GASTROENTEROLOGY 2013;145:1110–1120
A Gene Expression Signature of Epithelial Tubulogenesis and a Role for
ASPM in Pancreatic Tumor Progression
WEI–YU WANG,1,2,* CHUNG–CHI HSU,1,* TING–YUN WANG,1,* CHI–RONG LI,5,* YA–CHIN HOU,3 JUI–MEI CHU,1
CHUNG–TA LEE,2 MING–SHENG LIU,1 JIMMY J.–M. SU,1 KUAN–YING JIAN,1 SHENQ–SHYANG HUANG,1
SHIH–SHENG JIANG,1,6 YAN–SHEN SHAN,3 PIN–WEN LIN,3 YIN–YING SHEN,7 MICHAEL T.–L. LEE,8 TZE–SIAN CHAN,9
CHUN–CHAO CHANG,9 CHUNG–HSING CHEN,1,6 I–SHOU CHANG,1,6 YEN–LING LEE,1 LI–TZONG CHEN,1 and
KELVIN K. TSAI1,4,9
1
Laboratory for Tumor Epigenetics and Stemness, National Institute of Cancer Research and Translational Center for Glandular Malignancies, National Health Research
Institutes, Tainan, Taiwan; Departments of 2Pathology,3Surgery, and 4Medicine, National Cheng-Kung University Hospital and College of Medicine, National Cheng
Kung University, Tainan, Taiwan; 5Department of Medical Education and School of Nursing, Chung Shan Medical University and Hospital, Taichung, Taiwan; 6Taiwan
Bioinformatics Core, National Health Research Institutes, Zhunan, Taiwan; 7Pathology Core Laboratory, National Health Research Institutes, Tainan, Taiwan;
8
Department of Information Engineering, Kun Shan University, Tainan, Taiwan; and 9Graduate Institute of Clinical Medicine, School of Medicine, and Division of
Gastroenterology and Hepatology, Department of Internal Medicine, Taipei Medical University and Hospital, Taipei, Taiwan
BASIC AND
TRANSLATIONAL
PANCREAS
BACKGROUND & AIMS: Many patients with pancreatic ductal adenocarcinoma (PDAC) develop recurrent or
metastatic diseases after surgery, so it is important to
identify those most likely to benefit from aggressive
therapy. Disruption of tissue microarchitecture is an early
step in pancreatic tumorigenesis and a parameter used in
pathology grading of glandular tumors. We investigated
whether changes in gene expression during pancreatic
epithelial morphogenesis were associated with outcomes
of patients with PDAC after surgery. METHODS: We
generated architectures of human pancreatic duct
epithelial cells in a 3-dimensional basement membrane
matrix. We identified gene expression profiles of the cells
during different stages of tubular morphogenesis (tubulogenesis) and of PANC-1 cells during spheroid formation.
Differential expression of genes was confirmed by immunoblot analysis. We compared the gene expression profile
associated with pancreatic epithelial tubulogenesis with
that of PDAC samples from 27 patients, as well as with
their outcomes after surgery. RESULTS: We identified a
gene expression profile associated with tubulogenesis that
resembled the profile of human pancreatic tissue with
differentiated morphology and exocrine function. Patients
with PDACs with this profile fared well after surgery.
Based on this profile, we established a 628 gene
tubulogenesis-specific signature that accurately determined the prognosis of independent cohorts of patients
with PDAC (total n ¼ 128; accuracy ¼ 81.2%95.0%). One
gene, ASPM, was down-regulated during tubulogenesis but
up-regulated in human PDAC cell lines and tumor samples; up-regulation correlated with patient outcomes (Cox
regression P ¼ .0028). Bioinformatic, genetic, biochemical,
functional, and clinical correlative studies showed that
ASPM promotes aggressiveness of PDAC by maintaining
Wnt-b-catenin signaling and stem cell features of PDAC
cells. CONCLUSIONS: We identified a gene expression
profile associated with pancreatic epithelial tubulogenesis and a tissue architectureLspecific signature of
PDAC cells that is associated with patient outcomes
after surgery.
Keywords: Pancreatic Cancer; Tumor Progression; Prognostic Factor; Biomarker.
P
ancreatic ductal adenocarcinoma (PDAC) is a
devastating malignancy. Because of the paucity of
symptoms in early diseases and the aggressive behaviors
of the tumors, <20% of patients with PDAC present with
localized and resectable diseases at the time of diagnosis.
Even with curative-intent surgery, the majority of patients
with initially localized tumors developed recurrent or
metastatic diseases, and only a small subset (18%26%) of
the patients could attain long-term survival.1,2 Additional
improvements in the prognosis of patients with localized
PDAC might rely on elucidating the pathogenesis
underlying tumor recurrence and clinically reliable
prognostic prediction that can guide patient-tailored
treatment plans.
Loss of tissue architectures is one of the hallmark
features of malignant tumors.3 The extent to which a
tumor forms tissue microarchitecture reflects the degree of
malignant transformation, which dictates the clinical
behavior of the tumor. Glandular differentiation has been
widely used in the histopathological assessment for glandderived malignancies, including prostate cancer, breast
cancer, and PDAC.4–6 Nevertheless, assessing tissue
architectures by morphological criteria is partly subjective
and can fail to provide in-depth mechanistic insights.
Recently, comparative genomic analysis on clinical tumor
materials has led to the identification of gene profiles or
tumor molecular subtypes of PDAC that carried prognostic
*Authors share co-first authorship.
Abbreviations used in this paper: ASPM, abnormal spindle-like
microcephaly associated; C-index, concordance index; CSC, cancer
stem cell; 3D, 3-dimensional; Dvl, dishevelled; HPDE, human pancreatic
ductal epithelial; PDAC, pancreatic ductal adenocarcinoma; rBM,
reconstituted basement membrane; shRNA, short-hairpin RNA.
