Oncogene (1998) 17, 1969 ± 1978
 1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00
Disruption of the antiproliferative TGF-b signaling pathways in human
pancreatic cancer cells
Alberto Villanueva1, Carmen GarcõÂ a1, Ana BeleÂn Paules2, MoÂnica Vicente1, Montse MegõÂ as1,
GermaÂn Reyes1, Priam de Villalonga2, Neus Agell2, Felix LluõÂ s1, Oriol Bachs2 and
Gabriel CapellaÂ1
1
Laboratori d'Investigacio Gastrointestinal, Institut de Recerca, Hospital de la Santa Creu i Sant Pau; and 2Departament de
Biologia Cellular, Facultat de Medicina, Universitat de Barcelona, Barcelona, Spain
Resistance to TGF-b1 occurred in pancreatic cancer cells
suggesting that inactivation of TGF-b inhibitory signaling pathways may play an important role in human
pancreatic cancer. The aim of our study was to
determine the presence of alterations in the main putative
components of the TGF-b inhibitory signaling pathways
(p15, Smad4, Smad2, TGFb-RII, CDC25A). A panel of
human carcinomas of the exocrine pancreas orthotopically implanted and perpetuated in nude mice and
pancreatic cancer cell lines were studied. p15 gene
alterations, mainly homozygous deletions that involved
exons 1 and/or 2, were found in the 62.5% (5 of 8) of
pancreatic xenografts whereas Smad4 gene aberrations
were found in one of eight xenografts and in two of seven
cell lines. Additional aberrations in these genes were
acquired during in vivo perpetuation and distal dissemination. Paradoxically, TGFb-RII overexpression and a
decrease in CDC25A protein levels were found in all
tumors and cell lines. In one cell line, resistance to TGFb1 occurred in the absence of alterations in the genes
analysed so far. We conclude that all human pancreatic
tumor cells analysed herein have non-functional TGF-b
pathways. The majority of cells harbor alterations in at
least one of the putative components of TGF-b pathways, mainly in p15 and Smad4 genes. These results
suggest that inactivation of TGF-b signaling pathways
plays an important role in human pancreatic tumorigenesis.
Keywords: pancreatic cancer; p15; Smad4; Smad2;
TGFb-RII, CDC25A
Introduction
TGF-b mediates growth inhibitory signals in di€erent
cell types. It acts as an adapter that binds and brings
together members of two subfamilies of transmembrane serine/threonine kinases named TGF-b type I
and type II receptors (MassagueÂ, 1996). In the
resulting complex the type II receptor phosphorylates
the type I, which then propagates the signal (Ebner et
al., 1993; Wrana et al., 1994). SMAD2, a member of
the recently identi®ed SMAD protein family, has been
proposed to be directly phosphorylated by the
Correspondence: G CapellaÂ
The ®rst two authors contributed equally to this study
Received 20 February 1998; revised 7 May 1998; accepted 8 May
1998
activated TGF-b receptors (Liu et al., 1996; MaciasSilva et al., 1996; Nakao et al., 1997). Upon
phosphorylation, SMAD2 binds to SMAD4 (Lagna
et al., 1996), which plays a central role as the shared
hetero-oligomerization partner of the other SMAD
proteins (Shi et al., 1997). These speci®c complexes
translocate to the nucleus (Macias-Silva et al., 1996;
Nakao et al., 1997; Lagna et al., 1996; Eppert et al.,
1996), where they become associated with DNA
binding proteins such as Fast-1, and regulate
transcription (Chen et al., 1996).
TGF-b1 inhibits growth by causing cell cycle arrest
after increasing the concentration of the CDK4/6
inhibitor P15 (Hannon and Beach, 1994). The
association of P15 to Cyclin D1-CDK4/6 complexes
releases P27 that subsequently binds to, and inhibits
Cyclin A-CDK2 complexes (ReynisdoÂttir and MassagueÂ, 1997). Thus, it has been suggested that P15
expression would be under the transcriptional control
of SMAD2/4 complexes formed after the activation of
TGF-b receptors (Macias-Silva et al., 1996; Nakao et
al., 1997; Lagna et al., 1996). Until now, no other
extracellular signals have been shown to modify p15
levels. Recently, it has been described that TGF-b can
also cause inhibition of CDK4/6 activities by increasing their level of tyrosine phosphorylation. Tyrosine
phosphorylation and inactivation of CDK4/6 result
from the ability of TGF-b1 to inhibit the expression of
CDC25A, a CDK tyrosine phosphatase (Galaktionov
and Beach, 1991; Galaktionov et al., 1995). Since
repression of CDC25A and cell-cycle arrest induced by
TGF-b1 also occur in cells lacking p15 gene it has been
proposed that induction of P15 and repression of
CDC25A are independent mechanisms mediating the
inhibition of CDK4/6 activities (Iavarone and MassagueÂ, 1997).
TGF-b1 is involved in a variety of cell functions
including cell growth, cell di€erentiation and cell ± cell
or cell ± substrate interaction. Since TGF-b1 inhibits
cell proliferation, the disruption of its signaling
pathways could predispose to, or participate in the
genesis of cancer. Alterations in single members of the
TGF-b signaling pathways have been detected in a
growing number of tumors. The TGF-b type II
receptor is inactivated in a subset of gastrointestinal
cancers (Akiyama et al., 1997; Markowitz et al., 1995;
Souza et al., 1997) whereas it is overexpressed in
human pancreatic cancer (Friess et al., 1993). Although
the TGF-b type I receptor can occasionally be altered
in cancer cells (Kim et al., 1996a) little evidence is
available to support an important role of TGF-b1
alterations in tumor progression. Moreover, inactivat-
Disruption of TGF-b signaling pathways in pancreatic cancer
A Villanueva et al
1970
ing mutations in Smad4 and Smad2 genes have also
been identi®ed. Smad4 (previously known as DPC4)
(Derynck et al., 1996), a tumor-suppressor gene located
in chromosome 18q21.1, is inactivated in up to 40% of
pancreatic cancers (Hahn et al., 1996a,b) and in a
lower proportion of other tumors (Schutte et al., 1996;
Kim et al., 1996b; Thiagalingam et al., 1996; Nagatake
et al., 1996; Barret et al., 1996). Smad2 (previously
MADR-2) (Derynck et al., 1996), also located at 18q21,
has been found mutated in a proportion of colon
cancer (Eppert et al., 1996). Finally, while the incidence
of P15 protein inactivation in primary pancreatic
tumors remains unknown, p15 gene has been found
altered in approximately 50% of pancreatic cancer cell
lines (Naumann et al., 1996).
