(Received for publication, August 12, 1996, and in revised form, October 30, 1996)
From the Ludwig Institute for Cancer Research, Box
595, S-751 24 Uppsala, Sweden and ¶ Laboratory of Cancer
Genetics, Department of Pathology, Göteborg University,
Sahlgrenska Hospital, S-413 45 Göteborg, Sweden
Transforming growth factor- (TGF-
)
superfamily members are multifunctional cytokines that exert their
effects via heteromeric complexes of two distinct serine and threonine
kinase receptors. Drosophila mothers against decapentaplegic
and related genes in Caenorhabditis elegans, Xenopus,
and mammals were shown to function downstream in the intracellular
signaling pathways of TGF-
superfamily members. Here we report the
cloning of a Mad-related protein, termed Sma- and Mad-related protein 2 (Smad2). TGF-
stimulated the phosphorylation and nuclear
translocation of Smad2 in nontransfected Mv1Lu cells. In addition, we
demonstrated that TGF-
and activin mediated phosphorylation of Smad2
after its overexpression with appropriate type I and II receptors in
COS cells. Smad2 and Smad1 were found to be broadly expressed in human
tissues. Smad2 is closely linked to DPC4 on
chromosome 18q21.1, a region often deleted in human cancers.
Cells that lack Smad2 may escape from TGF-
-mediated growth
inhibition and promote cancer progression.
Transforming growth factor 1
(TGF-
1)1 is the prototype of a family of
structurally related cytokines, which also includes activins and bone
morphogenetic proteins (BMPs). TGF-
superfamily members are
multifunctional and have been shown to control proliferation, differentiation, migration, and apoptosis of many different cell types
(1). Signaling of these proteins occurs via ligand-mediated hetero-oligomerization of type I and II receptors, which are both endowed with intrinsic serine and threonine kinase activities (2). The
TGF-
and activin type II receptors are constitutively active kinases
that transphosphorylate their specific type I receptors on association.
The type I receptors thereby become activated, which is necessary and
sufficient for most signaling responses (3-5). The functionally active
receptor complex may be heterotetrameric, consisting of two molecules
of each receptor type (6).
Targets downstream of serine and threonine kinase receptors that
provide the link with the transcriptionally induced effects of TGF-
superfamily members are poorly understood. TGF-
-like signaling
pathways are present in Drosophila and Caenorhabditis elegans. Genetic analysis of signaling mediated by
decapentaplegic, a member of the TGF-
superfamily in
Drosophila, has led to the identification of mothers
against dpp (Mad) (7), a gene the activity of which is
controlled by Dpp. Genetic evidence suggests that Mad is an essential
downstream component for all Dpp-dependent signaling events
(7-10). In addition, Mad homologous genes have recently
been identified in C. elegans (sma genes) (11)
and Xenopus (12) and mammals (Smad genes) (10,
13-15). A putative tumor suppressor gene function in pancreatic cancer
was ascribed to DPC4 (14). The Mad family of proteins lacks
known sequence motifs but has conserved N-terminal (MH1) and C-terminal
(MH2) domains, which are linked by a sequence rich in serine, threonine and proline residues. Smad1 is phosphorylated and translocated to the
nucleus after stimulation of BMP-2 (10). In addition, the
C-terminal domain of Smad1 was found to have transcriptional activity
when fused to a heterologous DNA binding domain. The whole Smad1
protein was found to be transcriptionally latent and became activated
on BMP receptor-mediated phosphorylation (13). In Xenopus,
Smad1 induced ventral mesoderm, a BMP-like response, and Smad2 induced
dorsal mesoderm, a Vg1- and nodal-like response (12). These results
suggest that Mad family members may encode transcription factors
that mediate distinct responses to TGF-
family members.
Here we demonstrate the TGF--induced phosphorylation and nuclear
translocation of a ubiquitously expressed human Mad-related protein,
Smad2. The chromosomal location of Smad2 is close to DPC4 on chromosome 18q21.1, a region often deleted in
cancer, suggesting a putative tumor suppressor function for Smad2.
