The transforming growth factor-β (TGF-β) family consists of more than 30 secreted structurally related polypeptides including TGF-βs, activins, and bone morphogenetic proteins (BMPs). This family of secreted dimeric proteins regulates pivotal biological functions, including cell proliferation, differentiation, apoptosis, migration, and extracellular matrix production (Roberts and Sporn, 1990). Signaling by these cytokines occurs via ligand-induced heteromeric complex formation of distinct type I and type II serine/threonine kinase receptors, after which the type I receptor is phosphorylated by the constitutively active type II receptor (Massagué, 1998). The activated type I receptor kinase propagates the signal within the cell through the phosphorylation of specific receptor-regulated (R-) Smad proteins at their extreme carboxyl-terminal serine residues. Whereas Smad2 and Smad3 act downstream of TGF-β and activin type I receptors, Smad1, Smad5, and Smad8 are phosphorylated by BMP type I receptors. Activated R-Smads form heteromeric complexes with common-partner (Co-) Smads, which accumulate in the nucleus, where they control gene expression in a cell-type specific manner through interaction with other transcription factors, co-activators and co-repressors (Heldin et al., 1997; Derynck et al., 1998; Attisano and Wrana, 2000; Massagué, 2000; Massagué and Wotton, 2000). R- and Co-Smads share two conserved domains at their amino-terminal and carboxyl-terminal ends termed Mad Homology (MH)1 and MH2 domains, respectively. Inhibitory (I-) Smads, i.e., Smad6 and Smad7, form a distinct subclass among Smads by acting opposite from the signal transducing R- and Co-Smads. Whereas I-Smads have a conserved MH2 domain, their N-terminal regions only show weak similarity to MH1 domain of R- and Co-Smads (Heldin et al., 1997; Attisano and Wrana, 2000; Massagué and Wotton, 2000).
Tumor cells often escape from TGF-β-induced growth arrest and apoptosis. Smad2 and Smad4 are frequently mutated in particular tumor subsets, suggesting that they may act as tumor suppressors. In addition, several oncogenic proteins can inhibit the function of Smad proteins through their interaction with Smad proteins (Massagué et al., 2000; de Caestecker et al., 2000a; Derynck et al., 2001). Analysis of Smad knock-out mice have also revealed that their functional inactivation may play a role in tumorigenesis (Goumans and Mummery, 2000). However, TGF-β has a biphasic role in tumorigenesis; whereas in the initial phases it acts as a tumor suppressor, during late phases it has been shown to act as a tumor promoter (Derynck et al., 2001). Not only can TGF-β, often produced in high amounts by tumors, stimulate tumorigenesis indirectly by stimulating angiogenesis and via its potent immune suppressive action but also by directly affecting tumor cell behavior. These tumor cells may have selectively escaped TGF-β-induced growth inhibitory/apoptotic response but retained or gained certain TGF-β-responsive properties. Interestingly, these direct tumor effects by TGF-β appear to be mediated via Smad-independent pathways, or by perturbed Smad-dependent pathways. In this review, we will focus on recent new insights into the mechanism of action of Smads, the regulatory function of Smads in cell proliferation, and how inhibition or re-direction of TGF-β/Smad signaling may contribute to tumorigenesis.
REGULATION OF SMAD ACTIVATION
TGF-β receptor-induced Smad activation
The first step in the intracellular TGF-β/Smad pathway is the recruitment of Smads to the TGF-β receptor complex (Fig. 1). Several proteins with anchoring, scaffolding, and/or chaperone activity have now been identified which regulate and facilitate this process. Smad anchor for receptor activation (SARA) has been shown to regulate the subcellular distribution of Smad2 and Smad3 (Tsukazaki et al., 1998). SARA is associated to the inner leaflet of the plasma membrane via its FYVE domain, which interacts with phospholipid (Tsukazaki et al., 1998). In addition, SARA can simultaneously interact with (non-activated) R-Smads and the TGF-β receptor complex through its Smad-binding domain (SBD) and its C-terminal region, respectively. After type I receptor activation, two serine residues in the C-terminal sequence of Smad2 and Smad3 are phosphorylated by the type I receptor kinase, followed by the dissociation from SARA and type I receptor. The activated Smad2 forms subsequently a heteromeric complex with Smad4 (Fig. 1); the binding of SARA and Smad4 with activated R-Smad are mutually exclusive. Hrs/Hgs, another FYVE domain containing protein, was also shown to participate in Smad presentation to receptor and to synergize with SARA in stimulating TGF-β/Smad signaling (Miura et al., 2000). Hrs/Hgs binds phosphatidylinositol 3-phosphate (PtdIns(3)P) and is localized on early endosomes (Burd and Emr, 1998; Gaullier et al., 1998; Patki et al., 1998). These properties might be shared with SARA, which like Hrs/Hgs, has a punctate cytoplasmic staining pattern and binds PtdIns(3)P (Tsukazaki et al., 1998; Itoh et al., 2002). Thus, TGF-β receptor internalization is required for Smad presentation by SARA to TGF-β receptor. TGF-β receptors can interact with AP2, which is a component of clathrin-coated pits. Interestingly, dominant negative dynamin, a specific inhibitor for the formation of clathrin-coated pits, diminishes the phosphorylation of Smad2 upon TGF-β stimulation (E. Leof, personal communication). In contrast, Zaagstra et al. (2001) reported that the internalization of TGF-β receptor complexes is not mediated by clathrin-mediated endocytosis. Caveolin-1 (Cav-1), a principal component of caveolae membranes, was found to be cofractionated with TGF-β receptors and Smad2 but not with Smad4. Overexpression of Cav-1 was reported to reduce TGF-β-induced Smad activation (Razani et al., 2001). Furthermore, the family of sorting nexins (SNXs) that are intracellular trafficking molecules of receptor tyrosine kinases (Haft et al., 1998), also interact with TGF-β receptors (Parks et al., 2001).
The adaptor molecule disabled-2 (Dab-2) was also found to serve a bridging function between TGF-β receptor complex and Smads. Cells expressing a mutant Dab-2 are TGF-β/Smad signaling defective; the cellular response to TGF-β can be restored upon stable expression of Dab-2. Similar as found for SARA, Dab-2 can directly interact with Smad2 and Smad3 and the TGF-β receptor complex. However, in contrast with SARA, the interaction of Dab-2 with the TGF-β receptor complex is constitutive, and the interaction between Dab-2 and Smads is ligand dependent (Hocevar et al., 2001).
