Members of the transforming growth factor-β (TGF-β) superfamily bind to two different serine/threonine kinase receptors, i.e. type I and type II receptors. Upon ligand binding, type I receptors specifically activate intracellular Smad proteins. R-Smads are direct substrates of type I receptors; Smads 2 and 3 are specifically activated by activin/nodal and TGF-β type I receptors, whereas Smads 1, 5 and 8 are activated by BMP type I receptors. Nearly 30 proteins have been identified as members of the TGF-β superfamily in mammals, and can be classified based on whether they activate activin/TGF-β-specific R-Smads (AR-Smads) or BMP-specific R-Smads (BR-Smads). R-Smads form complexes with Co-Smads and translocate into the nucleus, where they regulate the transcription of target genes. AR-Smads bind to various proteins, including transcription factors and transcriptional co-activators or co-repressors, whereas BR-Smads interact with other proteins less efficiently than AR-Smads. Id proteins are induced by BR-Smads, and play important roles in exhibiting some biological effects of BMPs. Understanding the mechanisms of TGF-β superfamily signalling is thus important for the development of new ways to treat various clinical diseases in which TGF-β superfamily signalling is involved.
Introduction: the transforming growth factor-β (TGF-β) superfamily
Cytokines of the TGF-β superfamily are dimeric proteins with conserved structures, and have pleiotropic functions in vitro and in vivo (Kawabata et al. 1998a). The TGF-β superfamily includes nearly 30 proteins in mammals, e.g. TGF-βs, activins and inhibins, nodal, myostatin, bone morphogenetic proteins (BMPs), growth/differentiation factors (GDFs), and anti-Müllerian hormone (AMH, also called Müllerian inhibiting substance or MIS) (Fig. 1). TGF-β/BMP-like proteins are found in various species, including Xenopus, Caenorhabditis elegans and Drosophila melanogaster.
TGF-βs are the prototype of the TGF-β superfamily (Massagué 1998). They act as potent growth inhibitors for most types of cells, and induce the apoptosis of epithelial cells. In addition, TGF-βs stimulate the production of extracellular matrix proteins, inducing fibrosis in various tissues in vivo. Inhibins were originally identified as cytokines that inhibit the secretion of follicle-stimulating hormone (FSH) from the pituitary gland. Inhibins are composed of disulphide-bonded dimers of inhibin-α and -β chains. In contrast, dimers composed of inhibin-β chains (also called activin-β chains) are activins, which stimulate the production of FSH by the pituitary gland. Activins also play important roles in the induction of dorsal mesoderm during early embryogenesis. In addition, they act on epithelial cells and haematopoietic cells and regulate their growth, differentiation and apoptosis. Nodal plays critical roles in the induction of dorsal mesoderm, anterior patterning, and formation of left–right asymmetry. Myostatin, also known as GDF-8, is produced by cells of the skeletal muscle lineage, and inhibits their growth. Although these TGF-β superfamily proteins (Fig. 1; coloured in red and orange) have divergent biological effects in vivo, they transmit similar, although not identical, intracellular signals in target cells.
BMPs were originally identified as cytokines that induce bone and cartilage tissues in vivo (Rosen & Wozney 2002). BMPs were then shown to have diverse effects on various cells, and to induce the ventral mesoderm during early embryogenesis. GDF-5, also known as cartilage-derived morphogenetic protein 1 (CDMP-1), is structurally related to BMPs, and induces cartilage-like tissue in vivo. AMH induces Müllerian duct regression during embryogenesis and inhibits the transcription of gonadal steroidogenic enzymes. The biological effects of AMH appear to differ from those of BMPs and GDF-5; however, the intracellular signals activated by these cytokines (Fig. 1; coloured in dark and light blue) have been shown to be similar.