Previous studies have shown that the use of
pancreatic xenografts increases the sensitivity of
detection of allelic losses and enables the identification of homozygous deletions (i.e., p16 and DPC4
genes) that otherwise would have not been detected in
primary tumors (Caldas et al., 1994; Hahn et al.,
1996b). Although tumor cell selection probably occurs
during in vivo perpetuation, orthotopic xenografts are
close to their corresponding human pancreatic
primary tumors and have proved useful for the study
of those genetic events underlying distal dissemination
(Reyes et al., 1996).
Although scattered information is available, no
studies have attempted to simultaneously analyse
putative members of the TGF-b pathways in cancer
cells. The aim of this study was to determine whether
the TGF-b signaling pathways are disrupted in human
pancreatic cancer. We have analysed the presence of
concurrent alterations in members associated so far
with these pathways (TGFb-RII, Smad2, Smad4, p15,
and CDC25A) in a panel of orthotopic xenografts of
human pancreatic carcinomas and their corresponding
metastases, as well as in a series of pancreatic tumor
cell lines. Here we show that TGF-b signaling
pathways are disrupted in the majority of human
pancreatic xenografts and their metastases. p15 gene is
frequently inactivated in these tumors while Smad4
gene alterations are less common. Moreover, alterations in p15 and Smad4 genes may occur concomitantly
and can be acquired during distal dissemination.
Results
E€ect of TGF-b1 on DNA synthesis and proliferation
The e€ect of TGF-b1 on DNA replication was
analysed in ®ve cell lines (NP9, NP18, NP29, NP31
and PANC-1). Cells were treated with a high dose of
TGF-b1 (5 ng/ml) in a medium containing 10% FBS,
or in serum-free conditions, and cell proliferation was
assessed as described in Materials and methods. While
TGF-b1 clearly inhibited proliferation of peripheral
blood lymphocytes (data not shown), only a modest
inhibition was observed in NP18, NP31 and PANC-1
cells (Table 1). Lack of TGF-b1 response in NP9 and
NP29 cells was observed both in serum-free conditions
and in the presence of FBS (Table 1). Interestingly,
TGF-b1 induced changes in the morphology of NP31
cells (Table 1) in the absence of serum.
Genetic status and expression of TGF-bRII gene
The analysed repeats were normal in all xenografts and
metastases (Table 2; Figure 1a) and in all cell lines
(Table 3). RT ± PCR analysis showed that the TGFbRII transcript was present at high levels in all samples
(Table 2; Figure 1b). Moreover, all pancreatic xenografts
and cell lines showed increased TGF-bR11 mRNA levels
when compared to normal pancreatic tissues as assessed
by Northern blotting (Figure 1c). This overexpression
was also observed in metastases obtained during in vivo
dissemination in nude mice (Table 2; Figure 1c).
Alterations and expression of Smad4 and Smad2 genes
Three types of genetic aberrations in the Smad4 gene
were found: homozygous deletions, an abnormal SSCP
pattern and loss of expression. STS markers showed
that one (NP29) of the eight xenografts (Table 2; Figure
2a), and two (NP29, BxPC3) of the seven cell lines
(Table 3; Figure 2a) harbored homozygous deletions
encompassing the complete Smad4 locus. On the other
hand, in xenograft NP18 (Figure 2b) and its derived cell
line (data not shown) an abnormal SSCP pattern was
found when using primer set #2 (Table 4). While an
inactivating mutation or a polymorphism may be
present in this fragment, it could not be detected by
direct sequencing. RT ± PCR analysis con®rmed that
full-length Smad4 transcripts were absent in samples
containing homozygous deletions (Tables 2 and 3).
Finally, in three of the xenografts that retained all DNA
markers analysed (NP9, NP40 and NP43), expression of
Smad4 was progressively lost in advanced (more than 3)
passages in nude mice (Figure 2c). It is worth
mentioning that NP9 cell line retained normal levels
of Smad4 expression, whereas its xenograft of origin lost
it in advanced passages. Quantitation of Smad4
transcripts by Northern blot analysis was not technically feasible suggesting that it might be expressed at
very low levels.
One or more Smad4 STS markers were lost in all four
metastases derived from NP37 tumor and in one liver
metastasis from NP18 tumor (Table 2, Figure 3a). Smad4
transcript was also absent in these samples (Figure 3).
Table 1 Inhibitory e€ect of TGF-b1 on proliferation and DNA synthesis of human pancreatic cancer cells.
Cell line
MTT
Thymidine incorporation
Morphology
NP9
NP18
NP29
NP31
Panc-1
0%
15%
0%
20%
20%
0%
10%
0%
15%
15%
no change
no change
no change
cell elongationa
no change
Data are referred to untreated controls. All experiments were performed in the absence and in the presence of 10% FBS. aThis e€ect was
evidenced in the absence of serum
3
1
4b
Metastasis
+++
+++
+++
+++
+++
25
+
+
+
+
+
+
+
+
+
+
7
+
+
+
+
+
+
+
+
+
7
+
+
+
+
+
+
+
+
DPC4±
Exon 1
+
+
+
+
+
+
+
+
+
p0630±
H5-T7
7
+
+
+
+
+
+
+
+
+
+
(7)
(7)
(7)
+
+
+
+
+
p128±
N21
7
+
+
+
+
+
(7)
+
+
+
+
(7)
(7)
(7)
+
+
+
+
+
7
+
+
+
+
+
(7)
+
+
+
+
(7)
(7)
(7)
+
+
+
+
+
Smad4
c917±
DPC4±
46-T7
Exon 11
7
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
y747±
A6R
7
+c
+c
+
+
+
(7)
+
+
+
+
(7)
(7)
(7)
+c
+
+
+
+
RNA
wt
wt
wt
wt
wt
wt
wt
wt
wt
wt
wt
wt
wt
wt
abnormald
abnormald
±
abnormald
abnormald
abnormald
7
wt
wt
wt
wt
wt
wt
wt
wt
wt
7
7
7
wt
wt
wt
wt
wt
+
+
+
+
+
+
+
+
7
7
wtf
mutef
wtf
wtf
7
7
7
7
+
+
7
7
7
7
+f
+f
+f
+f
(7)
(7)
(7)
(7)
(7)
7
7
7
7
+f
(7)
7
7
7
7
7
7
7
+
nd
nd
nd
nd
nd
+
nd
nd
nd
7
nd
7
nd
wt
+/7
+/7
++h
+/7
++
+/7
+/7
++g
++
++
++
++
nd
nd
++
+
+
+
+
p15
p16
CDC25A2
Exon 2 RT ± PCR Protein
Protein
wtf
(7)
(7)
(7)
(7)
(7)
7
7
7
7
wtf
(7)
7
7
7
Smad2
Mutation RT ± PCR Exon 1
+++ overexpression compared to normal pancreatic tissues. The intensity of normal pancreas was set as + and progressive increases were semiquantitatively scored, ++ to ++++. 2Normal pancreatic
tissues showed high levels of CDC25A; aPancreatic xenografts were analysed in passages #1, 3 and/or 4; bAcquired genetic alterations during dissemination are highlighted using bold italic letters and
parentheses; cLoss of expression occurred in advanced passages that was not evident in the ®rst passage. dAn abnormal SSCP pattern was observed using primer set #2 (Table 4); eCGG(a)ATG to CGG(g)ATG;
f
Advanced passages showed homozygous deletions with concordant loss of expression that were not evidenced in the ®rst passage after implantation; gOne peritoneal metastasis showed a decrease of CDC25A
expression; hLoss of expression occurred in advanced passages. wt, wild type; 7, absence of PCR product; +, presence of PCR product
1
Number of samples: 15
Lymph node involvement and malignant ascitic cells
+++
NP37
2
+++
mediastinal lymph node mets. 2b
+++
celiac trunk lymph node met. 1b
b
ascitic ¯uid cells
1
+++
Hematogenous and peritoneal dissemination
+++
NP18
2
+++
liver mets.