Smad1 was cloned by reverse
transcription-polymerase chain reaction (RT-PCR) from placenta tissue
using the primers 5-TATAACTAGTGCTGTCATTATG-3
and
5
-TGCAAGGCTCAATAGTTTTCCA-3
that were based on the expression sequence tags (ESTs) R19015[GenBank] and R10104[GenBank] from the Washington
University-Merck EST Project. Smad2 was identified by RT-PCR from human
erythroleukemia cells using a degenerated sense primer,
5
-CCGAATTCTG(TC)AT(CT)AA(CT)CC(ACT)AC-3
, and an antisense primer,
5
-CCGGATCC(AG)TT(AGCT)C(GT)(AG)TT(AGCT)AC(AG)TT-3
, that were based on
regions of sequence similarity between the Drosophila Mad
and C. elegans Sma genes. The 0.5-kb insert of the Smad2 PCR
recombinant was used as a probe in the screening of a human brain
cDNA library, and the sequence of a 2-kb Smad2 cDNA clone was
determined.
Northern blot filters with mRNAs from different tissues were obtained from Clontech. cDNA restriction fragments encoding complete coding regions of Smad1 and Smad2 were used as probes. The filters were hybridized and washed as described previously (16).
ConstructsExpression constructs for TR-I, T
R-II,
activin type IB receptor (ActR-IB), and activin type II receptor
(ActR-II) were previously described (16). Smad1-Flag in pCMV5 was
obtained from Dr. J. Wrana (10). For expression of Smad2, its cDNA
was subcloned in pcDNA3. The Smad2-His/Flag construct was made by
PCR using a sense primer encoding 6 histidine residues and subcloning
the PCR product into pcDNA3-Flag.
COS cells and Mv1Lu mink lung epithelial cells were obtained from the American Type Culture Collection. The R4.2 mutant of Mv1Lu cells was provided by Dr. J. Massagué (Memorial Sloan Kettering Cancer Center, New York). Cells were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) with 10% fetal bovine serum, 100 units/ml penicillin, and 50 µg/ml streptomycin.
Preparation of Polyclonal AntibodyA synthetic peptide corresponding to amino acid residues 227-244 in Smad2 was coupled to keyhole limpet hemocyanin (Calbiochem) using glutaraldehyde, mixed with Freund's adjuvant, and used to immunize rabbits from which the SED antiserum was obtained.
Transient Transfection, Metabolic Labeling, Immunoprecipitation, and SDS-PAGETransfection, metabolic labeling of cells, immunoprecipitation, and SDS-PAGE were performed as described previously (16).
[32P]Orthophosphate Labeling of Cells and Two-dimensional Phosphoamino Acid AnalysisMv1Lu and COS cells were labeled for 2 h in phosphate-free medium supplemented with 0.5% dialyzed fetal bovine serum, 15 mM HEPES, pH 7.2, and 4.0 mCi/ml or 1 mCi/ml [32P]orthophosphate, respectively. Cells were incubated in the absence or presence of ligand, lysed, and subjected to immunoprecipitation with SED antiserum. The precipitate was analyzed by SDS-PAGE. Phosphorylated Smad2 was cut out from the gel, and the protein was extracted and hydrolyzed. Phosphoamino acids were separated by two-dimensional thin layer electrophoresis on cellulose plates (17).
Nuclear TranslocationSubcellular localization of Smad2 in Mv1Lu cells was determined as previously reported (18). The SED antiserum was used at a 50-100-fold dilution.
Somatic Cell Hybrid and Fluorescence in Situ Hybridization (FISH)Two panels of human-rodent somatic cell hybrids, mapping
panel 2, and the regional mapping panel for chromosome 18 (19) were
used to map the Smad2 gene. Southern blots were hybridized with a Smad2 cDNA probe containing the complete coding region. The
mapping of Smad2 was also confirmed and refined by FISH. The yeast artificial chromosome (YAC) Y739A3 was cohybridized with a
chromosome 18 satellite probe (D18Z1; Oncor) as described previously (20).
Three Centre d'Etude du
Polymorphisme Human YACs, Y747A6, Y945B11, and Y739A3 (obtained from
Genome Systems, Inc.), known to contain DPC4 and the
Mad-related gene JV18-1 (15, 21), were analyzed
by PCR for the presence of Smad2 sequences using two primer
pairs (sense primer 1, 5-TCGAAAAGGATTGCCACAT-3
; antisense primer 1, 5
-AGGGTTTACACATACTTCATCC-3
; and sense primer 2, 5
-CGAAGGCAGACGGTAACA-3
; antisense primer 2, 5
-GCTTGAGCAACGCACTGA-3
)
corresponding to exons 2 and 11 of Smad2.