© 2013 by the AGA Institute
0016-5085/$36.00
http://dx.doi.org/10.1053/j.gastro.2013.07.040
November 2013
TUBULOGENESIS SIGNATURE IN PANCREATIC CANCER
significance.7,8 As opposed to molecular patterns identified
from a developed tumor, which might reflect the accumulative effect of the malignant transformation process,
knowledge-based and biology-informed approaches offer
an opportunity to identify biomarkers or classifiers that
additionally
provide
pathogenetic
information.
For example, gene expression patterns associated with
glandular morphogenesis have been linked to clinical
prognosis of patients with breast or prostate cancer.9,10
In this study, we investigated whether the molecular
changes associated with the formation of pancreatic
epithelial architectures can provide prognostic information
in PDAC. We identified a gene signature/subtype of PDAC
that is associated with pancreatic epithelial tubulogenesis
and the oncogenic role of ASPM (abnormal spindle-like
microcephaly associated), which provides a mechanistic
link between tissue architecture, Wnt signaling pathway,
cancer stemness, and tumor aggressiveness.
Immunohistochemistry
1111
Formalin-fixed, paraffin-embedded tissues of human
PDAC from 53 patients who received tumor resection at National
Cheng-Kung University Hospital (Tainan, Taiwan; the National
Cheng-Kung University Hospital cohort; Supplementary Table 1)
were acquired and used in conformity with Institutional Review
Boardapproved protocols. All immunohistochemical staining
was evaluated by expert pathologists (C. Lee and Y. Shen) and the
staining patterns were quantified using the histological score.13
Gene Expression Manipulations
Sustained ASPM knockdown in PDAC cells was achieved
by lentivirus-mediated RNA interference using validated shorthairpin RNA (shRNA) oligonucleotides (MISSION shRNA
lentiviruses; Sigma-Aldrich, St Louis, MO). The S33Y mutant of
b-catenin was subcloned from pcDNA3 (Addgene, Cambridge,
MA) into the pQCXIH retroviral vector (Clontech, Mountain
View, CA).14
Cell Migration Assay
Detailed Materials and Methods are described in the
Supplementary Material.
Cell Culture and Staining
The sources and culture of human pancreatic ductal
epithelial (HPDE) cells and other cells are described in the
Supplementary Material. For organotypic cultures, the cells were
grown on top of a thick layer of 3-dimensional (3D) reconstituted basement membrane (rBM; Matrigel, BD Biosciences,
San Jose, CA), as described previously.11
Gene Expression Profiling
Gene expression analysis was performed on an Affymetrix
Human Genome U133A 2.0 Plus GeneChip platform according to
the manufacturer’s protocol (Affymetrix, Santa Clara, CA). The
gene expression data have been deposited in NCBIs Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) and are
accessible through GEO Series accession number GSE42270.
Quantitative Real-Time Polymerase
Chain Reaction, Immunoblot, and
Co-immunoprecipitation Analyses
Quantitative reverse transcription polymerase chain
reaction analysis was performed using the LightCycler
System (Roche Diagnostics GmbH, Mannheim, Germany). The
procedure and antibodies used for immunoblot and
co-immunoprecipitation analysis are described in the
Supplementary Material.
Gene Expression Data Sets of PDAC
Sources of different PDAC tumor transcriptome datasets
are described in the Supplementary Material.
Construction of Prognostic Predictors
To identify from the differentially expressed genes, a set
of gene markers that optimally predicted survival, we used a
previously described supervised approach,12 as described in the
Supplementary Material.
The migration capacity of cells was measured using the
modified Boyden chamber assay.
Luciferase Reporter Assay
Cells were transduced with Cignal Lenti TCF/LEF
Reporter (Qiagen, Taipei, Taiwan) according to the manufacturer’s
protocol. After stimulation of the cells with recombinant human
Wnt-3a (R&D Systems, Minneapolis, MN) or vehicle, the reporter
activity was measured by using the ONE-Glo Luciferase Assay
System (Promega, Madison, WI).
Orthotopic Pancreatic Tumorigenesis Model
and Bioluminescence Imaging
AsPC-1 cells were retrovirally transduced with a vector
encoding green fluorescence protein and firefly luciferase. Cells
(1 106 cells) were inoculated into the pancreatic body of 8-weekold nonobese diabetic/severe combined immunodeficient mice,
and tumor mass and distribution were assessed by bioluminescence (IVIS Imaging System, Caliper Life Sciences, Waltham, MA).
Flow Cytometry
Cells were dissociated, antibody-labeled and resuspended
in Hank’s balanced salt solution/2% fetal bovine serum as
described previously.15 The procedure of flow cytometry and
antibodies used are described in the Supplementary Material.
Statistical Analysis
We used the statistical programming language R (cran.
r-project.org) and SPSS 10.0 software (SPSS, Inc., Chicago, IL) to
conduct the statistical analysis of the data. The concordance index
(C-index) was used to evaluate the predictive accuracy in the
survival analysis.16 Statistical significance was considered P < .05.