In this review, we discuss the signalling pathways activated by members of the TGF-β superfamily. Although the biological effects of these cytokines appear to be complicated, two major pathways involving Smad proteins are activated by members of the TGF-β superfamily (Miyazono et al. 2001). Accordingly, the TGF-β superfamily cytokines can be classified into two subfamilies depending on the Smad signalling pathways they activate. Understanding the mechanisms of TGF-β superfamily signalling will aid in the development of new strategies for the treatment of various clinical diseases to which the TGF-β superfamily signalling is tightly linked.
Receptors for TGF-β superfamily proteins
Members of the TGF-β superfamily bind to two distinct receptor types, known as type II and type I receptors (Wrana et al. 1994; Heldin et al. 1997). Both type II and type I receptors are required for signal transduction. In addition, some cell surface proteins, including betaglycan (also known as TGF-β type III receptor), endoglin, and the EGF-CFC family proteins, containing a divergent EGF-like motif and a novel cysteine-rich CFC motif, act as co-receptors for certain members of the TGF-β superfamily.
Both type II and type I receptors contain serine/threonine kinase domains in their intracellular portions. The type II receptor kinases are constitutively active; upon ligand binding, hetero-tetrameric complexes composed of two molecules each of type II and type I receptors are formed (Kirsch et al. 2000; Qin et al. 2002). In the tetrameric receptor complexes, type II receptor kinases transphosphorylate the GS domain of type I receptors, which are located between the transmembrane domain and the kinase domain of type I receptors (Fig. 2). Following phosphorylation of the GS domain, type I receptor kinases are activated and phosphorylate intracellular substrates. Thus, type I receptors act as downstream components of type II receptors in the signalling pathways, and determine the specificity of the intracellular signals induced by the TGF-β superfamily cytokines.
Five type II receptors and seven type I receptors are present in mammals. The TGF-β type II receptor (TβR-II) is the specific receptor for TGF-βs. Activin type II and type IIB receptors (ActR-II and ActR-IIB) serve as type II receptors for activins, but are shared with other TGF-β superfamily members, including nodal and BMPs. BMP type II receptor (BMPR-II) and AMH type II receptor (AMHR-II) specifically bind to BMPs and AMH, respectively.
Of the seven type I receptors, the activin receptor-like kinases (ALK)-4, -5, and -7 are structurally similar to each other. ALK-5, also known as TβR-I, serves as a specific receptor for TGF-βs. ALK-4 binds actvins and nodal, whereas ALK-7 binds only nodal (Reissmann et al. 2001). Receptors for myostatin have not yet been determined. ALK-3 and ALK-6 (also termed BMPR-IA and BMPR-IB, respectively) are structurally similar, and function as BMP type I receptors. In addition, ALK-2, which is distantly related to ALK-3 and -6, acts as a type I receptor for certain BMPs, including BMP-6 and -7 (Macías-Silva et al. 1998; Aoki et al. 2001). ALK-1 is structurally very similar to ALK-2; in contrast to ALK-2, however, ALK-1 is expressed in endothelial cells, and serves as a type I receptor for TGF-β (see below; Oh et al. 2000).
Signalling by three types of Smad proteins
Smad proteins are major signalling molecules acting downstream of the serine/threonine kinase receptors (Heldin et al. 1997; Moustakas et al. 2001). Smads are classified into three subclasses, i.e. receptor-regulated Smads (R-Smads), common-partner Smads (Co-Smads), and inhibitory Smads (I-Smads). R-Smads are further divided into two subclasses; Smad2 and Smad3 (referred to as activin/TGF-β activated R-Smads or AR-Smads in this review) are activated by activin, nodal and TGF-β type I receptors, ALK-4, -5 and -7. Smad1, Smad5 and Smad8 (referred to as BMP activated R-Smads; BR-Smads) are activated by BMP type I receptors, ALK-3 and ALK-6. Smads 1 and 5 are also activated by ALK-1 and -2, although they are less similar to ALK-3/6 than the ALK-4/5/7 are. In contrast, only Smad4, which was originally isolated as the product of the tumour suppressor gene DPC4 (deleted in pancreatic carcinoma, locus 4; Hahn et al. 1996), serves as the Co-Smad in mammals. Smad6 and Smad7 function as I-Smads.
