5 out of 19 cell
lines were considered as differentially expressed between control
and tumor samples.
In addition, a one-sample Student’s t-test was performed for
each of these genes in the replicate data to test whether the mean
normalized expression ratio for the gene is statistically different
between HPDE and cancer cell lines. Genes with no significant
p-value were excluded as they were not considered to demonstrate
consistent differential expression.
Statistical class comparison
The data were also analyzed using BRB-ArrayTools (version
3.0; http://linus.nci.nih.gov/BRB-ArrayTools.html) developed by
Dr. Richard Simon and Amy Peng in order to identify genes
differentially expressed between cell lines of different origin. The
groups consisted of cell lines derived from lymph node metastasis
(Colo357, Hs766T and T3M4), primary tumors (PT45, Capan2,
FA6, IMIM-PC-2, MiaPaca2, Paca3, Panc1, PaTu1 and PaTu2),
liver metastases (CFPAC-1, MDAPanc-3, RWP and SUIT-2) and
ascites (A818-4, AsPC-1 and HPAF-II). We performed an F-test
using the randomized variance model26 in order to improve the
statistical power of the test for detecting differentially expressed
genes, while random permutations of the class labels were em-
ployed to assess the reliability of the p-value. Moreover, a multi-
variate permutation test was used to evaluate the false discovery
rate in order to control for multiple testing errors.27
101
The hierarchical cluster analysis tool from gCLUTO was used to
cluster both cell lines and genes. In order to remove the noise
generated by genes that are not differentially expressed between
cell lines of different origin, we selected genes that showed sta-
tistically significant differences in the class comparison analysis of
both normalized data sets. Genes and cell types were clustered by
average linkage and correlation coefficient metrics using an ag-
glomerative method. The gene dendrogram was cut at the point
that generates 10 partitions in order to produce clusters of genes
that share similar expression profiles among the cell lines under
analysis.
Gene ontology analysis
In the first step, all the terminal nodes belonging to the biologic
process associated with the genes present in the microarray were
obtained from the Human Ensembl database (release 15.33.1).
Then, for each gene, we recovered all the terms between the nodes
and the end node (biologic process) following the term-to-term
relationships present in the Gene Ontology database (July 2003
release) using a Perl script available on request. Terms that were
represented more then 1,000 times in the microarray, obsolete
terms and the “biologic process unknown” term were not consid-
ered useful for discrimination and eliminated.
A binary matrix (0, absent call; 1, present call) was built using
the identified biologic process terms (columns) and the genes used
in the expression clustering analysis for which biologic process
terms were available (rows). This matrix was analyzed in the
gCLUTO program with cosine as distance, repeated bisection as
method and 7 partitions in order to cluster together genes involved
in similar biologic processes.
We employed this strategy to evaluate possible relationships
between the gene ontology and gene expression partitions that
might identify group of genes sharing expression profile as well as
biologic process. The presence of statistically over- and underrep-
resented associations were estimated using Fisher’s exact test.
In parallel, since the gene expression cluster generated 3 major
groups of genes, we analyzed them separately to identify over- or
underrepresented biologic process terms among their genes using
the FatiGO Data mining program28 (http://fatigo.bioinfo.cnio.es/).
The comparison was done at different levels using 1,000 randomly
selected genes from our microarray after removing the genes under
analysis.
Serial analysis of gene expression (SAGE)
We searched the SAGEmap database (http://www.ncbi.nlm.
nih.gov/SAGE) using the Gene-to-Tag online analysis tools. This
tool allows exploration of the expression levels of selected SAGE
tags in all presently available SAGE libraries. The analysis was
focused on SAGE libraries of 4 pancreatic cancer cell lines (CA-
PAN1, CAPAN2, HS766T and Panc1), 2 primary pancreatic ad-
enocarcinomas (Panc 91-16113 and Panc 96-6252) and 2 short-
term cultures of normal pancreatic ductal epithelium (HX and
H126).
Immunohistochemical analysis
Immunohistochemical analysis was performed on 3 formalin-
fixed, paraffin-embedded pancreas-specific tissue microarrays (PT-
MAs) containing a total of 305 cores (290 ductal pancreatic
adenocarcinomas and 15 normal pancreata). The construction of
the TMAs was performed using a tissue arrayer (Beecher Instru-
ments, Silver Spring, MD) as described previously by Kallioniemi
et al.29 The tissue sections were deparaffinized and rehydrated by
standard procedures and were treated with 1% hydrogen peroxide
to inactivate endogenous peroxidase. Sections were blocked in 5%
normal serum for 30 min and incubated for 1 hr at room temper-
ature with the appropriate primary antibody. The antibody used
against AGR2 (1:10 dilution) was kindly provided by Dr. D.A.
Thompson.30 A monoclonal antibody against hepatocyte nuclear
factor 3/forkhead homologue 4 (FOXJ1; 1:10 dilution) was ob-
tained from NeoMarkers (Fremont, CA). In addition, immunohis-
 102

TABLE I – GENES UPREGULATED IN PANCREATIC CARCINOMA CELL LINES

Ensembl ID Gene Product Subcellular location Biological function Average ratio t-test p-value Reference SAGE

ENSG00000106541 AGR2 Anterior gradient 2 homologue Secreted Regulation of receptor adhesion and 9.6 ⬍0.001 36 confirmed
functioning
ENSG00000023445 BIRC3 Baculoviral IAP repeat containing 3 Cytoplasmic Antiapoptosis 4.9 ⬍0.001 31 confirmed
ENSG00000125378 BMP4 Bone morphogenic protein 4 Secreted Mesoderm development 116.8 ⬍0.01 confirmed
ENSG00000038427 CSPG2 Chontroitin sulfate proteoglycan 2 Extracellular Heterophilic cell adhesion 24.2 ⬍0.05 31 confirmed
matrix
2
ENSG00000014914 CRA Cisplatin resistance associated beta Integral 3.9 ⬍0.001 37 confirmed
protein membrane
2
Homologous to CLDN3 Claudin 3 Integral Promotes activation of pro-MMP2 22.3 ⬍0.001 36 confirmed
membrane
2
Homologous to CLDN4 Claudin 4 Integral Promotes activation of pro-MMP2 17.3 ⬍0.001 36, 39 confirmed
membrane
ENSG00000184113 CLDN5 Claudin 5 Integral Promotes activation of pro-MMP2 6.7 ⬍0.001 Not confirmed
membrane
3
Homologous to CRIP1 Cystein-rich intestinal protein 1 Cytoplasmic Cell proliferation 12.3 ⬍0.001 36 confirmed
ENSG00000129654 FOXJ1 Forkhead box J1 Nuclear Regulation of transcription, DNA- 29.6 ⬍0.05 Not confirmed
dependent
ENSG00000131981 LGALS3 Galectin-3 Nuclear/ Heterophilic cell adhesion 3.2 ⬍0.001 39 confirmed
cytoplasmic
ENSG00000115738 ID2 Id-2 Nuclear Cell proliferation 18.8 ⬍0.001 39 Not confirmed
ENSG00000146674 IGFBP3 Insulin-like growth factor binding Secreted Regulation of cell growth 63.7 ⬍0.01 37, 38 confirmed
protein 3
ENSG00000141753 IGFBP4 Insulin-like growth factor binding Secreted Regulation of cell growth 3.6 ⬍0.001 37 confirmed
protein 4
ENSG00000167779 IGFBP6 Insulin-like growth factor binding Secreted Regulation of cell growth 3.1 ⬍0.001 Not confirmed
protein 6
ENSG00000082781 ITGB5 Integrin, beta 5 Membrane Integrin-mediated signaling pathway 4.4 ⬍0.001 35 confirmed
MISSIAGLIA ET AL.