Interestingly, TGF-β receptor-associated protein-1 (TRAP-1) was found to specifically interact with Smad4 and to aid in the recruitment of Smad4 to the TGF-β and activin receptor complex (Wurthner et al., 2001). TRAP-1 associates with the inactive TGF-β receptor complex, and upon receptor activation, TRAP-1 dissociates from the receptor complex and associates with Smad4. The TRAP-1/Smad4 interaction is transient and disrupted by activated R-Smads. Wurthner et al. (2001) suggested that TRAP-1 may function as a chaperone to reduce the auto-inhibitory interactions between N- and C-terminal domains of Smad4 and thereby facilitate the interaction of Smad4 with activated R-Smads.
Microtubules were identified as another subcellular regulators for Smads. Smad2, Smad3, and Smad4 were found to colocalize and interact with tubulin. After phosphorylation of Smad2 by TGF-β, Smad2 dissociates from tubulin and translocates into the nucleus. Treatment of cells with nocodazole, which destabilizes the microtubule network, disrupts the interaction between Smads and microtubules and increases TGF-β-induced Smad2 phosphorylation and Smad-mediated transcription (Dong et al., 2000). Filamin 1, a protein that connects the actin filament networks to membrane receptors and acts as a scaffold protein for signal transduction molecules, was also found to interact with Smads. TGF-β does not induce Smad2 phosphorylation in filamin 1-deficient cells and re-introduction of filamin 1 in these mutant cells restored TGF-β responsiveness (Sasaki et al., 2001). In addition, Axin, a negative regulator in Wnt signaling, was recently identified as an adaptor of Smad3 that may facilitate TGF-β receptor-induced Smad3 activation (Furuhashi et al., 2001).
In the non-activated state, R-Smads have been reported to predominantly exist as monomers or in several distinct oligomeric states. Smad4 was found to form homo-trimers in a concentration-dependent manner (Shi, 2001). Upon receptor-mediated phosphorylation, R-Smads can form homo-oligomers and hetero-oligomers with each other, as well as hetero-oligomers with Smad4 (Kawabata et al., 1998; Qin et al., 1999; Jayaraman and Massagué, 2000; Chacko et al., 2001; Wu et al., 2001). However, there is still controversy about the stoichiometry between R- and Co-Smads in the heteromeric Smad complex. The complex between R-Smads and Co-Smads has been proposed to be a heterohexamer (Shi et al., 1997), a heterotrimer (Kawabata et al., 1998; Qin et al., 1999; Chacko et al., 2001) and more recently, a heterodimer (Wu et al., 2001). Possibly R- and Co-Smads may oligomerize in multiple complexes with different stoichiometry (Jayaraman and Massagué, 2000).
Nuclear import and export of Smad proteins
Without stimulation of ligand, R- and Co-Smads predominantly exist in the cytoplasm, and ligand stimulation induces their nuclear accumulation (Fig. 1) (Heldin et al., 1997; Attisano and Wrana, 2000; Massagué and Wotton, 2000). An exception is Xenopus Smad4β (for which a mammalian homolog has not been identified) that predominantly exists in the nucleus and of which the subcellular distribution does not change after ligand stimulation (Masuyama et al., 1999). The N-terminal regions of R- and Co-Smads contain a nuclear localization-like sequence (NLS-like; Lys-Lys-Leu-Lys) (Xiao et al., 2000a; Kurisaki et al., 2001). This motif has been reported to be necessary for Smad3 nuclear accumulation after TGF-β stimulation. Smad3 directly binds to importin-β via its N-terminal region but not to importin-α (Xiao et al., 2000b; Kurisaki et al., 2001), and Ran GTPase promotes nuclear translocation of activated Smad3 to the nucleus (Kurisaki et al., 2001). Smad2, when compared with Smad3, has a unique insert of exon3 in the N-terminal domain, which prevents its association with importin-β (Kurisaki et al., 2001). Nuclear import of non-phosphorylated Smad2 is blocked by SARA which masks the intrinsic nuclear import activity of Smad2, which was reported to have a NLS in its C-terminal region (Xu et al., 2000).
In contrast with Smad2 and Smad3, Smad4 has a leucine-rich nuclear export signal (NES) in its linker region (Pierreux et al., 2000; Watanabe et al., 2000). Xenopus Smad4β lacks the NES providing an explanation for its constitutive nuclear localization (Pierreux et al., 2000; Watanabe et al., 2000). The nuclear export of Smad4 is mediated by CRM1; inhibition of CRM1-dependent nuclear export by treatment with leptomycin B results in rapid accumulation of Smad4 (but not Smad2 and Smad3) in the nucleus (Pierreux et al., 2000). Smad4, which in its basal state is predominantly located in the cytoplasm, is proposed therefore to rapidly and constitutively shuttle between the nucleus and the cytoplasm. Upon heteromeric complex formation between R-Smad and Co-Smad, the nuclear accumulation of the complex may be stimulated by the shielding of the NES and/or unmasking the NLS on R-and/or Co-Smad (Pierreux et al., 2000; Watanabe et al., 2000). Recently, Smad1, but not Smad2 and Smad3, was found to contain a functional NES in the MH2 domain (Xiao et al., 2001) Thus, Smad1 may also constantly shuttle between nucleus and cytoplasm, and receptor-induced Smad1 phosphorylation shifts the balance towards nuclear accumulation of Smad1. Within the nucleus, the Smad complex can bind directly to the specific DNA sequences or associate indirectly with DNA through transcriptional factors (see below), upon which specific gene transcriptional responses mediated by TGF-β family member are activated (Fig. 1) (Heldin et al., 1997; Derynck et al., 1998; Attisano and Wrana, 2000; Massagué and Wotton, 2000).