Results
Molecular Profiling of Pancreatic Epithelial
Tubulogenesis
We recapitulated the process of pancreatic epithelial morphogenesis by using a physiologically relevant 3D
BASIC AND
TRANSLATIONAL
PANCREAS
Materials and Methods
1112 WANG ET AL
organotypic culture model.11 When pancreatic ductal
epithelial HPDE cells were seeded on top of 3D rBM for a
short length of time (3648 h), they grew into unorganized cellular clusters or cords (Figure 1A
and Supplementary Figure 1A). Intriguingly, HPDE cells
subsequently underwent structural reorganization and
formed branching and tortuous tubules after 68 days in
culture. Confocal imaging analysis revealed that these tubules consisted of a single layer of polarized cells, indicated
by the polarized expressions of the basal surface marker
a6-integrin and the adherens junction protein b-catenin,
and a cell-free lumen (Figure 1B and Supplementary
Figure 1B). The HPDE tubulogenesis is a differentiationspecific process, as malignant PDAC cells, such as PANC1 cells, only formed disorganized spheroids that lacked
discernible structural organization within the same context
(Supplementary Figure 1C and data not shown).
To molecularly dissect the pancreatic epithelial
morphogenetic process, we performed comparative transcriptomic analysis and thereby identified a list of 620
unique genes, the transcript levels of which varied
GASTROENTEROLOGY Vol. 145, No. 5
significantly during the tubular morphogenesis, or tubulogenesis, of HPDE cells (Figure 1C). In contrast, we found
surprisingly few (n ¼ 18) genes that displayed differential
expression during the formation of malignant PANC-1
spheroids. Functional clustering analysis on these 620
genes revealed a significant enrichment of the Gene
Ontology terms related to cell cycle as well as epidermis
developments, wound healing, and inflammatory response
(Supplementary Figure 2). Importantly, several genes that
specify the exocrine functions of pancreas, including CEL
(bile salt-stimulated lipase), CA9 (carbonic anhydrase 9),
MUC1 (mucin 1), AGR2 (anterior gradient homolog 2), and
MUC20 (mucin 20), were profoundly up-regulated during
epithelial tubulogenesis, and their expressions remained
unaltered during the formation of malignant tumor
spheroids (Figure 1D). Immunoblotting analysis confirmed
the tubulogenesis-specific expressional changes in these
functional markers (Figure 1E). These data support our
tissue organization model as a valid way to capture the
molecular signals related to the structural and functional
differentiation of exocrine pancreatic epithelium.
BASIC AND
TRANSLATIONAL
PANCREAS
Figure 1. Pancreatic epithelial tubulogenesis and the related molecular alterations. (A, B) Representative confocal images of HPDE cell clusters (A) or
tubules (B) formed in 3D rBM. The structures were immunostained with a6-integrin (red) and b-catenin (green). Nuclei were counterstained with
40 ,6-diamidino-2-phenylindole (DAPI) (blue). Asterisks: cell-free lumen. Scale bars ¼ 100 mm. (C) Heat map showing expression patterns of 620
differentially expressed genes (DEGs) during HPDE tubulogenesis or PANC-1 spheroid formation. The heat map depicts high (red) and low (green)
relative levels of medium-centered gene expression in log space. (D) Fold changes in the transcript levels of CEL, CA9, MUC1, AGR2, and MUC20 in
HPDE or PANC-1 organoids as measured by quantitative reverse transcription polymerase chain reaction analysis. (E) Western blot analysis of bile
saltstimulated lipase, carbonic anhydrase 9, or mucin-1 in HPDE or PANC-1 3D organoids. b-tubulin was included as a loading control.
November 2013
Tubulogenesis-Specific Gene Signature of
PDAC
1113
rtubules the “tubule-like” PDAC. Intriguingly, we found that
the patients carrying the tubule-like tumors had much
longer postoperative survival than those with lower rtubules
(Figure 2B, left panel). We repeated this analysis in 2
independent series of patients with PDAC and consistently
found a favorable clinical prognosis in patients with
tubule-like tumors (the Johns Hopkins and the Northwestern/NorthShore cohorts; Figure 2B, right panels).7
Next, to identify a smaller set of genes that could optimally predict the clinical outcomes of patients with PDAC,
we constructed a risk score based on a Cox’s model to
predict patients’ survival after surgery. We selected a set of
28 genes whose performance in survival prediction reached
the maximum as assessed by C-index (Figure 2C and
Supplementary Table 2). We found that patients in the
high risk-score group had poor postoperative prognosis,
BASIC AND
TRANSLATIONAL
PANCREAS
As disruption of tissue microarchitectures is one of
the hallmark features of glandular cancers, including
PDAC,3–6 we investigated whether the gene expression
profile associated with pancreatic epithelial tubulogenesis
can carry prognostic information in PDAC. To this end, we
interrogated the transcriptomes of microdissected, cancer
cell-enriched tumor samples from a cohort of 27 patients
with localized PDAC (the University of California San
Francisco dataset).8 We determined the degree of resemblance between these tumors and HPDE tubules by calculating Pearson’s correlation coefficients (rtubules) based on
the expression of the 620 tubulogenesis-associated genes
(Figure 2A). We divided the patients into 2 subgroups
according to rtubules and designated the tumors with higher
TUBULOGENESIS SIGNATURE IN PANCREATIC CANCER
Figure 2. Identification of a tubulogenesis-specific gene signature in PDAC. (A) An illustration depicting the derivation of rtubules. (B) Kaplan-Meier
survival curve comparing postoperative survival of PDAC patients with high or low rtubules of their tumors. (C) Selection of a 28-gene gene set
with the highest accuracy (C-index) for the prediction of postoperative survival in PDAC. (D) Kaplan-Meier survival curves comparing postoperative
survival in PDAC. The patients were stratified into 2 groups based on predicted risk of relapse (risk score; RS) calculated by the 28-gene signature. (E)
Forest plots showing hazard ratios (with 95% confidence limits) of death according to the RS and clinicopathological criteria in a Cox proportionalhazards analysis. *P < .05; **P < .01. JHMI, Johns Hopkins Medical Institutions; NW/NSU, Northwestern Memorial Hospital/NorthShore University
Health System; UCSF, University of California San Francisco.
