Notch4 intracellular domain binding to Smad3 and inhibition of the TGF-b
Youping Sun1,2, William Lowther2, Katsuaki Kato2, Caterina Bianco1, Nicholas Kenney1,
Luigi Strizzi1, Dina Raafat1, Morihisa Hirota1, Nadia I Khan1, Sharon Bargo2, Brenda Jones1,
David Salomon*,1and Robert Callahan*,2
1The Tumor Growth Factor Section, Mammary Biology and Tumorigenesis Laboratory, Center for Cancer Research, National Cancer
Institute, National Institute of Health, Bethesda, MD 20892, USA;2Oncogenetics Section, Mammary Biology and Tumorigenesis
Laboratory, Center for Cancer Research, National Cancer Institute, National Institute of Health, Bethesda, MD 20892, USA
We present evidence that Notch4ICD attenuates TGF-b
signaling. Cells expressing the activated form of the
Notch4 receptor (ICD4) were resistant to the growth-
inhibitory effects of TGF-b. Notch4ICD was found to
bind to Smad2, Smad3 and Smad4 but with higher affinity
to Smad3. Deletion analysis showed that binding of
Smad3 to ICD4 was mediated by its MH2 domain and
was not dependent on the presence of the RAM23 region
in ICD4. Using two TGF-b/Activin reporter luciferase
assays, RT–PCR and Western blot analysis, we demon-
strate that ICD4 and ICD4 dRAM23 inhibit Smad-
binding element and 3TP luciferase reporter activity and
PAI-1 gene expression. MCF-7 human breast cancer cells
express Notch4ICD (ICD4) and are resistant to the
growth-inhibitory effects of TGF-b. Blockage of Notch4
processing to ICD4 by c-secretase inhibitor renders MCF-
7 cells sensitive to growth inhibition by TGF-b. The
interplay between these two signaling pathways may be a
significant determinant during mammary tumorigenesis.
Oncogene (2005) 24, 5365–5374. doi:10.1038/sj.onc.1208528;
published online 20 June 2005
Keywords: Notch4ICD; Smad3; TGF-b
In mammals, the Notch transmembrane receptor gene
family is comprised of four members, Notch1/TAN-1,
Notch2, Notch3 and Notch4/Int3 (Callahan and Egan,
2004). The extracellular domain (ECD) is comprised of
between 29 and 36 epidermal growth factor (EGF)-like
repeats and three copies of a lin-12/Notch Glp motif.
The intracellular domain (ICD) has a RAM23 region,
seven CDC10/Ankyrin repeats and a PEST (proline,
glutamate, serine, threonine-rich) region (Jeffries and
Capobianco, 2000; Baron et al., 2002). Notch4/Int3
(ICD4) is smaller than Notch1ICD and lacks the
transcriptional activation domain (TAD) and cytokine
response (NCR) sequence found in the Notch1ICD
(Uyttendaele et al., 1996; Allenspach et al., 2002). Upon
ligand activation, the intracellular domain of Notch is
proteolytically cleaved releasing the ICD domain. The
ICD is subsequently translocated to the nucleus (Kopan
et al., 1996) where it binds to CBF-1, displacing it from
a corepressor complex comprised of histone deacetylases
(HDAC) and leading to the expression of members of
the Hes (Hairy Enhancer of split) (Jarriault et al., 1995;
Tamura et al., 1995) and Herp (Hes-related repressor
protein) gene families (Iso et al., 2001). The Notch4
promoter (but not Notch2 or Notch3) like Hes1
contains CSL-binding elements and therefore, is also a
target gene of Notch1 (Weijzen et al., 2002). Genetic
alterations of Notch1 or Notch4 that lead to the
deregulated expression of the Notch ICD represent gain
of function mutations that are associated with T-lympho-
(Ellisen et al., 1991; Jhappan et al., 1992; Zagouras
et al., 1995; Gallahan et al., 1996; Kiaris et al., 2004).
The TGF-b super family contains several growth and
differentiation factors, such as TGF-b 1–3, Activins,
bone morphogenic proteins (BMP) and Nodal. TGF-b,
Activin and Nodal bind to and activate type I and type
II TGF-b serine–threonine kinases, which then activate
downstream regulatory Smad proteins by phosphoryla-
tion. Phosphorylated Smad2 and Smad3 trimerize with
Smad4 and translocate from the cytoplasm to the
nucleus where they act as transcriptional activators of
TGF-b target genes (Kawabata et al., 1998; Massague,
2000). TGF-b performs a pivotal role during embryo-
genesis, development and tumorigenesis. TGF-b is a
potent growth inhibitor of most types of epithelial cells,
and perturbation of TGF-b signaling has been shown to
contribute to the progression of various tumors (Blobe
et al., 2000). Oncogenes, such as c-Ski/SnoN (Akiyoshi
et al., 1999), c-Myc (Akiyoshi et al., 1999) and Evi-1
(Kurokawa et al., 1998) cause malignant transformation
in part by repressing the TGF-b signaling pathway.
TGF-b functions as a tumor suppressor in epithelial
Received 28 September 2004; revised 5 January 2005; accepted 5 January
2005; published online 20 June 2005
*Correspondence: D Salomon; E-mail: firstname.lastname@example.org;
R Callahan; E-mail: Callahro@mail.nih.gov
Oncogene (2005) 24, 5365–5374
& 2005 Nature Publishing Group All rights reserved 0950-9232/05 $30.00
cells and during the early stages of tumor formation
(Wakefield and Roberts, 2002). Smad3 has been shown
to interact with different signal transduction pathways
by binding to various transcriptional proteins in these
pathways (Moustakas et al., 2001). For example, it has
recently been demonstrated that Smad3 can interact
with p53 (Cordenonsi et al., 2003), Lef-1 (Labbe et al.,
2000), estrogen receptor (Moustakas et al., 2001; Wu
et al., 2003), Disabled-2 (Hocevar et al., 2001) and HEF-
1 (Liu et al., 2000).