TRANSCRIPTIONAL CONTROL BY SMADs
Direct binding of Smads to DNA
R-Smads (except for Smad2) and Co-Smads can recognize specific DNA sequences (5′-AGAC-3′), termed Smad binding elements (SBEs), via their MH1 domains in the promoters of target genes (Fig. 1). Whereas full length Smad4 can interact with SBE, a prerequisite for Smad3-SBE association is the relief of the auto-inhibitory interaction between N- and C-terminal domains through C-terminal phosphorylation of Smad3 or artificial truncation of the MH2 domain (Dennler et al., 1998; Johnson et al., 1998; Jonk et al., 1998; Kawabata et al., 1998; Shi et al., 1998; Zawel et al., 1998). Multimers of SBE when placed in front of a minimal promoter reporter construct provide a strong enhancer function for TGF-β family members. SBE-like sequences have been shown to be critically important for TGF-β-inducibility of multiple TGF-β responsive genes (Dennler et al., 1998; Vindevoghel et al., 1998; Hanai et al., 1999; Nagarajan et al., 1999; Brodin et al., 2000; Chen et al., 2000; Pardali et al., 2000; Taylor and Khachigian, 2000). The crystal structure of Smad3 MH1 domain with SBE showed that the β-hairpin loop in Smad3 is in contact region with the SBE (Shi et al., 1998). The DNA-binding β-hairpin is highly conserved among R-Smads and Smad4, which suggests that the MH1 domain of other Smads might also use the same motif to recognize SBE. Interestingly, Smad2 has 30-amino acid insertion immediately prior to the DNA-binding β-hairpin, which inhibits the capacity of Smad2 to bind to DNA (Shi et al., 1998; Dennler et al., 1999; Yagi et al., 1999). BMP R-Smads (and also Smad3 and Smad4) have also been shown to bind to GC-rich sequence motifs (Kim et al., 1997; Labbé et al., 1998; Ishida et al., 2000; Kusanagi et al., 2000), suggesting that DNA binding specificity of Smads is not so strict (Qing et al., 2000). GC-rich sequences have been found to be critically important for BMP-induced activation of Smad6 (Ishida et al., 2000). However, whether in vivo the interaction of BMP R-Smads with these GC-rich sequences is direct, and perhaps mediated via a region distinct from β-hairpin loop, or through other transcription factors, remains to be elucidated (Shi, 2001). The TGF-β inhibitory element (TIE) in the c-Myc promoter, which is composed of 5′-GGCTTGGCGG-3′, can bind Smad3 and Smad4 upon TGF-?β stimulation. It is not clear whether Smads bind directly to TIE. The inhibitory effect by TGF-β appears not due to the recruitment of Smad co-repressors (see below) (Chen et al., 2001).
Smad proteins can also act as repressors of transcription, as is found for example for the TGF-β/Smad3-induced repression of RUNX2/Cbfa1 transcription factor activity in mesenchymal cells. Smad3 physically interacts with RUNX2/Cbfa1 (Alliston et al., 2001). Interestingly, the repression is cell type dependent. In contrast with mesenchymal cells, Smad3 and RUNX2/Cbfa1 cooperate in TGF-β-induced transcription in epithelial and leukemic cells (Hanai et al., 1999; Pardali et al., 2000; Zhang and Derynck, 2000).
Transcriptional co-activators and co-repressors of Smads
The transactivation properties of Smads became apparent from studies in which the C-terminal domains of R- and Co-Smads were fused to heterologous GAL4-DNA binding domains (Liu et al., 1996). Subsequently, the structurally related transcriptional co-activators, p300 and CBP, were found to be recruited in a ligand-dependent manner to the R-Smads via their MH2 domains (Fig. 1) (Feng et al., 1998; Janknecht et al., 1998; Nishihara et al., 1998; Pouponnot et al., 1998; Shen et al., 1998; Topper et al., 1998; Nakashima et al., 1999; Pearson et al., 1999). p300 and CBP have intrinsic histone acetyltransferase activity (HAT), which facilitates transcription by decreasing chromosome condensation through histone acetylation and by increasing the accessibility of Smads with the basal transcription machinery. In addition, P/CAF, another HAT-containing transcriptional co-activator, has been shown to associate with Smad3 upon TGF-β receptor activation and to augment TGF-β/Smad3 signaling (Itoh et al., 2000a). The Smad4 interacting protein MSG1, which lacks an intrinsic DNA binding ability, can recruit p300/CBP to Smad4 via its so-called Smad activation domain and function as a co-activator of Smad4 (Shioda et al., 1998; Yahata et al., 2000; de Caestecker et al., 2000b). Smad nuclear interacting protein 1 (SNIP1) which has a forkhead-associated domain interacts with Smad4 in a TGF-β-dependent manner and suppresses the TGF-β/Smad pathway by inhibiting the ability of p300/CBP to interact with Smad4 (Kim et al., 2000). In addition, E1A, an adenovirus protein, that antagonizes TGF-β signaling, competes with p300/CBP for the binding to Smad3, thereby inhibiting TGF-β/Smad signaling (Nishihara et al., 1999). p300/CBP, which binds a whole range of different transcription factors, can also act as a bridge between Smads and other transcription factors. For example, Stat3 activated by LIF was found to indirectly associate with BMP-activated Smad1 via p300, and both factors synergistically cooperate in the differentiation of the neural progenitor cells to astrocytes (Nakashima et al., 1999).
The co-repressor TGIF, a homeodomain-containing protein which recruits histone deacetylases (HDACs), was found to interact with Smad2 and Smad3 upon ligand stimulation. Recently, TGIF was shown to recruit mSin3 to the TGF-β-activated Smad complex (Wotton et al., 1999b; 2001). The TGIF interacting co-repressor, C-terminal binding protein (CtBP), can enhance the inactivating effect of TGIF on TGF-β-induced transcriptional responses (Melhuish and Wotton, 2000). In addition, binding of TGIF to Smads occurs via the C-terminal domain and is mutually exclusive with p300/CBP interaction, and will lead to inactivation of gene expression (Fig. 1). Thus, a balance of expression levels of TGIF (co-repressors) and p300/CBP (co-activators) within the cell determines the intensity of TGF-β/Smad signaling responses (Wotton et al., 1999a). Recently, Smad3 was also found to associate with HDAC activity via its MH1 domain (Liberati et al., 2001). The proto-oncogene Ski, and its structurally related protein SnoN, have also been found to exhibit transcriptional co-repressor activity for Smads. Ski can indirectly recruit HDACs by binding to N-CoR and Sin3A (Akiyoshi et al., 1999; Luo et al., 1999; Sun et al., 1999a). Like TGIF, Ski also competes with p300 for interaction with the activated Smad complex (Akiyoshi et al., 1999). The stability of SnoN and Ski is regulated by TGF-β. SnoN, and to a lesser extend Ski, are rapidly degraded after TGF-β treatment (Stroschein et al., 1999; Sun et al., 1999b). The ubiquitin-mediated degradation of SnoN is mediated by Smad ubiquitination regulatory factor (Smurf)2, which is recruited into a common complex through its interaction with activated Smad2 (Bonni et al., 2001). SnoN, which interacts with Smads in the non-activated state, has been proposed to silence TGF-β responsive genes in the absence of ligand (Luo et al., 1999). Upon TGF-β-induced SnoN degradation, signaling can be initiated. TGF-β also induces SnoN and it may thus therefore also participate in negative feedback control of TGF-β signaling (Stroschein et al., 1999; Sun et al., 1999b).
