Crossing Smads

See allHide authors and affiliations

Science's STKE  14 Mar 2000:
Vol. 2000, Issue 23, pp. re1
DOI: 10.1126/stke.2000.23.re1


The transforming growth factor–β (TGF-β) superfamily of secreted polypeptide growth factors exerts extensive control over all aspects of development and homeostasis, and components of this pathway are often mutated in cancers and in several hereditary disorders. Apart from TGF-β, the superfamily also includes the activins and the bone morphogenetic proteins. These factors signal through heteromeric complexes of type II and type I serine-threonine kinase receptors, which activate the downstream Smad signal transduction pathway. Three classes of Smads have been defined: the receptor-regulated Smads (R-Smads), the common-mediator Smads (co-Smads), and the antagonistic or inhibitory Smads (I-Smads). Receptor complexes activate the Smad pathway by interacting and phosphorylating specific R-Smads. Phosphorylation of the R-Smads causes dissociation from the receptor and induces assembly into complexes with Smad4, a co-Smad. This heteromeric complex then translocates into the nucleus, where the Smads function as transcriptional comodulators by recruiting coactivators or corepressors to Smad DNA binding partners. Thus, Smads transmit signals directly from the receptor kinase into the nucleus. Crosstalk between Smads and other signaling pathways occurs both in the cytosol and in the nucleus. In the cytosol, Smad translocation might be inhibited by mitogen-activated protein kinase–dependent phosphorylation, whereas in the nucleus Smads interact with a number of transcription factors that themselves are primary targets of other signaling pathways. Furthermore, Smad-dependent regulation of these targets often requires input from the primary signaling pathway. In these examples, Smad signaling may represent a secondary signal that modifies the output of the primary pathway. Consequently, the transcriptional response to TGF-β family ligands may be dependent on what other signals are being received by the cell. Crosstalk may thus provide one explanation for the long-standing observation that the biological response to TGF-β is often dependent on the extracellular environment of the cell.


Transforming growth factor–β (TGF-β) is the prototype of a large family of secreted polypeptide growth factors that signal through the transmembrane serine-threonine kinase class of receptors [reviewed in (1-5)]. Signaling from these receptors is initiated when ligand induces the assembly of heteromeric complexes of type II and type I receptor serine-threonine kinases (Fig. 1). Within the receptor complex, the type II receptor phosphorylates the type I receptor in a glycine- and serine-rich motif just upstream of the kinase domain. This region, termed the GS domain, is important in controlling type I receptor kinase function, and a threshold of phosphorylation here activates the type I kinase, which subsequently initiates the downstream Smad signaling pathways.

Fig. 1.

Mechanism of activation of TGF-β superfamily receptors. TGF-β superfamily ligands signal through type I and type II serine-threonine kinase receptors. TGF-β and activin binding to receptor II leads to recruitment of receptor I. The constitutively active type II phosphorylates the type I receptor in the highly conserved juxtamembrane region, known as the GS domain. The activated type I receptor then phosphorylates its downstream targets, the members of the Smad family of signal transducers. Although signaling through a heteromeric receptor complex is likely to be universal, for some BMPs, ligand can bind directly to type I receptors and recruit type II receptors.

Members of the Smad family of signal transduction molecules are components of a critical intracellular pathway that transmits TGF-β signals from the cell surface into the nucleus. Three distinct classes of Smads have been defined (Fig. 2): the receptor-regulated Smads (R-Smads), the common-mediator Smads (co-Smads), and the antagonistic or inhibitory Smads (I-Smads). [The core aspects of Smad signaling have been reviewed extensively (1-5) and are presented in this review in summary form as schematic figures; they can be viewed as a movie at] Briefly, activated type I receptors associate with specific R-Smads and phosphorylate them on a conserved SSXS motif (where S is serine and X can be any amino acid) at the COOH-terminus of the proteins (Fig. 3). The phosphorylated R-Smad dissociates from the receptor and forms a heteromeric complex with the co-Smad Smad4, and together the heteromeric complex moves to the nucleus. Once in the nucleus, Smads can target a variety of DNA binding proteins to regulate transcriptional responses. Analysis of an activin response element (ARE) in the promoter of the Xenopus homeobox gene Mix.2 led to the identification of forkhead activin signal transducer (FAST), the first nuclear target defined for Smads (6). FAST contains a forkhead DNA binding domain and has a Smad interaction domain (SID) near the COOH-terminus. On its own, FAST does not regulate transcription of the Mix.2 ARE. However, upon initiation of activin signaling, Smad complexes entering the nucleus bind to FAST via specific interactions between an R-Smad (Smad2 or Smad3) and the SID. Once recruited to FAST, Smad4 stabilizes the ternary Smad-FAST DNA binding complex by binding to DNA at a Smad binding element (SBE) that lies adjacent to the FAST binding site (6-11). Because the bone morphogenetic protein (BMP)–regulated Smads 1, 5, and 8 do not interact with FAST, specificity in activin- and TGF-β-dependent induction of FAST target genes is maintained. By functioning both as a substrate of the type I receptor and as an usher for Smad4, the R-Smad has a central role in maintaining specificity in the TGF-β pathway (Fig. 3).