TGF-b or BMP induces the expression of several
genes that are also known target genes for Notch1, such
as Hes-1 (Blokzijl et al., 2003), Hey1 (Zavadil et al.,
2004), Hes-5 (Takizawa et al., 2003), Herp2 (Itoh et al.,
2004) and the Notch ligand Jagged1 (Zavadil et al.,
2004). However, information as to whether Notch1 or
Notch4 can modify the TGF-b signaling pathway and
the response of cells to TGF-b is limited. Rao and
Kadesch (2003) have recently shown that overexpression
of Notch1ICD in mink lung Mv1Lu epithelial cells but
not other cells such as C2C12 or NIH 3T3 indirectly
deregulates c-Myc expression and can thereby render
Mv1Lu epithelial cells resistant to TGF-b growth-
inhibitory signals. Blokzijl et al. (2003) have reported
that Notch1ICD worked synergistically with TGF-b to
activate a multimerized CAGA element from the PAI-1
promoter in C2C12 cells. However, it was not shown
that a target gene of TGF-b was synergistically
upregulated by Notch1ICD at the mRNA and/or
We present evidence in the present study demonstrat-
ing that Notch4ICD attenuates TGF-b signaling in
several cell types. Cells expressing the activated form of
the Notch4 receptor (ICD4) were resistant to the
growth-inhibitory effects of TGF-b. This inhibition is
mediated through the interaction of ICD4 with Smad3
and Smad4 and the subsequent attenuation of the ability
of TGF-b to activate promoters that contain a Smad-
binding element (SBE). A consequence of these interac-
tions is the inhibition of the TGF-b signaling pathway
that may significantly predispose mammary epithelial
cells to the oncogenic activity of ICD4. Consistent
with this hypothesis is our demonstration that treatment
of MCF-7 breast cancer cells with a g-secretase inhibitor
(GSI), which can block presenillin-induced cleavage
of the endogenous Notch4 receptor to ICD4, can
sensitize MCF-7 cells to the growth-inhibitory effects
Activated ICD4 attenuates TGF-b-induced growth
inhibition of mammary epithelial cells
TGF-b is a tumor suppressor, which can inhibit cell
proliferation (Miyazono, 2000). As a first step towards
understanding the molecular events involved in ICD4-
induced tumorigenesis, we examined the effect of TGF-b
on cells expressing ICD4. The effect of TGF-b1 on the
growth of EpH4 mouse mammary epithelial cells was
compared with EpH4 cells stably transfected with a
vector expressing ICD4 (EpH4-ICD). After 7 days of
TGF-b treatment, EpH4 cells showed a dose-dependent
inhibition (approximately 60%) of cell growth, whereas
growth of EpH4-ICD-expressing cells were significantly
less inhibited (30%) (Figure 1a). Colony-forming assays
on plastic using these same cell lines showed a similar
effect after TGF-b treatment (data not shown) with the
EpH4-ICD cells being refractory (approximately 50%)
to the ability of TGF-b to inhibit colony formation on
To further define the regions of the ICD4 that are
required to attenuate the growth-inhibitory effect of
TGF-b in mammary epithelial cells, we compared the
growth of HC-11 mouse mammary epithelial cells that
were expressing either a full-length ICD4 or a trun-
cated form of ICD4 that lacks the RAM23 region
(dRAM-ICD4). Both cell lines were more resistant to
the growth-inhibitory effects of TGF-b compared to
wild-type HC-11 cells or empty vector-transfected HC-
11 cells (Figure 1b). This suggests that the RAM23
region of ICD4 is not necessary to confer resistance to
TGF-b-associated cell growth inhibition.
We treated EpH4 control cells and EpH4-ICD cells
with or without TGF-b and used bromodeoxyuridine
(BrdU) to measure DNA synthesis to study the effects of
ICD4 on cell proliferation (Cordenonsi et al., 2003)
(Figure 1c). As expected, TGF-b inhibited DNA
synthesis in control EpH4 cells by approximately 70%
(Figure 1c). However, TGF-b had only a marginal effect
on inhibition (approximately 10%) of DNA synthesis in
EpH4-ICD4 cells indicating that ICD4 expression
significantly interferes with the growth arrest induced
ICD4 inhibits TGF-b-induced PAI-1 gene expression
Transcription and translation of PAI-1 is strongly
induced by TGF-b and is often used as a marker for
TGF-b responsiveness in mammalian cells (Dennler
et al., 1998; Feng et al., 1998). To determine whether
ICD4 inhibits TGF-b target gene expression, we treated
HC-11 and HC-11-ICD4 cells with or without TGF-b
(2ng/ml) for 0–4h. As expected, TGF-b induced PAI-1
mRNA expression in HC-11 cells (Figure 2a). However,
ICD4-expressing HC-11 cells showed only 30% PAI-1
expression as compared to HC-11 wild cells (Figure 2a)
after treatment with TGF-b.
To determine whether ICD4 can affect TGF-b
signaling in nonmammary epithelial cells, we cloned
the ICD4 or dRAM-ICD4 into the GeneSwitch System
(Invitrogen) and stably transfected these vectors into
NIH3T3 cells. ICD4 or dRAM-ICD4 protein expression
can be induced after 16h treatment with 10nM
mifepristone (Invitrogen). dRAM-ICD4 NIH3T3 cells
exhibited 10% PAI-1 protein expression after TGF-b
treatment as compared to NIH3T3 cells (Figure 2b).
ICD4 NIH3T3 cells gave a similar result (data not
shown), indicating that ICD4 attenuation of TGF-b
signaling is independent of the cell type.
Notch4 inhibition of TGF-b signaling
Y Sun et al
ICD4 interacts with Smad2, Smad3 and Smad4
Since ICD4 attenuates the TGF-b-induced growth
arrest and appears to do so through a RAM23-
independent mechanism, we speculated that one or
more members of the Smad family directly interact with
ICD4. Interactions between the Smads and ICD4 was
assessed by coimmunoprecipitation using a Flag-tagged
Smads and immunobloting for V5-antibody in Bosc23
cells transiently transfected with the appropriate expres-
sion vectors. As shown in Figure 3a, Smad2, Smad3 or
Smad4 could bind to ICD4. However, it appeared that
Smad3 bound with a higher affinity to ICD4 than to
either Smad2 or Smad4. Similar binding was also
detected in a reciprocal experiment, in which the V5-
tagged ICD4 were used to pull down the Flag-tagged
Smad2, Smad3 and Smad4 (data not shown).