ENSG00000167755 KLK6 Kallikrein 6 Secreted Proteolysis and peptidolysis 23.7 ⬍0.01 confirmed
ENSG00000148346 LCN2 Lipocalin 2 Cytoplasmic Transport of small lipophilic substances 11.7 ⬍0.001 31, 38, 39 confirmed
ENSG00000175130 MLP MARCKS-like protein Cytoplasmic Coupling the protein kinase c and 4.4 ⬍0.001 37 Not confirmed
calmodulin signal transduction
ENSG00000163460 MUC1 Mucin 1 Membrane Cell-cell interactions 61.9 ⬍0.01 37 confirmed
ENSG00000166741 NNMT Nicotinamide N methyltransferase Cytoplasmic Biotransformation of many drugs 29.1 ⬍0.05 31 confirmed
ENSG00000104368 PLAT Plasminogen activator Secreted Proteolysis and peptidolysis 5.7 ⬍0.01 31 confirmed
ENSG00000167653 PSCA Prostate stem cell antigen Membrane Tumor antigen 25.2 ⬍0.001 31 confirmed
ENSG00000118849 RARRES1 Retinoic acid receptor responder 1 Membrane Negative regulation of cell proliferation 43.0 ⬍0.01 Not confirmed
ENSG00000160673 S100A4 S100 calcium-binding protein A4 Cytoplasmic Invasive growth 181.5 ⬍0.01 36 confirmed
ENSG00000163993 S100P S100 calcium-binding protein P Cytoplasmic Cell life, proliferation/growth 13.9 ⬍0.001 31, 35, 37, confirmed
38, 39
ENSG00000118785 SPP1 Secreted phosphoprotein 1 Extracellular Antiapoptosis 16.2 ⬍0.01 31 confirmed
(osteopontin) matrix
ENSG00000131171 SH3BGRL SH3 domain binding glutamic acid- Cytoplasmic Protooncogene 4.4 ⬍0.001 Not confirmed
rich-like
ENSG00000155465 SLC7A7 Solute carrier family 7, member 7 Integral Protein complex assembly 13.1 ⬍0.01 31 Not confirmed
membrane
ENSG0000035862 TIMP2 Tissue inhibitor of metalloproteinase 2 Extracellular Complexes with metalloproteinases 4.3 ⬍0.001 confirmed
matrix
ENSG00000101418 TGM2 Transglutaminase 2 Membrane Protein modification 41.0 ⬍0.001 31, 38 confirmed
ENSG00000160182 TFF1 Trefoil factor 1 Secreted Cell growth and/or maintenance 6.7 ⬍0.01 39 confirmed
ENSG00000160181 TFF2 Trefoil factor 2 (spasmolytic protein1) Secreted Stimulator of cell motility 30.1 ⬍0.01 39 confirmed
ENSG00000026025 VIM Vimentin Kinesin 21.7 ⬍0.01 35 confirmed
complex
ENSG00000101443 WFDC2 WAP 4 disulfide domain 2/HE4 Secreted Proteolysis and peptidolysis 35.2 ⬍0.001 confirmed
1
No HUGO symbol.–25⬘ and 3⬘ sequences available.–35⬘ and 3⬘ sequence available only.
 TABLE II – GENES DOWNREGULATED IN PANCREATIC CANCER CELL LINES

Average
Ensembl ID Gene Product Subcellular location Biological function t-test p- Reference
ratio value
ENSG00000162772 ATF3 Activating transcription factor 3 Nucleus Regulation of transcription, DNA- 0.21 ⬍0.001
dependent
ENSG00000169594 BNC Basonuclin Nucleus Regulation of transcription, DNA- 0.09 ⬍0.001
dependent
1
Homologous to BNIP3 BCL2/adenovirus E1B 19kDa interacting protein 3 Mitochondrion Antiapoptosis 0.44 ⬍0.001 34
ENSG00000151914 BPAG1 Bullous pemphigoid antigen Intercellular junction Cell adhesion and cell cycle arrest 0.10 ⬍0.001
ENSG00000151465 C10orf7 Protein D123 Cell cycle arrest 0.22 ⬍0.001
ENSG00000118971 CCND2 Cyclin D2 Nucleus Regulation of cell cycle 0.01 ⬍0.001 39
ENSG00000062038 CDH3 Cadherin 3 Integral to Cell adhesion 0.08 ⬍0.001 33, 39
membrane
ENSG00000124762 CDKN1A Cyclin-dependent kinase inhibitor 1A (p21, Cip1) Nucleus Cell cycle arrest 0.30 ⬍0.001
ENSG00000084636 COL16A1 Collagen, type XVI, alpha 1 Collagen Cell adhesion 0.21 ⬍0.001
ENSG00000114270 COL7A1 Collagen, type VII, alpha 1 Basement membrane Cell adhesion 0.32 ⬍0.001
ENSG00000175183 CSRP2 Cysteine and glycine-rich protein 2 Nucleus Cell proliferation 0.24 ⬍0.001
ENSG00000121552 CSTA Cystatin A (stefin A) Epidermal development and 0.22 ⬍0.001 39
maintenance
ENSG00000142871 CYR61 Cysteine-rich, angiogenic inducer, 61 Extracellular Cell proliferation 0.38 ⬍0.001
ENSG00000170477 EBS2 Keratin 4 Intercellular junction Cytoskeleton organization and 0.12 ⬍0.001
biogenesis
ENSG00000134531 EMP1 Epithelial membrane protein 1 Integral to Cell death 0.30 ⬍0.001
membrane
2
ENSG00000086570 FAT2 FAT tumor suppressor homolog 2 (Drosophila) Integral to Homophilic cell adhesion 0.19 ⬍0.001
membrane
ENSG00000138829 FBN2 Fibrillin 2 Extracellular matrix Formation/maintenance of extracellular 0.21 ⬍0.001
microfibrils
ENSG00000161958 FGF11 Fibroblast growth factor 11 Extracellular space Cell-cell signaling 0.28 ⬍0.001
ENSG00000182106 GIP2 Interferon, alpha-inducible protein (clone IFI-15K) Extracellular space Cell-cell signaling 0.34 ⬍0.001
ENSG00000116717 GADD45A Growth arrest and DNA-damage-inducible, alpha2 Nucleus Cell cycle arrest 0.21 ⬍0.001
ENSG00000152661 GJA1 Gap junction protein, alpha 1, 43kDa (connexin 43) Connexon complex Cell-cell signaling 0.25 ⬍0.001
ENSG00000063660 GPC1 Glypican 1 Extracellular matrix Cell division and growth regulation 0.36 ⬍0.001 39
ENSG00000117318 ID3 Inhibitor of DNA binding 3 Nucleus Development 0.45 ⬍0.001
ENSG00000163453 IGFBP7 Insulin-like growth factor binding protein 7 Extracellular Negative regulation of cell 0.42 ⬍0.001
proliferation
ENSG00000125538 IL1B Interleukin 1, beta Extracellular space Negative regulation of cell 0.18 ⬍0.001
proliferation
ENSG00000185479 KRT6E Keratin 6E Intercellular junction Cytoskeleton organization and 0.14 ⬍0.001
PANCREATIC TUMORIGENESIS AND METASTASIS