A data base search for mammalian sequences related to the Mad gene revealed the existence of multiple human and mouse ESTs corresponding to different genes. PCR using placenta cDNA with two Mad-related EST primers resulted in the isolation of Smad1 (10, 13). The closely related Smad2 was identified by RT-PCR using human erythroleukemia cell mRNA with two degenerate primers that were based on short stretches of sequence identity between Drosophila Mad (7) and C. elegans Sma genes (11). A cDNA encoding the complete Smad2 protein was obtained from a human brain cDNA library. As reported previously by others (10, 13, 15, 22), the cDNA sequences predict that Smad1 and Smad2 are proteins of 465 and 467 amino acid residues, respectively.
The distribution of Smad1 and Smad2 in various human tissues was
determined by Northern blot analysis of mRNA (Fig.
1). One Smad1 transcript of approximately 3.2 kb was
detected in all tissues that were examined, except in peripheral blood
leukocytes. When reprobing the Northern blots with a Smad2 probe, two
major ubiquitously expressed transcripts of approximately 3.4 and 2.9 kb were detected.
Smad2 Encodes a 58-kDa Protein in Mv1Lu Cells
After
transfection of Smad2-Flag/His in COS cells and metabolic labeling with
[35S]methionine and [35S]cysteine, a cell
extract was prepared and subjected to immunoprecipitation using Flag
antiserum or a Smad2 antiserum, termed SED, which was raised against a
peptide corresponding to amino acid residues 227-244 in the
proline-rich region of Smad2. This region is divergent in sequence
among the Smads. With both Flag and SED antisera a component of 62 kDa
was specifically immunoprecipitated, which was not seen when preimmune
serum was used or when excess cognate SED peptide was added together
with the SED antiserum (Fig. 2A). Moreover,
it was not detectable in samples derived from untransfected COS cells.
When overexpressed in COS cells, Smad1 and DPC4 were not
immunoprecipitated with SED antiserum (data not shown), indicating that
SED antiserum does not cross-react with these proteins. The SED
antiserum precipitated a 58-kDa component from Mv1Lu cells. This
component was not observed when preimmune serum was used or when
blocking peptide was added together with the SED antiserum. The
observed mass of 58 kDa for Smad2 is somewhat higher than the predicted
size of 52.5 kDa predicted from the cDNA sequence.
Ligand-induced Phosphorylation of Smad2
The effect of TGF-
on the phosphorylation of Smad2 was analyzed using
[32P]orthophosphate-labeled Mv1Lu cells. Smad2 was not
appreciably phosphorylated in the absence of ligand in Mv1Lu cells. The
phosphorylation of Smad2 was strongly induced after stimulation with
TGF-
1 (Fig. 3A). Significant
phosphorylation was observed 15-30 min after addition of TGF-
1 and
plateaued after about 2 h of incubation at 37 °C (Fig.
3B). Activin A stimulated no (or in some experiments weak)
increase in Smad2 phosphorylation in Mv1Lu cells (Fig. 3A). BMP-2 and BMP-7 did not stimulate the phosphorylation of Smad2 in Mv1Lu
cells (data not shown). Phosphoamino acid analysis of phosphorylated
Smad2 after TGF-
stimulation revealed that phosphorylation was
mainly on serine, little on threonine, and not on tyrosine residue(s)
(Fig. 3C).