To determine whether this interaction with Smad3
could also be detected in vivo, we prepared protein
extracts from mammary tissue of pregnant FVB/N mice
and from mammary tumors of WAP-ICD4 mice. The
WAP-ICD4 mice express ICD4 under the control of the
whey acidic protein (WAP) promoter (Smith et al.,
1995). Protein was coimmunoprecipitated using an anti-
ICD4 antibody, 341 (Smith et al., 1995) and probed with
an anti-Smad3 antibody. As seen in Figure 3b, both
control (FVB/N) and WAP-ICD4 express Smad3, while
only WAP-ICD4 mice express ICD4. Similar to the in
vitro data, Smad3 was also found to make a complex
with the ICD4 in the WAP-ICD4 transgenic mammary
RAM23 domain of ICD4 is not responsible for binding
As expression of the truncated dRAM-ICD4 in HC-11
cells was also able to promote resistance to the growth-
EpH4 and HC-11 mammary epithelial cells. (a) Cell growth
inhibition as a function of varying concentrations of TGF-b
(expressed as a percentage of untreated cells). ICD4-expressing
EpH4 cells and control cells were treated with varying concentra-
tions of TGF-b for 7 days as indicated. All values represent the
mean (7s.e., n¼3). (b) Cell growth inhibition as a function of
varying concentrations of TGF-b (expressed as a percentage of
untreated cells). ICD4- or dRRM-ICD4-expressing HC-11 cells
and control cells were treated with varying concentrations of TGF-
b for 7 days, as indicated. All values represent the mean (7s.e.,
n¼3). (c) Effects of TGF-b on proliferation of EpH4 and EpH4-
ICD4 cells as measured by BrdU incorporation. After 16h TGF-b
treatment (2ng/ml), the EpH4 and EpH4-ICD4 cells were pulsed
with BrdU and harvested 1h later. All values represent the mean
ICD4 blocks the TGF-b growth-inhibitory activity in
11 and NIH3T3 cells. (a) RT–PCR was used to detect PAI-1
mRNA expression in HC-11 PEF-1 and HC-11 ICD4 PEF-1 stably
transfected cells. The cells were treated with TGF-b (2ng/ml) for 1–
4h. RT–PCR was performed using specific primers as described in
Material and methods. (b) Western blot analysis was used to detect
PAI-1 (upper panel), dRAM-ICD4 (middle panel) and control a-
tubulin (bottom panel) in NIH3T3 and NIH3T3 dRAM-ICD4
cells. The cells were treated with or without TGF-b (2ng/ml) for 18
ICD4 inhibits TGF-b-induced PAI-1 expression in HC-
Notch4 inhibition of TGF-b signaling
Y Sun et al
inhibitory effects of TGF-b, we confirmed that this
region is not involved in binding to Smad3. Flag-tagged
Smad3 was cotransfected together with the V5-tagged
dRAM-ICD4 expression vector (Figure 4a) in Bosc 23
cells. As seen in Figure 4b, both the full-length ICD4
and dRAM-ICD4 could bind to Smad3. Thus, the
RAM23 domain is not required for the binding of
Smad3 to the ICD4.
ICD4 interacts with the MH2 domain of Smad3
Smad3 has two conserved domains, the N-terminal Mad
homology 1 (MH1) domain and the C-terminal Mad
homology 2 (MH2) domain (Figure 5a). The MH1
domain regulates nuclear import and transcription by
binding to DNA (Moustakas et al., 2001). It also
interacts with nuclear proteins. The MH2 domain
regulates Smad oligomerization, recognizes Type I
receptors and interacts with cytoplasmic adaptors and
several transcriptional factors such as p300 (Moustakas
et al., 2001). To ascertain which domain of Smad3 binds
to ICD4, we cotransfected a HaHis-tagged ICD4
expression vector (Wu et al., 2001) with a Flag-tagged
MH1 Smad3 or a Flag-tagged MH2 Smad3 expression
construct into Bosc 23 cells. The extracts were im-
munoprecipitated with an anti-Flag antibody to pull
down the Smad3 proteins and immunobloted with anti-
ICD4 antibody. As seen in Figure 5b, Smad3 interacts
with ICD4 through its MH2 domain.
tion of V5-his-ICD4 with Flag-Smad2, Smad3 or Smad4 in Bosc23
cells was examined by coimmuoprecipitation (IP) followed by
immunobloting (Blots). The top panel shows the immuoprecipita-
tion, and the lower two panels show the level of expression by
Western blot analysis. (b) The interaction between ICD4 and
Smad3 was detected by IP (top panel) in protein extracted from
mammary glands of day 15 pregnant FVB/N (lane1) and Wap-
ICD4 (lane 2) transgenic mice. The bottom two panels are Western
blots of total lysates of the samples used for IP with antibodies for
Notch4ICD (341) and Smad3
Interaction between Smads and ICD4. (a) The interac-
RAM23 is the CBF-1-binding domain. CDC10 corresponds to
seven contiguous Ankryn-like repeats. PEST is a domain rich in
proline, glutamate, serine and threonine. The V5 epitope is
indicated. (b) The interaction between Smad3 and a deletion
mutant of ICD4 in Bosc23 cells. The interaction between Smad3-
Flag and ICD4-V5 or dRAM-ICD4-V5 was detected by IP (Anti-
Flag, top panel) followed Western blot analysis with Anti-V5 (top
panel). The next panels are Western blots of total lysates of the
samples used for IP with Anti-V5 for ICD4 and Anti-Flag for
Smad3. The Bosc 23 cells were transfected with the Smad3, ICD4,
and ICD4dRAM expression vectors are indicated
Domains of ICD4 that interact with Smad3. (a)
(a) Functional domains of the full-length (F-Smad3), N0terminal
MAD homology domain 1 (MH1), and C0terminal MAD
homology domain 2 (MH2) proteins. Each has the Flag epitope
at its C-terminus. (b) The interaction between ICD4 and deletion
mutants of Smad3 was examined in Bosc23 cells. The interaction of
Smad3-Flag deletion mutants with ICD4 was detected by IP (Anti-
Flag, top panel) followed Western blot analysis with anti-ICD4
(top panel). The bottom two panels are Western blots of total
lysates of the samples used for IP with Anti-ICD4 and anti-Flag for
Smad3. The Bosc 23 cells were transfected with the Smad3 and/or
ICD4 expression vectors are indicated
5DomainsofSmad3that interact withICD4.
Notch4 inhibition of TGF-b signaling
Y Sun et al
ICD4 expression has no effect on TGF-b-induced
phosphorylation of Smad2 or Smad3
TGF-b rapidly induces phosphorylation of Smad2 and
Smad3. This is essential for activation of these proteins
so that they can trimerize with Smad4 and is translo-
cated to the nucleus. To determine whether ICD4
attenuates TGF-b signaling by impairing phosphoryla-
tion of Smad2 or Smad3, we examined their phosphor-
ylation state in HC-11 and HC-11 ICD4 cells treated
with TGF-b for 30min to 24h. We found that there was
no significant difference in TGF-b-induced phosphor-
ylation of either Smad2 or Smad3 in cells with or
without ICD4 expression (Figure 6).