biogenesis
ENSG00000178939 LGALS7 Lectin, galactoside-binding, soluble, 7 (galactin 7) Extracellular space Heterophilic cell adhesion 0.16 ⬍0.001
ENSG00000169688 MT1J Metallothionein 1J Cytoplasm Modulate apoptosis 0.22 ⬍0.001
ENSG00000125144 MT1K Metallothionein 1K Modulate apoptosis 0.31 ⬍0.001
ENSG00000087250 MT3 Metallothionein 3 Cell proliferation 0.24 ⬍0.001
ENSG00000112245 PTP4A1 Protein tyrosine phosphatase type IVA, member 1 Nucleus Protein amino acid dephosphorylation 0.35 ⬍0.001
ENSG00000160675 S100A2 S100 calcium-binding protein A2 Tumor suppressor function 0.17 ⬍0.001
ENSG00000163220 S100A9 S100 calcium binding protein A9 (calgranulin B) Extracellular space Cell-cell signaling 0.20 ⬍0.001
ENSG00000166401 SERPINB2 Serine proteinase inhibitor, clade B (ovalbumin), member 2 Antiapoptosis 0.15 ⬍0.001
ENSG00000175793 SFN Stratifin Extracellular space Regulation of cell cycle 0.21 ⬍0.001
ENSG00000129194 SOX15 SRY-box 15 Chromatin Regulation of transcription, DNA- 0.07 ⬍0.001 39
dependent
ENSG00000113140 SPARC Secreted protein, acidic, cysteine-rich (osteonectin) Basement membrane Collagen binding 0.08 ⬍0.001
ENSG00000105825 TFP12 Tissue factor pathway inhibitor 2 Extracellular matrix Blood coagulation 0.44 ⬍0.001
Homologous to THBS22 Thrombospondin 2 Extracellular matrix Cell adhesion 0.06 ⬍0.001 39
ENSG00000073282 TP73L Tumor protein p73-like Nucleus Induction of apoptosis 0.19 ⬍0.001
ENSG00000008838 TRAP1003 Thyroid hormone receptor-associated protein Nucleus Regulation of transcription, DNA- 0.23 ⬍0.001
dependent
1
5⬘ and 3⬘ sequences available.–25⬘ and 3⬘ sequence available only.–3No HUGO symbol.
103
104 MISSIAGLIA ET AL.
tochemistry (IHC) was performed for 2 members of the family of
insulin-like growth factor binding proteins, IGFBP3 (1:900 dilu-
tion; Upstate Biotechnology, Upstate, NY) and IGFBP4 (1:40
dilution; R&D Systems, Minneapolis, MN). After application of
primary antibody for 30 min at room temperature, the slides were
washed 3 times in phosphate-buffered saline (PBS). The sections
were then incubated for 30 min with the appropriate biotinylated
secondary antibody followed by 3 washes in PBS, and then for
another 30 min with a streptavidin/horseradish peroxidase com-
plex followed by application of diaminobenzidine chromogen. The
slides were finally counterstained with hematoxylin and cover-
slipped. All the reagents used were obtained from Dako (Carpin-
teria, CA).
Scoring of IHC results
The immunohistochemical results were reviewed by specialist
pathologists (N.R.L. and Y.-H.W.). Both intensity of staining and
the percentage of positively stained tumor cells (extent) were
recorded. A graded score was obtained by multiplying the intensity
of staining on an arbitrary 0 –3 scale by the extent of staining on a
1– 4 scale (where 1 is ⬍ 25%; 2, 26 –50%; 3, 51–75%; and 4, ⬎
76% of stained tumor cells). A score of 0 – 4 corresponds to low
immunoreactivity, a score of 5– 8 to moderate and 9 –12 to high
immunoreactivity. Statistical analysis was performed using Fish-
er’s exact test in Stata7 software (Stata, College Station, TX).
RT-PCR
cDNAs from 20 pancreatic cell lines were synthesized from 1
␮g of total RNA using an oligo dT primer and the Superscript II
reverse transcription kit (Gibco). The primers used were LIMK2
sense, 5⬘-TCCTCCTCCCCATTTCCG-3⬘; LIMK2 antisense,
5⬘-AGGCCCTGTCATCAGCAGG-3⬘; STC-1 sense, 5⬘-TGAGGT-
CGTCCAGCTGCCCAATC-3⬘; STC-1 antisense, 5⬘-GGCACAG-
TGGTCTGTCTGCAGGATG-3⬘; P8 sense, 5⬘-GGCACGATGGC-
CACCTTCCCACC-3⬘; P8 antisense, 5⬘-CTCATCTCCAGC-
TCTGTCTCAGCG-3⬘.
These amplify a 260 bp LIMK2 fragment, a 172 bp STC-1
fragment and a 273 bp P8 fragment. One ␮l of the cDNA was used
for each PCR in 20 ␮l reaction volume containing 200 ␮M dNTPs,
1 ␮M sense and antisense primers, 1.25 Mm MgCl2, 2 ␮l of 10 ⫻
PCR buffer and 0.5 unit of Taq DNA polymerase (Roche Diag-
nostics, East Sussex, U.K.). PCR was performed on a DNA ther-
mal cycler (MJ Research, Waltham, MA) and the PCR conditions
for LIMK2 and P8 were as follows: 1 cycle of denaturation at 95°C
for 1 min, followed by 33 cycles of 95°C for 1 min, 55°C for 1 min
and 72°C for 1 min before a final incubation at 72°C for 10 min.
The PCR conditions were the same for STC-1, except that the
primers were annealed at 62°C. The 18S primers were used as the
endogenous control. The PCR products were analyzed by electro-
phoresis on a 2% agarose gel, followed by staining with ethidium
bromide.
RESULTS
Differential gene expression
The gene expression profile of each of the 19 pancreatic cancer
cell lines was compared to that of the immortalized human pan-
creatic ductal epithelial cell line HPDE. In total, 173 clones (of
which 170 are unique genes) were overrepresented more than
4-fold in at least 5 out of 19 cancer cell lines as well as statistically
significant using a one-sample t-test, while 177 clones (167 unique
genes) were underrepresented using the same criteria. Tables I and
II represent a selection of those genes, while full lists of all
expressed genes are available online (http://sci.cancerresearchuk.
org/axp/mphh/ijc04/).
In order to obtain independent validation of our results at the
RNA level, we used the Gene-to-Tag online analysis tool (http://
www.ncbi.nlm.nih.gov/SAGE) to compare the differential gene
expression profiles with SAGE data and found concordance in
about 75% of cases (data not shown). The Stanford Online Uni-
versal Resource for cDNA Clones and expressed sequence tags
(ESTs) (SOURCE; http://genome-www.stanford.edu/source) was
also mined for additional gene information and where the gene has
been found differentially expressed in previous studies using mi-
croarray technology, the reference is cited in the tables.
Genes with a wide variety of functions were identified among
the overrepresented transcripts, ranging from tight junction pro-
teins (claudins 3, 4 and 5), ion homeostasis regulators (S100P,
S100A4, cysteine-rich heart protein), transcription factors (fork-
head box J1, Id2), to extracellular matrix proteins (MMP2, MMP7,
TIMP2, plasminogen activator). Equally, among the underrepre-
sented genes there are several putative tumor suppressor genes
(FAT tumor suppressor homologue 2, IGFBP7, S100A2, TP55) as
well as cell cycle-related gene GADD45A and several cell adhe-
sion genes (cadherin 3, cysteine-rich angiogenic inducer 61, pla-
kophilin 1).
Validation by IHC on PTMAs
Immunohistochemical analysis on PTMAs comprising in total
305 pancreatic cancer and normal cores was used to correlate the
level of protein expression for selected genes (anterior gradient 2
homologue, AGR2, insulin-like growth factor binding protein 3
and 4, and forkhead box J1). For all these proteins, the normal
pancreatic epithelium showed absent or very weak staining (Fig.
1a, c, e and g), while the majority of the scoreable tumor cores
demonstrated elevated levels of expression (Table III, Fig. 1b, d, f
and h).
Expression of AGR2 was found in all of the scoreable tumor cores
and was scored as high in the majority (63%). Although this protein
was mainly localized in the cytoplasm, there were a few cases show-
ing intense nuclear immunoreactivity (Fig. 1b). Immunoreactivities
for both insulin-like growth factor binding proteins 3 and 4 were
confined to the cytoplasm with weak or moderate intensity for the
majority of the cancer cores (Fig. 1d and f). The forkhead box J1
(FOXJ1) protein was localized mainly in the cytoplasm but occasion-
ally showed some nuclear staining (Fig. 1h).
Identification of metastasis-related genes
Using the lowess-normalized and variance stabilization-trans-
formed datasets, the class comparison analysis identified, respec-
tively, 175 and 221 genes to be differentially expressed between
cell lines derived from different sites of origin (p ⬍ 0.001 in the
univariate test, and a maximum allowed number of false positive
genes ⫽ 1). A total of 123 genes (representative of 80 unique
genes and 21 ESTs) were differentially expressed independently
from the normalization applied, and these were used for further
analysis.
Expression clustering
As expected, the hierarchical clustering performed using these
selected subsets of genes was able to separate the cell lines from
different origins, with the exception of RWP1 (originated from
liver metastases), which fell into the cluster with cell lines origi-
nated from ascites. All the replicates were close to each other, as
well as the cell lines PaTu1 and PaTu2 (derived from the same
patient), indicating the overall reproducibility of the array tech-
nique and clustering analysis (Fig. 2).
Restricting the dendrogram to 10 clusters of genes produced 3
major groups with different profile patterns between cell lines.
Cluster A represents genes more expressed in cell lines originating
from primary tumors and some from liver metastases, cluster B
FIGURE 1 – Validation by IHC on pancreatic tissue microarray of 4
genes found to be overexpressed in pancreatic cancer cell lines: AGR2
(human homologue of the Xenopus laevis cement gland gene XAG-2)
in normal pancreas (a) and pancreatic cancer (b); IGFBP3 in normal
pancreas (c) and pancreatic cancer (d); IGFBP4 in normal pancreas (e)
and pancreatic cancer (f); FOXJ1 in normal pancreas (g) and pancre-
atic cancer (h).
FIGURE 1.
106 MISSIAGLIA ET AL.

TABLE III – FREQUENCY OF SAMPLES THAT HAD WEAK (1– 4), MODERATE (5– 8) AND INTENSE (9 –12) SCORE OR SHOWED NO IMMUNOREACTIVITY
IN IHC ANALYSIS WITH APPROPRIATE ANTIBODIES FOR 4 PROTEINS WHOSE, RNA LEVELS WERE INCREASED IN PANCREATIC
CANCER CELL LINES IN THE CDNA MICROARRAY ANALYSIS

Number of
Protein Tissue type Low Moderate High Negative p-value
tissue cores

AGR2 Tumor 51 (30%) 13 (7%) 107 (63%) 0 (0%) 171

Normal 0 (0%) 0 (0%) 0 (0%) 10 (100%) 10 ⬍0.0001
IGFBP3 Tumor 113 (55%) 57 (27%) 27 (13%) 11 (5%) 208
Normal 0 (0%) 0 (0%) 0 (0%) 10 (100%) 10 ⬍0.0001
IGFBP4 Tumor 50 (29%) 59 (34%) 63 (36%) 0 (0%) 172
Normal 10 (100%) 0 (0%) 0 (0%) 0 (0%) 10 ⬍0.0001
FOXJ1 Tumor 17 (20%) 15 (18%) 50 (60%) 2 (2%) 84
Normal 1 (10%) 0 (0%) 0 (0%) 9 (90%) 10 ⬍0.0001
The number of cancer cores scored was determined by those that remained intact through the slide processing and actually contained clearly
identifiable tumor tissue.
IHC was performed on TMAs consisting of tissue cores from clinical pancreatic adenocarcinoma specimens (numbers in parentheses indicate
percentage of total).