Fig. 2.

Summary of the Smad family. The Smads thus far identified in vertebrates, Drosophila, and C. elegans are summarized as indicated, according to the three functional classes of Smads. A second form of Smad4, Smad4β/Smad10, has been identified only in Xenopus.

Fig. 3.

Two distinct Smad signaling pathways. TGF-β binds to its receptors TβRII and TβRI while activins interact with the type II receptors ActRII or ActRIIB and the type I receptor ActRIB. In contrast, BMPs and possibly other TGF-β family ligands act through the type II receptors ActRII, ActRIIB, or BMPRII and the type I receptors ALK1, 2, 3, and 6; these receptors propagate the signal to Smads 1, 5, and 8. The activated type I receptors phosphorylate a receptor-regulated Smad (R-Smad; blue) at a conserved SSXS motif present at the COOH-terminus of the protein. In the case of TGF-β/activin, Smad2 and Smad3 are phosphorylated, whereas for BMPs, Smads 1, 5, and 8 are targeted. This phosphorylation causes dissociation of the R-Smad from the receptor complex and induces formation of a heteromeric Smad complex with the co-Smad Smad4 (red). The R-Smads then mediate both nuclear translocation of Smad4 and interaction of the Smad complex with specific nuclear targets to maintain specificity in the transcriptional response.

In opposition to the R-Smads, the I-Smads function as potent antagonists of TGF-β signaling [reviewed in (1-5)]. Smad7 binds to either TGF-β or the related BMP receptor complexes and prevents the access and phosphorylation of the respective R-Smads (Fig. 4). Smad6, on the other hand, appears to preferentially inhibit BMP signaling by interacting with the receptor, or by binding to phosphorylated Smad1 and preventing formation of a heteromeric complex with Smad4. Transcription of the I-Smad genes is stimulated by TGF-β family members, providing for potential negative feedback of the pathway. In addition, the expression of Smad6 and Smad7 is enhanced by multiple signals including epidermal growth factor (EGF), the phorbol ester 12-O-tetradecanoylphorbol 13-acetate (TPA), and interferon-γ (12, 13), which could provide an important mechanism whereby these pathways negatively regulate Smad activation.

Fig. 4.

Intracellular antagonism of TGF-β superfamily signaling. TGF-β superfamily signals can be blocked in a variety of ways. Smurf1, a HECT domain containing ubiquitin ligase, can induce degradation of Smad1, whereas the I-Smads Smad6 and Smad7 can prevent R-Smad interaction with the receptor complex and (in the case of Smad6) might also prevent formation of R-Smad–Smad4 complexes. Phosphorylation of the nonconserved linker region of Smads by MAP kinases can also antagonize signaling by preventing Smad nuclear accumulation.

The identification of Smads in multiple phyla led to the identification of two domains that are particularly well conserved. These are termed the MAD homology 1 (MH1) and MAD homology 2 (MH2) domains (also called the N and C domains, respectively). These are found at the NH2- and COOH-terminal regions of the molecule, respectively, and are joined by a linker region (Fig. 5). Inactivating mutations in Smads identified in human cancers and in Drosophila and Caenorhabditis elegans have all been mapped to either of these two domains, with the majority found in the MH2 domain (14). Smad MH1 and MH2 domains do not appear to have any intrinsic enzymatic activity, but they function as effectors in TGF-β superfamily signaling by controlling protein-protein interactions, protein-DNA interactions, and subcellular localization.

Fig. 5.

Domain structure of Smads. Smad proteins comprise three domains: the highly conserved MH1 and MH2 domains, joined by a nonconserved linker region. A number of distinct functions have been ascribed to each of these domains, as indicated. SAD refers to the Smad activation domain identified in Smad4 (65).

Protein-protein interactions in Smad signaling.