ICD4 inhibits TGF-b signaling pathway through Smad3
We next determined whether the interaction between
ICD4 or dRAM-ICD4 and Smad3 could affect canoni-
cal TGF-b signaling. A TGF-b-sensitive luciferase
construct, 3TP-luc, which contains TGF-b-response
elements from the plasminogen activator inhibitor-1
and collagenase promoters (Wrana et al., 1992), as well
as an artificial promoter, 4?SBE-luc, containing four
GTCTAGAC) (Zawel et al., 1998) were transiently
transfected together with ICD4 or dRAM-ICD4 expres-
sion constructs into either EpH4 or 293T/17 cells.
Expression of either ICD4 or dRAM-ICD4 inhibited
the TGF-b-induced 3TP-luc activity by approximately
50% in both EpH4 (Figure 7a) and inhibited the TGF-
b-induced 4?SBE activity by 70% in 293 T/17 cells
(Figure 7b). Similar results were also observed in HC-11
and HC-11 ICD4 cells (Figure 7c).
To determine whether ICD4 and Smad3 act synergis-
tically or competitively on TGF-b/Activin-dependent
signaling, we transfected Smad3 or ICD4 and 4?SBE-
luc expression vectors into the HC-11 cells. As shown in
Figure 7c, Smad3 activates 4?SBE-luc expression while
amounts of ICD4 or Smad3 expression vectors were
cotransfected with the same amount of 4?SBE-luc into
HC-11 cells. We found that ICD4 could inhibit by 70%
4?SBE-luc activities that was induced by TGF-b and
induced by TGF-b. HC-11 and HC-11-ICD4 cells were treated with
TGF-b (2ng/ml) for 0.5, 1, 5 and 24h. Cell lysates were assayed by
Western blot analysis for Smad2 and Smad3 phosphorylation using
antiphosphorylated Smad2 (P-Smad2) and Smad3 (P-Smad3)
antibody and for total Smad2 and Smad3 expression
ICD4 fails to block phosphorylation of Smad2 or Smad3
vectors on transactivation by TGF-b was examined by a dual luciferase assay in EpH4 mouse mammary epithelial cells and using the
3TP-luciferase reporter (3TP-luc) (Figure 7a) or 293 T/17 cells using the 4?SBE-luc (Figure 7b). The values represent fold induction
relative to the basal 3TP-luc or SBE-luc activity in a dual luciferase assay. Cells were treated with or without 2ng/ml TGF-b treatment
for 16h before lysing cells, as indicated, and dual luciferase were performed 48h after transfection. (c) The effect of ICD4 on
transactivation by TGF-b was examined in HC-11 or HC-11 ICD4 stably transfected cells with the 4?SBE-luc reporter vector. The
values represent fold induction as compared with the basal 4?SBE-luc activity in a dual luciferase assay. (d) ICD4 inhibition of TGF-
b-dependent signaling can by rescued by Smad3. The effect of ICD4 and Smad3 on transactivation by TGF-b was examined in HC-11
cells with the 4?SBE-luc reporter vector. The mg of ICD4 and Smad3 expression vectors used is indicated. The values represent fold
induction as compared with the basal 4?SBE-luc activity in a dual luciferase assay
ICD4 inhibits TGF-b-dependent signaling. (a and b) The effect of ICD4 (0.5mg) and dRAM-ICD4 (0.5mg) expression
Notch4 inhibition of TGF-b signaling
Y Sun et al
that this inhibitory effect of ICD4 on TGF-b induction
of 4?SBE-luc could be rescued by expression of Smad3
in a dose-dependent manner (Figure 7d).
Blockage of Notch signaling by GSI can induce MCF-7
cells to become sensitive to TGF-b treatment
Xie et al. (2002) have reported that the growth of eight
human breast tumor cell lines, including MCF-7, are not
arrested by TGF-b treatment. The result for MCF-7 is
confirmed in Figure 8a (compare the upper two panels).
As shown in Figure 8b, MCF-7 cells also express
endogenous Notch4 and ICD4. To test whether
endogenous Notch4 signaling was associated with the
resistance of MCF-7 cells to TGF-b-induced growth
arrest, we blocked the processing of wild-type Notch4 to
the ICD4 form with GSI (Figure 8b). As expected,
treatment of MCF-7 cells with GSI increased their
sensitivity to TGF-b-induced growth arrest (Figure 8a,
compare lower two panels and 8c).
Differences between Notch1ICD (ICD1) and Notch4ICD
(ICD4) signaling in C2C12 cells
To determine whether there are biological differences
between ICD1 and ICD4 on modulating the TGF-b
signaling pathway, we cotransfected human ICD1 or
mouse ICD4 with TGF-b luciferase reporter constructs
SBE or P(CAGA)9 into C2C12 cells. We found that
ICD4 suppressed both SBE and P(CAGA)9 luciferase
activities in C2C12 cells, whereas Notch1 does not
significantly affect TGF-b signaling in these cells (Figure
9a and b). However, ICD1 was found to attenuate TGF-
b signaling in HC-11 mammary epithelial cells but not in
NIH3T3 cells, suggesting that the inhibitory effects of
ICD1 and ICD4 are cell line-dependent (data not
MCF-7 breast cancer cells by inhibiting the formation of ICD4
with g-secretase inhibitor (GSI) makes the cells sensitive to TGF-b-
induced cell growth suppression. (a) Representative fields of MCF-7
cells treated with TGF-b, g-secretase inhibitor and their combina-
tion. (b) Notch4 and ICD4 protein expression in MCF-7 cells
treated for 24h with 5mM GSI. (c) Cells growth inhibition as a
function of TGF-b and g-secretase inhibitor in MCF-7 breast
cancer cells. All values represent the mean (7s.e., n¼3)
Silencing endogenous wild-type Notch4 signaling in
Human ICD1 (0.5mg) and ICD4 (0.5mg) expression vectors on
transactivation by TGF-b was examined by a dual luciferase assay
in C2C12 cells and using the 4?SBE-luciferase reporter (panel a)
or the p(CAGA)9-luc (panel b). The values represent fold induction
relative to the basal SBE-luc or p(CAGA)9-luc activity in a dual
luciferase assay. Cells were treated with or without 5ng/ml TGF-b
for 16h before lysing cells and the dual luciferase were performed
48h after transfection
ICD1 and ICD4 signaling in C2C12 cells. The effect of
Notch4 inhibition of TGF-b signaling
Y Sun et al
Effects of TGF-b1 or BMP-2 on Notch signaling during
Two publications (Blokzijl et al., 2003; Takizawa et al.,
2003) have recently shown that TGF-b1 can regulate
expression of the Notch1 target genes, Hes1 (Blokzijl
et al., 2003) and Hey1 (Zavadil et al., 2004) and the
Notch ligand, Jagged1 (Zavadil et al., 2004) during
embryonic development and during TGF-b-induced
epithelial-to-mesenchymal transition (EMT). In Blok-
zijl’s study (2003), TGF-b1 was shown to be a positive
regulator of Hes1 expression in the C2C12 mouse
myoblast cells. A dominant negative CSL mutant
blocked the effect of TGF-b on the activation of Hes1.