FIGURE 2 – Cluster of expression of 123 features in 19 pancreatic cancer cell lines. Each row represents an individual clones and each column a cell
line. The log ratio of the abundance of transcripts of each gene to its median abundance across all cell lines is represented by color: green (transcript
level below median), white (equal to median), red (greater than median). Color intensity reflects the magnitude of the ratio. Capital letters indicate the
3 major cluster of clones obtained using 10 partitions in the dendogram of clones. The dendrogram of cell lines has a branch that includes only cell
lines derived from lymph node metastases of pancreatic adenocarcinomas (in green; lymph node metastasis cluster). A group of clones (named cluster
C) can be seen that shows increased expression in these cell lines compared to the rest (metastasis-related gene cluster). Cell lines derived from liver
metastases are highlighted in red, the ones from ascitic cells in pink, while the ones from primary tumors in blue.

those upregulated in cell lines from ascites and liver metastases, Gene ontology
while cluster C contains genes overexpressed in cell lines from Cluster analysis was applied to the lists filtered for gene ontol-
lymph node metastases. ogy in order to generate groups of genes that belong to similar
 PANCREATIC TUMORIGENESIS AND METASTASIS 107

biologic processes. The results are represented in Figure 3 and DISCUSSION

summarized in Table IV.
A significant correlation was found between gene ontology
clusters and expression clusters (p ⬍ 0.05). This observation was
due to the presence of several genes in expression cluster A that
also belong to the gene ontology cluster for cell proliferation and
cell cycle terms. The analysis by FatiGO confirmed the overrep-
resentation of these 2 terms in cluster A with a significant adjusted
p-value of 0.002. No association was observed between cluster
expression B and any of the gene ontology clusters, while FatiGO
found the terms “lipid metabolism” and “catabolism” statistically
overrepresented, but not after adjustment.
In cluster C, terms such as “response to extracellular stimulus,”
“response to biotic stimulus,” “catabolism” and “cell-matrix adhe-
sion” were again found significant with the FatiGO tool only at the
unadjusted level. In the same cluster, 8 genes were connected to
the “cell communication” term, but were also associated with other
terms such as “transport” or “internal signaling cascade.”
RT-PCR
The different expression levels of P8 (candidate of metastasis
1), Stanniocalcin 1 (STC1) and LIM kinase 2 (LIM2) observed in
microarray analysis between cell lines originated from different
sites was confirmed by RT-PCR (Fig. 4). It is evident that all 3
genes show a very high level of expression in cell lines derived
from lymph node metastases.
In this study, we attempted to identify potential biomarkers for
pancreatic cancer by analyzing pancreatic cancer cell lines using
high-density cDNA microarrays. The comparison of gene expres-
sion between cell lines of different origin was also applied to
identify genes potentially involved in the tumor progression to
metastasis.
One of the advantages of using cancer cell lines is that pure
tumor cells are tested without any contamination from surrounding
stromal elements. Genes identified as differentially expressed in
this study were assessed further for differential expression by
mining the SAGE and SOURCE databases as well as the wealth of
microarray results that are now publicly available. Importantly, the
expression signatures for the pancreatic cell line panel showed
similarity to those of clinical specimens of pancreatic neoplasia
despite differences in expression profiling platforms used. A num-
ber of genes previously reported as upregulated in pancreatic
adenocarcinoma31–38 and preinvasive tumoral lesions (IPMTs)39
such as S100P, S100A4, claudin 4, PSCA, lipocalin 1, TFF1 and 2
and galectin 3 were also found to be more abundant in the cancer
cell lines than in the immortalized HPDE pancreatic cell line. Note
that while the HPDE line is not fully normal, a comprehensive
analysis has shown that the cells have a near-normal genotype and
phenotype except for the loss of the p53 functional pathway.40 This
supports the validity of using cancer cell lines as models of

FIGURE 3 – Cluster of the 218 gene ontology terms related to the 56 genes differentially expressed for which informative biologic process terms
were available. Only the first 2 gene ontology terms are reported in order to describe the biologic processes that better represent the genes
belonging to the 7 partitions of the cluster. Cell prolif, cell proliferation; transp, transport; sig trans, signal transduction; nucl met, nucleobase,
nucleoside, nucleotide and nucleic acid metabolism; reg trans, regulation of transcription DNA dependent; cell comm, cell communication; carb
acid met, carboxylic acid metabolism; org acid met, organic acid metabolism; prot met, protein metabolism; catab, catabolism.
108 MISSIAGLIA ET AL.

TABLE IV – DIFFERENTIALLY EXPRESSED CLONES IDENTIFIED BY CLASS COMPARISON ANALYSIS BETWEEN CELL LINES OF DIFFERENT ORIGIN

HUGO gene Adjusted Lymph node Primary

Clone ID Ensembl ID Cluster1 GO cluster2 Ascites4 Liver4
symbol p-value3 metastasis4 tumor4