1114 WANG ET AL
and patients in the low risk-score group fared well
(Figure 2D). A multivariate Cox proportional-hazards
analysis confirmed that this tubulogenesis-specific signature was the strongest prognostic predictor and significantly outperformed clinicopathological criteria across
independent PDAC datasets (C-index ¼ 0.8000.899;
Figure 2E and Supplementary Tables 3 and 5). To further
enhance clinical utility, we sought to refine the prognostic
model and found that the 6 top-ranked genes from the
28-gene signature (Cox regression P < .005; Supplementary
Table 2 and Supplementary Figure 3), including ATP9A,
ASPM, ACOX3, CDC45L, SLC40A1, and AGR2, could form a
more condensed 6-gene signature that performed as
excellently as the 28-gene signature in the prognostic
prediction (C-index ¼ 0.8120.950; Supplementary
Figure 4 and Supplementary Tables 4 and 5).
ASPM as a Tissue-ArchitectureLSpecific
Prognostic Marker in PDAC
Among the constituent genes in the tubulogenesisspecific signature, ASPM exhibited the most prominent
GASTROENTEROLOGY Vol. 145, No. 5
transcriptional change during pancreatic tubulogenesis
(28.6-fold down-regulation; Figure 3A, arrow). This,
together with a strong correlation of ASPM with the
prognosis of PDAC patients (Cox regression P ¼ .0028),
incited us to investigate its prognostic and biological roles
in PDAC. We first asked if ASPM expression was regulated
in a tissue architecturedependent manner. We found
that the transcript level of ASPM markedly decreased when
HPDE cells were transferred from cell monolayers to 3D
rBM, and the level decreased further upon tubulogenesis
(Figure 3B). Interestingly, when we enzymatically digested
HPDE tubules and cultured the recovered cells in 3D rBM,
the transcript level of ASPM in the resultant “secondary
cell clusters” was restored to a level similar to that previously measured in the original cellular clusters, indicating
that the expression of ASPM is regulated in a contextdependent and reversible manner.
The human ASPM gene encodes a large (409.8 kDa)
and multifunctional protein that plays a critical role in
neurogenesis, neuronal migration, and the expansion of
glioma stem cells.17,18 Recently, ASPM expression has been
BASIC AND
TRANSLATIONAL
PANCREAS
Figure 3. ASPM as a tissue architecturespecific prognostic marker in PDAC. (A) The transcript levels of selected top-ranked genes in the
tubulogenesis-specific signature as measured by quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis. (B) Schematic
representation of the experimental protocol for growing and manipulating different HPDE tissue assemblies. Right, the transcript levels of ASPM as
measured by qRT-PCR analysis. Data are represented as mean SEM (n ¼ 3). *P < .05; **P < .01; ***P < .001 in (A) and (B). (C) Overall survival of
patients stratified based on the expression levels of ASPM in the University of California San Francisco (UCSF) (C[i]), Johns Hopkins Medical Institutions (JHMI) (C[ii]), or Northwestern/NorthShore (NW/NSU) (C[iii]) cohort of patients with PDAC. (D) Representative immunostaining of ASPM in
PDAC tissues (400 magnification). Shown are tumors with moderate (PDAC #1) or undetectable (PDAC #2) ASPM staining. (E) Distribution of the
staining intensities of ASPM in tumors in the National Cheng-Kung University Hospital (NCKUH) cohort. (F) Kaplan-Meier survival curves comparing
postoperative survival of patients in the NCKUH cohort stratified according to the staining intensities of ASPM.
linked with poor clinical prognosis in ovarian cancer and
hepatocellular carcinoma.17,19,20 In accordance with the
poor prognostic role of ASPM in human malignancies, we
found that PDAC patients with their tumors expressing
high transcript levels of ASPM fared poorly across 3
independent cohorts (Figure 3C). To corroborate these
findings, we carried out immunohistochemical staining of
the tumor tissues from another cohort of 53 patients with
PDAC (National Cheng-Kung University Hospital cohort;
Supplementary Table 1). We found that more than two
thirds of tumors exhibited weak (1þ) or moderate (2þ)
staining intensities for ASPM in cancer cells (Figure 3D
and E). In line with the tumor transcriptome data,
patients with their tumors exhibiting moderate ASPM
staining had a significantly higher risk of death after
surgery than those with weak or no ASPM staining
(Figure 3F).
ASPM Promotes PDAC Aggressiveness
To assess if ASPM plays a role in pancreatic
tumorigenesis, we surveyed a panel of PDAC cell lines and
Figure 4. The roles of ASPM
in the malignant behaviors of
PDAC cells. (A) The transcript
levels of ASPM in HPDE cells
and various PDAC cell lines
as measured by quantitative
reverse transcription polymerase chain reaction analysis. (B)
Western blot analysis of ASPM
in HPDE and PDAC cells.
b-tubulin was included as
a loading control. (C) ASPM
immunostaining in a representative PDAC tissue and the
adjacent
normal
exocrine
pancreatic
ducts
(400
magnification).
Right,
the
intensities of ASPM were
quantified using histological
score. (D) Immunoblots (left) or
confocal
images
(right)
showing effect of ASPM
knockdown on PDAC cells.
Cells were immunostained with
ASPM (red) with nuclei counterstained with 40 ,6-diamidino2-phenylindole (blue). Scale
bars ¼ 40 mm. (E) Line
graphs showing the rate of
growth of control- or ASPMshRNAtransduced
PDAC
cells. (F) Silencing of ASPM
attenuated the migratory capacity of PDAC cells. Data are
represented as mean SEM
(n ¼ 36). *P < .05; **P < .01;
***P < .001 vs HPDE (A), PDAC
(C) or control (E and F).