Transcription factors as Smad partners
TGF-β family members activate different sets of genes in each cell type (Derynck et al., 1998; Attisano and Wrana, 2000; Massagué and Wotton, 2000; Derynck et al., 2001). Since the intrinsic DNA binding ability of Smads is relatively weak (the dissociation constant between Smad3 MH1 and SBE is 1.14 × 10−7 M) (Shi et al., 1998), Smads must cooperate with other transcription factors to activate or repress target genes (Fig. 1). The first identified Smad transcription partner was Xenopus forkhead activin signal transducer (FAST)-1, which forms a complex with Smad2 in an activin–dependent manner to bind to the activin response element (ARE) in the Mix.2 promoter (Chen et al., 1996). Smad2 also interacts with the paired-like homeodomain proteins mixer and milk (Germain et al., 2000). Interestingly, a common Smad interaction motif, containing the PPNK amino acid sequence, was identified in FAST and mixer/milk (Germain et al., 2000). A large number of transcriptional factors have been found to interact and co-operate with Smads to positively or negatively regulate the transcriptional responses (Fig. 1) (Derynck et al., 1998; Attisano and Wrana, 2000; Massagué and Wotton, 2000; Itoh et al., 2000b; Derynck et al., 2001). For example, TGF-β is known to activate the IgA promoter and induce IgA class switching in B cells (Lin and Stavnezer, 1992). The IgA promoter contains SBE and AML1 binding sites located closely to each other. TGF-β stimulation triggers the interaction between Smads and AML1 and their binding to the IgA promoter, upon which the IgA promoter is activated (Hanai et al., 1999; Pardali et al., 2000; Zhang and Derynck, 2000). Smad4 was found to interact and functionally cooperate with β-catenin and LEF1/TCF, which are downstream components of the Wnt signaling pathway, in inducing Xenopustwin (Xtwn) gene expression (Nishita et al., 2000). Smad3 has also been reported to functionally interact with LEF1/TCF in a TGF-β-dependent manner independent of β-catenin (Labbé et al., 2000).
Smad proteins can also activate gene expression by silencing the function of transcriptional repressor proteins. For example, the Hoxc-8 transcriptional repressor has been identified as a Smad partner (Shi et al., 1999). Hoxc-8 binds constitutively to its binding site in osteopontin promoter and keeps osteopontin gene expression in an inactive status. BMP can stimulate the association of Smad1 with Hoxc-8 and thereby dislodge Hoxc-8 from the promoter sequence, upon which the osteopontin gene is activated (Shi et al., 1999). Thus, Smad proteins can not only induce transcription by cooperating with transcriptional activators but also by interfering with the function of transcriptional repressors (Fig. 1).
NEGATIVE REGULATION OF R- AND CO-SMADs
Mechanism of action of I-Smads
The duration and intensity of TGF-β/Smad responses need to be tightly regulated. One way by which this occurs is through the action of inhibitory (I-) Smads (Fig. 1) (Hayashi et al., 1997; Imamura et al., 1997; Nakao et al., 1997; Hata et al., 1998a; Ishisaki et al., 1998, 1999; Souchelnytskyi et al., 1998; Lebrun et al., 1999). Whereas Smad7 acts as a general inhibitor of TGF-β family member signaling pathway, Smad6 preferentially blocks BMP signaling (Itoh et al., 1998; Ishisaki et al., 1999). The first described mechanism for I-Smads antagonism was through their competition with R-Smads for the interaction with the activated type I receptor (Hayashi et al., 1997; Imamura et al., 1997; Nakao et al., 1997; Souchelnytskyi et al., 1998). More recently, another mechanism by which I-Smads inhibit TGF-β/Smad signaling has been reported. Smad7 (via its PY-motif) has been found to constitutively interact with (WW-domains in) HECT-domain ubiquitin ligase, Smurf 2, and more recently Smurf 1 (Kavsak et al., 2000; Ebisawa et al., 2001). Upon recruitment of the Smad7/Smurf complex to the activated TGF-β receptor, Smurf1 or Smurf2 induces TGF-β receptor degradation via proteasomal and lysosomal pathways. Thus, similar as found for Smad2 which mediates the degradation of SnoN via the recruitment of Smurf2 (Bonni et al., 2001), Smad7 may also function as an adaptor protein to mediate the selective degradation of a Smad interacting protein (Kavsak et al., 2000; Ebisawa et al., 2001). Other inhibitory mechanisms of I-Smads have been reported, including competition of Smad6 with Smad4 for complex formation with phosphorylated Smad1 (Hata et al., 1998a). AMSH, a protein previously implicated in interleukin signaling, can compete with activated type I receptors or R-Smads for interaction with I-Smads. Ectopic expression of AMSH promotes TGF-β family signaling by inhibiting I-Smad function (Itoh et al., 2001).
I-Smads are potently induced by TGF-β family members, and may thus participate in a negative feedback loop to control the intensity and duration of TGF-β signaling (Nakao et al., 1997; Afrakhte et al., 1998; Ishisaki et al., 1998; 1999, Takase et al., 1998). Regulation of I-Smad expression is emerging as a frequently utilized mechanism for cross-talk of other signaling pathways with the TGF-β/Smad pathway. Interferon-γ via the Jak1/STAT1 pathway (Ulloa et al., 1999) and tumor necrosis factor-α and interleukin-1 through NF-κB/RelA (Bitzer et al., 2000) have been able to negatively regulate TGF-β/Smad signaling via the upregulation of I-Smads. In addition, the activation of CD40 can also induce I-Smad mediated through NFκB activation (Patil et al., 2000). Moreover, ultraviolet irradiation induces the expression of Smad7 to block the cellular responses to TGF-β (Quan et al., 2001).
Degradation of R-Smad proteins
In addition to the I-Smads, Smurfs have been found to interact with R-Smads (Zhu et al., 1999; Lin et al., 2000; Zhang et al., 2001), thereby directly targeting R-Smads for ubiquitin-mediated degradation via the proteasome pathway. Whereas Smurf 1 preferentially interacts with BMP R-Smads, Smurf 2 can associate with TGF-β/activin R-Smads as well as BMP R-Smads (Zhu et al., 1999; Lin et al., 2000; Zhang et al., 2001). WW motifs in Smurf recognize specific PY motifs in the linker region of R-Smads except for Smad8, which does not possess a PY motif in its linker region (Zhu et al., 1999; Ebisawa et al., 2001). Smurf-mediated degradation of R-Smads induces a decrease in the cellular competence to TGF-β family-induced responses (Zhu et al., 1999; Zhang et al., 2001).