6b
b
+++
liver mets.
1
b
+++
peritoneal mets.
4
+++
retroperitoneal mets.
1b
renal implant
1b
+++
Absence of distal dissemination
NP29
1
+++
NP40
2
++++
NP43
2
+++
NP46
2
+++
Peritoneal dissemination
NP9
2
peritoneal mets.
NP31
2
peritoneal mets.
retroperitoneal mets.
Xenograft
a
TGF-bRII1
RT ± PCR
and/or
Northern
Table 2 Analysis of perpetuated pancreatic xenografts and their metastasis obtained during distal dissemination in nude mice
Disruption of TGF-b signaling pathways in pancreatic cancer
A Villanueva et al
1971
Disruption of TGF-b signaling pathways in pancreatic cancer
A Villanueva et al
1972
Finally, no point mutation or loss of expression was
detected in the conserved domains of the Smad2 gene in
any of the pancreatic tumors, metastases or cell lines
analysed (Tables 2 and 3).
p15 gene aberrations
CGC(g)ATG) that does not a€ect RNA processing.
In the remaining three tumors (NP9, NP18 and
NP29) homozygous deletion of exon 1 could be
evidenced only in advanced passages. Thus, in
contrast with the high degree of genetic stability
regarding K-ras, p53, p16 genes status (Reyes et al.,
Four out of eight (50%) tumors have homozygous
deletions of exons 1 and/or 2 of the p15 gene (NP31,
NP37, NP43 and NP46). A point mutation in exon 1
was detected in NP40 tumor (Table 2; Figure 4a); the
signi®cance of this mutation is currently unknown
since it is a single base substitution at nucleotide 71
just before the initiating codon (CGC(a)ATG to
Figure 1 Analysis of the TGFb-RII gene in perpetuated human
pancreatic tumors orthoimplanted in nude mice and their
metastases. (a) SSCP/PCR analysis of DNA at (A)10 and (GT)3
repeats. Xenograft NP18 in di€erent passages and their
metastases showed a normal mobility pattern. Co-ampli®cation
of mouse DNA occurred in (GT)3 repeats (arrowhead) but SSCP
analysis allowed identi®cation of the human fragment. (b and c)
RT ± PCR and Northern blot analysis of total RNA from
xenografts NP9, NP18 and NP40 in di€erent passages and their
metastases. All xenografts and their metastases showed increased
TGFb-RII mRNA levels when compared with normal human
pancreas. cDNA ampli®cation of the p53 gene and b-actin
hybridization were used as a control for RNA quality. X#1,
xenograft in passage 1; X#3, xenograft in passage 3; PM,
peritoneal metastasis; LM, liver metastasis; Mouse Pa, normal
mouse pancreas; Human Pa, normal human pancreas; M, DNA
size marker
Figure 2 Genetic aberrations and expression of the Smad4 gene
in pancreatic xenografts and cell lines. (a) Xenograft NP29 and its
corresponding cell line, as well as BxPC3 cell line, harbored
homozygous deletion of DPC4 Exon 1, p128N21, c917-46-T7,
and DPC4 Exon 11 STS markers. Co-ampli®cation of mouse
DNA occurred in DPC4 Exon l and DPC4 Exon 11 markers, but
the SSCP analysis allowed identi®cation of the human pattern.
Deletions were absent in xenografts NP40 and NP43. (b) SSCP/
PCR analysis using primer set #2 (Table 3) showed an abnormal
mobility pattern in xenograft NP18 (arrowhead). In contrast, NP9
and NP37 showed normal mobility patterns. (c) RT ± PCR
ampli®cation of the whole coding region of the Smad4 gene.
Expression of Smad4 gene was progressively lost in advanced
passages of NP40 and NP43 xenografts. X#1 to 5, xenografts in
passage 1 to 5; Line, NP29 cell line; H2O, unpurposefully added
DNA. Other abbreviations as in Figure 1
Table 3. Analysis of pancreatic cancer cell lines.
TGF-bRII
RT ± PCR
Cell lines
NP9
++
NP29
++
NP31
++
BxPC3
++
MiaPaCa-2
++
PANC-1
++
NP18
++
Normal human tissues
Pancreas
+
a
Smad4
DNA markers RT ± PCR
Smad2
RT ± PCR
DNA
p15
RT ± PCR
Protein
p16
Protein
p27
Protein
CDC25A
Protein
No loss
Homoa
No Loss
Homoa
No loss
No loss
No loss
wt
7
wt
7
nd
nd
abnormalb
wt
wt
wt
wt
wt
wt
wt
wt
wt
Homoc
Homocd
Homocd
Homocd
wt
+
+
7
7
7
7
+
+
+
7
7
7
7
+
7
7
7
nd
nd
nd
+
+
++
++
nd
nd
nd
+
+++
++
++
nd
nd
nd
+++
No loss
+
wt
wt
+
nd
+
+
+++
Homozygous deletion at Smad4 gene; bAn abnormal SSCP pattern was observed using primer set #2 (Table 4); cHomozygous deletion of exon
1 and 2; dWhen cell lines were inoculated in nude mice the resulting tumors showed the same homozygous deletion; wt, wild type allele;+,
presence of RT ± PCR product or protein;7 absence of RT ± PCR product or protein; nd, not done
Disruption of TGF-b signaling pathways in pancreatic cancer
A Villanueva et al
Table 4
Gene
Primer
set
Amino
acid
Amplicon
size(bp)
97 ± 559
1386
5'-ACTAGAGACAGTTTGCCATGA-3'
5'-GCCGTCTTCAGGAATCTTCT-3'
1
1 ± 120
360
2
100 ± 225
375
3
212 ± 335
369
4
323 ± 442
364
5
429 ± 553
372
5'-ATGGACAATATGTCTATTA-3'(*)
5'-GTCAAACGCATACTGACA-3'
5'-GGCCTGATCTTCACAAAA-3'
5'-AGGCTGACTTGTGGAAG-3'
5'-GCCAACTTTCCCAACATT-3'
5'-TACCTGAACATCCATTTCA-3'
5'-GGTGTTCCATTGCTTAC-3'
5'-CTGACGCAAATCAAAGAC-3'
5'-GATCTACCCAAGTGCATA-3'
5'-TCAGTCTAAAGGTTGTGG-3' (*)
1 ± 137
411
TGF-bRII
Smad4
p15
Primer sets designed for PCR ampli®cation
Primer sequences
5'-ATGCGCGAGGAGAACAG-3'
5'-TCAGTCCCCCGTGGCTGT-3'
(*) When these primers were used the whole coding region of the Smad4 gene (1659bp) was ampli®ed
Figure 3 Acquisition of genetic aberrations and expression of
Smad4 gene during distal dissemination. (a) Two lymph node
metastases (LN) and cells in ascitic ¯uid (AS) from NP37
xenograft harbored an additional homozygous deletion involving
p128-N21, c917-46-T7, DPC4 Exon 11 and y747A6R markers.