Smads are composed of the N-terminal Mad homology (MH1) domain, followed by linker regions and the C-terminal MH2 domains. MH2 domains are conserved in all three subclasses of Smads, whereas MH1 domains are conserved in R-Smads and Co-Smads, but not in I-Smads. The amino acid sequences of linker regions diverge between Smads. In addition, R-Smads have Ser-Ser-X-Ser motifs in their most C-terminal parts, which are phosphorylated by type I receptors.
R-Smads are directly phosphorylated and activated by type I receptor kinases. Without receptor activation, the MH1 and MH2 domains are physically associated with each other, and R-Smads are anchored as dimers to the plasma membrane through SARA and other molecules (Fig. 2, Table 1) (Qin et al. 2002). Following the receptor activation and phosphorylation of R-Smads, the interaction between the MH1 and MH2 domains is disrupted, and R-Smads form hetero-oligomers with the Co-Smad through their MH2 domains. Although the exact structures of the Smad oligomers have not been fully determined, and hetero-trimer and hetero-dimer models have been suggested (Kawabata et al. 1998b; Qin et al. 2001; Wu et al. 2001), Qin et al. (2001) proposed that the R-Smad-Co-Smad complexes are hetero-trimers, containing two R-Smad molecules and one Co-Smad molecule. The R-Smad-Co-Smad complexes translocate into the nucleus and regulate the transcription of target genes.
In the nucleus, R-Smads and Co-Smad interact with various DNA-binding proteins (Table 1), which bind to promoter regions of target genes together with the Smads. In addition, R-Smads and Co-Smad bind to transcriptional co-activators and co-repressors, which induce the acetylation and de-acetylation of histones, respectively, and play important roles in transcriptional regulation (Miyazono et al. 2000).
I-Smads associate with activated type I receptors and prevent the activation of R-Smads by the receptors. In addition, Smad6 interacts with activated R-Smads and interferes with the formation of a complex with Smad4. Smad7 inhibits both TGF-β activin and BMP signalling, whereas Smad6 preferentially inhibits BMP signalling (Hanyu et al. 2001). The N-terminal domain of Smad7 physically interacts with its MH2 domain, and plays an important role in the inhibition of TGF-β signalling (Hanyu et al. 2001). The HECT type E3 ubiquitin ligases, Smurf1 and Smurf2, support the inhibitory activities of I-Smads on TGF-β superfamily signalling. When Smurfs bind to Smad7, they induce the nuclear export of Smad7, and enhance the degradation of TGF-β receptors, resulting in the inhibition of TGF-β signalling (Kavsak et al. 2000; Ebisawa et al. 2001). Since I-Smads are induced by various signals, including TGF-β and BMPs, the levels of expression of I-Smads modulate the magnitude and duration of TGF-β superfamily signalling.
Signalling by AR-Smads and BR-Smads
The MH1 domains of Smads are responsible for their DNA-binding. One of the AR-Smads, Smad3, and Co-Smad, Smad4, directly bind to ‘CAGA’ and related DNA sequences, and activate the transcription of target genes together with DNA-binding proteins (Dennler et al. 1998; Zawel et al. 1998). Smad2 does not directly bind DNA, but by forming complexes with Smad3 and/or Smad4, it is also able to bind to the CAGA sequence. In addition, AR-Smads bind to certain other DNA sequences, e.g. the TGF-β inhibitory element (TIE), which is present in the promoter regions of c-myc (Chen et al. 2001; Yagi et al. 2002). In contrast, BR-Smads do not efficiently bind the CAGA sequence, but bind other sequences, e.g. GC-rich sequences, with low affinities (Ishida et al. 2000; Kusanagi et al. 2001). Smad4 binds the CAGA sequence in vitro, but does not efficiently activate the transcription of reporter genes containing CAGA sequences when complexed with BR-Smads. Exceptionally, however, a promoter-reporter construct containing the CAGAC sequence from the junB gene strongly responds to TGF-β, as well as to BMPs (Jonk et al. 1998). In general, if some ligands activate AR-Smads and transduce activin/TGF-β-like signals, they strongly transactivate the promoter-reporter constructs containing CAGA sequences. If some ligands activate BR-Smads and transmit BMP-like signals, they activate reporters containing GC-rich sequences. Members of the TGF-β superfamily can be classified based on their abilities to activate CAGA-reporters or GC-rich reporters, even if their receptors have not been determined.