Activated nuclear Smad2 has also been found to be targeted for proteasome degradation by selective multi-ubiquitination. Proteolysis of the activated Smad2 involves E2 enzymes UbcH5b/c and Ubc3, and is not dependent on C-terminal phosphorylation of Smad2 but requires nuclear localization of Smad2 (Lo and Massagué, 1999). Smad3 was reported to be degraded by E3 ligase complex ROC1-SCFFbw1a (Fukuchi et al., 2001). The MH2 domain of Smad3 is necessary to bind to ROC1-SCFFbw1a. SCFFbw1a efficiently induces the ubiquitination of Smad3. After interaction of Smad3 with the E3 ligase complex, which is promoted by p300 transcriptional coactivator, this nuclear complex is subsequently exported to the cytoplasm for degradation (Fukuchi et al., 2001).
Inhibition of Smad pathway by proto-oncogene activity
Growth factors such as EGF and HGF can inhibit TGF-β-induced growth inhibition (Massagué and Chen, 2000). Activation of Erk by EGF, HGF or activated Ras can induce the phosphorylation of S/TP or PXS/TP motifs in the linker region of R-Smads, and thereby inhibit the ligand-induced nuclear accumulation of R-Smads (Kretzschmar et al., 1997, 1999; Calonge and Massagué, 1999). In addition, activated Ras has been shown to promote the degradation of Smad4 through the proteasome (Saha et al., 2001) and to induce an increase in TGIF expression level (Lo et al., 2001). Ectopic expression of c-Myc makes cells insensitive to the antiproliferative action of TGF-β (Alexandrow et al., 1995). Recently, c-Myc was found to inhibit the TGF-β/Smad-mediated activation of p15INK4b gene, a cyclin-dependent kinase inhibitor (see below). Whether Smad proteins can also interact directly with c-Myc is not known.
In addition to transcriptional repressors c-Ski/SnoN, the winged-helix containing brain factor (BF)-1 proto-oncogene, has been reported to interfere with TGF-β/Smad pathway through its constitutive interaction with the MH2 domain of R- and Co-Smads (Rodriguez et al., 2001), or by competition with Smad2 to interact with the FAST-2 transcriptional factor (Dou et al., 2000). Evi-1, a transcriptional regulator containing two zinc finger domains, and its fusion product with AML1 (AML1/Evi-1) in leukemia interact with Smad3 and abrogate its transcriptional activity by blocking its DNA binding (Kurokawa et al., 1998a,b), or by recruiting the repressor, CtBP (Izutsu et al., 2001). Tax encoded in human T-cell leukemia virus-I represses Smad-dependent TGF-β signaling pathway through its competition with Smads to associate with p300/CBP (Mori et al., 2001).
In addition to the cooperative effects between Smad3 and AP-1 family members (Zhang et al., 1998; Liberati et al., 1999: Qing et al., 2000), c-Jun as well as Jun B has been reported to suppress the transcriptional activity of Smad3 (Verrecchia et al., 2000, 2001a,b; Dennler et al., 2000). The inhibitory action of c-Jun to Smad3 may be caused by inducing a dissociation of Smad3 from DNA (Verrecchia et al., 2000, 2001a,b). In addition, c-Jun can apparently enhance the association of Smad2 with TGIF through its direct interaction to TGIF, thereby interfering with the assembly of Smad2 and p300 upon TGF-β stimulation (Pessah et al., 2001).
SMAD FUNCTION IN TGF-β-INDUCED RESPONSES
Cell cycle regulation by Smads
TGF-β inhibits the proliferation of many cell types by arresting cells in the G1 phase of the cell cycle (Fig. 2) (Massagué et al., 2000). TGF-β mediates its anti-proliferative effect through multiple mechanisms and in a cell type dependent manner. For example, in keratinocytes, TGF-β potently upregulates the expression of cyclin dependent kinase inhibitors p15INK4b and p21CIP/WAF1, which inhibit the CDK4/6-cyclin D and CDK2-cyclin E-mediated phosphorylation of the retinoblastoma protein (pRB). In another cell type, however, TGF-β strongly increases transcription of P15INK4b, which then sequesters CDK4. Subsequently, p27 bound to cyclin D1-CDK4 is displaced and binds to and inhibits cyclinE-CDK2 (Fig. 2) (Massagué et al., 2000). TGF-β has been shown to upregulate p15INK4b and p21CIP/WAF1 genes via Sp1 binding sites (Datto et al., 1995; Li et al., 1995). Smads can physically and functionally interact with Sp1 (Feng et al., 2000; Pardali et al., 2000). Analysis of p15INK4b-defective human cancer cell lines, including MCF 10A, revealed that the anti-proliferative effect of TGF-β is also mediated by inhibiting the expression of CDK tyrosine phosphatase cdc25A (Iavarone and Massagué, 1997). The downregulation of cdc25A by TGF-β is mediated by an E2F-p130 complex which recruits HDAC to the promoter (Fig. 2) (Iavarone and Massagué, 1999). It is unclear whether Smad proteins contribute directly to suppression of cdc25A gene.
TGF-β inhibits the expression of c-Myc via a TIE in the Myc promoter capable of binding Smad3 and Smad4 (Chen et al., 2001). Interestingly, TGF-β-induced downregulation of c-myc protein was shown to be required for TGF-β-induced p15INK4b expression (Fig. 2) (Massagué et al., 2000). Recently, c-Myc and Miz-1 were directly implicated in the activation of the p15INK4b promoter by TGF-β (Seoane et al., 2001; Staller et al., 2001). Miz-1 can activate the p15INK4b gene by binding to the transcriptional initiator (Inr) element of the p15INK4b gene. In the absence of TGF-β, c-Myc and its partner Max stably interact with Miz-1 to block the transcriptional activity of Miz-1 on the Inr element. The binding of Myc to Miz-1 interferes with the recruitment of transcriptional co-activator, p300/CBP. Upon TGF-β-induced reduction of c-Myc protein, Miz-1 through its zinc finger domains is able to interact with the MH1 domain of Smad3 and Smad4, which recognize a SBE in the p15INK4b promoter. In addition, Miz-1 is also capable of associating with Sp1 in a TGF-β-dependent manner when Smad4 is present. Thus, TGF-β-induced complex formation of Smads, Sp1 and Miz-1, each factor binding to its specific binding site in the promoter, activates the p15INK4b gene (Seoane et al., 2001; Staller et al., 2001).