110503_A FOSL16 A ⬍0.001 0.42 0.79 0.50 1.21

121357_A ENSG00000100526 CDKN3 A Cell cycle/cell prolif ⬍0.001 0.77 1.02 0.46 1.28
124345_A ENSG00000117724 CENPF A Cell cycle/cell prolif ⬍0.001 0.79 1.07 0.76 1.40
1353930_A ENSG00000171608 PIK3CD A Transp/sig trans ⬍0.001 0.62 1.09 0.81 1.55
1894964_A ENSG00000115163 CENPA A Nucl met/trans reg ⬍0.001 0.50 0.65 0.41 0.85
259596_A ENSG00000126787 DLG7 A Cell comm/sig trans ⬍0.001 0.71 0.93 0.54 1.39
292577_A ENSG00000117650 NEK2 A Cell cycle/cell prolif ⬍0.001 0.71 1.09 0.60 1.28
293274_A ENSG00000100526 CDKN3 A Cell cycle/cell prolif ⬍0.001 0.79 0.96 0.57 1.26
295286_A ENSG00000175063 UBE2C A Cell cycle/cell prolif ⬍0.001 0.73 0.91 0.68 1.17
300169_A ENSG00000113810 SMC4L1 A ⬍0.001 0.97 1.30 0.87 1.53
301111_B ENSG00000175792 RUVBL1 A Nucl met/trans reg ⬍0.001 0.58 0.72 0.70 0.93
322146_A ENSG00000134531 EMP1 A Death/cell death ⬍0.001 0.72 1.02 0.69 2.22
327575_A ENSG00000056678 KIFC1 A Cell cycle/cell prolif ⬍0.001 0.62 1.01 0.63 1.00
357373_A ENSG00000126787 DLG7 A Cell comm/sig trans ⬍0.001 0.71 1.03 0.60 1.55
358777_A ENSG00000013810 TACC3 A ⬍0.001 0.75 0.85 0.70 1.18
36498_A ENSG00000119326 CTNNAL1 A Death/cell death ⬍0.001 0.74 0.58 0.40 1.12
36869_A NSA1 A ⬍0.001 0.54 0.75 0.54 1.00
376599_A ENSG00000175315 CST6 A ⬍0.001 0.15 1.19 0.18 1.23
376757_A ENSG00000163171 CDC42EP3 A Cell comm/sig trans ⬍0.001 0.39 1.30 0.58 1.23
427857_A ENSG00000145386 CCNA2 A Cell cycle/cell prolif ⬍0.001 0.43 0.68 0.46 0.87
667598_C ENSG00000164985 PSIP1 A ⬍0.001 0.73 1.16 0.64 1.19
758484_A ENSG00000175063 UBE2C A Cell cycle/cell prolif ⬍0.001 0.67 0.87 0.67 1.15
769921_A ENSG00000175063 UBE2C A Cell cycle/cell prolif ⬍0.001 0.70 0.94 0.68 1.18
785707_A ENSG00000185308 PRC1 A Cell cycle/cell prolif ⬍0.001 0.67 1.15 0.72 1.24
132508_A LOC2537827 B ⬍0.001 1.87 1.27 0.41 0.88
1341530_A B ⬍0.001 2.43 1.45 1.43 1.13
136154_A ENSG00000164120 HPGD B Carb acid met/org acid met ⬍0.001 4.87 4.04 1.35 1.05
136836_A ENSG00000157110 RBPMS B Nucl met/trans reg ⬍0.001 2.76 2.05 0.91 0.95
149832_A ENSG00000054983 GALC B Carb acid met/org acid met ⬍0.001 4.53 2.00 1.07 1.58
153361_B ENSG00000090920 FCGBP B ⬍0.001 7.05 1.95 1.14 1.02
2012025_A ENSG00000169903 TM4SF4 B Death/cell death ⬍0.001 11.58 1.48 1.21 1.09
201787_A SLC22A36 B ⬍0.001 1.84 1.29 0.71 0.58
230010_A ENSG00000132437 DDC B Carb acid met/org acid met ⬍0.001 4.91 1.70 1.43 1.23
23710_B ENSG00000054983 GALC B Carb acid met/org acid met ⬍0.001 3.59 1.77 0.67 1.29
247987_A ENSG00000114859 CLCN2 B Transp/sig trans ⬍0.001 2.03 1.21 0.95 0.96
249287_A ENSG00000085545 SMURF18 B ⬍0.001 4.16 1.40 1.70 1.35
271629_A ENSG00000141542 RAB40B B Transp/sig trans ⬍0.001 2.95 1.42 1.35 0.97
272966_A ENSG00000175130 MLP B ⬍0.001 3.35 3.10 0.69 1.56
27659_A ENSG00000176595 KIAA07118 B ⬍0.001 2.64 0.73 0.75 0.67
28370_A ENSG00000115525 SIAT9 B Death/cell death ⬍0.001 4.47 2.57 1.18 1.54
290867_A ENSG00000175130 MLP B ⬍0.001 3.72 3.12 0.73 1.61
293626_A ENSG00000144852 NR112 B Cell comm/sig trans ⬍0.001 2.22 0.82 0.84 0.82
294203_A ENSG00000109062 SLC9A3R1 B Transp/sig trans ⬍0.001 1.96 0.86 1.24 0.78
296332_A SLC22A37 B ⬍0.001 1.82 1.36 0.68 0.45
301888_C NSA B ⬍0.001 2.67 1.69 1.06 0.87
306948_A ENSG00000115980 ANXA4 B ⬍0.001 3.30 2.67 1.11 0.97
321351_A ENSG00000100600 LGMN B Prot met/catab ⬍0.001 2.10 1.56 0.96 0.77
321953_A ENSG00000100600 LGMN B Prot met/catab ⬍0.001 2.44 1.78 0.97 0.76
360643_A ENSG00000013455 ARPCIA B ⬍0.001 3.14 1.32 1.14 1.21
362795_A ENSG00000016082 ISL1 B Nucl met/trans reg ⬍0.001 2.01 2.66 1.12 1.02
36962_A ENSG00000178892 B ⬍0.001 2.67 1.44 0.81 1.02
36981_A ENSG00000100600 LGMN B Prot met/catab ⬍0.001 2.29 1.50 0.91 0.77
381219_A ENSG00000016082 ISL1 B Nucl met/trans reg ⬍0.001 2.16 3.34 1.19 1.15
38922_A ENSG00000105854 PON2 B ⬍0.001 3.48 1.34 1.50 0.95
417087_A ENSG00000149260 CAPN5 B Prot met/catab ⬍0.001 5.60 2.79 0.89 1.30
44373_B ENSG00000175130 MLP B ⬍0.001 3.23 3.08 0.66 1.50
471115_A ENSG00000164120 HPGD B Carb acid met/org acid met ⬍0.001 5.96 4.25 1.26 0.88
47858_A ENSG00000070277 TRRAP B Cell comm/sig trans ⬍0.001 3.26 1.25 1.19 1.25
49502_A ENSG00000106665 CYLN2 B ⬍0.001 2.87 2.18 1.52 1.18
51565_A ENSG00000105854 PON2 B ⬍0.001 3.45 1.33 1.45 0.95
731649_A ENSG00000114859 CLCN2 B Transp/sig trans ⬍0.001 2.04 1.14 0.99 0.92
785833_A ENSG00000152268 SPON1 B ⬍0.001 5.52 1.31 1.69 1.38
795498_A ENSG00000140526 ABHD2 E ⬍0.001 1.93 0.83 0.70 0.82
795498_C ENSG00000140526 ABHD2 B ⬍0.001 1.98 1.01 0.79 0.95
810137_A NSA B ⬍0.001 9.68 5.20 2.02 1.32
138907_A ENSG00000173391 OLR1 C Prot met/catab ⬍0.001 1.09 1.27 3.98 0.91
145340_A ENSG00000173391 OLR1 C Prot met/catab ⬍0.001 0.97 0.74 2.58 0.88
148677_A ENSG00000130600 H197 C ⬍0.001 0.98 1.39 10.74 1.51
153589_A ENSG00000159167 STC1 C Cell comm/sig trans ⬍0.001 0.87 0.75 5.01 1.05
180304_A ENSG00000144544 SRGAP2 C ⬍0.001 1.35 0.76 2.26 0.95
233296_A ENSG00000087303 NID2 C Cell comm/sig trans ⬍0.001 0.85 0.68 2.40 0.82
23333_A ENSG0000008853 RHOBTB2 C Transp/sig trans ⬍0.001 0.85 0.56 1.43 0.77
 PANCREATIC TUMORIGENESIS AND METASTASIS 109
TABLE IV – DIFFERENTIALLY EXPRESSED CLONES IDENTIFIED BY CLASS COMPARISON ANALYSIS BETWEEN
CELL LINES OF DIFFERENT ORIGIN (CONTINUED)

HUGO gene Adjusted Lymph node Primary

Clone ID Ensembl ID Cluster1 GO cluster2 Ascites4 Liver4
symbol p-value3 metastasis4 tumor4

23819_A ENSG00000160179 ABCG1 C Transp/sig trans ⬍0.001 1.06 0.68 2.29 0.55
244303_A ENSG00000168405 CMAH C ⬍0.001 0.94 0.94 1.75 0.85
24984_A ENSG00000159403 CIR C Prot met/catab ⬍0.001 0.76 0.75 3.41 0.78
257329_A CMAH6 C ⬍0.001 0.84 0.82 1.75 0.85
293325_C ENSG00000082074 FYB C Transp/sig trans ⬍0.001 0.96 0.98 2.88 0.95
298914_A PDZK37 C ⬍0.001 0.75 0.87 2.63 0.94
300481_A ETS26 C ⬍0.001 0.79 0.65 1.77 0.51
30063_A ENSG00000160179 ABCG1 C Transp/sig trans ⬍0.001 1.18 0.66 2.12 0.56
30587_A ENSG00000182541 LIMK2 C Transp/sig trans ⬍0.001 0.38 0.45 1.22 0.37
306798_A ENSG00000138386 NAB1 C Nucl met/trans reg ⬍0.001 1.52 1.63 2.38 1.04
30850_A ENSG00000177129 MMP17 C Prot met/catab ⬍0.001 0.53 0.70 2.36 0.78
31093_B ENSG00000177129 MMP17 C Prot met/catab ⬍0.001 0.67 1.02 2.22 0.90
31528_B ENSG00000177129 MMP17 C Prot met/catab ⬍0.001 0.80 1.11 2.55 0.95
323618_B ENSG00000111801 BTN3A3 C ⬍0.001 0.95 1.44 2.48 1.01
324169_A NSA C ⬍0.001 0.69 0.86 1.62 0.53
340860_A EGFL36 C ⬍0.001 1.87 3.00 6.10 1.67
345996_A ENSG00000185885 IFITM1 C Cell comm/sig trans ⬍0.001 1.48 2.18 10.29 2.58
361834_A PDZK37 C ⬍0.001 1.21 0.95 3.08 1.24
380849_A ENSG00000134107 BHLHB2 C Nucl met/trans reg ⬍0.001 0.86 1.19 1.76 0.97
50483_A ENSG00000140092 FBLN5 C Cell comm/sig trans ⬍0.001 1.30 0.95 2.69 0.98
727272_A ENSG00000176046 P8 C Death/cell death ⬍0.001 0.59 0.42 3.52 0.55
741699_A ENSG00000182541 LIMK2 C Transp/sig trans ⬍0.001 0.37 0.36 1.30 0.41
772208_A ENSG00000182326 C1S C Prot met/catab ⬍0.001 0.70 0.72 5.64 0.78
789273_A ENSG00000138386 NAB1 C Nucl met/trans reg ⬍0.001 1.55 1.43 2.23 1.08
810521_A ENSG00000135363 LMO2 C ⬍0.001 0.89 1.18 4.79 1.25
1033805_A NSA D ⬍0.001 2.12 1.07 2.46 0.75
135221_A ENSG00000163993 S100P D ⬍0.001 10.08 3.31 11.65 1.11
175182_B NSA D ⬍0.001 1.28 0.95 1.39 0.72
2113456_A ENSG00000124107 SLP1 D ⬍0.001 0.78 1.29 3.57 0.64
262390_A ENSG00000182240 BACE2 D Prot met/catab ⬍0.001 0.77 1.02 1.49 0.70
293270_A ENSG00000151632 AKR1C2 D Transp/sig trans ⬍0.001 8.65 2.21 6.73 0.71
305151_A ENSG00000085871 MGST2 D Cell comm/sig trans ⬍0.001 2.07 1.50 2.56 1.18
310800_A ENSG00000143463 DAF D Prot met/catab ⬍0.001 3.21 1.65 6.68 1.18
47238_A ENSG00000116209 Clorf8 D ⬍0.001 2.12 1.86 2.38 1.28
490969_A ENSG00000163993 S100P D ⬍0.001 8.18 2.74 8.93 1.04
756580_A ENSG00000153234 NR4A2 D Cell comm/sig trans ⬍0.001 2.86 1.25 2.74 1.07
110529_A ENSG00000170545 LOC572288 E ⬍0.001 1.27 1.42 0.53 0.96
130698_A ENSG00000170545 LOC572288 E ⬍0.001 1.37 1.53 0.39 0.99
161878_A ENSG00000131016 AKAP12 E Transp/sig trans ⬍0.001 6.44 2.22 0.80 4.07
stSG89529 ASLL7 E ⬍0.001 2.50 1.22 0.79 1.04
132469_B ATP1B37 F ⬍0.001 1.00 0.96 2.26 0.92
42669_A ENSG00000118515 SGK F Transp/sig trans ⬍0.001 0.41 0.74 3.40 1.21
472138_A NSA F ⬍0.001 0.53 0.95 4.15 1.53
282740_A ENSG00000147044 CASK G Transp/sig trans ⬍0.001 1.41 1.82 1.12 0.82
809715_A DAPK17 G ⬍0.001 1.45 10.04 4.30 1.43
161000_A ENSG00000185340 GAS2L1 G Cell cycle/cell prolif ⬍0.001 0.75 0.84 1.34 1.22
241085_A ENSG00000154640 BTG3 G Cell cycle/cell prolif ⬍0.001 0.39 0.82 0.88 0.96
246304_C ENSG00000154640 BTG3 G Cell cycle/cell prolif ⬍0.001 0.43 0.77 0.85 0.91
146229_A ATP6V1B26 H ⬍0.001 0.71 0.54 1.03 0.90
796284_A ENSG00000169047 IRS1 I Cell comm/sig trans ⬍0.001 1.42 0.96 0.51 1.17
1 2
Expression cluster.– Gene ontology cluster. In the column are reported only the first 2 descriptive gene ontology terms out of 4 obtained by
gCLUTO clustering program. These terms are the most observed in the cluster and should better describe the biologic process involved. cell
prolif, cell proliferation; transp, transport; sig trans, signal transduction; nucl met, nucleobase, nucleoside, nucleotide and nucleic acid
metabolism; reg trans, regulation of transcription DNA; cell comm, cell communication; carb acid me, carboxylic acid metabolism; org acid met,
organic acid metabolism; prot met, protein metabolism; catab, catabolism.–3Adjusted p values.–4Geometric mean of ratios against HPDE cell
line.–5No sequence data available.–65⬘ and 3⬘ sequences available.–75⬘ or 3⬘ sequence available only.–8No HUGO symbol.