TUBULOGENESIS SIGNATURE IN PANCREATIC CANCER
1115
found that ASPM expression was significantly upregulated in most cancer lines relative to HPDE cells
(Figure 4A and B). Consistent with the cell line data,
ONCOMINE and immunohistochemistry analyses
revealed significantly increased transcript and protein
levels of ASPM in human PDAC tissues compared with
adjacent normal pancreatic ducts (Figure 4C and
Supplementary Figure 5B).21
To investigate the functional role of ASPM in PDAC
cells, we stably down-regulated its expression using
lentivirus-mediated RNA interference (Figure 4D). We
found that knockdown of endogenous ASPM expression
in primary tumor-derived PANC-1 cells by 2 different
shRNA vectors could significantly attenuate cellular proliferation (Figure 4E and data not shown) and migration
(Figure 4F). Silencing of ASPM also compromised the
proliferative and migratory capacities of metastatic AsPC-1
cells (Figure 4E and F, right panels). To further address the
oncogenic role of ASPM in vivo, we stably expressed a
luciferase reporter in AsPC-1 cells and orthotopically
implanted them into the pancreatic tail of nonobese
BASIC AND
TRANSLATIONAL
PANCREAS
November 2013
1116 WANG ET AL
diabetic/severe combined immunodeficient mice. As
shown in Figure 5, the tumors expressing the
ASPMshRNA grew significantly slower than the control
tumors and were associated with attenuated formation of
malignant ascites. The mice carrying ASPM-deficient
tumors exhibited significantly prolonged survival so that
they survived, on average, 37% (16.5 days) longer than the
control animals (log-rank test P < .001; Figure 5D).
Together, these functional studies indicate an important
role of ASPM in PDAC aggressiveness.
ASPM Maintains WntLb-catenin Signaling
in PDAC
We performed gene set enrichment analysis
comparing the transcriptomes of control- and ASPMshRNAtransduced PDAC cells and found the KEGG
Wnt signaling pathway significantly enriched (P < .001)
in the differential gene expression profile (Supplementary
Figures 6A and Supplementary Figure 7). To assess if
ASPM regulates Wnt pathway activity, we expressed a
Wnt reporter construct in PANC-1 and AsPC-1 cells and
found that silencing of ASPM dramatically blunted
Wnt-mediated luciferase reporter activation (Figure 6A).
b-catenin is an essential mediator of canonical Wnt
signaling and its active form frequently accumulates in
PDAC tissues and contributes to PDAC maintenance and
metastasis.22–24 We found that silencing of ASPM led to a
GASTROENTEROLOGY Vol. 145, No. 5
substantial reduction in b-catenin protein level
(Figure 6B), raising the possibility that ASPM can
maintain Wnt pathway activity by regulating b-catenin.
To address this possibility, we conducted a series
of co-immunoprecipitation experiments, thereby
identifying several upstream positive regulators of
b-catenin, including dishevelled (Dvl)-2, axin, and
protease-activated receptor-1, which interacted with
endogenous ASPM (Supplementary Figure 8 and
Figure 6C). Interestingly, silencing of ASPM resulted in a
dramatic decrease in the protein level of Dvl-2, but not
that of axin or protease-activated receptor-1, without
affecting its transcript level (Figure 6B and
Supplementary Figure 9B and 9C). Given that Dvl stimulates canonical Wnt signaling principally by attenuating
b-catenin degradation,25 it follows that ASPM can
promote Wnt pathway activity by increasing the protein
stability of Dvl-2. Consistently, confocal imaging and
reciprocal co-immunoprecipitation analyses confirmed
that endogenous ASPM interacted with and vastly
co-localized with endogenous Dvl-2 in the cytosolic
compartment near the plasma membrane in PDAC cells
(Figure 6C; Supplementary Figure 9D and E). Importantly, we found that silencing of ASPM markedly
increased the protein polyubiquitination of Dvl-2
(Supplementary Figure 9F), indicating that ASPM
might inhibit the proteasome-dependent degradation of
BASIC AND
TRANSLATIONAL
PANCREAS
Figure 5. ASPM promotes
pancreatic cancer aggressiveness.
(A)
Representative
bioluminescence images (BLI)
of nonobese diabetic severe
combined
immunodeficient
mice implanted in the pancreatic tails with firefly luciferase
labeled, control-, or ASPMshRNAtransduced AsPC-1
cells at the indicated time
points after cell inoculation. (B)
Tumor bulk quantified as BLI
normalized photon counts as a
function of time. (C) Representative images showing
presence of bloody ascites in
mice implanted with controlshRNAtransduced AsPC-1
cells, but not in animals
implanted with ASPM-shRNA
transduced cells. Right, the
amounts of ascites measured
at 6 weeks after cell implantation. Data are represented
as mean SEM (n ¼ 69).
*P < .05; **P < .01 vs control
in (B) and (C). (D) Percent survival as a function of time in
mice described in (A).
November 2013
TUBULOGENESIS SIGNATURE IN PANCREATIC CANCER
1117
Dvl-2. Indeed, treatment with a specific proteasome
inhibitor or ectopic expression of Dvl-2 in ASPMdeficient cells could restore Dvl-2 and b-catenin protein
levels (Figure 6D).