TGF-β receptor-initiated signaling that is Smad-independent signaling has also been implicated in the anti-proliferative effect on EpH4 polarized mammary epithelial cells by TGF-β (Petritsch et al., 2000). The Bα subunit of phosphatase 2A (PP2A) interacts with TGF-β type I receptor (Griswold-Prenner et al., 1998) and p70S6k (Petritsch et al., 2000). Upon stimulating EpH4 cells with TGF-β, PP2A dephosphorylates and inactivates p70S6k, a serine/threonine kinase that is essential for G1/S progression. Both inactivation of p70S6k and activation of Smad pathway may cooperate in TGF-β-induced-G1 arrest in EpH4 cells.
Regulation of apoptosis by Smads
TGF-β family members can induce programmed cell death in many cell types and often this is accompanied with growth inhibition. Ectopic expression of Smads has been shown to enhance TGF-β-induced apoptosis in certain cells, suggesting an involvement of Smads in TGF-β-induced apoptosis (Fig. 3) (Yanagisawa et al., 1998). Stimulation and inhibition of expression of pro- and anti-apoptotic Bcl family members, respectively, with subsequent caspase activation has been implicated in TGF-β-induced apoptosis (Fig. 3) (Motyl et al., 1998; Francis et al., 2000; Chipuk et al., 2001; Kim et al., 2001). In addition, TGF-β-mediated induction of Id3 expression in primary B lymphocyte progenitors (Kee et al., 2001), TIEG in mink lung epithelial and Hep3B cells (Chalaux et al., 1999; Ribeiro et al., 1999), and TGF-β-stimulated clone 22 (TSC-22) in gastric carcinoma cells (Ohta et al., 1997), have been implicated in apoptosis. However, whether Smads are directly regulating the expression of these genes with role in apoptosis is not known. Recently, death associating protein (DAP) kinase, which induces apoptosis in certain cell types, was found to be transcriptionally induced by the TGF-β/Smad pathway. Inhibition of DAP kinase activity protected the cells from TGF-β-induced apoptosis (Jang et al., 2002).
TGF-β-induced activation of TGF-β-activated kinase-1 (TAK-1), a protein of the MAP kinase kinase kinase family, which can activate the p38 and JNK pathways has been shown to be involved in TGF-β family-induced apoptosis (Fig. 3) (Yamaguchi et al., 1995). The upstream activator of TAK1, TAB1 (Shibuya et al., 1996), may be linked to receptor activation via HPK1 or X-linked inhibitor of apoptosis (XIAP) in case of TGF-β or BMP, respectively (Wang et al., 1997; Yamaguchi et al., 1999). Recently, ARTS, apoptosis-related protein in the TGF-β signaling pathway (Larisch et al., 2000), a protein belonging to the septin family, was found to enhance cell death by TGF-β. ARTS is localized to mitochondria and translocates to nucleus when apoptosis occurs (Fig. 3). Moreover, Daxx, a Fas-receptor-associated protein, has been proposed to act as a mediator of TGF-β-induced apoptosis in a B-cell lymphoma via its interaction with TβR-II. Daxx acts downstream of TβR-I, and activates the JNK pathway and programmed cell death induced by Fas (Fig. 3) (Perlman et al., 2001). The involvement of the Smad pathway in TGF-β-induced apoptosis through XIAP, ARTS, or Daxx is unclear. Interestingly, there are several reports that Smad7, a downstream target gene of TGF-β family members, can efficiently induces apoptosis in epithelial cells (Landström et al., 2000; Lallemand et al., 2001; Mazars et al., 2001; Schiffer et al., 2001) independent of its antagonistic effect on TGFβ/Smad signaling (Mazars et al., 2001). Smad7 may induce apoptosis by inhibiting the survival NFκB (Lallemand et al., 2001) or by activation of JNK (Mazars et al., 2001).
Role of Smads in epithelial-mesenchymal transdifferentiation
In late stages of tumorigenesis, TGF-β can act as a promoter of tumor progression by acting directly on tumor cells (Oft et al., 1998; Yin et al., 1999). TGF-β can induce the transdifferentiation of mammary and skin epithelial cells into fibroblast-like cells with a highly invasive and metastatic phenotype (Fig. 4) (Miettinen et al., 1994; Portella et al., 1998; Piek et al., 1999). Epithelial to mesenchymal differentiation (EMT) is a complex process that involves a disruption of polarized epithelial morphology into cells with spindle-shaped morphology with formation of actin stress fibers, reduced cell–cell junctions through delocalization and downregulation of E-cadherin, and increased cellular motility. Induction of EMT by TGF-β can be seen in many different epithelial cell types including non-transformed mouse mammary cell line (NMuMG) and human keratinocytes (HaCat) cells (Miettinen et al., 1994; Portella et al., 1998; Piek et al., 1999; Zavadil et al., 2001), and is promoted by activated Ras (Oft et al., 1996), activated Raf (Lehmann et al., 2000) or by serum treatment (Piek et al., 1999). TGF-β can rapidly induce the activation of RhoA, a regulator of actin cytoskelton and adhesion junctions in NMuMG cells (Fig. 4) (Bhowmick et al., 2001). Rho kinase inhibitors block TGF-β-induced stress fiber formation and relocalization of E-cadherin (Bhowmick et al., 2001). Moreover, TGF-β-induced activation of Akt/PKB in a RhoA and phosphatidyklinositol 3-kinase-dependent manner was shown to be required for the rapid effects of TGF-β on relocalization of ZO-1 from tight junctions and changes in cell morphology (Fig. 4) (Bakin et al., 2000). Thus, these studies implicate Smad-independent pathways in TGF-β-induced EMT. However, the induction of stress fibers by TGF-β in NMuMG (and many other) cells occurs relatively late (6–12 h) which is not consistent with the rapid and transient activation of RhoA. In addition, TGF-β-induced EMT requires new protein synthesis (Bhowmick et al., 2001). This suggests that Smad signaling is required of TGF-β-induced EMT. Indeed, there are several reports that implicate Smads in certain aspects in TGF-β-induced EMT. Net1, a RhoA specific guanine exchange factor, is rapidly induced by TGF-β in a Smad-dependent manner in many different cells, including HaCat and Swiss 3T3 cells (Shen et al., 2001; Zavadil et al., 2001). Net1 was shown to be essential for TGF-β-induced stress fiber formation (Fig. 4) (Shen et al., 2001). Whether Net1 is also upregulated in NMuMG cells and mediates TGF-β-induced stress fiber formation in these cells, is not known. In addition, Piek et al. (1999) found that co-transfection of Smad2, Smad3, and Smad4 in NMuMG weakly induces stress fiber formation and changes in cell morphology, although full EMT required TGF-β receptor activation. Moreover, high constitutive expression of Smad7 or dominant negative Smad3 in NMuMG cells was found to inhibit TGF-β-induced stress fiber formation and TGF-β-induced cell morphology (A. Moustakas, personal communication). TGF-β-induced downregulation of E-cadherin expression in NMuMG cells is also likely a Smad-dependent process. SIP1 and Slug, transcriptional repressors of E-cadherin, are induced by TGF-β in NMuMG and HaCat cells, respectively (Fig. 4) (Comijn et al., 2001; Zavadil et al., 2001). However, whether SIP1 or Slug is involved in TGF-β/Smad-mediated E-cadherin downregulation is not known.