One liver metastasis (LM) from NP18 xenograft harbored a
homozygous deletion of c917-46-T7, DPC4 Exon 11 and
y747A6R markers. Co-ampli®cation of mouse DNA occurred in
DPC4 Exon 1 and DPC4 Exon 11 markers; SSCP analysis
showed that the human fragment was ampli®ed in Exon 1
whereas the mouse fragment was ampli®ed in DPC4 Exon 11
(data not shown; see also Figure 1). (b) RT ± PCR ampli®cation
of the whole coding region showing that the full-length transcript
of Smad4 gene was absent in metastases which harbored
homozygous deletions. Abbreviations as in previous ®gures
1996) and allelotype (Hahn et al., 1995), all pancreatic
xenografts acquired p15 gene aberrations after several
passages. All metastases derived from early passages
of NP9 and NP18, which initially harbored a normal
p15 gene, also acquired exon 1 homozygous deletions
Figure 4 Genetic aberrations and expression of the p15 gene. (a)
PCR analysis showed homozygous deletions of Exons 1 and 2 of
p15 gene. Xenograft NP31 harbored homozygous deletion of both
exons; xenograft NP40 that initially had a point mutation in Exon
1 of p15 gene (*) , acquired an additional loss of Exon 1 in
advanced passages. Metastases derived from early (53) passages
of xenograft NP18 acquired additional homozygous deletions of
Exon 1. (b) p15 transcript was present in initial passages of
xenograft NP40 but expression was lost in advanced passages. p15
transcript was absent in all passages of xenografts NP37 and
NP31. Expression was progressively lost in advanced passages as
well as in all metastases obtained during in vivo dissemination of
NP9 and NP18 xenografts. Abbreviations as in previous ®gures
(Table 2; Figure 4a). RT ± PCR analysis con®rmed
that full-length p15 transcript was absent in these
samples (Figure 4b).
It is of interest that NP40 tumor (that initially had a
point mutation before the initiating codon of exon 1),
and NP43 and NP46 tumors (that initially had lost
exon 2 of the p15 gene) acquired additional losses in
exon 1 in advanced passages. Since p15 and p16 genes
are located in the 9p21 region, P16 protein analysis,
although not directly related to the TGF-signaling
pathway, was also included to compare their expression pattern. It is noteworthy that in two tumors
(NP18, NP37) p15 aberrations occurred independently
of p16 gene status (Table 2).
1973
Disruption of TGF-b signaling pathways in pancreatic cancer
A Villanueva et al
1974
note that levels of expression of CDC25A were
independent on the status of p15 and Smad4 genes.
In agreement with genetic analysis, NP31, MiaPaCa2, PANC-1 and BxPC3 cell lines have lost expression
of P15 protein (Table 3; Figure 5a). Finally, levels of
P27 protein, also involved in the response to TGF-b1,
were similar to those of normal pancreas in all cell lines
analysed (Table 3).
The modest inhibition of growth observed in NP18,
NP31 and Panc-1 cells and morphological changes in
NP31 cells, prompted us to analyse changes in P27
and CDC25A levels after TGF-b1 treatment. Synchronized serum-deprived cell cultures were treated for
24 h with 5 ng/ml of TGF-b1, and protein levels were
analysed at di€erent time intervals. As shown in
Figure 5a, no e€ect of TGF-b1 on P27 or CDC25A
expression was observed in any of the three cell lines,
independently of the status of the TGF-b pathways.
Similarly, expression of P15 was analysed in NP18 cell
line. Results showed that P15 protein was not
expressed in synchronized NP18 cells, even in the
presence of TGF-b1 (Figure 5a), suggesting that other
pathways could regulate P15 expression in these cells.
Figure 5 Expression of CDC25A, P15 and P27 proteins. (a)
Asynchronously growing cells: In the absence of TGF-b P15
protein was undetectable in NP31 and BxPC3 cells which
harbored homozygous deletions in p15 gene; expression of
CDC25A protein in NP9 and NP18 was similar to human
pancreas, whereas it was lower in NP29 and NP31 cells.
Synchronously growing cells: No e€ect of TGF-b1 on P15, P27
and CDC25A expression was observed in any cell line at 0, 6 and
12 h of TGF-b treatment. Note that P15 protein was not
expressed in NP18 cells when synchronously growing. (b)
CDC25A protein expression was lower in all xenografts when
compared to normal human pancreas. Additional decreases were
observed during tumor perpetuation and distal dissemination.
Abbreviations as in previous ®gures
Four (NP31, MiaPaCa-2, PANC-1 and BxPC3) out
of seven cell lines analysed harbored p15 gene
homozygous deletions that involved both exons
(Table 3). Interestingly, an intact p15 gene was
maintained in NP9 and NP18 cell lines after 60
passages (data not shown), whereas the corresponding
perpetuated tumors progressively lost it.
Expression of CDC25A, P15 and P27
Since it has recently been reported that CDC25A is a
positive cell cycle e€ector of the TGF-b pathways that
is overexpressed in breast cancer (Iavarone and
MassagueÂ, 1997; Galaktionov et al., 1995), the levels
of this protein were analysed by Western blotting.