To further address if down-regulation of Dvl-2 and/or
b-catenin mediates the cellular effects induced by ASPM
deficiency, we ectopically expressed Dvl-2 or a constitutively active b-catenin (S33Y) mutant in ASPM-silenced
PANC-1 cells.14 Indeed, forced expression of Dvl-2 or
functional activation of b-catenin could significantly
increase Wnt reporter activity and the migratory potential
of ASPM-deficient cells (Figure 6E). The clinical relevance
of ASPM-dependent regulation of Wnt/b-catenin pathway
activity was credentialed by a strong positive correlation
between the staining intensities of ASPM and
cytoplasmic/nuclear b-catenin, which indicate active Wnt
signaling (Supplementary Figure 10),23 in human PDAC
tissues (n ¼ 27; r ¼ 0.539, P ¼ .004; Figure 6F).
ASPM Maintains Pancreatic Cancer Stemness
Studies have indicated a role of ASPM in regulating
neural stem cells.18,26 Consistently, a core stem cells-like
gene module that is activated in human cancers was
significantly enriched by gene set enrichment analysis in
the gene profile associated with silencing of ASPM in PDAC
cells (Supplementary Figure 6B).27 This finding, together
with the roles of Wnt signaling in stem-like cells in PDAC
and gastrointestinal malignancies,23,24,28 prompted us to
investigate whether ASPM regulates pancreatic cancer
stemness. To this end, we measured the proportion of
BASIC AND
TRANSLATIONAL
PANCREAS
Figure 6. ASPM maintains Wnt pathway activity in PDAC. (A) Fold Wnt-mediated luciferase expression in control- or ASPM-shRNA (shRNA)
transduced PANC-1 or AsPC-1 cells. Data are represented as mean SEM (n ¼ 6). ***P < .001 vs control. (B) Western blot analysis of b-catenin and
Dvl-2 in control- or ASPM-shRNAtransduced HEK293 or PANC-1 cells. (C) Left, endogenous Dvl-2 co-immunoprecipitated with endogenous ASPM in
HEK293 or PANC-1 cells. Right, representative confocal images showing co-localization of ASPM (green) and Dvl-2 (red) in PANC-1 cells. Cell nuclei
were counterstained with 40 ,6-diamidino-2-phenylindole (blue). Scale bars ¼ 40 mm. (D) Control- or ASPM-shRNAtransduced PANC-1 cells were
treated with MG132 (10 mM for 12 h) or co-transfected with Dvl-2, after which b-catenin and Dvl-2 were detected by immunoblotting. b-tubulin
was included as a loading control in (B), (C), and (D). (E) Fold changes in the Wnt reporter activity or cellular migration of control- or
ASPM-shRNAtransduced PANC-1 cells with or without co-expression of Dvl-2 or b-catenin (S33Y). Data are represented as mean SEM (n ¼ 6).
*P < .01 vs control shRNA; †P < .01 vs ASPM shRNA plus empty vector. (F) Representative immunostaining of human PDAC tissues with moderate
ASPM/high cytoplasmic or nuclear b-catenin staining (PDAC #1) or undetectable ASPM/low cytoplasmic or nuclear b-catenin staining (PDAC #2; 400
magnification). Bottom, scatterplot of the staining intensities of ASPM vs cytoplasmic/nuclear b-catenin of 27 PDAC tissues with the linear regression line
shown.
1118 WANG ET AL
GASTROENTEROLOGY Vol. 145, No. 5
Figure 7. ASPM maintains pancreatic cancer stemness. (A) Representative plots showing patterns of CD133, CD44, CD24, and epithelialspecific antigen (ESA) staining of control- or ASPM-shRNAtransduced PANC-1 cells, with the frequency of the boxed CD44hiCD133hi (left)
or CD44hiCD24hiESAhi (right) cell population as a percentage of cancer cells shown. (B) Percentages of CD44hiCD133hi or CD44hiCD24hiESAhi
cell subpopulation in control- or ASPM-shRNAtransduced PANC-1 or AsPC-1 cells. (C) Transcript levels of ASPM and other stem cell markers
in CD44hiCD133hi and CD44loCD133lo PANC-1 cells as measured by quantitative reverse transcription polymerase chain reaction analysis.
(D) Representative phase-contrast images of tumorspheres formed by control- or ASPM-shRNAtransduced CD44hiCD133hi PANC-1 cells.
Bars ¼ 100 mm. (E) Bar graphs showing diameters of tumorspheres in (D). (F) The percentages of CD44hiCD133hi cell subpopulation in control- or
ASPM-shRNAtransduced PANC-1 cells and those co-transduced with Dvl-2 or b-catenin (S33Y). Data are represented as mean SEM (n ¼ 3).
**P < .01; ***P < .001 vs control in (B), (C), and (E). *P < .01 vs control shRNA; †P < .01 vs ASPM shRNA plus empty vector in (F).
BASIC AND
TRANSLATIONAL
PANCREAS
PDAC cells that co-expressed the surface markers CD44
and CD133 or CD44, CD24, and epithelial-specific antigen,
which are known to contain the enriched cancer stem cells
(CSCs) in PDAC (Figure 7A).15,29 We confirmed that these
pancreatic CSCs were capable of forming death-resistant
tumorspheres in vitro, exhibited tumor-initiating potentials in vivo, and profoundly responded to Wnt pathway
activation or inhibition (Supplementary Figure 11).
Importantly, silencing of ASPM led to a substantial reduction of the CSC subpopulation in different PDAC cells
(Figure 7B). Complementing these results, we detected
higher ASPM expression in pancreatic CSCs than their
non-CSC counterparts (Figure 7C) and found that depletion of ASPM markedly blunted CSC tumorsphere growth
in a nonadherent condition (Figure 7D and E). Importantly, forced expression of Dvl-2 or b-catenin (S33Y) in
ASPM-deficient PDAC cells could restore their CSC
subpopulations (Figure 7F), pointing to a pivotal role
of Wnt–Dvl-2–b-catenin signaling in pancreatic cancer
stemness regulated by ASPM.