Phosphorylation of Smad2 was also investigated using
[32P]orthophosphate-labeled COS cells, transfected with
Smad2 alone or cotransfected with TR-II and T
R-I, in the absence
or presence of TGF-
. Analysis of the Smad2 immunoprecipitates showed
that Smad2 in the absence of overexpressed receptor and ligand was not
(or weakly) phosphorylated. The observed increase in Smad2 phosphorylation after cotransfection with receptors is most likely due
to ligand-independent complex formation of a T
R-I and T
R-II complex induced by their high overexpression. Further receptor activation by TGF-
addition led to a strong increase in the Smad2 phosphorylation (Fig. 3D). The phosphoamino acid analysis
pattern and two-dimensional tryptic phosphopeptide maps of
phosphorylated Smad2 from transfected COS cells and Mv1Lu cells were
nearly identical.2 In an analogous manner
the effect of activation on Smad2 phosphorylation induced by activin
was investigated using COS cells transfected with Smad2, ActR-II, and
ActR-IB. We found that activin-mediated signaling induced Smad2
phosphorylation. In addition, we found that TGF-
receptor activation
induced the phosphorylation of Smad1 in COS cells, albeit weaker than
BMP receptor activation (Fig. 3D).2 Taken
together with our unpublished results, which indicate that TGF-
does
not induce Smad1 phosphorylation in Mv1Lu
cells,3 these data suggest that
overexpression of Smads and receptors in COS cells may lead to
interactions that are not physiologically relevant.
The subcellular
organization of Smad2 in Mv1Lu cells in the absence or presence of
ligand was analyzed by immunofluorescence using the SED antiserum. In
the absence of ligand, Smad2-specific staining was predominant in the
cytoplasm of the cells, whereas after stimulation with TGF-1 for
1 h the staining was concentrated in the nucleus (Fig.
4). Thus, TGF-
1 induced a translocation of Smad2 from
the cytoplasm to the nucleus.
Smad2 and DPC4 are Closely Linked on Chromosome 18q
The
chromosomal localization of the Smad2 gene was initially
determined by analysis of its segregation in mapping panel 2 (19). The
Smad2-specific fragments segregated only with human chromosome 18. Smad2 was subsequently mapped to an interval
between 18cen and 18q21.3 by analysis of a regional mapping panel for chromosome 18 (Fig. 5A). This localization
places Smad2 in the same region of 18q as the related
DPC4 gene (14). To test whether the two genes are closely
linked, we analyzed two DPC4-containing YACs, Y747A6 and
Y945B11, for the presence of Smad2 sequences by PCR. Both
YACs were negative for Smad2, indicating that
Smad2 is not located in the immediate vicinity of
DPC4. In contrast, YAC Y793A3, which was recently shown to
contain the Mad-related gene JV18-1 (15), was
positive for Smad2 (data not shown). This YAC was mapped by
FISH to 18q21.1 (Fig. 5B). All three YACs are known to map
within the Whitehead-MIT YAC contig WC18.4. Based on the sequence tag
site map of chromosome 18, Smad2 is located approximately 3 Mb proximal to DPC4 (23).
Recently, a conserved cytoplasmic family of Mad-like proteins,
termed Smads, was identified and shown to play a pivotal role in the
downstream signal transduction pathway of various TGF- superfamily
members (24). In the present study we report the identification of a
human Mad-related protein, Smad2, which acts in the TGF-
signaling
pathway.
Smad1 and Smad2 have 66% overall sequence identity, with the highest
level of similarity in the MH1 and MH2 domains. The proline-rich intervening region is divergent, and a SED peptide derived from this
region was successfully used to raise a specific antisera useful in the
detection of Smad2 in nontransfected cells. Smad1 and Smad2 were found
to be broadly expressed (Fig. 1), suggesting that they may be
downstream of TGF- superfamily members with effects on many cell
types.
TGF- specifically stimulated a phosphorylation of Smad2 in
nontransfected Mv1Lu cells (Fig. 3). Consistent with the notion that
type I receptors appear to act downstream of type II receptors (5), we
found that Mv1Lu cells, lacking functional T
R-I, do not
phosphorylate Smad2 on TGF-
challenge (data not shown). The kinetics of our TGF-
-induced phosphorylation of Smad2 is slower than
previously reported for BMP-2-induced phosphorylation of Smad1 (10). It
needs to be investigated whether this is due to the use of transfected
versus nontransfected cells. TGF-
and activin induced
Smad2 phosphorylation when overexpressed with appropriate type I and II
receptors in COS cells. In addition, we observed that TGF-
induced
Smad1 phosphorylation, which is likely to have occurred due to the
supraphysiological levels of T
Rs and Smad1 in COS cells. Supportive
of this notion, we found that TGF-
did not induce the
phosphorylation of Smad1 in Mv1Lu cells, whereas Smad1 could be
immunoprecipitated from metabolically labeled cells using an antisera
that recognizes Smad1, but not Smad2.3 Together these data
indicate the physiological relevance of using nontransfected cells
compared with transfected cells for the biochemical characterization of
Smad activation.