Expression of CDC25A protein was diminished in all
xenografts and cell lines with respect to normal
pancreas (Table 2; Figure 5a); decrease was more
evident in tumors NP29, NP31, NP40 and NP46 (Table
2). Since immunocytochemistry was not performed it
cannot be ruled out that most of CDC25A expression
in normal pancreas is present in acinar cells. It must be
emphasized, however, that metastases derived from
early passages of NP9 tumor, and all liver and some
peritoneal metastases from NP18 tumor, showed
additional decreases in CDC25A level (Figure 5b)
suggesting that cells containing low levels of CDC25A
are selected for during in vivo dissemination. It is of
Discussion
The presence of genetic aberrations and changes in
protein expression in some of the known components
of the TGF-b signaling pathways (TGF-bRII, Smad2,
Smad4, p15 and CDC25A) have been simultaneously
analysed in a panel of human orthotopic pancreatic
xenografts and their corresponding metastases in nude
mice, as well as in their derived pancreatic tumor cell
lines. The results reveal that all xenografts and their
derived metastases and cell lines (with exception of
NP9 cell line) harbor alterations in at least one of the
putative components of TGF-b pathways, mainly in
p15 and Smad4 genes. Since growth of the NP9 cell
line was not inhibited by TGF-b1, we conclude that
all human pancreatic xenografts, metastases and cell
lines analysed herein have non-functional TGF-b
pathways. These results suggest that inactivation of
TGF-b signaling pathways plays an important role in
human pancreatic tumorigenesis.
Interestingly, two of the analysed putative components of the TGF-b pathways, p15 and Smad4 genes,
are inactivated in the pancreatic tumors studied. p15
gene homozygous deletions that involve exons 1 and/
or 2, were found in 50% of early passages of
pancreatic xenografts. The incidence of p15 aberrations in primary carcinomas of the human pancreas
remains unknown. The use of xenografted tumors has
allowed the detection of homozygous deletions that
otherwise would not have been detected in primaries
because of the presence of contaminating stromal
cells. The identi®cation of p15 aberrations in early
passages of the majority of samples suggests that p15
gene, a well-established cell cycle e€ector of TGF-b in
some model systems, plays an important role in
human pancreatic tumorigenesis. While it cannot be
ruled out that p15-de®cient tumors may be overrepresented in our xenografts, it is likely that its
inactivation may be present in a signi®cant proportion
of primary tumors. The identi®cation of p15 aberrations in early passages of the majority of samples
Disruption of TGF-b signaling pathways in pancreatic cancer
A Villanueva et al
suggests that p15 gene, a well-established cell cycle
e€ector of TGF-b in some model systems, plays an
important role in human pancreatic tumorigenesis.
While it cannot be ruled out that p15-de®cient tumors
may be overrepresented in our xenografts, it is likely
that its inactivation may be present in a signi®cant
proportion of primary tumors. Alterations in the p15
gene might be responsible for the resistance to TGF-b
inhibitory e€ects in a high proportion of pancreatic
tumors. It must be emphasized that p15 is a universal
cyclin/cdk inhibitor and that could also be the cellcycle e€ector of other signaling pathways.
Both p15 and p16 genes are located in the 9p21 region
and losses of this region occur at high frequency in
pancreatic carcinomas (Reyes et al., 1996; Hahn et al.,
1995). Co-deletions of p15 and p16 genes have been
previously reported in pancreatic cancer cell lines and it
has been suggested that p16 is more likely the target gene
in this region (Naumann et al., 1996; Huang et al., 1996).
In agreement with other studies (Caldas et al., 1994), we
showed that the majority of pancreatic xenografts
contained p16 genetic aberrations (Reyes et al., 1996).
Interestingly, the concurrent analysis of p15 and p16
genes in the xenografts demonstrates that p15 gene
aberrations occur independently of p16 status, and
suggests that p15 may be also an important target gene
in the 9p21 region.
Homozygous deletions and inactivating intragenic
mutations in the Smad4 gene have been previously
reported in approximately 40% of pancreatic xenografts
(Hahn et al., 1996a,b), and in a lower proportion of
tumors arising from other organs (Schutte et al., 1996;
Kim et al., 1996b; Thiagalingam et al., 1996; Nagatake
et al., 1996; Barret et al., 1996). We have detected a
homozygous deletion in the Smad4 gene in only one of
the eight xenografts suggesting that these alterations
may not be as frequent as previously reported. It could
be speculated that the geographical di€erences could
account for these di€erences as has been previously
shown for ras mutations (Scarpa et al, 1994). On the
other hand, Smad2 gene was not altered in pancreatic
tumorigenesis, in contrast with the presence of Smad2
gene mutations in sporadic colon tumors (Eppert et al.,
1996). The di€erential pattern of genetic aberrations in
Smad4 and Smad2 may re¯ect some degree of tissuespeci®city among SMAD proteins.
Orthotopic implantation of human pancreatic
cancers in the nude mice facilitates the reproduction,
in the experimental animal, of their distal dissemination patterns. Using this model system we have shown
that p15 and Smad4 genes aberrations accumulate
during distal dissemination of early passages of
orthotopic xenografts indicating that these alterations
may confer selective growth advantage in vivo.
Previously we showed that ras and p53 gene
mutations but not p16 alterations are acquired during
dissemination (Reyes et al., 1996). Altogether these
®ndings reinforce the role of p15 and Smad4 genetic
alterations in the more advanced stages of human
pancreatic tumorigenesis. The fact that concomitant
inactivation of both genes have been found in
experimental metastases suggests that P15 and
SMAD4 proteins may play complementary roles in
the regulation of human pancreatic cancer cell growth,
and re¯ects the complexity of the TGF-b signaling
pathways.
In previous reports it has been shown that human
pancreatic xenografts show a high degree of genetic
stability regarding K-ras, p53, p16 genes status (Reyes
et al., 1996). In contrast, additional p15 and/or Smad4
genes alterations accumulate in advanced passages of
xenografts. P15 is absent from all xenografts in their
advanced passages. It is noteworthy that additional
exon 1 losses can occur in the presence of previous
exon 2 losses. This might re¯ect the loss of p10, a
recently identi®ed transcript that shares exon 1 with
p15 gene (Tsubari et al., 1997), similar to what has
been described for the INK4a locus (Serrano et al.,
1996). Finally, loss of expression, a novel mechanism
of Smad4 inactivation in vivo, has been observed in
advanced passages of several of our pancreatic
xenografts lines suggesting that Smad4 gene may be
rendered inactive by transcriptional regulation or by
RNA instability. Finally, it should be noted that p15
and Smad4 genes are simultaneously inactivated in the
majority (5 out of 8) of advanced passages of our
orthotopic pancreatic xenografts reinforcing the
putative complementary roles of both proteins.
The pattern of genetic alterations present in
established tumor cell lines may di€er from primaries
or perpetuated tumors regarding speci®c genes (i.e. p16
or p53) (Huang et al., 1996). The incidence of Smad4
gene alterations in the panel of cell lines analysed (3 of
7) is higher than in xenografts indicating that these
alterations may be selected for during in vitro
immortalization. On the other hand, the percentage
of alterations in p15 gene in our tumor cell lines is
somewhat lower than in the corresponding xenografts
(43%) suggesting that p15 loss does not confer selective
growth advantage in vitro.