Discussion
By using a physiologically relevant tissue organization model, we identified a transcript profile specific to
pancreatic epithelial tubulogenesis, and we presented
evidence for the favorable outcomes of PDAC patients
with tumors carrying this molecular pattern. This
biology-informed approach led to the identification of a
biologically tractable and clinically instructive molecular
signature for PDAC and ASPM-mediated Wnt pathway
activity as a novel mechanistic link between tissue
architecture, cancer stemness, and PDAC aggressiveness.
Our results offer insight toward the pathogenetic
significance of tissue-architectural formation in PDAC.
One of the major challenges in the care of patients with
localized PDAC is the ability to identify those patients
who are at risk for early relapse after surgery, such that
they can benefit from more aggressive therapy. However,
traditional pathological grading of PDAC fails to provide
an accurate prediction of tumor relapse. Recent efforts in
improving classification and outcomes prediction of
PDAC have been the molecular characterizations of the
excised tumors. For instance, Stratford and colleagues
identified a 6-gene metastasis-associated signature that
was predictive of survival in patients with early-stage
PDAC.7 By comparing the transcriptomes of 27 microdissected PDAC tissues, Collisson and colleagues identified a 62-gene signature that could be used to define
molecular subtypes in PDAC.8 In this classification
scheme, tumors classified as the “classical subtype,” which
expressed epithelium-associated markers, were associated
with favorable clinical outcomes. In this study, we
discovered a novel tubule-like subtype of PDAC based on
criteria informed by tissue architecture. The tubule-like
PDAC might be partially similar to the classical subtype
of PDAC, as they both exhibit the epithelial and secretory
properties of exocrine pancreas. Importantly, the biologyinformed approach used in our study could provide
additional mechanistic insights into PDAC differentiation
or de-differentiation, permitting identification of molecular pathways related to its pathogenesis.
ASPM was initially identified as a centrosomal protein
that regulates neurogenesis and brain size, and it was
later known to be widely expressed in a variety of normal
or malignant tissues.19,30 For example, ASPM expression
positively correlated with the pathological grade of
glioma and was up-regulated in recurrent tumors.17
ASPM expression also correlated with the pathological
grade and poor survival in patients with ovarian cancer or
hepatocellular carcinoma.19,20 Interestingly, ASPM was
both cytoplasmic and nuclear localized in interphase and
its cytoplasmic expression levels were highly variable
among tumors,19 suggesting that it might have diverse
biological functions in malignant tissues. Consistently,
we provided the first evidence demonstrating that ASPM
is a robust poor prognostic factor in PDAC and plays a
critical role in the malignant behaviors of PDAC cells as
well as pancreatic cancer aggressiveness. We dissected the
mechanisms underlying the oncogenic potential of
ASPM in PDAC, which was attributed to its ability to
promote Wnt pathway activity and cancer stemness by
positively regulating Dvl-2 and b-catenin. As Wnt
signaling is commonly activated in human PDAC and
plays a crucial role in cancer cell proliferation, survival,
and metastasis,22–24 it is plausible that ASPM-mediated
TUBULOGENESIS SIGNATURE IN PANCREATIC CANCER
1119
Wnt–Dvl-2b-catenin signaling plays a principal role in
its oncogenic functions in PDAC. Despite these findings,
additional studies are required to explore how expression
of ASPM is regulated by tissue architecture at the cellular
and molecular levels in normal or malignant pancreatic
epithelial tissues.
In conclusion, this is the first study to link tissue
architectureassociated molecular pattern with the
clinical behavior of PDAC. The study illustrates that
biology-informed molecular markers can greatly enhance
outcomes prediction in malignant tumors and provide
valuable diagnostic and therapeutic information.
Supplementary Material
Note: To access the supplementary material
accompanying this article, visit the online version of
Gastroenterology at www.gastrojournal.org, and at http://
dx.doi.org/10.1053/j.gastro.2013.07.040.
References
1. Ahmad NA, Lewis JD, Ginsberg GG, et al. Long term survival after
pancreatic resection for pancreatic adenocarcinoma. Am J Gastroenterol 2001;96:2609–2615.
2. Oettle H, Post S, Neuhaus P, et al. Adjuvant chemotherapy with
gemcitabine vs observation in patients undergoing curative-intent
resection of pancreatic cancer: a randomized controlled trial. JAMA
2007;297:267–277.
3. Potter JD. Morphogens, morphostats, microarchitecture and malignancy. Nat Rev Cancer 2007;7:464–474.
4. Gleason DF. Histologic grading of prostate cancer: a perspective.
Hum Pathol 1992;23:273–279.
5. Rakha EA, El-Sayed ME, Lee AH, et al. Prognostic significance of
Nottingham histologic grade in invasive breast carcinoma. J Clin
Oncol 2008;26:3153–3158.
6. Adsay NV, Basturk O, Bonnett M, et al. A proposal for a new and
more practical grading scheme for pancreatic ductal adenocarcinoma. Am J Surg Pathol 2005;29:724–733.
7. Stratford JK, Bentrem DJ, Anderson JM, et al. A six-gene signature
predicts survival of patients with localized pancreatic ductal adenocarcinoma. PLoS Med 2010;7:e1000307.
8. Collisson EA, Sadanandam A, Olson P, et al. Subtypes of pancreatic
ductal adenocarcinoma and their differing responses to therapy. Nat
Med 2011;17:500–503.
9. Fournier MV, Martin KJ, Kenny PA, et al. Gene expression signature
in organized and growth-arrested mammary acini predicts good
outcome in breast cancer. Cancer Res 2006;66:7095–7102.