The MH2 domain of Smads appears to exert its function in the pathway by controlling protein-protein interactions. In the case of the R-Smads, the MH2 domain mediates binding to the activated type I receptor, and three residues within loop 3 of the MH2 domain appear to mediate interactions with specific type I receptors (15). On the receptor side, a loop between β sheets 4 and 5 (loop 45) appears to be an important determinant for Smad interactions (15). How loop 3 on the Smad and loop 45 on the receptor interact with each other to generate specificity in Smad-receptor interaction is unclear. A structure of an R-Smad bound to an activated type I receptor has not been solved, but in the crystal structure of the Smad2 MH2 domain, the divergent loop 3 resides adjacent to a basic pocket that is likely conserved in all R-Smads (16). Thus, loop 45–loop 3 interactions, which specify the Smad-receptor interaction, may be facilitated by binding between the phosphorylated GS domain on type I receptors and the basic pocket on R-Smads.

The TGF-β-regulated R-Smads, Smad2 and Smad3, also bind to the Smad anchor for receptor activation (SARA) (17). This interaction occurs between a Smad-binding domain (SBD) on SARA and the MH2 domain of the unphosphorylated R-Smad (Fig. 6) The SBD of SARA lies adjacent to a FYVE domain. FYVE domains interact specifically with the membrane phospholipid phosphatidylinositol 3´-phosphate (18). In SARA, the FYVE domain mediates its subcellular localization to punctate regions within the cell, and SARA can recruit Smad2 to these domains. TGF-β receptors are also localized to this domain and SARA can also interact with the receptors, potentially scaffolding the receptor kinase to its Smad substrate. Thus, SARA functions to recruit Smad2 to the TGF-β receptor complex by controlling the subcellular localization of Smad2. The crystallographic structure of the Smad2 MH2 domain was solved in a complex with the SBD from SARA (16). This structure reveals that a coiled region, an α helix, and a β strand on the SARA SBD are involved in an extended interaction surface with the Smad2 MH2 domain. Furthermore, an asparagine residue in the Smad2 MH2 domain makes several important contacts with the SARA SBD. Mutating this residue to a serine, which is found in Smad1, appears to interfere with Smad2-SARA interactions and suppresses Smad2-dependent signaling. SARA mutants that cause mislocalization of Smad2 also suppress TGF-β-dependent signaling, and together these data suggest that regulation of the subcellular localization of unactivated R-Smads is an important feature in TGF-β-mediated signal transduction.

Fig. 6.

SARA, a FYVE domain protein, promotes TGF-β signaling. Membrane-localized SARA binds unphosphorylated Smad2 (or Smad3) and recruits it to the membrane. Upon initiation of TGF-β signaling, SARA, which can also interact with the receptor complex, presents Smad2 to the activated receptor complex. Phosphorylation of Smad2 frees it from SARA and induces formation of a Smad2-Smad4 heteromeric complex. SARA can then recruit additional Smad2 to the membrane.

Once Smads are phosphorylated at the COOH-terminal SSXS motif, they dissociate from the receptors and SARA. Phosphorylation further induces association with Smad4 and nuclear translocation of the complex. How phosphorylation regulates these events is unknown; however, it may function to relieve an intramolecular inhibition imposed by the physical interaction of the MH1 domain with the MH2 domain (26). Binding of the R-Smad to Smad4 is also mediated by their respective MH2 domains. Initial crystallographic studies on the COOH-terminal MH2 domain of Smad4 led to the suggestion that the complex is a hexamer (19), although more recent studies indicate that in vivo the R-Smad–Smad4 heteromer may be a trimer (20). The MH2 domains of Smads also bind to a host of downstream nuclear targets in the TGF-β and BMP pathways (Fig. 7). Each of these protein-protein interactions in the nucleus can control transcriptional responses to TGF-β. Consequently, they provide the basis for crosstalk between TGF-β signaling and numerous other regulatory pathways (see below).

Fig. 7.

Summary of Smad partners. A variety of protein-protein interactions have been defined for the MH1 and MH2 domains and the linker region. Listed below each domain are known protein interactors with their Smad partners, followed by reference numbers for descriptions of these interactions.

The linker region that connects the MH1 and MH2 domains is less well conserved than the MH domains. Nevertheless, it contains a number of important peptide motifs that can function to control Smad activity. In particular, there are consensus phosphorylation sites for proline-directed kinases, and these may be important targets for crosstalk with the mitogen-activated protein kinase (MAPK) pathway (21). In addition, the linker regions of the R-Smads and the I-Smads contain a conserved proline-tyrosine (PY) motif that mediates interactions with WW domains. The protein Smurf1 contains a C2 domain, two WW domains, and a homology to E6 COOH-terminus (HECT) ubiquitin ligase domain. Smurf1 binds to Smad1 or Smad5 via their PY motifs and mediates the ubiquitination and degradation of the Smad protein (22) (Fig. 4). Consistent with this, Smurf1 is a potent antagonist of BMP signaling. Interestingly, activated R-Smad2 is also targeted for ubiquitin-mediated degradation (23). In this case it is the entrance of Smad2 into the nucleus, not phosphorylation per se, that triggers Smad degradation. Although the nuclear E3 activity responsible for this pathway has not been identified, it may not function through the PY motif of R-Smad2, because the MH2 domain alone is also targeted by a similar pathway.

In contrast to the Smad MH2 domain, the MH1 domain mediates DNA binding as well as protein-protein interactions. The Smad3 and Smad4 MH1 domains bind the transcription factor ATF2, whereas the Smad3 MH1 domain and linker region bind Jun and can bind the basic helix-loop-helix transcription factor TFE3 (24, 25). In the case of ATF2 and Jun, these protein-protein interactions provide for crosstalk between the TGF-β signaling pathway and the c-Jun NH2-terminal kinase (JNK) and p38 MAPK cascades.

DNA binding in Smad signaling.

Initial work on the Smads suggested that the MH1 domain functioned to inhibit the activity of the MH2 domain through direct physical interactions (26). However, it has now become apparent that the MH1 domain also provides important effector functions for Smads, mediating binding to DNA and to other proteins.

Binding of the MH1 domain of Smads to DNA was first described for Drosophila MAD, the prototypic member of the Smad family (27). This led to the identification of a MAD-binding element in the quadrant enhancer of the dpp target gene, vestigial, that when mutated abolished dpp-dependent regulation of a reporter gene (27). Binding sites for Smad1, the vertebrate homolog of MAD, have not been defined in BMP-regulated promoters. However, extensive work with Smad4 and Smad3 has led to the definition of a host of DNA binding motifs in various TGF-β and activin target elements (3). Furthermore, in most cases, as in Drosophila, mutation of the Smad binding sites is sufficient to reduce activation of the promoter in transient transfection assays, which suggests that Smad DNA binding is important for activating target genes. DNA binding site selection and extensive analysis by mutagenesis has shown that the consensus DNA site for Smads is GNC (28-30). Thus, Smads bind DNA with low specificity, and biochemical studies suggest that this interaction is also of low affinity (31). Consequently, DNA binding alone is unlikely to be sufficient for Smad-dependent regulation of specific target genes. Indeed, there are a growing number of examples in which Smads are found to cooperate with DNA binding partners to regulate transcription in what appears to be a general model for how Smads control target gene activity (Fig. 8). In this model, the nuclear complex of R-Smad–Smad4 binds to specific promoter elements through the interaction of R-Smad with DNA binding partners. This promotes the binding of Smads to DNA at adjacent Smad elements, which in turn stabilizes the higher order DNA binding complex. The Smads then positively regulate transcriptional activity at the target elements by recruiting transcriptional coactivators, such as the cAMP response element–binding protein (CBP), p300, and melanocyte-specific gene 1 (MSG1) (3). Alternatively, in some instances Smads can negatively regulate transcription by recruiting corepressors such as TGIF or the oncoproteins Ski and SnoN, which can all associate with histone deacetylase (32-37). As discussed below, this general model has important implications for the ways in which signals from distinct pathways crosstalk in the nucleus in the manifestation of transcriptional responses to TGF-β.

Fig. 8.

Nuclear functions of Smads. (A) The R-Smad–Smad4 complex interacts with various DNA-binding protein partners to regulate the activation of specific target genes. In some cases, DNA binding of Smads is required for transcriptional regulation and may function to stabilize the higher order DNA binding complex. Smads positively or negatively regulate transcription by recruiting the coactivators CBP, p300, or MSG, or the corepressors TGIF or Ski, which in turn recruit histone deacetylases. (B) Activation of mouse goosecoid and Xenopus Mix.2 occurs through the association of Smad2-Smad4 complexes with FAST, a winged/helix forkhead transcription factor that specifically binds to the ARE in the promoter. Smad2 directly interacts with FAST while Smad4 binds DNA at a site adjacent to the ARE. Smad complexes comprising Smad3 and Smad4 can prevent activation of mouse goosecoid through the ability of Smad3 to bind to the Smad4 DNA binding site, thereby possibly preventing Smad4 access to DNA.

In the case of Smad2, a unique insert in its MH1 domain abolishes DNA binding, making Smad2 functionally distinct from the other TGF-β and activin R-Smad, Smad3 (38). This difference between Smad2 and Smad3 may have important ramifications at target genes. For instance, in the mouse goosecoid promoter, Smad2 activates the gene in cooperation with Smad4 and FAST2, whereas Smad3 does not activate the element and actually blocks Smad2-dependent activation (8). This functional difference appears to result from the DNA binding activity of the Smad3 MH1 domain, which may compete with Smad4 for binding to the Smad site that lies adjacent to the FAST site (Fig. 8). Thus, in addition to promoting the activation of certain target genes, Smad DNA binding can also block activation. This may not occur at all FAST target elements, because the activin response element of the Xenopus Mix.2 gene appears to be activated by either Smad2 or Smad3 (11, 38). The presence of the Smad2 MH1 domain insert is controlled by alternative splicing (38). However, the relevance of the two isoforms to the function of the Smad2 gene in vivo is unclear.

At some TGF-β-responsive elements, the Smad MH1 domain may use both DNA and protein binding to modulate the activity of the promoter. In the case of the basic helix-loop-helix transcription factor TFE3, the MH1 domain interacts directly with the transcription factor and can bind to DNA at a Smad element that lies 3 base pairs downstream from the TFE3 binding site (39). This cooperativity is essential for TGF-β-dependent regulation of the element, and if the Smad element is moved away from the TFE3 binding site, TGF-β responsiveness of the element is lost. Similarly, the Smad3 MH1 domain also binds Jun, and Smad regulation depends on a Smad3 binding site that lies very close to the Jun-binding AP1 site (24). Thus, protein-protein and protein-DNA interaction by the MH1 domain can function cooperatively to mediate TGF-β responsiveness at target promoters.

Crosstalk at the Membrane and in the Cytosol

Crosstalk in Smad activation.

There is remarkable specificity in the regulation of R-Smad phosphorylation by the type I receptor kinase. Thus, in most cells that have been examined, BMPs regulate Smad1-, Smad5-, and Smad8-dependent responses, including increased transcription of BMP-specific target genes, whereas TGF-β and activin regulate Smad2 and Smad3 and their target genes [(1-5); see also Fig. 3]. There have been very few reports of any crosstalk between these two parallel pathways, although in some cells TGF-β may induce phosphorylation of Smad1 (40). The molecular basis for this observation has not been explored. However, such relaxed specificity is not observed upon overexpression of activated type I receptors or receptor complexes. This suggests that potential crosstalk between TGF-β and the BMP-regulated Smads may not be due to the capacity of the receptor kinase to recognize both Smads in vivo. Alternative type I receptors, such as ALK1, that potentially bind TGF-β and mediate activation of Smad1 (41, 42) could explain these observations. In contrast to the R-Smads, the co-Smad Smad4 functions as a component of both the TGF-β/activin and BMP pathways. This raises the interesting possibility that Smad4 can mediate crosstalk between the TGF-β/activin and BMP pathways. Indeed, studies in Xenopus suggest that BMP and activin pathways antagonize each other, and that this antagonism can be relieved by increasing the pool of Smad4 in the cell (43). Thus, the TGF-β/activin and BMP pathways may antagonize each other through competition by their respective R-Smads for the co-Smad Smad4.

Activation of Smads by phosphorylation at the SSXS motif might also be mediated by other pathways. Treatment of cells with EGF or a variant of Hepatocyte Growth Factor (HGF) can lead to transient phosphorylation of R-Smad2 (21, 44). Although some of this phosphorylation likely occurs at MAPK or ERK (extracellular signal-regulated kinase) sites present in the linker region (see below), activation of Smad2, possibly via phosphorylation at the SSXS motif, was also reported (44). Overexpression of constitutively activated mitogen-activated protein kinase kinase kinase–1 (MEKK-1) in endothelial cells may also lead to activation of Smads (45). It is currently unclear whether Smad activation by these pathways involves serine-threonine kinase receptor signaling or occurs through some as yet uncharacterized mechanism.

MAPK regulation of Smad function.

In cells overexpressing R-Smads, Ras-dependent activation of MAPK or ERK can lead to phosphorylation of R-Smads in the linker region (21, 46). In the case of Smad1, this phosphorylation occurs on four MAPK consensus phosphorylation sites present in the linker region and can be induced by stimulation of cells with EGF or HGF. This phosphorylation inhibits nuclear accumulation of the R-Smads (Fig. 4) and causes inhibition of BMP-dependent activation of a GAL4-Smad1 fusion protein (21). Four sites in the Smad2 and Smad3 linker regions are similarly regulated by activated ERK, leading to inhibition of Smad2 nuclear accumulation, with a concomitant block in TGF-β signaling (46). Currently it is unclear how phosphorylation in the linker region inhibits nuclear accumulation of the R-Smads, but it could function to interfere with nuclear import or stimulate either nuclear export or nuclear degradation of the R-Smad.

Although BMP and receptor tyrosine kinase (RTK) pathways are antagonistic in several developmental systems, they can also function synergistically. For instance, in Drosophila, branching morphogenesis in the trachea involves antagonistic interactions between EGF receptor (DER) signaling and the dpp pathway (dpp is the Drosophila ortholog of mammalian BMP2) (47). However, during endoderm induction the DER pathway cooperates synergistically with DPP to induce expression of labial (48). Furthermore, in many cell types TGF-β antiproliferative activity is dominant over Ras-activated mitogenic pathways, and in the frog embryo, activin, which signals through a TGF-β-like pathway, appears to require Ras activity for its function (49). These results indicate that in some biological contexts, direct antagonism of Smad signaling by MAPK pathways may not occur, and thus the interplay between Smad signaling and Ras-activated kinase cascades is likely to be quite complex.

One important level of complexity is likely provided by the temporal aspects of Ras versus Smad activation. In particular, the rapid and transient activation of Ras observed for many agonists sharply contrasts with the sustained activation of Smad pathways by BMPs and TGF-β that is typically observed in a number of systems. This may limit the degree to which MAPK or ERK activation can interfere with Smad signaling, particularly in situations where Ras is activated transiently. In contrast, when sustained activation of Ras is achieved, as in nerve growth factor (NGF)–induced differentiation of neuronal precursor cells (50, 51), antagonism of Smad signaling may be manifested. Furthermore, in human cancers, Ras is often constitutively active, thus providing for sustained MAPK activation. Crosstalk with the Smads in such cells could thus provide an important mechanism whereby tumor cells lose responsiveness to TGF-β.

Nuclear Crosstalk

The involvement of Smads as transcriptional coregulators provides a number of interesting examples of nuclear crosstalk between Smads and several other cell signaling pathways.


Early experiments in which Smad MH2 domains were fused to GAL4 DNA binding domains revealed that Smads could activate transcription. This activation likely occurs through interaction of Smads with the transcriptional coactivators p300 and CBP (3). Recruitment of CBP by Smad3 is the most extensively characterized interaction and occurs between the MH2 domain and a region that corresponds to residues 1891 to 2175 in CBP. Association of CBP or p300 with Smad1 or Smad2 has also been documented, whereas direct binding to Smad4 has been difficult to demonstrate and in vivo may only occur through the R-Smad. Smad4 may thus cooperate with R-Smads to enhance transcription by recruiting additional coactivators such as MSG1 (52).

CBP and p300 function as coactivators for a host of transcription factors that are endpoints for a wide range of signaling systems (53). These include CREB, AP1, steroid hormone or nuclear receptors, signal transducers and activators of transcription (STATs), MyoD, nuclear factor κB, and p53. CBP may thus provide an important crossroad in the nucleus for interaction between Smads and these other pathways. Indeed, the viral oncoprotein E1A, which binds to a region of CBP adjacent to the Smad binding site, blocks Smad-CBP interactions and suppresses TGF-β signaling (54). Furthermore, as discussed below, Smads may also modulate STAT signaling through their interactions with p300. Finally, because CBP is also a target for kinases, such as Ca2+-calmodulin–dependent kinase IV (55), it is possible that the posttranslational modification of CBP may also play a role in mediating crosstalk with the Smad pathway.

DNA binding partners for Smads.

Interactions have been described between Smads and a diverse array of DNA binding proteins that are themselves targets for regulation by other signaling pathways (Fig. 9). These studies have brought about some important insights into understanding how the diversity and complexity of TGF-β superfamily biology is mediated at the molecular level.

Fig. 9.

Summary of nuclear crosstalk between Smads and DNA binding partners. Smads have been found to interact with a variety of nuclear partners that are themselves regulated by other signaling pathways. The Smad partners that have been identified to date are shown together with the pathways that regulate their activity. Smads can modulate transcriptional activity of these partners by recruiting either coactivators (green box) or corepressors (red box). The Smad interactions that have been characterized thus far are as indicated (S1, S2, S3, and S4 refer to Smad1, Smad2, Smad3, and Smad4, respectively).

Many TGF-β family target genes are also regulated by TPA, including those encoding plasminogen activator–1 (PAI-1), osteopontin, and Smad7. Several TGF-β-responsive elements contain AP1 sites, and in artificial promoters the introduction of AP1 binding sites often enhances TGF-β-dependent activation. When Smad3 was first identified, functional studies showed that overexpression of the protein could strongly activate AP1-containing promoters, which perhaps suggested that Smads might cooperate with AP1 binding proteins to regulate activation at these elements (56, 57). Jun and Fos are important activators of AP1 sites and form a heterodimeric complex that binds to the DNA with high affinity. Smad3 can directly bind Jun via its MH1 domain and linker region and may bind Fos indirectly through its MH2 domain, and in mammalian nuclear extracts, Smad3-Smad4 heteromers can form DNA binding complexes with Jun-Fos heterodimers on AP1 sites that contain overlapping Smad binding sites (24). Furthermore, overexpression of Smad3 can potentiate transcription from AP1-containing reporter genes. Thus, it appears that Jun-Fos heterodimers are functional targets for Smad3 in TGF-β signaling. Because AP1 sites are targets of TPA, these observations may explain some of the overlap that has been observed in TGF-β- and TPA-mediated gene regulation. Jun-Fos activity is also regulated by a number of other external stimuli. In particular, Jun is a target for JNK, which is part of a kinase cascade regulated by a host of external cues that include ultraviolet radiation, mitogenic stimuli, tumor necrosis factor, and others. JNK stimulates the transcriptional activity of Jun by phosphorylating it at Ser63 and Ser73 in the transactivation domain (58). Thus, Smad signaling can converge with these signaling pathways at the level of AP1-containing regulatory elements. However, in the absence of Jun and Fos binding, TGF-β and Smad3 only weakly activate AP1-containing elements (24, 59). This suggests that Smads require active Jun-Fos dimers as DNA binding partners in order to stimulate AP1-containing regulatory elements. Consistent with this finding, the phosphorylated, activated form of Jun preferentially interacts with Smads (59). This has important biological implications, because it suggests that TGF-β-dependent activation of specific target genes through AP1 sites may occur only in cells that contain Jun-Fos activity.

Another downstream target of JNK as well as p38 MAPK is the transcription factor ATF2, which, like Jun and Fos, is a member of the b-ZIP family of DNA binding proteins. ATF2 can bind to cAMP response elements (CREs) as a homodimer or a heterodimer with Jun. ATF2 also binds the MH1 domains of Smad3 or Smad4, and these Smads can cooperate with ATF2 to activate CRE-containing elements in transient assays (25). Inhibition of ATF2 activation strongly suppresses the TGF-β-responsiveness of CRE reporter elements (25, 60). Thus, Smads appear to interact with the ATF2 family. ATF2-like factors may also function as a focus for crosstalk in Drosophila, because activation of a DPP-responsive element in the Ultrabithorax promoter was found to be dependent on an intact CRE-containing element (61).

Smad signaling can also interact with signals for the vitamin D receptor (VDR). Analysis of a vitamin D response element (VDRE) linked to a reporter gene revealed that the VDRE was responsive to TGF-β (62). In the absence of vitamin D, a VDRE reporter gene was unresponsive to TGF-β, but in the presence of vitamin D, which activated the promoter, TGF-β induced additional activity. Consistent with these observations, the Smad3 MH1 domain interacted weakly with the VDR in the absence of vitamin D. However, vitamin D treatment caused a significant enhancement in the association of VDR with Smad3. Vitamin D binding to VDR also promotes interaction of the transcriptional coactivators SRC-1 or TIF2 with the AF-2 activation domain that resides at the COOH-terminus of the receptor. Current evidence suggests that the SRC-1 and TIF2 binding domain is required for Smad3 interaction with the VDR. Smad4 appears not to bind to the VDR and does not affect activation of the VDRE, which suggests that enhancement of vitamin D–regulated promoters by TGF-β may not require Smad4 activity. How Smad3 functions to enhance vitamin D–dependent transcription is still unclear, but because Smad3 can bind to CBP/p300, it may act by bringing more coactivators into the DNA-bound complex.

During differentiation of neuroepithelial cells, BMP can function to promote astroglial cell formation. This activity is also observed with leukemia inhibitory factor (LIF). BMP and LIF can synergistically enhance formation of astroglial cells with a concomitant induction of glial fibrillary acidic protein (GFAP), which is a marker of the astroglia lineage. The induction of GFAP expression during differentiation can be recapitulated with a portion of the GFAP promoter and required STAT binding sites (63). This suggested the possibility that Smad1 in the BMP signaling pathway and STATs in LIF signaling might cooperate with each other. No direct interaction of STAT and Smad1 was found, although overexpression of STAT and Smad1 together with p300 allowed formation of STAT-Smad1 complexes, presumably bridged by p300. It is currently unclear how Smad1 binding to p300 might enhance STAT signaling. Nevertheless, the data suggest that Smad1 might modulate STAT activity. Because STATs are controlled by cytokine signaling, these data again suggest the possibility that Smads might modulate preexisting signals but would have no activity on their own on STAT-regulated promoters.

Smads as Transcriptional Comodulators

All of the above examples of nuclear crosstalk involve Smad interaction with proteins that are themselves targeted by other signaling systems (Fig. 9). This mode of interaction differs from the regulation of the first Smad nuclear target identified, the forkhead protein FAST (6). FAST has not been reported to function in any signaling pathway other than that mediated by Smads. Thus, we might envision two distinct ways in which Smads mediate transcriptional responses to TGF-β family signaling (Fig. 10). In the example of FAST-dependent target genes, FAST may represent a unique Smad target that on its own binds DNA with high affinity and specificity but is unable to activate transcription. Smads can directly target FAST to positively or negatively regulate transcription independent of other signaling pathways. Thus, the Smads may function as the primary regulatory signal at FAST target elements. Because FAST is only expressed during gastrulation, this may provide for "hard-wired" induction of a specific TGF-β/activin-regulated transcriptional program required at this stage of development.

Fig. 10.

Smads as primary or secondary transcriptional comodulators. At FAST target elements (upper panel), transcription is silent in the absence of the Smad partner. Smads that translocate into the nucleus function as primary signals that can regulate transcription at these elements independently of other signaling pathways. In the known examples of nuclear crosstalk (lower panel), Smads are not required for the activity of the element but can modulate the signal. In the known examples, Smads appear to function secondarily to modify the primary signals generated by JNK, p38, vitamin D, and STAT activation. Once recruited to these partners, Smads may then modulate the transcriptional activity at these elements by recruiting coactivators (co-A, green) or corepressors (co-R, red).

Other Smad partners are DNA binding proteins that are directly regulated by other extracellular stimulae. Furthermore, in most cases, efficient transcriptional cooperation with Smad signaling requires some activation of the DNA binding partner by other pathways. In these cases, the Smads do not function as the primary signal but require other inputs to regulate the activity of these pathways. The Smads then function as components of a secondary signaling system that modifies the output of the primary signal by recruiting additional coactivators or corepressors (Figs. 9 and 10). In these examples, it might be useful to think of Smads as general transcriptional comodulators that are controlled through regulation of their subcellular localization. Because the subcellular localization of Smads is directly regulated by serine-threonine kinase receptors, the nuclear concentration of Smads can provide a precise readout of the extracellular concentration of TGF-β family ligand. In this context, Smads as comodulators would play important roles in positively or negatively regulating preexisting gene expression patterns, depending on whether they recruit coactivators or corepressors to particular DNA binding partners. During development, and particularly during organogenesis, such a mechanism would allow a cell to interpret its position within a morphogenetic gradient and modify its gene expression program appropriately.

In the 1980s a host of studies showed that TGF-β could elicit very different biological responses that were dependent on the cellular context (64). The mechanisms underlying these observations were never clear. However, we now understand that a variety of signaling pathways can crosstalk with Smads to control nuclear concentrations of Smads through regulation of R-Smad activation, steady-state protein levels, and translocation. This, coupled with the ability of Smads to function in the nucleus as transcriptional comodulators for a number of DNA binding partners, creates a dynamic pathway in which the activity is highly dependent on the extracellular environment of the cell. Thus, the investigation of crosstalk in the Smad pathway is beginning to shed light on how TGF-β family biology is dependent on cellular environment. It will be interesting to see what other pathways intersect with Smad signaling and how they do so.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
View Abstract

Stay Connected to Science Signaling

Navigate This Article