TGF-b has been also shown to inhibit CDC25A,
whose main action is to activate CDK4 by desphosphorylating a speci®c site of this kinase (Galaktionov
and Beach., 1991). Previously, it has been reported that
CDC25A may act as an oncogene when overexpressed
(Galaktionov et al., 1995). Surprisingly, expression of
CDC25A was found decreased in all xenografts and in
some cell lines analysed when compared to normal
human pancreas. Moreover, cells with low CDC25A
levels were selected for during in vivo tumor
perpetuation and dissemination. Since decreased levels
of CDC25A should lead to a reduced activity of
CDK4, the signi®cance of this ®nding is not clear. One
can speculate that this may re¯ect a cellular response
to hyperproliferation with the aim of braking the cell
cycle. If this was the case, CDC25A could be a relevant
component of a cell cycle checkpoint involved in the
regulation of CDKs activity.
In agreement with previous reports (Friess et al.,
1993; Baldwin and Korc, 1993), TGF-bRII is
consistently overexpressed in our pancreatic xenografts and cell lines. Moreover, overexpression is
maintained during distal dissemination. Since it always
occurs in the presence of aberrations in Smad4 and/or
p15 genes (with the possible exception of NP9 cell line),
TGF-bRII overexpression might be a cellular escape
from hyperproliferation, with the aim of arresting the
cell cycle by activation of the TGF-b inhibitory
pathways. Alternatively, TGF-bRII could be involved
in signaling pathways other than those involving
SMAD4 and P15 proteins, which could confer
selective growth advantage to pancreatic tumor cells.
1975
Disruption of TGF-b signaling pathways in pancreatic cancer
A Villanueva et al
1976
The ®nding of unaltered components in the NP9 cell
line, refractory to TGF-b treatment, could re¯ect the
limited sensitivity of the experimental procedures
utilized. Alternatively, it may also indicate that, other
known members of the TGF-b signaling pathways (i.e.
TGF-bRI) could be altered, although alterations in
these genes are not common in cancer cells (Kim et al.,
1996). The recent description of new Smad genes, that
are apparently not mutated in pancreatic cancer
(Riggins et al., 1997), emphasizes the increasing
complexity of these pathways and add new candidate
genes that could be altered. In conclusion, results
reported herein indicate that the TGF-b signaling
pathways are disrupted in human pancreatic cancer
cells. Disruption mainly occurs by inactivation of p15
and Smad4 genes, suggesting that alterations of these
genes may be critical for human pancreatic tumorigenesis and dissemination to occur. Paradoxical changes in
the expression of TGF-bRII and CDC25A have been
detected, although its meaning remains unclear.
Materials and methods
Human pancreatic cancer xenografts and cell lines
Eight pancreatic cancer xenografts orthotopically implanted
in nude mice, that retained the histological appearance of
the primary tumors and a high degree of genetic stability
(Reyes et al., 1996), were analysed in di€erent passages. A
total of 25 metastases from di€erent locations (Table 2)
obtained during distal dissemination of four perpetuated
tumors were also analysed (Reyes et al., 1996).
Human pancreatic tumor cell lines Panc-1. MiaPaCa2,
BxPC3 were obtained from the American type Culture
Collection (Rockville, MD, USA). Human pancreatic tumor
cell lines NP9, NP29, NP31 and NP18 were established in our
laboratory from perpetuated xenografts orthotopically
implanted in nude mice (GarcõÂ a et al., submitted)
In vitro proliferation assays
Thymidine incorporation In vitro studies were performed
using exponentially growing monolayers of NP9, NP18,
NP29, NP31 and Panc-1 cells. Peripheral blood lymphocytes were used as a control. Cells were grown in 6-well
tissue plates at an initial seeding density of 30 000 cells/
well in culture medium (RPMI or DME : Ham's F12)
containing 10% FBS for 48 h. After this period of
preincubation, TGF-b1 (5 ng/ml) was added to the
medium in the presence or absence of 10% FBS for 48 h
and 72 h. At the end of this period cells were incubated for
1 h with [3H]thymidine (1 mCi of 50 Ci/mmol) (Amersham
International, Buckinghamshire, UK) and the incorporation into acid-insoluble fraction was assayed (Baldwin and
Korc, 1993).
MTT Cells (15 000 ± 30 000 per well) were seeded into 96well plates and grown for 48 h in medium containing 10%
FBS. Cells were then treated as described above. Following
an incubation period of 72 ± 96 h at 378C, 10 ml of MTT
(Sigma, St. Louis) (10 ng/ml) were added to each well for
4 h. Optical density was measured at 550 nm using an
ELISA microplate reader (Behring EL311). Data were
expressed as percentage of stimulated versus control cells
(Baldwin and Korc, 1993).
Analysis of TGF-bRII gene
Point mutations Two segments of the TGF-bRII gene, the
(A)10 region spanning codons 97 ± 153 and the (GT)3 region
extending from codons 510 ± 559, that comprise the
majority of reported somatic mutations of this gene were
analysed (Akiyama et al., 1997). Electrophoresis was
carried out in a 8% polyacrylamide/10% glycerol nondenaturing sequencing gel at room temperature under 7 W
for 12 h. When concomitant mouse DNA ampli®cation
occurred, SSCP analysis allowed the characterization of the
human fragments.
RT ± PCR and Northern blot analysis Total RNA was
extracted from cell pellets and tissues by a single-step
method (Chomczyski and Sachi, 1987). Two mg of total
RNA was reverse transcribed to CDNA as previously
described (Mato et al., 1993) and a 1386 bp fragment,
encompassing codons 97 ± 559, was ampli®ed using speci®c
primers (Table 4). Northern blot was used to analyse cell
lines, pancreatic xenografts, and normal tissues. Approximately 10 mg of total RNA was electrophoresed in 1.2%
agarose gels containing formaldehyde and blotted to nylon
membranes (Hybond N+, Amersham International, Buckinghamshire, UK) following manufacturer's instructions.
The 1386 bp fragment generated by RT ± PCR was used as
a probe. Membranes were hybridized overnight, and
washed under stringent conditions (658C and 0.1% SSC/
0.1% SDS). Internal control for amount of RNA in the
®lter was obtained by hybridization with a human cDNA
b-actin probe.
Detection of Smad4 gene aberrations
Analysis of homozygous deletions of the Smad4 gene and
neighboring regions Six sequence-tagged site (STS) markers (Hahn et al., 1996a,b), that map centromeric to the
DCC gene, were used (Hahn et al., 1996b): y747A6R,
DPC4Exon 1, c917-46-T7, p128-N21, DPC4Exon 11,
p0630-HS-T7. All STSs were ampli®ed using 50 ng of
genomic DNA in the presence of radioactive nucleotide
([32P]CTP, 2 mCi/PCR tube) in a mixture containing PCR
bu€er [10 mM Tris-HCl (pH 8.3), 2.5 mM MgCl2,50 mM
KCl, 0.01 % (w/v) gelatine], 100 mM of each dNTPs, 1 mM
of each primer, and 1 unit of Taq DNA polymerase (LifeTechnologies, Gaithersburg, MD, USA) in a ®nal volume
of 20 ml. Annealing temperature, extension time, and
concentration of MgCl2 were optimized for each primer
set. All reactions were performed in a Hybaid thermal
cycler for 35 cycles. A homozygous deletion was de®ned by
the absence of PCR product in three independent
experiments. DNA quality was tested by amplifying a
1.2 kb fragment of the p53 gene. In all experiments,
controls for DNA contamination were included every
other sample. Co-ampli®cation of mouse DNA was
evidenced with markers DPC4 Exon 1 and DPC4 Exon
11; SSCP analysis allowed characterization of the human
PCR product.