10. Li CR, Su JJ, Wang WY, et al. Molecular profiling of prostatic acinar
morphogenesis identifies PDCD4 and KLF6 as tissue architecturespecific prognostic markers in prostate cancer. Am J Pathol 2013;
182:363274.
11. Weaver VM, Petersen OW, Wang F, et al. Reversion of the malignant
phenotype of human breast cells in three-dimensional culture
and in vivo by integrin blocking antibodies. J Cell Biol 1997;
137:231–245.
12. Wang Y, Klijn JG, Zhang Y, et al. Gene-expression profiles to predict
distant metastasis of lymph-node-negative primary breast cancer.
Lancet 2005;365:671–679.
13. Budwit-Novotny DA, McCarty KS, Cox EB, et al. Immunohistochemical analyses of estrogen receptor in endometrial adenocarcinoma
using a monoclonal antibody. Cancer Res 1986;46:5419–5425.
14. Kolligs FT, Hu G, Dang CV, et al. Neoplastic transformation of RK3E
by mutant beta-catenin requires deregulation of Tcf/Lef transcription
but not activation of c-myc expression. Mol Cell Biol 1999;
19:5696–5706.
BASIC AND
TRANSLATIONAL
PANCREAS
November 2013
1120 WANG ET AL
15. Li C, Heidt DG, Dalerba P, et al. Identification of pancreatic cancer
stem cells. Cancer Res 2007;67:1030–1037.
16. Pencina MJ, D’Agostino RB. Overall C as a measure of discrimination
in survival analysis: model specific population value and confidence
interval estimation. Stat Med 2004;23:2109–2123.
17. Bikeye SN, Colin C, Marie Y, et al. ASPM-associated stem cell proliferation is involved in malignant progression of gliomas and constitutes an attractive therapeutic target. Cancer Cell Int 2010;10:1.
18. Buchman JJ, Durak O, Tsai LH. ASPM regulates Wnt signaling pathway activity in the developing brain. Genes Dev 2011;25:1909–1914.
19. Bruning-Richardson A, Bond J, Alsiary R, et al. ASPM and microcephalin expression in epithelial ovarian cancer correlates with
tumour grade and survival. Br J Cancer 2011;104:1602–1610.
20. Lin SY, Pan HW, Liu SH, et al. ASPM is a novel marker for vascular
invasion, early recurrence, and poor prognosis of hepatocellular
carcinoma. Clin Cancer Res 2008;14:4814–4820.
21. Grutzmann R, Pilarsky C, Ammerpohl O, et al. Gene expression
profiling of microdissected pancreatic ductal carcinomas using highdensity DNA microarrays. Neoplasia 2004;6:611–622.
22. Wang L, Heidt DG, Lee CJ, et al. Oncogenic function of ATDC in
pancreatic cancer through Wnt pathway activation and beta-catenin
stabilization. Cancer Cell 2009;15:207–219.
23. Pasca di Magliano M, Biankin AV, Heiser PW, et al. Common activation of canonical Wnt signaling in pancreatic adenocarcinoma.
PLoS One 2007;2:e1155.
24. Yu M, Ting DT, Stott SL, et al. RNA sequencing of pancreatic
circulating tumour cells implicates WNT signalling in metastasis.
Nature 2012;487:510–513.
25. Clevers H, Nusse R. Wnt/beta-catenin signaling and disease. Cell
2012;149:1192–1205.
26. Horvath S, Zhang B, Carlson M, et al. Analysis of oncogenic signaling
networks in glioblastoma identifies ASPM as a molecular target. Proc
Natl Acad Sci U S A 2006;103:17402–17407.
27. Wong DJ, Liu H, Ridky TW, et al. Module map of stem cell genes
guides creation of epithelial cancer stem cells. Cell Stem Cell 2008;
2:333244.
GASTROENTEROLOGY Vol. 145, No. 5
28. Vermeulen L, De Sousa EMF, van der Heijden M, et al. Wnt activity
defines colon cancer stem cells and is regulated by the microenvironment. Nat Cell Biol 2010;12:468–476.
29. Hermann PC, Huber SL, Herrler T, et al. Distinct populations of
cancer stem cells determine tumor growth and metastatic activity in
human pancreatic cancer. Cell Stem Cell 2007;1:313–323.
30. Kouprina N, Pavlicek A, Collins NK, et al. The microcephaly ASPM
gene is expressed in proliferating tissues and encodes for a mitotic
spindle protein. Hum Mol Genet 2005;14:2155–2165.
Author names in bold designate shared co-first authorship.
Received December 5, 2012. Accepted July 24, 2013.
Reprint requests
Address requests for reprints to: Kelvin K. Tsai, MD, PhD, National
Institute of Cancer Research, National Health Research Institutes, 367
Shengli Road, Tainan 70456, Taiwan. e-mail: [email protected];
fax: þ886-6-208-3427.
Acknowledgments
Transcript Profiling: The gene expression data have been deposited in
NCBIs Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/
geo/) and are accessible through GEO Series accession number
GSE42270.
Conflicts of interest
The authors disclose no conflicts.
Funding
Supported in part by grants CA-100-PP-19 and CA-101-PP-19 from
National Health Research Institutes (K. K. Tsai); National Core Facility
Program for Biotechnology Grants from National Science Council (NSC
101-2319-B-400-001; I.-S. Chang); NSC 101-2628-B-400-003-MY2 (K. K.
Tsai); and Development of Cancer Research System Excellence Program
from Department of Health, Taiwan (DOH 100-TD-C-111-004, DOH 101TD-C-111-004, and DOH 102-TD-C-111-004).
BASIC AND
TRANSLATIONAL
PANCREAS