TGF-βs bind to ALK-5, and activins and nodal bind to ALK-4 and/or -7; these ligands thus activate AR-Smads (Fig. 2A). Although the receptors for myostatin have not been determined, it activates CAGA-reporters (Zimmers et al. 2002), and may thus transduce signals similar to actvins and TGF-βs through AR-Smads. In contrast, most BMPs, including BMP-2, BMP-4, BMP-6, BMP-7 and GDF-5, activate BR-Smads (Aoki et al. 2001). AMH binds to ALK-2 and ALK-6, thus activating BR-Smads (Fig. 2B) (Gouédard et al. 2000; Visser et al. 2001; Clarke et al. 2001). BMP-3 was originally isolated as a member of the BMP family. BMP-3, however, does not have the ability to induce bone formation in vivo, but transmits activin/TGF-β-like signals through ActR-II and ALK-4 (Daluiski et al. 2001).
Certain members of the TGF-β superfamily, including some GDFs, have been isolated by virtue of their amino acid sequence similarities with other members of the TGF-β superfamily. Receptors and downstream signals have not been determined for most of these ligands (Fig. 1); it will be important to determine whether they activate AR-Smads or BR-Smads to perform their activities. Although the inhibin-α chain is a member of the TGF-β superfamily, it is not known whether inhibins display their effects by antagonizing activins, or by binding to specific signalling receptors (Bernard et al. 2002). Inhibins bind to betaglycan; however, it is not known whether they bind to serine/threonine kinase receptors.
Although both TGF-βs and activins activate Smad2 and Smad3, their biological activities are in most cases not identical. Analyses by gene targeting in mice have suggested that Smad2 plays important roles in embryogenesis, whereas Smad3 is of critical importance in cell growth and migration after birth (Nomura & Li 1998; Zhu et al. 1998; Yang et al. 1999; Datto et al. 1999). Thus, the effects of Smad2 may be more closely linked to activins, and those of Smad3 to TGF-βs. It is not currently known whether ALK-4 and ALK-5 differentially activate Smad2 and Smad3 in vivo. It will also be important to determine whether Smad-independent pathways are involved in the signalling pathways of TGF-β and other factors.
In classifying TGF-β superfamily cytokines based on their intracellular signals, an important exception is observed in TGF-β signalling in endothelial cells (Fig. 2C). TGF-β binds to ALK-5 in almost all TGF-β responsive cells, and transmits typical TGF-β-like signals. In addition, however, TGF-β binds to ALK-1 in endothelial cells and possibly in certain other types of cells (Oh et al. 2000). ALK-1 is structurally similar to ALK-2 and activates BR-Smads. TGF-β is thus able to transmit intracellular signals through AR-Smads and BR-Smads in endothelial cells. The balance of activation of ALK-5/AR-Smads and ALK-1/BR-Smads determines the activation status of endothelial cells in the development and homeostasis of vascular tissues (Goumans et al. 2002).