RT ± PCR assay and mutation analysis The whole coding
region of the Smad4 gene was ampli®ed in ®ve overlapped
reactions. All primers used for PCR ampli®cation were
designed based on the published cDNA sequence (Hahn et
al., 1996b) (Table 4). Five ml of the RT reaction was added
to a mixture containing PCR bu€er, 0.3 mM of each
dNTPs, 2.5 mM MgCl2, 1 mM of primers and 1.5 U of
Taq polymerase in the presence of radioactive nucleotide
([32P]dCTP, 2 mCi/PCR tube). The reaction was performed
in a Perkin-Elmer thermal cycler for 30 cycles (578C,
1 min; 728C, 2 min; 948C, 1 min) in a ®nal volume of 30 ml.
No PCR product was obtained when genomic DNA was
used as template. P53 cDNA gene ampli®cation was
performed to test the eciency of RT reaction. Mouse
RNA was included in all experiments to rule out
contamination by nude mouse products. All reactions
were performed independently at least three times using
di€erent RNA isolates. The identity of the ampli®ed
Disruption of TGF-b signaling pathways in pancreatic cancer
A Villanueva et al
fragments was con®rmed by cycle sequencing using the
AmpliCycle Sequencing Kit (Perkin Elmer, Branchburg,
NJ, USA) following manufacturer's directions. Mutational
analysis of the Smad4 gene was performed by means of the
SSCP method. Electrophoresis was carried out in a 7%
polyacrylamide/10% glycerol non-denaturing sequencing
gel at room temperature under 5 W for 12 ± 15 h, and at
48C under 30 W for 7 ± 10 h. When concomitant mouse
DNA ampli®cation occurred, SSCP analysis allowed the
characterization of the human fragment. Northern blot
analysis using total cellular RNA was also attempted
following essentially the same protocol described for TGFbRII analysis. Probes encompassing the whole coding
region of the Smad4 gene and the di€erent ampli®ed
fragments described above were utilized.
Analysis of Smad2 gene
cDNA was obtained as described above. The coding
regions of the conserved domains of SMAD2 protein
were ampli®ed using primers previously reported (Eppert et
al., 1996). Annealing temperature, extension time, and
concentration of MgCl2 were optimized for each primer set.
Mutational analysis was performed following essentially
the same protocol used for the Smad4 gene analysis.
Detection of p15 gene aberrations
Mutational analysis One hundred 7200 ng of genomic
DNA was used to amplify exons 1 and 2 of the p15 gene in
three independent PCR reactions using human intronic
primers Jen et al., 1994). SSCP analysis was essentially
performed as described above with the only exception that
electrophoresis was carried out at room temperature under
7 W for 18 h using radioactive nucleotide or silver staining.
Identity of the fragments was con®rmed by cycle
sequencing adding 10% DMSO to the sequencing mixture.
Allelic losses They were analysed by means of two
microsatellite length polymorphism (D9S171, IFNA) as
described (Reyes et al., 1996). The products were separated
on a 6% polyacrylamide/8 M urea gel, and autoradiography was performed.
RT ± PCR The levels of p15 mRNA were analysed by
means of RT ± PCR using speci®c exonic primers (Table 4).
Western blotting
Expression of P15, P16, P27 and CDC25A proteins was
analysed by Western blotting in pancreatic xenografts,
metastases and cells growing asynchronously. Samples
were frozen, lysed in 80 mM Tris-HCl (pH 6.8) containing
2% SDS, sonicated and ®nally resuspended in Laemmii
bu€er. Fifty mg of protein were then electrophoresed using
10% (P27, CDC25A) or 15% (P15, P16) SDS-polyacrylamide gel and transferred onto Immobilon P membranes
(Millipore, Bedford, MA, USA). Western blotting was
performed using speci®c polyclonal antibodies against P15
(C-20, Santa Cruz Biotechnology, Santa Cruz, CA, USA;
dilution 1 : 200); P16 (C-20, Santa Cruz Biotechnology,
Santa Cruz, CA, USA; dilution 1 : 200); P27 (a kind gift of
Dr MassagueÂ, New York, USA; dilution 1 : 100); and
CDC25A (144, Santa Cruz Biotechnology, Santa Cruz,
CA, USA; dilution 1 : 2000). Additional experiments were
performed in cell lines NP18, NP31 and PANC-1 that
showed a modest inhibitory response to TGF-b1 or
morphological changes in the presence of TGF-b1. Cells
were synchronized in serum-deprived media for 72 h and
then exposed to TGF-b1 (5 ng/ml) (Calbiochem, Cambridge, MA, USA). Proteins were analysed at di€erent time
intervals (0, 6, and 12 h) after exposure to TGF-b1.
Abbreviations
TGF-bRII Transforming Growth Factor b receptor type
II; LOH, loss of heterozygosity; SSCP, single strand
conformation polymorphism; STS, Sequence-tagged site.
Acknowledgements
This work was supported in part by grants from ComisioÂn
Interministerial de Ciencia y TecnologõÂ a (SAF95-0281),
(SAF96-0187-CO2-01) and (SAF97-0214); Marato TV3
(11-95) and (26-95); AgeÁncia d'Avaluacio de Tecnologia
MeÁdica; Comissionat per Universitats i Recerca (GRQ939501) de Catalunya; Laboratorios Dr Esteve SA and
Fundacio Catalana de Gastroenterologia. The authors are
indebted to Lourdes Farre and Gemma Tarafa for their
technical assistance and expertise in caring for the nude
mice.
References
Akiyama Y, lwanaga R, Saitoh K, Shiba K, Ushio K, Ikeda
E, Iwama T, Nomizu T and Yuasa Y. (1997). Gastroentererology, 112, 33 ± 39.
Baldwin RL and Korc M. (1993). Growth Factors, 8, 23 ± 34.
Barrett MT, Schutte M, Kern SE and Reid BJ. (1996).
Cancer Res., 56, 4351 ± 4353.
Caldas C, Hahn SA, da Costa LT, Redston MS, Schutte M,
Seymour AB, Weinstein CL, Hruban RH, Yeo CJ and
Kern SE. (1994). Nat. Genet., 8, 27 ± 32.