pancreatic cancer, for which it is difficult to obtain large numbers
of clinical specimens and for which RNA quality usually is not of
a sufficient quality for large-scale gene expression profiling stud-
ies.
We have selected 4 genes that have not previously been studied
in pancreatic adenocarcinoma for validation and measured expres-
sion of their protein products in PTMAs. Anterior gradient 2
(AGR2) is a homologue of the Xenopus laevis XAG2 gene, which
is expressed in a gradient in the ectoderm during early develop-
ment and differentiation of the mucus-secreting cement gland. The
human tissues in which AGR2 has been found to be expressed
(trachea, lung, stomach, colon, prostate and small intestine) all
contain mucus-secreting cells. AGR2 has been found coexpressed
with the estrogen receptor (ER) in breast cancer cell lines using
suppression subtractive hybridization, not clear what was the ex-
periment30 but has not previously been otherwise implicated in
neoplasia.
Two other overexpressed genes selected for validation were
IGFBP3 and IGFBP4, which belong to a family of at least 6
structurally homologous proteins that bind IGF-I and IGF-II with
high affinity, thereby modulating their bioavailability and bioac-
tivity.41 IGFBPs are widely expressed in normal and malignant
tissues with variable expression. The number of different binding
proteins, the variation in physical IGF-binding properties, and their
ability to either enhance or inhibit IGF activity all suggest complex
mechanisms for the control of cell growth via modulation of IGF
110 MISSIAGLIA ET AL.
FIGURE 4 – RT-PCR analysis of expression of p8, STC1 and LIMK2 genes across the panel of 20 pancreatic cell lines.
activity. Recently, several studies reported increased IGFBP3 gene
expression in lung and colon carcinomas42,43 as well as increased
IGFBP3 levels in the serum that were associated with progression
of breast cancer.44 Increased levels of IGFBP4 have been observed
in prostate cancer cell lines and have been shown to be involved in
tumor growth.45 It seems likely that a coordinate regulation of
IGFBPs may exist, such that altered production of one binding
protein may profoundly influence the levels of others.
The protein levels of hepatocyte nuclear factor 3/forkhead ho-
mologue 4 (FOXJ1) were also examined on PTMAs. This gene is
a member of the forkhead/winged-helix family of transcription
factors that are expressed in tissue-specific patterns and play
critical roles in development and cell differentiation.46 This gene
has not been implicated in the development of any type of tumor
previously, and its role in cancer needs to be further explored.
As described above, the class comparison analysis was able to
isolate a set of genes that could separate the cell lines on the basis
of their origin. In particular, the hierarchical clustering divided
these genes into 3 major subsets characterized by different expres-
sion behavior in the cell lines derived from primary tumors, liver
metastasis, ascites and lymph node metastasis. Cluster A contains
several genes significantly downregulated in cell lines originated
from lymph node metastasis and ascites and shows a correlation
with a gene ontology cluster characterized by biologic process
terms such as cell proliferation and cell cycle. Indeed, genes such
as Cyclin A2, CDKN3, DLG7 and UBE2C (control of cell cycle
progression), NEK2, PRC1, KIFC1, CENPA, CEMPF, TACC3 and
SMC4L1 (chromosomal assembly and stabilization, spindle forma-
tion and maintenance) were present in cluster A and are all
involved in the G2/M phase of the cell cycle. Their downregulation
may be a consequence of a lower proliferation rate in the cell lines
derived from ascites and lymph nodes in comparison to the ones
from primary tumors.47 In a recent study,48 the authors described
a mutation in BUB1 that is able to induce a defective mitotic
spindle checkpoint in the cell line Hs766T. Inhibition of BUB1
gene has also been associated with genome instability and anchor-
age-independent growth.49 We can speculate that similar alter-
ations in key genes involved in the control of the G2/M phase may
be able to induce chromosomal instability and facilitate tumor
progression.
Expression cluster C was characterized by genes upregulated in
cell lines that originated from lymph nodes. Although no correla-
tion was found with any specific biologic process, several genes
were annotated with cell communication and signal transduction
terms (FYB, IFITM1, SRGAP2, NID2, RHOBRB2, ABCG1, SRC1,
LIMK2, LMO2 and p8).
We have validated by RT-PCR the high expression of p8, STC1
and LIMK2 in metastatic lymph node-derived cell lines. The p8
gene encodes a transcription factor and several studies have al-
ready demonstrated its overexpression in acute pancreatitis as well
as in pancreatic cancer and human pancreatic cancer cell lines.50 –52
In addition, one of these studies has shown that p8 overexpression
is significantly correlated with nodal involvement.50 The authors
speculated that activation of p8 protein in ischemic conditions may
prevent apoptosis and allow cancer cell to proliferate. Another
study demonstrated that expression of candidate of metastasis 1
gene (com1), which is identical to p8, could mediate the growth of
tumor cells after metastatic establishment in a secondary organ.53
Overexpression of the stanniocalcin 1 (STC1) gene has previ-
ously been described in breast cancer where its expression level
significantly correlated with the number of involved lymph
nodes.54 STC1 was also shown to be induced by hypoxia and
hypertonicity, increasing resistance to ischemic situations55 prob-
ably by mediating inflammation or angiogenesis.56
The third gene confirmed by RT-PCR, LIMK2, has 2 LIM
domains. It is a kinase that controls actin cytoskeletal reorganiza-
tion: filopodial formation by inducing Cdc42, and stress fiber
formation by Rho induction. LIMK2 is specifically activated by
ROCK and its function is mediated by cofilin.57–59
The gene BHLHB2 (also known as DEC1) was also present in
cluster C. This gene has previously been described as being in-
duced in response to hypoxia in pancreatic cancer cell lines.60 In
addition, cluster C genes with RhoGTPase activity (SRGAP2 and
RHOBTB2) were also found upregulated, indicating a possible
 PANCREATIC TUMORIGENESIS AND METASTASIS
involvement of this pathway in the progression of pancreatic tumor
toward malignancy.
Finally, cluster B is characterized by genes overexpressed in
ascites and/or liver metastasis and appears to be a heterogeneous
group characterized by different biologic processes. Notable is the
tetraspanin gene, TM4SF4. The tetaspanins are a superfamily
involved in cell motility, metastasis, cell proliferation and differ-
entiation. In particular, the control of cell migration seems to be
mediated by interaction with integrin proteins.61 Furthermore,
MLP (MARCKS-like protein) that was also found upregulated in
cluster B is able to interact with tetraspanins and is a substrate for
protein kinase C and may function as a regulator of cortical
cytoskeleton dynamics in cell spreading and migration.62
Additional evaluation of these candidate genes in clinical tissue
specimens is required in order to establish their involvement in