Smad3 could be recruited to CSL-binding sites on DNA
in the presence of CSL and ICD1. The MH2 region of
Smad3 was found to bind to the Notch1ICD (ICD1).
Similarly, Zavadil et al. (2004) reported that Hey1 is not
only a Notch target gene (Maier and Gessler, 2000) but
also a target gene of TGF-b, because the Hey1 promoter
region contains two SBE sites and two CSL-binding
elements. During TGF-b-induced EMT in epithelial
cells, Hey1 and Jagged1 were induced. Hey1 is an early
response gene of TGF-b and Jagged1 is a late response
gene of TGF-b. RNA silencing of Hey1 or Jagged1
blocked TGF-b-induced EMT. There are two other
reports concerned with how BMP activates the expres-
sion of the Notch target genes, Hes5 (Takizawa et al.,
2003) and Herp2 (Itoh et al., 2004). Takizawa et al.
(2003) showed that BMP2 enhances the expression of
Hes5 in mouse neuroepithelial cells. The Hes5 promoter
has two BMP SBE sites. However, Smad1 was found to
only weakly interact with ICD1. Itoh et al. (2004)
reported that the ICD1 target gene Herp2 is synergis-
tically induced by BMP6 and ICD1. The Herp2
promoter has two CSL-binding elements and multiple
GC-rich palindromic sites that contain BMP co-SBE
sites (Korchynskyi and ten Dijke, 2002).
Effect of Notch on TGF-b signaling during mammary
development and tumorigenesis
In the present study we have shown that ICD4
expression attenuates TGF-b-induced growth inhibition
in HC-11 and EpH4 mammary epithelial cells as well as
inhibits PAI-1 mRNA and protein expression in HC-11
and NIH3T3 cells. In addition, ICD4 attenuates TGF-
b-induced transcription from the 3TP-luc reporter
vector and a 4?SBE reporter vector. Evidence was
presented that Smad2, Smad3 and Smad4 can physically
interact with the ICD4 both in vitro and in vivo and that
interaction between Smad3 and the ICD4 does not
block phosphorylation of Smad3. Further, this interac-
tion does not require the RAM23 region of the ICD4.
Others have reported that activated Notch1 or
Notch4 can affect TGF-b signaling. For instance,
Uyttendaele et al. (1998) have shown that ICD4 can
block TGF-b1-induced branching morphogenesis in
TAC-2 mouse mammary epithelial cells. In their study,
the minimal region of the ICD4 required for this activity
is the RAM23 and the CDC10/Ankrin repeats. The
molecular mechanism by which ICD4 signaling affects
branching morphogenesis in response to TGF-b1 is not
known and may be different than the pathway described
here with Smad3 since the RAM23 region of the ICD4
was required for blocking TGF-b-induced branching
morphogenesis in TAC-2 cells. Likewise, Rao and
Kadesch (2003) showed that overexpressing ICD1 in
Mv1Lu epithelial cells but not in other cell lines such as
C2C12 or NIH 3T3 could indirectly enhance c-Myc
expression and thereby render Mv1Lu epithelial cells
resistant to TGF-b growth-inhibitory signals. In the case
of HC-11 cells, we did not find that ICD4 expression
upregulated c-Myc expression compared to HC-11 cells
(data not shown). In Blokzijl et al. (2003), mouse
Notch1ICD functioned synergistically with TGF-b to
activate a multimerized CAGA element from the PAI-1
promoter in C2C12 cells. We found that Notch4ICD
suppressed both SBE and P(CAGA)9 in three different
cell lines. However, the Notch1ICD affect on TGF-b
signaling is cell line-dependent, Notch1ICD can inhibit
TGF-b signaling in HC-11 cells, but not in NIH3T3 cells
or C2C12 cells. We did not find results similar to
Blokzijl et al demonstrating that Notch1ICD can
P(CAGA)9-luc activity. A possible explanation for this
difference is that we used human Notch1ICD and not
differences between Notch1ICD and Notch4ICD. In
fact, Notch4ICD is smaller than Notch1ICD and lacks
the transcriptional activation domain (TAD) and
cytokine response (NCR) sequence found in the
Notch1ICD (Uyttendaele et al., 1996; Allenspach
et al., 2002). Shimizu et al. (2002) have reported that
there is functional diversity among the Notch receptors.
The transcriptional activities of Notch1ICD, Notch2-
ICD and Notch3ICD were markedly different from each
other, and the activities of Notch1ICD or Notch3ICD
were reduced by coexpression of Notch2ICD. Fan et al.
(2004) also showed that Notch1ICD and Notch2ICD
can have opposite effects on the growth of a single
tumor type. Moreover, Notch1 acts as an oncogene in
some carcinomas (Capobianco et al., 1997; Weijzen
et al., 2002; Kiaris et al., 2004), and as a tumor
suppressor (Nicolas et al., 2003; Fan et al., 2004) in
others. Our results taken together with these are
consistent with the conclusion that the Nocth effect on
TGF-b signaling is cell line-dependent and Notch-
Our results show that the RAM23 region of ICD4 is
not necessary to confer resistance to TGF-b-induced cell
growth inhibition or to bind Smad3. Since a RAM23-
independent pathway can disrupt or attenuate TGF-b
signaling, this may be a factor that is essential for
promoting early stages of tumorigenesis by activated
ICD4 in the mammary gland. Consistent with this
premise is our observation that blockage of endogenous
Notch4 signaling in MCF-7 breast cancer cells, which
are normally resistant to TGF-b cell growth inhibition
(Xie et al., 2002), with GSI treatment develop a
Notch4 inhibition of TGF-b signaling
Y Sun et al
Furthermore, preliminary results of a genetic cross
between WAP-ICD4 and WAP-TGF-b1 transgenic mice
(our unpublished data) in which we have compared the
latency and frequency of mammary tumors in WAP-
ICD4 mice with bitransgenic mice that express both
WAP-ICD4 and WAP-TGF-b1 are also consistent with
this speculation and further confirms our in vitro data.
These bitransgenic mice develop mammary tumors with
the same latency and frequency as the WAP-ICD4 mice,
demonstrating that the TGF-b signaling pathway is
compromised in ICD4-expressing mammary epithelial
cells. In conclusion, the inhibitory effect of ICD4 on
Smad3-dependent signaling may be biologically signifi-
cant since overexpression of ICD4 is a potent oncogene
in the mouse mammary gland and this oncogenic
activity may depend upon the ability of ICD4 to
function as antagonist of TGF-b signaling during the
early stages of tumor formation where TGF-b normally
functions as a tumor suppressor.
to TGF-b-inducedgrowth inhibition.
Materials and methods
Flag-tagged mouse Smad3, Smad3-N and Smad3-C expression
vectors were kindly provided by Tongwen Wang (Virginia
Mason Center, WA, USA) (Liu et al., 2000). Flag–tagged
human Smad2, Smad3 and Smad4 expression vectors and
adenoviral constructs encoding mouse Smad2, Smad3 and
Smad4 were gifts from Kohei Miyazono, University of Tokyo
(Fujii et al., 1999). Notch4/Int3 intracellular domain (ICD4)
and human dRAM23-ICD4 cDNA (Imatani and Callahan,
2000) were ligated into pcDNA 3.1 pEF1/V5His TOPO TA
expression vector (Invitrogen, CA, USA). The ICD4 and
dRAM-ICD4 were also cloned into the GeneSwitch System
(Invitrogen) and stably transfected into NIH3T3 cells. ICD4
and dRAM-ICD4 protein expression was induced for 16h with
10nM mifepristone in NIH3T3 cells. The HAHis-tagged
mouse ICD4 was kindly provided by Jan Kitajewski,
Columbia University (Wu et al., 2001). Human Notch1ICD
was kindly provided by Anthony J Capobianco, University of
Pennsylvania (Capobianco et al., 1997).
Cell culture, transfection and reagents
The Bosc 23 cell line, C2C12 and MCF-7 were purchased from
ATCC (Manassas, VA, USA). Stable transfectants of pEF-1
control vector, pEF-ICD4, pEF1-dRAM-ICD4 in HC-11
mouse mammary epithelial cells (a gift from Dr Nancy Hynes,
Friedrich Miescher Institute, Basel, Switzerland) and EpH4
mouse mammary epithelial cells were cultured and stably
transfected as described (Imatani and Callahan, 2000; Wech-
selberger et al., 2001). Colonies were selected and tested by
RT–PCR (Sun et al., 2001) and Western blot analysis (Bianco
et al., 2002) for transgene expression. TGF-b1 was purchased
from R&D (Minneapolis, MN, USA). g-Secretase inhibitor
(L-685, 458) was purchased from Calbiochem (San Diego,
Analysis of cell growth inhibition by TGF-b1 in vitro
In vitro growth inhibition by TGF-b1 was assessed in EpH4
wild-type and EpH4 ICD4 stable cell lines as well as HC-11,
HC-11pEF1, HC-pEF1-ICD4, HC-11 pEF1-dRAM-ICD4 cell
lines. Briefly, each of the cell lines were seeded into 12-well
plates at 1?104cells/well. The following day, the cells were
treated with 0.4, 2, 10 or 50ng/ml of TGF-b1 (R&D,
Minneapolis, MN, USA) for 7 days, and then counted. The
ratio of cell number of each treatment with TGF-b1 and the
cell number of control cells without TGF-b treatment is
expressed as the cell growth inhibition.
BrdU incorporation assay
EpH4 and EpH4-ICD4 cells were plated onto cover slips with
or without TGF-b1 (2ng/ml) for 48h. For BrdU incorpora-
tion, cells were labeled with 10mM BrdU (Roche, Indianapolis,
IN, USA) for 30min and processed according to BrdU
labeling and detection kit (Roche) instructions. Cells were
also treated with antifade/4,6-diamino-2-phenylindole (DAPI)
staining and then scored for BrdU-positive cells. The experi-
ments were performed twice, each time in duplicate.
RT–PCR analysis of PAI-1 and GAPDH in HC-11 pEF-1
and HC-11 ICD4 cells
Total RNA was isolated from HC-11 pEF-1 and HC-11 ICD4
cells treatment with TGF-b1, cDNA synthesis and PCR was
performed as described (Sun et al., 2001). RT–PCR primers
are described elsewhere (Bianco et al., 2002): mouse PAI-1:
50-atgagatcagtactgcggatgccatct and 5-gcacagagacggtgctgccat
(accession number: M16006); glyceraldehyde-3-
phosphate dehydrogenase (GAPDH): 50-cccttcattgacctcaac
Protein extraction, Western blotting and antibodies
Cells were lysed as previously described (Bianco et al., 2002;
Sun et al., 2002). Protein (50mg) was subjected to SDS–
polyacrylamide gel electrophoresis (Invitrogen, Carlsbad, CA,
USA) as described (Bianco et al., 2002; Sun et al., 2002). The
antibodies used were anti-V5 (Invitrogen, 1:5000), anti-Flag-
HRP (Sigma, St Louis, MO, USA; 1:1000), anti-Notch4
(Upstate, Charlottesville, VA, USA; 1:1000), anti-P-Smad2
(Upstate, 1:1000), anti-Smad2 (Zymed, 1:1000), anti-Smad3
(Zymed, South San Francisco, CA, USA; 1:1000), anti-P-
Smad3 (kindly provided by Dr Michael Reiss), anti-Smad2/
3(Santa Cruz, 1:250), anti-Lamin A/C (Santa Cruz Biotech-
nology, Inc., Santa Cruz, CA, USA; 1:250) and anti-PAI-1
(American Diagnostica Inc., 1:2000).
Coimmunoprecipitation of ICD4 and Smad2, Smad3 or Smad4
Bosc 23 cells (1?106cells on 60-mm-diameter plates) were
transfected with 2mg HaHis-ICD4 expression vector using
Lipofectamine reagent (Invitrogen) according to the manu-
facturer’s instruction. After 16h, cells were infected with an
empty adenovirus or an adenovirus containing mouse Smad2,
Smad3 or Smad4 genes at a multiplicity of infection of 5 in
medium containing 4mg/ml of polybrene (Sigma). At 72h after
transfection, the cells were lysed for 20min with 0.5ml lysis
buffer as previously described (Zhou et al., 2000) for 20min.
Total protein (1mg) was incubated with 30ml of anti-V5
antibody conjugated to agarose beads (Sigma) overnight at
41C. The Agarose was washed four times with 0.5ml lysis
buffer. Western blotting was analysed as described (Bianco
et al., 2002) with anti-V5 (Invitrogen) or anti-Flag horseradish
peroxidase-conjugated antibody (Sigma). Similarly, 1mg of
total protein lysates were incubated with 30ml anti-Flag
Agarose (Sigma) gel matrix overnight at 41C. The Agarose
Notch4 inhibition of TGF-b signaling
Y Sun et al
was then washed with 0.5ml lysis buffer four times, and the
samples were analysed as described above.
Coimmunoprecipitation of ICD4 and Smad3, N-Smad3 (MH1)
or C-Smad3 (MH2)
Bosc23 cells (1?106cells on 60-mm-diameter plates) were
transfected with 2mg V5His-ICD4 expression vector alone or
with 2mg Flag-Smad3, Flag-Smad3 (MH1) and Flag-Smad3
(MH2) using LipofectAMINE reagent (Invitrogen). Cells were
lysed 72h after transfection, and coimmunoprecipitation was
processed as described above.
Coimmunoprecipitation of ICD4 and Smad3 from Wap-ICD4
transgenic mammary gland
Mammary glands from pregnant (day 15) FVB/N or Wap-
ICD4 mice were lysed with lysis buffer as described above. The
341 rabbit anti-ICD4 antibody (Smith et al., 1995) was
conjugated to protein G-agarose (Roche) by adding 4mg of
341–50ml protein G-agarose in 1ml of lyses buffer at 41C with
rotation for 2h. Excess antibody was removed by washing in
lysis buffer and centrifugation at 12000g for 30s four times.
After the last wash, the supernatant was removed and a 1ml
sample containing 2mg total protein was added to the antibody
bound to the Protein G-agarose. This complex was rotated
overnight at 41C. The next day, the complexes were washed
three times. The Western blots were probed with Anti-Smad3
(Santa Cruz, 1:250) and anti-Notch4 antibody (Santa Cruz,
1:250) to detect Smad3 and ICD4 in 341 immunoprecipatates.
HC-11, EpH4 and 293T/17 cell lines were transfected with 1mg
ICD4 pEF-1 expression vector and 0.5mg 3TP-luc (Wrana
et al., 1992) or 4xSBE-luc (Zawel et al., 1998) using
lipofectamine or lipofectamine 2000 regent (Invitrogen). The
total amount of DNA for each transfection was adjusted to
2mg using pEF1 control vector (Invitrogen). Renilla luciferase
control reporter vector (pRL-TK) (Promega, Madison, WI,
USA) was cotransfected in the cells to normalize for
transfection efficiency. At 5h after transfection, complete
media were added to the cells. At 48h after transfection, cells
were lysed and luciferase activity was determined by using a
Dual-luciferase reporter assay system (Promega), according to
the manufacturer’s manual. All assays were performed in
triplicate and represent a mean (7s.e.) of three independent
We thank Dr Anita B Roberts and Dr Adam B Glick for
critical comments about this manuscript. We thank Dr Kohei
Miyazono, Dr Tongwen Wang and Dr Ying Zhang for
generous gifts of Smad constructs; Dr Jan Kitajewswi for his
generous gift of HaHis-Int3; Dr Anthony J Capobianco,
University of Pennsylvania for Human Notch1ICD construct
and Dr Michael Reiss for providing the phospho-Smad3
antibody. We thank Ahmed Raafat, Tatsuro Ohta, David
McCurdy, Xin liu, Vijayachandra Kinnimulki and Jun Zhou
for their helpful discussion and technical assistance.
Akiyoshi S, Inoue H, Hanai J, Kusanagi K, Nemoto N,
Miyazono K and Kawabata M. (1999). J. Biol. Chem., 274,
Allenspach EJ, Maillard I, Aster JC and Pear WS. (2002).
Cancer Biol. Ther., 1, 466–476.
Baron M, Aslam H, Flasza M, Fostier M, Higgs JE,
Mazaleyrat SL and Wilkin MB. (2002). Mol. Membr. Biol.,
Bianco C, Adkins HB, Wechselberger C, Seno M, Normanno
N, De Luca A, Sun Y, Khan N, Kenney N, Ebert A,
Williams KP, Sanicola M and Salomon DS. (2002). Mol.
Cell. Biol., 22, 2586–2597.
Blobe GC, Schiemann WP and Lodish HF. (2000). N. Engl. J.
Med., 342, 1350–1358.
Blokzijl A, Dahlqvist C, Reissmann E, Falk A, Moliner A,
Lendahl U and Ibanez CF. (2003). J. Cell Biol., 163, 723–728.
Callahan R and Egan SE. (2004). J. Mammary Gland Biol.
Neoplasia, 9, 145–163.
Capobianco AJ, Zagouras P, Blaumueller CM, Artavanis-
Tsakonas S and Bishop JM. (1997). Mol. Cell. Biol., 17,
Cordenonsi M, Dupont S, Maretto S, Insinga A, Imbriano C
and Piccolo S. (2003). Cell, 113, 301–314.
Dennler S, Itoh S, Vivien D, ten Dijke P, Huet S and Gauthier
JM. (1998). EMBO J., 17, 3091–3100.
Ellisen LW, Bird J, West DC, Soreng AL, Reynolds TC, Smith
SD and Sklar J. (1991). Cell, 66, 649–661.
Fan X, Mikolaenko I, Elhassan I, Ni X, Wang Y, Ball D, Brat
DJ, Perry A and Eberhart CG. (2004). Cancer Res., 64,
Feng XH, Zhang Y, Wu RY and Derynck R. (1998). Genes
Dev., 12, 2153–2163.