Targets of AR-Smads and BR-Smads
AR-Smads interact with various DNA-binding proteins (Table 1), which play important roles in exhibiting multiple biological activities of TGF-βs and activins. In contrast, only a few DNA-binding partners of BR-Smads have been identified, and some of them are shared with AR-Smads. Similarly, the transcriptional co-activators p300 and CBP, and the co-repressors c-Ski and SnoN, interact strongly with AR-Smads, but only weakly with BR-Smads. Why do BR-Smads interact with only a limited number of proteins compared to AR-Smads? As shown in Fig. 3, the surface of the trimer composed of the MH2 domain of Smad3 is rich in charged amino acid residues. The N-terminal upper side is rich in acidic amino acids, and the lower side is rich in basic amino acids. In contrast, the surface of the Smad1 MH2 trimer is much less charged than that of the Smad3 MH2 trimer. Although the Smad1/3 MH2 domain may interact with other proteins through hydrophobic amino acid residues in protein–protein interaction surfaces, this finding suggests that Smad3 MH2 is able to interact with other proteins more efficiently than Smad1 MH2.
Through interactions with various DNA-binding partners, AR-Smads transactivate various target genes, including plasminogen activator inhibitor-1 (PAI-1), type I collagen, junB, Smad7 and Mix.2. For inhibition of cell growth by TGF-β, AR-Smads induce the transcription of cyclin-dependent kinase (CDK) inhibitors p21 and p15 (Fig. 2A). In addition, Smad3 binds directly to the promoter region of c-myc through TIE, and represses the transcription of c-myc. In contrast, only a few target genes for BMPs have been identified, including Id (inhibitor of differentiation or inhibitor of DNA-binding) 1 through 3, Smad6, Vent-2, and Tlx-2 (Fig. 2B). Id proteins, however, appear to play important roles in the exhibition of the multiple biological activities of BMPs (Miyazono & Miyazawa 2002).
Id proteins act as negative regulators of cell differentiation and positive regulators of cell proliferation (Norton et al. 1998; Yokota & Mori 2002). Four Id proteins, Id1 through Id4, have been identified in mammals. They have overlapping profiles of expression, and elicit similar, but not identical, biological activities. Id proteins have a helix-loop-helix (HLH) dimerization domain, which binds to ubiquitously expressed transcription factors containing the basic HLH (bHLH) domain (Fig. 2B). Ubiquitously expressed bHLH transcription factors interact with tissue-specific bHLH transcription factors, and these complexes activate the transcription of genes containing an E-box in their promoter regions. In contrast to bHLH transcription factors, Id proteins lack the basic region responsible for binding to DNA. Id proteins thus sequester ubiquitous bHLH transcription factors and inhibit their transcriptional activities. Tissue-specific bHLH transcription factors include MyoD in muscle, and neurogenin and NueroD in neurones. Inhibition of the transcriptional activities of MyoD and neurogenin/NeuroD by Id proteins results in the inhibition of myogenesis and neurogenesis, respectively. These findings suggest that BMPs may elicit some of their biological activities through Id proteins (Nakashima et al. 2001; Goumans et al. 2002). In addition, Id2 interacts with the hypo-phosphorylated forms of Rb family proteins and inhibits their effects. Id2 may therefore play a role in the positive regulation of cell growth. BMP has been shown to both stimulate and inhibit the growth of cells, depending on cell type and culture conditions. Whether Id proteins play a role in the growth stimulation induced by BMPs is an important question to be answered in the future.
Conclusion: relation to clinical diseases
Members of the TGF-β superfamily exhibit various biological activities, and perturbations of their signalling are linked to certain clinical disorders. For example, TGF-β induces fibrosis in various tissues, including lung fibrosis, glomerulonephritis and liver fibrosis (Blobe et al. 2000). Inhibition of TGF-β signalling may therefore result in the prevention of fibrosis in these disorders. Recently, inhibitors of ALK-4, -5, and -7 kinases have been generated (Laping et al. 2002; Inman et al. 2002). SB-431542 inhibits ALK-4/5/7 signals, but not ALK-3/6 signals; it thus inhibits TGF-β signals without affecting BMP signals. However, these type I receptor kinase inhibitors should be used with care, since they may inhibit not only TGF-β, but also activins, nodal and myostatin. Understanding the signalling pathways of the TGF-β superfamily is therefore important for the clinical application of various agents which regulate the signalling activities of serine/threonine kinase receptors and Smads.
We thank Ms Hisako Hirano for preparing the list of references. This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan.