(Received for publication, January 16, 1997, and in revised form, March 21, 1997)
From the Laboratory of Chemoprevention, NCI, National Institutes of Health, Bethesda, Maryland 20892-5055
Smad proteins are a family of highly conserved,
intracellular proteins that signal cellular responses downstream of
transforming growth factor- (TGF-
) family serine/threonine kinase
receptors. One of these molecules, Smad4, originally identified as the
candidate tumor suppressor gene dpc-4, reconstitutes
TGF-
- and activin-dependent transcriptional responses in
Smad4 null cell lines and interacts in a ligand-dependent
manner with other Smad family members in both TGF-
, activin, and
bone morphogenetic protein-2/-4 pathways. Here, we used an assay based
on the restoration of ligand-dependent transcriptional
responses in a Smad4 null cell line to characterize functional domain
structures within Smad4. We showed that restoration of TGF-
-induced
transcriptional responses by Smad4 was inhibited by co-transfection
with a kinase dead TGF-
type II receptor and that constitutive
activation was blocked with TGF-
neutralizing antibodies, confirming
the essential role of Smad4 in TGF-
signaling. Using a series of
Smad4 mutation, deletion, and Smad1/Smad4 chimera constructs we
identified a 47-amino acid deletion within the middle-linker region of
Smad4 that is essential for the mediation of signaling responses. In
addition, we showed that the NH2-terminal domain of Smad4
augments ligand-dependent activation associated with the
middle-linker region, indicating that there is a distinct ligand-response domain within the N terminus of this molecule.
TGF-1 is the prototype for a family
of structurally related proteins that mediate a diverse array of
biological functions in organisms as disparate as the fruit fly and
man. Cloning and characterization of TGF-
family receptors revealed
that these ligands signal through two interacting transmembrane
serine/threonine kinase receptors (1). While it is recognized that the
ligand binds the type II receptor, which then recruits and
transphosphorylates the type I receptor to signal downstream pathways
(2), the mechanisms involved in mediating signaling cascades downstream of the type I receptor were, until recently, largely unknown.
Clues from Drosophila genetics have been applied to identify
a number of novel mammalian proteins involved in signaling downstream of TGF- family members (1). In Drosophila,
decapentaplegic (dpp) encodes a TGF-
superfamily
ligand that is essential for dorsal-ventral patterning in developing
fruit flies (3). Genetic screening for mutants acting in this pathway
identified a novel gene, Mothers against dpp
(mad), which encodes an essential cytoplasmic element in the
dpp signal transduction pathway (4, 5). Searches for Mad-related
proteins in other organisms have revealed an array of conserved
homologues in species from Caenorhabditis elegans to man
(1). Characteristically these proteins contain highly conserved
NH2- and COOH-terminal domains separated by a proline-rich linker and lack any known enzymatic or functional sequence motifs. Five
mammalian Mad-related proteins, Smad1 through 5, have been characterized and shown to participate in signaling responses induced
by TGF-
, activin, and BMP2/4.
The rudimentary mechanisms of action of the Smad family proteins are
now beginning to be understood. For example, Smad1 is phosphorylated
and undergoes nuclear translocation following treatment with BMP2, BMP4
(6, 7), and possibly TGF- (8, 9), while Smad2 is phosphorylated and
undergoes nuclear translocation in response to TGF-
and activin (10,
11). This phosphorylation event occurs in association with the
activated TGF-
receptors and results in dissociation of Smad2 from
the receptor complex and translocation to the nucleus (11). Recent
studies using Xenopus embryos indicate that Smad2 forms a
transcriptional complex with the winged-helix transcription factor
FAST1 when co-injected with activin mRNA (12), providing a
downstream mechanism whereby activin response elements may be activated
following ligand-induced phosphorylation of Smad2. Interestingly, Smad3
has also been shown to signal in the TGF-
pathway, to associate with
the activated TGF-
receptor complex, and to be phosphorylated
following activation (13), suggesting that receptor-mediated
phosphorylation and subsequent nuclear translocation may represent a
more generalized phenomenon. Furthermore, Smad1 displays
transcriptional activity in response to BMP4 when fused to a
heterologous Gal4 DNA-binding domain (7), indicating that other members
of Smad family may act as transcriptional activators once translocated
to the nucleus.
Smad4 is a Mad-related protein independently identified as the
candidate tumor suppressor gene dpc4, deleted or mutated in a proportion of pancreatic (14), breast, ovarian (15), and colorectal
tumors (16), with tumor suppressor activity attributable to its
participation in TGF- receptor-mediated signaling (13, 17). However,
while this molecule shares broad sequence similarity with the other
Smad family members, it displays significant variance in all three
molecular domains (1), suggesting that it may have distinct functional
characteristics. For example, while Smad4 has been clearly implicated
in both TGF-
and activin signaling pathways (13, 17), it is
phosphorylated only in response to activin (17) and does not associate
with activated TGF-
receptor complexes (11, 13). Smad4 also lacks
the COOH-terminal serine residues that are phosphorylated in Smad2
following activation of the TGF-
receptor complex (11) and are
present in all the other mammalian Smad family members. In addition,
Smad4 not only forms ligand-dependent heterodimeric
complexes with Smad2 following activation of TGF-
or activin
pathways, but also with Smad1 following activation through BMP2/4
receptors (17). These data further support a model in which Smad4 acts
in a manner distinct from Smad1, -2, and -3, and is central to
signaling pathways involving multiple TGF-
family ligands. However,
the mechanisms regulating ligand-dependent transmission of
signaling responses by Smad4 remain largely unknown.
In this study, we employ an assay based on the restoration of
ligand-dependent transcriptional responses in a Smad4 null
cell line to characterize the functional domains within Smad4 that are
essential for the mediation of TGF--dependent signaling
responses. Using a series of mutants, chimeras, and deletion
constructs, we have identified a 47-amino acid region within the
proline-rich middle linker domain of Smad4 which, when deleted,
disrupts the ability of the wild-type protein to restore
TGF-
-dependent signal transduction in the Smad4 null
cell line. We also show that the NH2-terminal domain of
Smad4 enhances TGF-
-dependent activation associated
with the middle-linker region, indicating that there is separate
ligand-response domain within the N terminus of this molecule.
Mutant, chimeric,
deletion, and COOH-terminal constructs were designed as shown in the
Figs. 5, 6, and 8 based on the known sequence similarity domains (8),
and cloned into the pCDNA3 expression vector. These were generated
by polymerase chain reaction amplification and primer extension with
proofreading polymerase, using Smad4 and Smad1 templates. Deletion
constructs were sequenced using the dideoxynucleotide method.
Cell Lines and Proliferation Assay
MDA-MB468 is a cell line
derived from a patient with breast carcinoma, which has a homozygous
deletion of the complete Smad4 coding region (15). A549 is a
lung carcinoma cell line with intact Smad4 genes. All cells
were obtained from the American Type Culture Collection and maintained
in Dulbecco's modified Eagle's medium containing 10% fetal bovine
serum and antibiotics. Cellular proliferation was determined by
[3H]thymidine incorporation as described previously (18),
after treatment with or without 10 ng/ml recombinant human TGF-1
over a 48-h period in medium containing 10% fetal bovine serum.
Recombinant human TGF-1 was
labeled with 125Iodine using the chloramine-T method (19).
Cells were seeded at 60% confluence in 100-mm plates and incubated for
4 h at 4 °C with 100 pM [125I]
TGF-
1 with or without 100-fold excess of unlabeled TGF-
1 competitor. Cross-linking was performed with disuccinimidyl suberate as
described previously (20), samples subjected to electrophoresis on a
4-12% gradient sodium dodecyl sulfate-polyacrylamide gel electrophoresis followed by autoradiography.
Cells were seeded in six-well dishes, incubated for
18 h in medium containing 0.2% fetal bovine serum with or without
1 ng/ml TGF-1, and labeled with 50 µCi/ml
[35S]methionine over the last 6 h. After metabolic
labeling, PAI-1 and fibronectin were extracted from the cell associated
extracellular matrix and supernatants, respectively, as described
previously (21). Both preparations were subjected to 10%
SDS-polyacrylamide gel electrophoresis followed by autoradiography.
PAI-1 and fibronectin were identified from their characteristic sizes
(45 and 240 kDa) and inducibility by TGF-
in a TGF-
-responsive
cell line.
MDA-MB468 cells were
transiently transfected using the polyamine transfection reagent LT-2
(Pan Vera Co.). Cells were seeded to 50% confluence in six-well plates
and incubated in serum-free medium with LT-2, cDNAs, reporter
constructs, and empty vector to equalize amounts of DNA. After 6 h, cells were washed and allowed to recover for 24 h in growth
medium. Cells were serum starved overnight prior to the addition of
TGF-1 for a further 24 h. Luciferase activity was determined in
the cell lysate using an assay kit (Analytic Luminescence Laboratory),
and a Dynatech Laboratories ML3000 luminometer. Activities were
normalized on the basis of
-galactosidase expression from pSV
-galactosidase in all luciferase reporter experiments. All
experiments were repeated at least three times with similar
results.
To develop a functional assay for
Smad4, we used the MDA-MB468 breast carcinoma cell line that has a
homozygous deletion of the complete Smad4 coding region
(15). Northern blot analysis shows that these cells express Smad1,
Smad2, Smad3, and Smad5 and confirms the absence of Smad4 mRNA
expression.2 Receptor cross-linking shows
that these cells express types I and II TGF- receptors (Fig.
1A), but show no growth inhibitory (Fig.
1B), PAI-1 or fibronectin induction responses to TGF-
(Fig. 1C). Having established that Smad4 deficiency is
associated with a lack of responsiveness to TGF-
, we performed a
series of experiments to characterize the role of this molecule in
TGF-
-dependent signaling responses. Transient
transfection with Smad4 and the reporter constructs p3TP-Lux (which
contains multiple TGF-
response elements (21)) or p800-Luc (which
contains only the PAI-1 gene promoter (22)), resulted in a clear
activation of luciferase reporter activity in response to TGF-
(Fig.
2A). Ligand-dependent activation of p800-Luc was lower than that of p3TP-Lux, although both were dependent on the dose of Smad4 DNA added to the transfection mix (Fig.
2A) and the concentration of exogenous TGF-
used to treat the cells (Fig. 2B).
To determine if these responses were TGF-
receptor-dependent, we evaluated reporter gene activation
after co-transfection of TGF-
type I and type II receptors with and
without Smad4 (Fig. 3A). Transfection of
TGF-
type I or type II receptors alone had no effect on the
transcriptional response to TGF-
, although co-transfection of Smad4
with the TGF-
type II receptor caused an apparent ligand-independent activation of the p3TP-Lux reporter gene (Fig. 3A).
Importantly, this effect was blocked almost entirely by treatment with
TGF-
neutralizing antibodies (Fig. 3B), indicating that
the apparently ligand-independent signaling response seen in cells
transfected with both the type II TGF-
receptor and Smad4 is the
result of stimulation by secreted TGF-
. In addition, while
transfection of the kinase inactive TGF-
type II receptor alone had
no effect on the transcriptional response to TGF-
, co-transfection
of the mutant receptor with Smad4 caused a nearly complete loss of
ligand-induced signaling (Fig. 4), confirming the
receptor specificity of the Smad4 induced response. Furthermore, basal
transcriptional activation of the p3TP-Lux reporter with Smad4 was also
inhibited by co-transfection with the kinase-inactive type II TGF-
receptor, indicating that stimulation of endogenous receptors by
autocrine or paracrine TGF-
may account for some of the constitutive
activation seen in the absence of additional exogenous ligand.
Functional Analysis of COOH-terminal Domains of Smad1 and Smad4
Studies on Smad1 and Smad2 have identified mutations within the COOH-terminal domains which block ligand-dependent phosphorylation (6, 10) and inhibit dorsal mesoderm induction by Smad2 in Xenopus (10). Furthermore missense mutations of a highly conserved glycine residue in the COOH-terminal domain of Drosophila Mad, and C. elegans Sma2, generate null phenotypes in both species (4, 23), indicating that the functional activity of these molecules may be disrupted by specific point mutations within this domain. The corresponding glycine to serine substitution at position 508 in Smad4 (Fig. 5A) did not abolish ligand-dependent augmentation of reporter gene activation, although it significantly reduced the maximal luciferase induction compared with the wild-type protein (Fig. 5B). This suggests that the biological inactivation resulting from mutation of this residue in other Mad-related proteins (4, 10, 23) may result from a decrease in specific transcriptional responses below critical levels rather than a total loss of function. Alternatively, other Mad family members may be completely inactivated by this mutation, while the Smad4 G508S mutation retains some signaling activity.
The COOH-terminal domains of Smad1 (234-465) and Smad4 (266-552) display constitutive transcriptional activity when fused to a heterologous Gal4 DNA-binding domain, while the full-length Smad1 fusion protein demonstrates ligand-dependent transcriptional activation (7). In addition, Lagna et al. (17) have shown a constitutive, ligand-independent activation of p3TP-Lux reporter activity when MDA-MB468 Smad4 null cells are transfected with the 266-552 COOH-terminal Smad4 construct. In our study, however, neither the Smad1 (270-465) nor the Smad4 (322-552) COOH-terminal domain constructs (Fig. 5A) induced constitutive or ligand-dependent transcriptional activation of the reporter gene (Fig. 5B). This indicates that there may be an activation domain between amino acids 266-321 in Smad4, which is involved in the regulation of signaling responses by this protein.
Mapping of a Smad4 Activation Domain (SAD) within the Middle-linker Region of Smad4To define the activation domain within the
middle-linker region of Smad4 further, we created a series of amino
acid deletions within this region (Fig. 6A).
When these constructs were transfected into MDA-MB468 cells, the M1
(136-181) and
M2 (182-227) deletions had similar functional
activity to the full-length molecule in restoring
TGF-
-dependent activation of the p3TP-Lux reporter. In
contrast, deletion of amino acids 228-273 in
M3 substantially reduced this response, while the
M4 (274-321) deletion had
virtually no signaling activity (Fig. 6b).
Smad1
alone does not restore reporter gene activation in MDA-MB468 cells, nor
does it augment Smad4-dependent activity (Fig. 5B; Fig. 7). For this reason, we were able to
extend our studies to identify regions of Smad4 that confer
ligand-specific responses, by generating a panel of domain-swap
chimeras between Smad4 and Smad1 (Fig. 8A).
Like the Smad4 G508S mutant, none of the chimeras reconstituted the
degree of luciferase expression seen with intact Smad4 (Fig.
8B), but the presence of a Smad4 proline-rich linker domain
was associated with reproducible ligand-responsive reporter gene
activation, regardless of the amino- or carboxyl-terminal domain
structures. Furthermore, while the N-terminal domain of Smad4 had no
effect on ligand-dependent transcriptional activity in the
4.1.1 and 4.1.4 chimeras, ligand-dependent activation was dramatically increased when the Smad4 NH2 terminus was
associated with the Smad4 middle-linker domain in the 4.4.1 chimera,
bringing the level of ligand-dependent activation to within
36 (±7)% of the full-length Smad4 (results from three separate
experiments expressed as mean (±S.D.)). This is comparable with the
ligand-dependent transcriptional activation seen with the
G508S mutation of Smad4 (Fig. 5B), suggesting that this
mutation may be functioning like the 4.4.1 chimera, with incomplete
activation of response elements in Smad4. These data indicate that
there is a ligand-response domain within the N terminus of Smad4 and
that this region requires the presence of a Smad4 middle-linker domain
to produce ligand-dependent transcriptional activation
responses to TGF-.
Smad4/DPC4 is a member of the Smad family of intracellular
signaling proteins and appears to be central to signaling pathways involving multiple TGF- family ligands (17). Despite this, the
mechanisms regulating the transmission of signaling responses by this
molecule are largely unknown. In these studies, we characterize separate activation and ligand-response domains within Smad4 using a
functional assay based on the restoration of TGF-
responsiveness in
a Smad4 null cell line by transient transfection with Smad4.
We have shown that restoration of cellular responses to TGF- by
Smad4 in MDA-MB468 cells is dependent on TGF-
receptor function, as
reporter gene activation is inhibited by co-transfection with a
kinase-inactive TGF-
type II receptor mutant shown previously to
exert a dominant negative effect on TGF-
-induced transcriptional responses in other systems (24). Furthermore, augmentation of basal
reporter gene activation seen following co-transfection of Smad4 and
the wild-type TGF-
type II receptor, and its reversal following
incubation of the transfected cells with TGF-
neutralizing antibodies, indicates that restoration of cellular responses following transfection with Smad4 is dependent both on the level of receptor expression and the degree of autocrine or paracrine stimulation of this
pathway resulting from the secretion of TGF-
into the culture
medium. In the same manner, basal reporter gene activation following
transfection with Smad4 alone could be inhibited by co-transfection
with a kinase inactive TGF-
type II receptor, indicating that
stimulation of endogenous receptors by secreted TGF-
could account
for some of the variability in basal activation of this system. This
also has implications for other assay systems. For example, the
apparently ligand-independent activation of a PAI-1 reporter construct
seen following co-transfection of Smad3 and Smad4 into Smad4 null SW480
cells (13) could result from the effects of endogenous TGF-
on this
cell line. Indeed these authors showed that constitutive reporter gene
activation associated with transfection of Smad3 and Smad4 was reduced
when their cells were co-transfected with a dominant negative TGF-
type II receptor construct, indicating that the observed basal reporter
gene activation is likely to have resulted from stimulation of the
TGF-
receptor signaling pathway by endogenous TGF-
.
Having characterized this assay for Smad4, we have used it to identify
functional domains within the Smad4 molecule that are involved in
regulating TGF--induced transcriptional responses. Comparison of our
initial studies, demonstrating that a 322-552 COOH-terminal Smad4
construct had no effect on constitutive or ligand-dependent
activation of the p3TP-Lux reporter gene (Fig. 5), with those of Lagna
et al. (17) demonstrating constitutive p3TP-Lux activation
in the same cell line following transfection with a 266-552 Smad4
construct (17), indicates that there may be critical residues within
the intervening amino acids 266-321 in Smad4 that are required to
activate signaling responses by this protein. We have delineated this
region further with the demonstration that deletion of amino acid
residues 274-321 in Smad4 prevents the restoration of TGF-
responsiveness by this molecule in our assay system (Fig. 6). While it
is possible that this is an indirect effect resulting from
conformational disruption of other domain structures within Smad4,
taken together with data derived using the Smad4 COOH-terminal
constructs, these results suggest that there is an SAD within the
COOH-terminal portion of the middle-linker region of Smad4 that is
necessary for this molecule to transduce signaling responses downstream
from TGF-
receptors. This domain of Smad4 shows minimal sequence
similarity with the other Smad family members, and it is of interest
that a COOH-terminal Smad1 construct 234-465, corresponding to
residues 270-552 in Smad4, does not restore constitutive activation of the p3TP-Lux reporter in this assay system (17), indicating that the
localization of the SAD between residues 274 and 322 in Smad4 may be
unique to this particular Smad family member.
Transfection with the full-length Smad4 induces a much lower basal
activation of the p3TP-Lux reporter gene than the 266-552 construct
(17). This suggests that the signaling response associated the SAD is
reversibly inhibited by either direct or indirect interaction with
structures upstream of residue 266 in the full-length Smad4. Interestingly, Baker and Harland (25) describe a constitutively active
NH2-terminal truncation of Smad2 184-467, corresponding to
residues 160-552 in Smad4, that may result from the deletion of a
similar inhibitory component in the NH2 terminus of this molecule. Our data indicate that this type of inhibitory domain is also
located within the NH2 terminus of Smad4, as none of the segmental deletions upstream of the SAD and within the middle-linker domain of Smad4 M1,
M2, and
M3 (137-273) were constitutively active in our assay system (Fig. 6). On this basis, ligand-activated signal transduction by Smad4 may result from dissociation of intra- or
extramolecular inhibitory structures from the SAD, themselves under the
regulation of a ligand-responsive region within the NH2
terminus of Smad4.
Experiments using the Smad1/Smad4 domain swap chimeras further
characterize the activation and ligand-responsive domains in Smad4.
Thus, while the proline-rich linker domain of Smad4 is essential for
the transduction of TGF- signaling responses by the Smad1/Smad4
chimeras, results with the 1.4.4 and 1.4.1 chimeras show that the
activity of this region is only weakly enhanced by the
NH2-terminal domain of Smad1. In contrast, TGF-
responses are significantly enhanced when the Smad1 NH2
terminus is swapped for the corresponding NH2-terminal
domain of Smad4 in the 4.4.1 chimera (Fig. 8). This indicates that
there is a ligand-response domain located within the NH2
terminus of Smad4 and that the ligand-dependent regulatory
activity of this domain requires the presence of the Smad4
middle-linker region.
Taken together, these data suggest a model for Smad4 functional domains that is reminiscent of that of other signaling molecules. In the absence of ligand, the NH2 terminus and/or middle-linker domains of Smad4 may obscure the SAD located at the extreme COOH terminus of the middle-linker domain, either by direct blockade of this region or by conformational interference. Following ligand-activation, there may be unfolding of the molecular structure resulting in exposure of the SAD, allowing it to interact with other signaling partners. This event is not associated with phosphorylation of Smad4 (17), but may result from intermolecular conformational interference with NH2-terminal domain structures following association between Smad4 and other, activated Smad family members (17). Identification of specific activation domains within the N terminus and how they interact with the rest of the molecule will help elucidate this mechanism.
We thank Scott Kern for advice regarding
dpc4 null cell lines and other members of the Laboratory of
Chemoprevention for their helpful advice and criticisms. Jeff Wrana and
Lilliana Attisano generously provided the TGF- receptor and p3TP-Lux
constructs. The Smad4 expression construct was a kind gift from
Seong-Jin Kim, and p800-Luc was provided by D. Luskutoff. Catherine
Bonham and China Eng provided technical assistance. We also thank Drs. Shin-Geon Choi, Young-Suk Yi, and Yong-Seok Kim for assistance with the
preparation of synthetic oligonucleotides