Fujii M, Takeda K, Imamura T, Aoki H, Sampath TK,
Enomoto S, Kawabata M, Kato M, Ichijo H and Miyazono
K. (1999). Mol. Biol. Cell, 10, 3801–3813.
Gallahan D, Jhappan C, Robinson G, Hennighausen L, Sharp
R, Kordon E, Callahan R, Merlino G and Smith GH.
(1996). Cancer Res., 56, 1775–1785.
Hocevar BA, Smine A, Xu XX and Howe PH. (2001). EMBO
J., 20, 2789–2801.
Imatani A and Callahan R. (2000). Oncogene, 19, 223–231.
Iso T, Sartorelli V, Chung G, Shichinohe T, Kedes L and
Hamamori Y. (2001). Mol. Cell. Biol., 21, 6071–6079.
Itoh F, Itoh S, Goumans MJ, Valdimarsdottir G, Iso T, Dotto
GP, Hamamori Y, Kedes L, Kato M and Dijke Pt P. (2004).
EMBO J., 23, 541–551.
Jarriault S, Brou C, Logeat F, Schroeter EH, Kopan R and
Israel A. (1995). Nature, 377, 355–358.
Jeffries S and Capobianco AJ. (2000). Mol. Cell. Biol., 20,
Jhappan C, Gallahan D, Stahle C, Chu E, Smith GH, Merlino
G and Callahan R. (1992). Genes Dev., 6, 345–355.
Kawabata M, Inoue H, Hanyu A, Imamura T and Miyazono
K. (1998). EMBO J., 17, 4056–4065.
Kiaris H, Politi K, Grimm LM, Szabolcs M, Fisher P,
Efstratiadis A and Artavanis-Tsakonas S. (2004). Am. J.
Pathol., 165, 695–705.
Kopan R, Schroeter EH, Weintraub H and Nye JS. (1996).
Proc. Natl. Acad. Sci. USA, 93, 1683–1688.
Korchynskyi O and ten Dijke P. (2002). J. Biol. Chem., 277,
Kurokawa M, Mitani K, Irie K, Matsuyama T, Takahashi T,
Chiba S, Yazaki Y, Matsumoto K and Hirai H. (1998).
Nature, 394, 92–96.
Notch4 inhibition of TGF-b signaling
Y Sun et al
Labbe E, Letamendia A and Attisano L. (2000). Proc. Natl. Download full-text
Acad. Sci. USA, 97, 8358–8363.
Liu X, Elia AE, Law SF, Golemis EA, Farley J and Wang T.
(2000). EMBO J., 19, 6759–6769.
Maier MM and Gessler M. (2000). Biochem. Biophys. Res.
Commun., 275, 652–660.
Massague J. (2000). Nat. Rev. Mol. Cell. Biol., 1, 169–178.
Miyazono K. (2000). J. Cell Sci., 113 (Part 7), 1101–1109.
Moustakas A, Souchelnytskyi S and Heldin CH. (2001). J. Cell
Sci., 114, 4359–4369.
Nicolas M, Wolfer A, Raj K, Kummer JA, Mill P, van Noort
M, Hui CC, Clevers H, Dotto GP and Radtke F. (2003).
Nat. Genet, 33, 416–421.
Rao P and Kadesch T. (2003). Mol. Cell. Biol., 23, 6694–6701.
Shimizu K, Chiba S, Saito T, Kumano K, Hamada Y and
Hirai H. (2002). Biochem. Biophys. Res. Commun., 291,
Smith GH, Gallahan D, Diella F, Jhappan C, Merlino G and
Callahan R. (1995). Cell Growth Differ., 6, 563–577.
Sun Y, Kuraishi T, Aoki F and Sakai S. (2001). Mol. Cell.
Endocrinol., 172, 177–184.
Sun Y, Nonobe E, Kobayashi Y, Kuraishi T, Aoki F,
Yamamoto K and Sakai S. (2002). J. Biol. Chem., 277,
Takizawa T, Ochiai W, Nakashima K and Taga T. (2003).
Nucleic Acids Res., 31, 5723–5731.
Tamura K, Taniguchi Y, Minoguchi S, Sakai T, Tun T,
Furukawa T and Honjo T. (1995). Curr. Biol., 5, 1416–1423.
Uyttendaele H, Marazzi G, Wu G, Yan Q, Sassoon D and
Kitajewski J. (1996). Development, 122, 2251–2259.
Uyttendaele H, Soriano JV, Montesano R and Kitajewski J.
(1998). Dev. Biol., 196, 204–217.
Wakefield LM and Roberts AB. (2002). Curr. Opin. Genet.
Dev., 12, 22–29.
Wechselberger C, Ebert AD, Bianco C, Khan NI, Sun Y,
Wallace-Jones B, Montesano R and Salomon DS. (2001).
Exp. Cell Res., 266, 95–105.
Weijzen S, Rizzo P, Braid M, Vaishnav R, Jonkheer SM,
Zlobin A, Osborne BA, Gottipati S, Aster JC, Hahn WC,
Rudolf M, Siziopikou K, Kast WM and Miele L. (2002).
Nat. Med., 8, 979–986.
Wrana JL, Attisano L, Carcamo J, Zentella A, Doody J, Laiho
M, Wang XF and Massague J. (1992). Cell, 71, 1003–1014.
Wu G, Lyapina S, Das I, Li J, Gurney M, Pauley A, Chui I,
Deshaies RJ and Kitajewski J. (2001). Mol. Cell. Biol., 21,
Wu L, Wu Y, Gathings B, Wan M, Li X, Grizzle W, Liu Z,
Lu C, Mao Z and Cao X. (2003). J. Biol. Chem., 278,
Xie W, Mertens JC, Reiss DJ, Rimm DL, Camp RL, Haffty
BG and Reiss M. (2002). Cancer Res., 62, 497–505.
Zagouras P, Stifani S, Blaumueller CM, Carcangiu ML and
Artavanis-Tsakonas S. (1995). Proc. Natl. Acad. Sci. USA,
Zavadil J, Cermak L, Soto-Nieves N and Bottinger EP. (2004).
EMBO J., 23, 1155–1165.
Zawel L, Dai JL, Buckhaults P, Zhou S, Kinzler KW,
Vogelstein B and Kern SE. (1998). Mol. Cell, 1, 611–617.
Zhou Y, Sun H, Danila DC, Johnson SR, Sigai DP, Zhang X
and Klibanski A. (2000). Mol. Endocrinol., 14, 2066–2075.
Notch4 inhibition of TGF-b signaling
Y Sun et al