Chen X, Rubock MJ and Witman M. (1996). Nature, 383,
691 ± 696.
Chomczyski P and Sachi N. (1987). Anal. Biochem., 162,
156 ± 159.
Derynck R, Gelbart WM, Harland RM, Heldin CH, Kern
SE, Massague J, Melton DA, Mlodzik M, Padgett RW,
Roberts AB, Smith J, Thomsen GH, Vogelstein B and
Wang XF. (1996). Cell, 87, 173.
Ebner R, Chen RH, Lawler S, Zionchenck T and Derynck R.
(1993). Science, 262, 900 ± 902.
Eppert K, Scherer SW, Ozcelik H, Pirone R, Hoodless P,
Kim H, Tsui LC, Bapat B, Gailinger S, Andrulis IL,
Thomsen GH, Wrana JL and Attisano L. (1996). Cell, 86,
543 ± 552.
Friess H, Yamanaka Y, Buchler M, Berger HG, Kobrin MS,
Baldwin RL and Korc M. (1993). Cancer Res., 53, 2704 ±
2707.
Galaktionov K and Beach D. (1991). Cell, 67, 1181 ± 1194.
Galaktionov K, Lee AK, Eckstein J, Draetta G, Meckler J,
Loda M and Beach D. (1995). Science, 269, 1575 ± 1577.
Hahn SA, Hoque AT, Moskaluk CA, da Costa LT, Schutte
M, Rozenblum E, Seymour AB, Weinstein CL, Yeo CJ,
Hruban RH and Kern SE. (1996a). Cancer Res., 56, 490 ±
494.
Hahn SA, Schutte M, Hoque AT, Moskaluk CA, da Costa
LT, Rozenblum E, Weinstein CL, Fischer A, Yeo CJ,
Hruban RH and Kern SE. (1996b). Science, 271, 350 ± 353.
1977
Disruption of TGF-b signaling pathways in pancreatic cancer
A Villanueva et al
1978
Hahn SA, Seymour AB, Shamsul Hoque ATM, Schutte M,
da Costa LT, Redston MS, Caldas C, Weinstein CL,
Fischer A, Yeo CJ, Hruban RH and Kern SE. (1995).
Cancer Res., 55, 4670 ± 4675.
Hannon GJ and Beach D. (1994). Nature, 371, 257 ± 261.
Huang L, Goodrow TL, Zhang SY, Klein-Szanto AJ, Chang
H and Ruggeri BA. (1996). Cancer Res., 56, 1137 ± 1141.
Iavarone A, and Massague J. (1997). Nature, 87, 417 ± 422.
Jen J, Harper JW, Bigner SH, Bigner DD, Papadopoulos N,
Markowitz S, Willson JK, Kinzler KW and Vogelstein B.
(1994). Cancer Res., 54, 6353 ± 6358.
Kim IY, Ahn H-J, Zeiner DJ, Shaw JW, Sensibar JA, Kim
JH, Kato M and Lee C. (1996a). Cancer Res., 56, 44 ± 48.
Kim SK, Fan Y, Papadimitrakopoulou V, Clayman G,
Hittelman WN, Hong WK, Lotan R and Mao L. (1996b).
Cancer Res, 56, 2519 ± 2521.
Lagna G, Hata A, Hemmati-Brivanlou A, and Massague J.
(1996). Nature, 383, 832 ± 836.
Liu F, Hata A, Baker JC, Doody J, Carcamo J, Harland RM
and Massague J. (1996). Nature, 381, 620 ± 623.
Macias-Silva M, Abdollah S, Hoodless PA, Pirone R,
Attisano L and Wrana JL. (1996). Cell, 87, 1215 ± 1224.
Markowitz S, Wang J, Myero€ L, Parsons R, Sun L,
Lutterbaugh J, Fan RS, Zborowska E, Kinzler KW,
Vogelstein B, Brattain M and Willson JKV. (1995).
Science, 268, 1336 ± 1338.
Massague J. (1996). Cell, 85, 947 ± 950.
Mato E, Santisteban P, Viader M, Capella G, Fornas O,
Puig-Domingo M and Webb SM. (1993). J. Pineal Res.,
15, 43 ± 45.
Nagatake M, Takagi Y, Osada H, Uchida K, Mitsudomi T,
Saji S, Shimokata K and Takahashi T. (1996). Cancer
Res., 56, 2718 ± 2720.
Nakao A, Roijer E, Imamura T, Souchelnytskyi S, Stenman
G, Heldin CH and ten Dijke P. (1997). J. Biol. Chem., 272,
2896 ± 2900.
Naumann M, Savitskaia N, Eilert C, Schramm A, Kaltho€ H
and Schmiegel W. (1996). Gastroenterology, 110, 1215 ±
1224.
Reyes G, Villanueva A, GarcõÂ a C, Sancho FJ, Piulats J, LluõÂ s
F and Capella G. (1996). Cancer Res., 56, 5713 ± 5719.
ReynisdoÂttir I and Massague J. (1997). Genes Dev., 11, 492 ±
503.
Riggins GJ, Kinzler KW, Vogelstein B and Thiagalingam S.
(1997). Cancer Res., 57, 2578 ± 2580.
Schutte M, Hruban RH, Hedrick L, Cho KR, Nadasdy GM,
Weinstein CL, Bova GS, Isaacs WB, Cairns P, Nawroz H,
Sidransky D, Casero RA, Meltzer PS, Hahn SA and Kern
SE. (1996). Cancer Res., 56, 2527 ± 2530.
Serrano M, Lee H, Chin L, CordoÂn-Cardo C, Beach D and
DePinho RA. (1996). Cell, 85, 27 ± 37.
Scarpa A, Capelli P, Villanueva A, Zamboui G, LluõÂ s F,
Accolla R, Manuzzi G and Capella G. (1994). Int. J.
Cancer, 57, 167 ± 171.
Shi Y, Hata A, Lo RS, Massague J and Pavietich NP. (1997).
Nature, 388, 87 ± 93.
Souza FR, Lei J, Jin J, Appel R, Zou TT, Zhou X, Wang S,
Rhyu MG, Cymes K, Chan O, Park WS, Krasna MJ,
Greenwald BD, Cottrell J, Abraham JM, Simms L,
Leggett B, Young J, Harpaz N and Meltzer SJ. (1997).
Gastroenterology, 112, 40 ± 45.
Thiagalingam S, Lengauer C, Leach FS, Schutte M, Hahn
SA, Overhausser J, Willson JK, Markowitz S, Hamilton
SR, Kern SE, Kinzler KW and Vogelstein B. (1996). Nat.
Genet., 13, 343 ± 346.
Tsubari M, Tiihonen E and Laiho M. (1997). Cancer Res.,
57, 2966 ± 2973.
Wrana JL, Attisano L, Wieser R, Ventura F, and MassagueÂ
J. (1994). Nature, 370, 341 ± 347.