Mv1Lu cells were shown to have TGF-, activin, and BMP receptors by
ligand affinity cross-linking, followed by immunoprecipitation with
receptor-specific antiserum, and to be responsive to all three ligand
subgroups (16, 26). However, activin did not stimulate (or only weakly
stimulated) Smad2 phosphorylation, and BMPs did not induce
phosphorylation of Smad2 in Mv1Lu cells (Fig. 3B). In
agreement with our results, Eppert et al. (22) recently demonstrated Smad2 regulation by TGF-
and not BMP, whereas Smad1 has
been implicated as a specific downstream target of BMP- but not
TGF-
-mediated signaling (10). Thus the specific signaling pathways
autonomously induced by Smad1 (ventral mesoderm) and Smad2
(dorsal mesoderm) in Xenopus (12) resemble the
ligand-specific activation of Smad1 and Smad2, as detected by
biochemical means.
However, Lechleider et al. (27) and Yingling et
al. (28) reported the phosphorylation of endogenous Smad1 (or
immunologically related Smads) in response to TGF-. The latter
result may suggest a functional redundancy of Smads in TGF-
and BMP
signal transduction. Differences may also prevail between different
cell types. In addition, there may be a need for multiple Smads in
TGF-
signaling. It will be interesting, therefore, to test whether
TGF-
can induce an association between Smad2 and other Smads,
e.g. Smad1 and DPC4. Involvement of multiple Mads in a
TGF-
-like signaling pathway was proposed for the C. elegans MAD homologs, as mutant phenotypes for sma-2,
sma-3, and sma-4 all closely resemble the mutant
phenotype of daf-4 (11).
TGF- stimulated the nuclear accumulation of Smad2 in Mv1Lu cells.
Recently, Baker and Harland (29) reported the induction of
nuclear localization of Smad2 by activin in Xenopus
explants. In addition, they observed that the N-terminally truncated
Smad2 is localized in the nucleus in the absence of ligand. It is
unclear whether phosphorylation provides the trigger for nuclear
accumulation, e.g. by unmasking a nuclear localization
signal for active nuclear entry or by inducing the dissociation from a
putative cytoplasmic retention protein.
Smad2 was localized to chromosome 18q21.1 by a combination of mapping techniques. Using YACs we could demonstrate that Smad2 and DPC4 are closely linked within a 3-Mb region, with Smad2 being proximal to DPC4. The 18q21 chromosomal region is frequently deleted or rearranged in a variety of human cancers. Recently, DPC4 was found to be frequently homozygously deleted or mutated in pancreatic cancers (14) but only rarely in other types of cancers with chromosome 18q alterations (25). The interesting possibility that Smad2 has a tumor suppressor function in these other cancers or in pancreatic cancer needs to be examined. In agreement with our results, Riggins et al. (15) and Eppert et al. (22) reported the chromosomal assignment of Smad2 to 18q21 by PCR analysis of YAC clones. In addition, both groups found that certain colorectal cancers contain somatic mutations in Smad2.
The elucidation of intracellular signal transduction pathways by which
TGF- executes its multifunctional effects is of great importance for
a better understanding of the (patho)physiological processes in which
TGF-
has been implicated and for future therapeutic interventions.
In this respect the identification of Smad2 as an intracellular
phosphorylation target in the TGF-
signaling pathway is an important
finding. The identification of the upstream Smad2 activating serine and
threonine kinase and the downstream Smad2 nuclear effectors that result
in the transcriptional modulation of Smad2 target genes will be the
subject of future studies.
We thank Kohei Miyazono, Johan de Winter, and
our colleagues for stimulating discussions, Susanne Grimsby for
excellent technical assistance, and Irie Nakao for encouragement. We
also thank Ryutaro Yamada and Yoshio Misumi for the human brain
cDNA library, M. Kawabata for pcDNA3-Flag, J. Wrana for
Smad1-Flag in pCMV5, H. Ohashi for TGF-1, Y. Eto for activin A,
T. K. Sampath for BMP-2 and BMP-7, and J. Massagué for R mutant
cells.