metastasis. A better understanding of the regulation of metastasis
is necessary for the discovery of novel therapeutic targets and for
better assessment of individual risk for tumor progression, espe-
REFERENCES
1. Landis SH, Murray T, Bolden S, Wingo PA. Cancer statistics, 1998.
CA Cancer J Clin 1998;48:6 –29.
2. Evans DB, Lee JE, Pisters PW, Charnsangavej C, Ellis LM, Chiao PJ,
Lenzi R, Abbruzzese JL. Advances in the diagnosis and treatment of
adenocarcinoma of the pancreas. Cancer Treat Res 1997;90:109 –25.
3. Ross DT, Scherf U, Eisen MB, Perou CM, Rees C, Spellman P, Iyer
V, Jeffrey SS, Van de Rijn M, Waltham M, Pergamenschikov A, Lee
JC, et al. Systematic variation in gene expression patterns in human
cancer cell lines. Nat Genet 2000;24:227–35.
4. Lehnert L, Trost H, Schmiegel W, Roder C, Kalthoff H. Hollow-
spheres: a new model for analyses of differentiation of pancreatic duct
epithelial cells. Ann NY Acad Sci 1999;880:83–93.
5. Chen WH, Horoszewicz JS, Leong SS, Shimano T, Penetrante R,
Sanders WH, Berjian R, Douglass HO, Martin EW, Chu TM. Human
pancreatic adenocarcinoma: in vitro and in vivo morphology of a new
tumor line established from ascites. In Vitro 1982;18:24 –34.
6. Kyriazis AA, Kyriazis AP, Sternberg CN, Sloane NH, Loveless JD.
Morphological, biological, biochemical, and karyotypic characteris-
tics of human pancreatic ductal adenocarcinoma Capan-2 in tissue
culture and the nude mouse. Cancer Res 1986;46:5810 –5.
7. Schoumacher RA, Ram J, Iannuzzi MC, Bradbury NA, Wallace RW,
Hon CT, Kelly DR, Schmid SM, Gelder FB, Rado TA, Frizzell RA.
A cystic fibrosis pancreatic adenocarcinoma cell line. Proc Natl Acad
Sci USA 1990;87:4012– 6.
8. Morgan RT, Woods LK, Moore GE, Quinn LA, McGavran L, Gordon
SG. Human cell line (COLO 357) of metastatic pancreatic adenocar-
cinoma. Int J Cancer 1980;25:591– 8.
9. Nagata N, Akatsu T, Kugai N, Yasutomo Y, Kinoshita T, Kosano H,
Shimauchi T, Takatani O, Ueyama Y. The tumor cells (FA-6) estab-
lished from a pancreatic cancer associated with humoral hypercalce-
mia of malignancy: a simultaneous production of parathyroid hor-
mone-like activity and transforming growth factor activity.
Endocrinol Jpn 1989;36:75– 85.
10. Metzgar RS, Gaillard MT, Levine SJ, Tuck FL, Bossen EH, Borowitz
MJ. Antigens of human pancreatic adenocarcinoma cells defined by
murine monoclonal antibodies. Cancer Res 1982;42:601– 8.
11. Owens RB, Smith HS, Nelson-Rees WA, Springer EL. Epithelial cell
cultures from normal and cancerous human tissues. J Natl Cancer Inst
1976;56:843–9.
12. Vila MR, Lloreta J, Schussler MH, Berrozpe G, Welt S, Real FX.
New pancreas cancers cell lines that represent distinct stages of ductal
differentiation. Lab Invest 1995;72:395– 404.
13. Frazier ML, Pathak S, Wang ZW, Cleary K, Singletary SE, Olive M,
Mackay B, Steck PA, Levin B. Establishment of a new human
pancreatic adenocarcinoma cell line, MDAPanc-3. Pancreas 1990;5:
8 –16.
14. Yunis AA, Arimura GK, Russin DJ. Human pancreatic carcinoma
(MIA PaCa-2) in continuous culture: sensitivity to asparaginase. Int J
Cancer 1977;19:218 –35.
15. Kloppel G, Lingenthal G, von Bulow M, Kern HF. Histological and
fine structural features of pancreatic ductal adenocarcinomas in rela-
tion to growth and prognosis: studies in xenografted tumours and
clinico-histopathological correlation in a series of 75 cases. Histopa-
thology 1985;9:841–56.
16. Lieber M, Mazzetta J, Nelson-Rees W, Kaplan M, Todaro G. Estab-
lishment of a continuous tumor-cell line (panc-1) from a human
carcinoma of the exocrine pancreas. Int J Cancer 1975;15:741–7.
17. von Bulow M, Scharfe T, Kloppel G, Kern HF. Establishment and
111
cially in types of cancer such as pancreatic adenocarcinoma with
extremely high occurrence of metastasis.
ACKNOWLEDGEMENTS
The authors thank the staff (David Vetrie, Cordelia Langford,
Adam Whitakker, Neil Sutton) of the Sanger Institute Microarray
Facility for the supply of arrays, laboratory protocols and technical
advice; the bioinformatics staff (Kate Rice, Rob Andrews, Adam
Butler, Harish Chudasama) of the Sanger Institute for assistance
with Quantarray/GeneSpring data files and all data analysis and
databases relating to elements on the arrays; Dr. D.A. Thompson
for kindly providing the antibodies against AGR2. The human
IMAGE cDNA clone collection was obtained from the MRC
HGMP Resource Centre (Hinxton, U.K.). All cDNA clone rese-
quencing was performed by Team 56 at the Sanger Institute. The
microarray consortium is funded by the Wellcome Trust, Cancer
Research U.K. and the Ludwig Institutes of Cancer Research.
characterizatin of continuous tumour-cell lines from human pancreatic
carcinoma in vivo. Digestion 1982;25:17– 8.
18. Kalthoff H, Schmiegel W, Roeder C, Kasche D, Schmidt A, Lauer G,
Thiele HG, Honold G, Pantel K, Riethmuller G, Scherer E, Maurer J,
et al. p53 and K-RAS alterations in pancreatic epithelial cell lesions.
Oncogene 1993;8:289 –98.
19. Dexter DL, Matook GM, Meitner PA, Bogaars HA, Jolly GA, Turner
MD, Calabresi P. Establishment and characterization of two human
pancreatic cancer cell lines tumorigenic in athymic mice. Cancer Res
1982;42:2705–14.
20. Iwamura T, Katsuki T, Ide K. Establishment and characterization of a
human pancreatic cancer cell line (SUIT-2) producing carcinoembry-
onic antigen and carbohydrate antigen 19-9. Jpn J Cancer Res 1987;
78:54 – 62.
21. Okabe T, Yamaguchi N, Ohsawa N. Establishment and characteriza-
tion of a carcinoembryonic antigen (CEA)-producing cell line from a
human carcinoma of the exocrine pancreas. Cancer 1983;51:662– 8.
22. Furukawa T, Duguid WP, Rosenberg L, Viallet J, Galloway DA, Tsao
MS. Long-term culture and immortalization of epithelial cells from
normal adult human pancreatic ducts transfected by the E6E7 gene of
human papilloma virus 16. Am J Pathol 1996;148:1763–70.
23. Yang YH, Dudoit S, Luu P, Lin DM, Peng V, Ngai J, Speed TP.
Normalization for cDNA microarray data: a robust composite method
addressing single and multiple slide systematic variation. Nucl Acids
Res 2002;30:E15.
24. Huber W, Von Heydebreck A, Sultmann H, Poustka A, Vingron M.
Variance stabilization applied to microarray data calibration and to the
quantification of differential expression. Bioinformatics 2002;
18(Suppl 1):S96 –104.
25. Ihaka R, Gentleman R. R: a language for data analysis and graphics.
J Comp Graph Stat 1996;5:299 –314.
26. Baldi P, Long AD. A Bayesian framework for the analysis of mi-
croarray expression data: regularized t-test and statistical inferences of
gene changes. Bioinformatics 2001;17:509 –19.
27. Reiner A, Yekutieli D, Benjamini Y. Identifying differentially ex-
pressed genes using false discovery rate controlling procedures.
Bioinformatics 2003;19:368 –75.
28. Al-Shahrour F, Dı̀az-Uriarte R, Dopazo J. FatiGO: a Web tool for
finding significant associations of gene ontology terms with groups of
genes. Bioinformatics, 2004;20:578 – 80.
29. Kallioniemi OP, Wagner U, Kononen J, Sauter G. Tissue microarray
technology for high-throughput molecular profiling of cancer. Hum
Mol Genet 2001;10:657– 62.
30. Thompson DA, Weigel RJ. hAG-2, the human homologue of the
Xenopus laevis cement gland gene XAG-2, is coexpressed with es-