Transcriptional profiling using DNA micro-arrays of TGF-β-induced EMT in HaCat cells revealed a dynamic change in expression of approximately 4,000 genes, including mesenchymal factors, and Wnt and Notch signaling components that may be associated with the initiation of EMT. Inhibition of TGF-β-induced (indirect) ERK activation inhibited TGF-β-induced disassembly of adhesions junctions and cell motility. Among the 80 genes that are activated by TGF-β in an ERK-dependent manner, many have roles in cell matrix interaction, cell motility, and endocytosis (Fig. 4) (Zavadil et al., 2001). Thus, TGF-β-induced EMT is likely to occur through Smad-independent as well as Smad-dependent pathways.
SMAD MUTATIONS IN CANCER
Tumor cells often become resistant to TGF-β-induced growth inhibition and apoptosis due to functional inactivation of TGF-β receptors and Smads. In addition to epigenetic mechanisms, Smad2 and Smad4 have been found inactivated due to deletion or mutation in their genes (Gold, 1999; Blobe et al., 2000; de Caestecker et al., 2000a; Massagué, 2000; Massagué and Chen, 2000; Derynck et al., 2001; Pasche, 2001). Smad4 gene, which is located in 18q21, is frequently mutated in pancreatic and colorectal cancers (Hahn et al., 1996; Schutte et al., 1996; Thiagalingam et al., 1996). The loss or mutation of the Smad4 gene in one allele is often accompanied by a genetic alteration of Smad4 in the other allele. The Smad2 gene has been found mutated in colorectal and lung cancers although this gene is a less frequent target than Smad4 gene (Eppert et al., 1996; Riggins et al., 1996; 1997; Uchida et al., 1996; Takagi et al., 1998). Mutations in the Smad3 gene have thus far not been discovered in tumors. The MH domain of Smad4 and Smad2 is often the target for point mutations and frameshift mutations leading to premature proteins. Mutations in the MH2 domain may disrupt the core structure of the protein, perturb the ability to form stable heteromeric or homomeric Smad complexes, block receptor-dependent R-Smad phosphorylation or result in unstable Smad proteins (Hata et al., 1998b). Interestingly, overexpression of two tumor-derived Smad2 mutants (Smad2.D450E and Smad2.P445H) in MDCK cells was found to promote TGF-β-induced cellular invasion (Prunier et al., 1999). Thus, these two tumor-derived mutants appear to have retained some TGF-β-dependent signaling function.
The missense mutations in MH1 domain of Smad4 caused a marked reduction of Smad4 DNA binding ability (Jones and Kern, 2000; Morén et al., 2000). These MH1 domain mutants map to residues distant from β-hairpin motif, which has been shown to make direct contact with DNA (Shi et al., 1998). This suggests that the MH1 domain as a whole is important for DNA binding. In addition, some Smad mutants showed decreased protein stability (Morén et al., 2000; Xu and Attisano, 2000). In particular, Smad4.R100T and Smad2.R133C are rapidly cleared through the ubiquitin-proteasome pathway (Xu and Attisano, 2000). Moreover, certain Smad4 mutants (Smad4.G65V and Smad4.P130S) showed an inefficient nuclear accumulation upon TGF-β stimulation, although they associated with Smad3 upon TGF-β receptor stimulation (Morén et al., 2000). Thus, tumorigenic missense mutations in Smad genes impair (or redirect) TGF-β signaling by a variety of mechanisms.
SMAD KNOCK-OUT MICE
Gene ablation studies in mice have revealed specific developmental and physiological functions of Smads (Goumans and Mummery, 2000). Whereas mice deficient in Smad1, Smad2, Smad4 or Smad5 die early during development, Smad3 and Smad6 null mice make it to term (Goumans and Mummery, 2000). Smad1−/− embryos die at E10.5 exhibiting defects in the morphogenesis and over-proliferation of extra-embryonic tissues. Loss of Smad1 led to a dramatic reduction in the size and differentiation of the allantois and a failure to form the umbilical connection to the placenta. In addition, Smad1 deficient embryos display a reduced number of primordial germ cells (PEGs) at E8.5. Analysis of chimeric embryos revealed that Smad1 is needed to coordinate the growth of extra-embryonic structures (Tremblay et al., 2001). Mice lacking the Smad2 gene show failure in egg cylinder elongation, mesoderm formation, gastrulation, and establishment of an anterior-posterior (A-P) axis and die between E7.5 and E8.5. Analysis of chimeric embryos demonstrated that Smad2 also plays an important role in extra-embryonic tissue formation (Nomura and Li, 1998; Waldrip et al., 1998; Weinstein et al., 1998; Heyer et al., 1999). In contrast to Smad2-knockout mice, Smad3-deficient mice do not die before birth. In one report, adult Smad3-knockout mice suffer from metastatic colorectal cancer (Zhu et al., 1998). However, this phenotype was not observed by others where the mice died between 1 and 8 months after birth due to chronic infections. Loss of Smad3 was found to result in an impaired mucosal immunity and the T-cell response to TGF-β was diminished (Datto et al., 1999; Yang et al., 1999b). Furthermore, mice lacking Smad3 exhibit an accelerated wound healing due to an increased rate of re-epithelialization and reduced local infiltration of monocytes (Ashcroft et al., 1999). Thus, the Smad3 pathway may play an important role in tissue repair and regulation of keratinocytes and monocyte function. In addition, Smad3 knockout mice displays skeletal abnormalities shortly after weaning and have been shown to develop degenerative joint disease, indicating an important role for Smad3-mediated signaling in cartilage maintenance (Yang et al., 2001).
Analysis of mouse embryo fibroblasts (MEFs) from either Smad2- or Smad3-knockout mice revealed a reduced sensitivity to TGF-β-induced growth arrest (Piek et al., 2001). Consistent with this notion, TGF-β was unable to induce the p15INK4b gene in both Smad2 and Smad3-knockout cells (Piek et al., 2001). Smad2- and Smad3-deficient MEFs revealed elective TGF-β-induced gene responses via Smad2 or Smad3; e.g., matrix metalloproteinase-2 gene is selectively induced through Smad2 and the immediately early responsive genes, c-fos and Smad7, are selectively induced by Smad3 (Piek et al., 2001).
Smad4-deficient embryos have a phenotype similar to Smad2-deficient mice. Smad4 knock-out mice die between E7.5 and E8.5, and show growth retardation, no mesoderm formation, no gastrulation, and abnormal visceral endoderm development. The gastrulation defect of Smad4 null mice was rescued when a mutant embryos were surrounded by wild-type extraembryonic tissue (Sirard et al., 1998; Yang et al., 1998). Heterozygous Smad4 mice survive and develop malignant intestinal tumors after 1 year of age (Takaku et al., 1999; Xu et al., 2000), most likely due to a loss of heterozygosity and reduplication of the mutated Smad4 allele (Takaku et al., 1999). In heterozygous mice carrying mutations of Smad4 and Apc (responsible for human familial adenomatous polyposis), loss of heterozygosity and reduplication of the gene carrying the mutations results in a more malignant intestinal polyposis phenotype than that in mice that are only heterozygous for Apc (Takaku et al., 1998). Interestingly, families with familial juvenile polyposis (FJV) were found to have a germ-line mutation in the Smad4 locus. FJV is an autosomal dominant disease characterized by hamartomatous polyps and gastrointestinal cancer (Howe et al., 1998).
Smad5 knockout mice die between E9.5 and E10.5 due to an impairment of the circulation system. Smad5-deficient embryos show defects in angiogenesis resulting in enlarged dilated vessels and decreased numbers of vascular smooth muscle cells. The Smad5−/− yolk sacs develop into a disorganized vitelline network. In addition, Smad5-deficient mice show cranial-facial abnormalities and increased apoptosis (Chang et al., 1999; Yang et al., 1999a). As has been described for Smad1, Smad5 deficient embryos form a disorganized allantois and the number of PGCs was greatly reduced (Chang and Matzuk, 2001). Furthermore, left-right asymmetry and embryo turning are also impaired in Smad5 deficient embryos (Chang et al., 2000).
Smad6-knockout mice show hyperplasia of cardiac valves and outflow tract septation defects. In addition, the development of aortic ossification and elevated blood pressure were observed in adult Smad6-deficient mice. Thus, Smad6 might function to maintain homeostasis in the cardiovascular system (Galvin et al., 2000).
Smads are pivotal intracellular mediators of TGF-β family members that are activated by serine/threonine kinase receptors at the plasma membrane and relay the signal to the nucleus, where they act as transcription factors (Fig. 1). Consistent with the multifunctional and context dependent action of TGF-β family members, Smad activity can be positively or negatively regulated by many other stimuli (Itoh et al., 2000b). Functional genomics and proteomics approaches (Padgett and Patterson, 2001) will be important to further delineate the gene and protein targets that mediate distinct biological responses of these cytokines in different cell types within different cellular environments.
The TGF-β/Smad pathway can induce growth inhibitory and apoptotic responses (Figs. 2 and 3) and inactivation of intracellular components of this pathway has been shown to contribute to tumorigenesis. TGF-β has also been shown to promote tumorigenesis (Derynck et al., 2001). Ectopic expression of dominant negative TGF-β type II receptor in tumor cells inhibits their invasion and metastasis (Oft et al., 1998; Yin et al., 1999). Consistent with this notion, colon cancers with microsatellite instability having inactivating mutations in TGF-β type II receptor (Markowitz and Roberts, 1996) show a reduced level of distant metastasis and increased patient survival (Gryfe et al., 2000). Invasive and highly metastatic tumor cells may have therefore retained certain TGF-β-induced responses.
Mutation of Smad2 or Smad4 that are found in particular tumor subsets, may subvert TGF-β signaling. Inactivation of Smad2 may leave the TGF-β/Smad3 signaling intact, and cause a redirection of TGF-β-induced responses. Many of the target genes for Smad2 and Smad3 are distinct (Piek et al., 2001). Of note, Smad3 has been shown to cooperate with AP1 family members that may activate genes important in invasion and metastasis (de Caestecker et al., 2000a). Certain TGF-β responses, including TGF-β-induced upregulation of fibronection, can occur in the absence of Smad4 (Hocevar et al., 1999; Sirard et al., 2000). Moreover, mutated Smads that are stably expressed may have retained certain signaling properties (Prunier et al., 1999). Alternatively, tumors may have become temporarily resistant to TGF-β/Smad signaling via epigenetic mechanisms. In fact, many tumors that are insensitive to TGF-β-induced growth inhibition have no genetic defects in TGF-β receptors or Smads. It is thus possible that during the initial phases of tumorigenesis, cells become (selectively) insensitive to TGF-β/Smad-induced growth inhibition and apoptosis, e.g., through (temporarily) downregulation of TGF-β receptors or functional Smad inactivation by active Ras. This may lead to uncontrolled cellular growth and predispose the cells to obtain mutations that promote their malignant potential. In late stages, TGF-β receptor/Smad signaling is restored (or remains redirected), and TGF-β (often produced in high amounts by tumor) actively promotes invasive and metastatic properties of tumor cells in a cell-autonomous manner (Fig. 4). Smads, like TGF-β, depending on cellular context, may thus act as tumor suppressor and tumor promoter (de Caestecker et al., 2000a). Further insights into the molecular mechanisms of TGF-β receptor/Smad (in)dependent pathways, which inhibit or promote tumorigenesis, have the promise of providing new targets for therapeutic intervention of cancer.
We are grateful to the Dutch Cancer Society (NKI 2000-22117 and NKI 2001-2481), the Netherlands Organization for Scientific Research (ALW 809.67.024 and MW 902-16-295), and the Ludwig Institute for Cancer Research for their support of our research on TGF-β signal transduction. We thank Dr. Scott Kern, Dr. Aristidis Moustakas, Dr. Ed Leof, and Dr. Mark Taketo for valuable discussion.