trogen receptor in breast cancer cell lines. Biochem Biophys Res
Commun 1998;251:111– 6.
31. Iacobuzio-Donahue CA, Maitra A, Shen-Ong GL, van Heek T, Ash-
faq R, Meyer R, Walter K, Berg K, Hollingsworth MA, Cameron JL,
Yeo CJ, Kern SE, et al. Discovery of novel tumor markers of pan-
creatic cancer using global gene expression technology. Am J Pathol
2002;160:1239 – 49.
32. Argani P, Rosty C, Reiter RE, Wilentz RE, Murugesan SR, Leach SD,
Ryu B, Skinner HG, Goggins M, Jaffee EM, Yeo CJ, Cameron JL, et
al. Discovery of new markers of cancer through serial analysis of gene
expression: prostate stem cell antigen is overexpressed in pancreatic
adenocarcinoma. Cancer Res 2001;61:4320 – 4.
33. Crnogorac-Jurcevic T, Missiaglia E, Blaveri E, Gangeswaran R, Jones
112 MISSIAGLIA ET AL.
M, Terris B, Costello E, Neoptolemos JP, Lemoine NR. Molecular
alterations in pancreatic carcinoma: expression profiling shows that
dysregulated expression of S100 genes is highly prevalent. J Pathol
2003;201:63–74.
34. Han H, Bearss DJ, Browne LW, Calaluce R, Nagle RB, Von Hoff DD.
Identification of differentially expressed genes in pancreatic cancer
cells using cDNA microarray. Cancer Res 2002;62:2890 – 6.
35. Tan ZJ, Hu XG, Cao GS, Tang Y. Analysis of gene expression profile
of pancreatic carcinoma using cDNA microarray. World J Gastroen-
terol 2003;9:818 –23.
36. Ryu B, Jones J, Blades NJ, Parmigiani G, Hollingsworth MA, Hruban
RH, Kern SE. Relationships and differentially expressed genes among
pancreatic cancers examined by large-scale serial analysis of gene
expression. Cancer Res 2002;62:819 –26.
37. Logsdon CD, Simeone DM, Binkley C, Arumugam T, Greenson JK,
Giordano TJ, Misek DE, Kuick R, Hanash S. Molecular profiling of
pancreatic adenocarcinoma and chronic pancreatitis identifies multi-
ple genes differentially regulated in pancreatic cancer. Cancer Res
2003;63:2649 –57.
38. Iacobuzio-Donahue CA, Maitra A, Olsen M, Lowe AW, van Heek
NT, Rosty C, Walter K, Sato N, Parker A, Ashfaq R, Jaffee E, Ryu B,
et al. Exploration of global gene expression patterns in pancreatic
adenocarcinoma using cDNA microarrays. Am J Pathol 2003;162:
1151– 62.
39. Terris B, Blaveri E, Crnogorac-Jurcevic T, Jones M, Missiaglia E,
Ruszniewski P, Sauvanet A, Lemoine NR. Characterization of gene
expression profiles in intraductal papillary-mucinous tumors of the
pancreas. Am J Pathol 2002;160:1745–54.
40. Ouyang H, Mou L, Luk C, Liu N, Karaskova J, Squire J, Tsao MS.
Immortal human pancreatic duct epithelial cell lines with near normal
genotype and phenotype. Am J Pathol 2000;157:1623–31.
41. Grimberg A, Cohen P. Role of insulin-like growth factors and their
binding proteins in growth control and carcinogenesis. J Cell Physiol
2000;183:1–9.
42. Beer DG, Kardia SL, Huang CC, Giordano TJ, Levin AM, Misek DE,
Lin L, Chen G, Gharib TG, Thomas DG, Lizyness ML, Kuick R, et al.
Gene-expression profiles predict survival of patients with lung ade-
nocarcinoma. Nat Med 2002;8:816 –24.
43. Kansra S, Ewton DZ, Wang J, Friedman E. IGFBP-3 mediates TGF
beta 1 proliferative response in colon cancer cells. Int J Cancer
2000;87:373– 8.
44. Vadgama JV, Wu Y, Datta G, Khan H, Chillar R. Plasma insulin-like
growth factor-I and serum IGF-binding protein 3 can be associated
with the progression of breast cancer, and predict the risk of recur-
rence and the probability of survival in African-American and His-
panic women. Oncology 1999;57:330 – 40.
45. Drivdahl RH, Sprenger C, Trimm K, Plymate SR. Inhibition of
growth and increased expression of insulin-like growth factor-binding
protein-3 (IGFBP-3) and -6 in prostate cancer cells stably transfected
with antisense IGFBP-4 complementary deoxyribonucleic acid. En-
docrinology 2001;142:1990 – 8.
46. Murphy DB, Seemann S, Wiese S, Kirschner R, Grzeschik KH, Thies
U. The human hepatocyte nuclear factor 3/fork head gene FKHL13:
genomic structure and pattern of expression. Genomics 1997;40:
462–9.
47. Sipos B, Moser S, Kalthoff H, Torok V, Lohr M, Kloppel G. A
comprehensive characterization of pancreatic ductal carcinoma cell
lines: towards the establishment of an in vitro research platform.
Virchows Arch 2003;442:444 –52.
48. Hempen PM, Kurpad H, Calhoun ES, Abraham S, Kern SE. A double
missense variation of the BUB1 gene and a defective mitotic spindle
checkpoint in the pancreatic cancer cell line Hs766T. Hum Mutat
2003;21:445.
49. Musio A, Montagna C, Zambroni D, Indino E, Barbieri O, Citti L,
Villa A, Ried T, Vezzoni P. Inhibition of BUB1 results in genomic
instability and anchorage-independent growth of normal human fibro-
blasts. Cancer Res 2003;63:2855– 63.
50. Su SB, Motoo Y, Iovanna JL, Xie MJ, Mouri H, Ohtsubo K, Yamagu-
chi Y, Watanabe H, Okai T, Matsubara F, Sawabu N. Expression of
p8 in human pancreatic cancer. Clin Cancer Res 2001;7:309 –13.
51. Su SB, Motoo Y, Iovanna JL, Berthezene P, Xie MJ, Mouri H,
Ohtsubo K, Matsubara F, Sawabu N. Overexpression of p8 is in-
versely correlated with apoptosis in pancreatic cancer. Clin Cancer
Res 2001;7:1320 – 4.
52. Mallo GV, Fiedler F, Calvo EL, Ortiz EM, Vasseur S, Keim V,
Morisset J, Iovanna JL. Cloning and expression of the rat p8 cDNA,
a new gene activated in pancreas during the acute phase of pancre-
atitis, pancreatic development, and regeneration, and which promotes
cellular growth. J Biol Chem 1997;272:32360 –9.
53. Ree AH, Tvermyr M, Engebraaten O, Rooman M, Rosok O, Hovig E,
Meza-Zepeda LA, Bruland OS, Fodstad O. Expression of a novel
factor in human breast cancer cells with metastatic potential. Cancer
Res 1999;59:4675– 80.
54. Wascher RA, Huynh KT, Giuliano AE, Hansen NM, Singer FR,
Elashoff D, Hoon DS. Stanniocalcin-1: a novel molecular blood and
bone marrow marker for human breast cancer. Clin Cancer Res
2003;9:1427–35.
55. Zhang K, Lindsberg PJ, Tatlisumak T, Kaste M, Olsen HS, Andersson
LC. Stanniocalcin: a molecular guard of neurons during cerebral
ischemia. Proc Natl Acad Sci USA 2000;97:3637– 42.
56. Kahn J, Mehraban F, Ingle G, Xin X, Bryant JE, Vehar G, Schoenfeld
J, Grimaldi CJ, Peale F, Draksharapu A, Lewin DA, Gerritsen ME.
Gene expression profiling in an in vitro model of angiogenesis. Am J
Pathol 2000;156:1887–900.
57. Arber S, Barbayannis FA, Hanser H, Schneider C, Stanyon CA,
Bernard O, Caroni P. Regulation of actin dynamics through phosphor-
ylation of cofilin by LIM-kinase. Nature 1998;393:805–9.
58. Sumi T, Matsumoto K, Takai Y, Nakamura T. Cofilin phosphoryla-
tion and actin cytoskeletal dynamics regulated by rho- and Cdc42-
activated LIM-kinase 2. J Cell Biol 1999;147:1519 –32.
59. Yang N, Higuchi O, Ohashi K, Nagata K, Wada A, Kangawa K,
Nishida E, Mizuno K. Cofilin phosphorylation by LIM-kinase 1 and
its role in Rac-mediated actin reorganization. Nature 1998;393:809 –
12.
60. Yoon DY, Buchler P, Saarikoski ST, Hines OJ, Reber HA, Hankinson
O. Identification of genes differentially induced by hypoxia in pan-
creatic cancer cells. Biochem Biophys Res Commun 2001;288:882– 6.
61. Berditchevski F. Complexes of tetraspanins with integrins: more than
meets the eye. J Cell Sci 2001;114:4143–51.
62. Wiederkehr A, Staple J, Caroni P. The motility-associated proteins
GAP-43, MARCKS, and CAP-23 share unique targeting and surface
activity-inducing properties. Exp Cell Res 1997;236:103–16.