Divergence and convergence of TGF-β/BMP signaling


  • Kohei Miyazono,

    Corresponding author
    1. Department of Molecular Pathology, Graduate School of Medicine, University of Tokyo, Hongo, Bunkyo-ku, Tokyo Japan
    2. Department of Biochemistry, The JFCR Cancer Institute, Kami-ikebukuro, Toshima-ku, Tokyo, Japan
    • Department of Biochemistry, The JFCR Cancer Institute, 1-37-1 Kami-ikebukuro, Toshima-ku, Tokyo 170-8455, Japan.
    Search for more papers by this author
  • Kiyoshi Kusanagi,

    1. Department of Biochemistry, The JFCR Cancer Institute, Kami-ikebukuro, Toshima-ku, Tokyo, Japan
    Search for more papers by this author
  • Hirofumi Inoue

    1. Department of Biochemistry, The JFCR Cancer Institute, Kami-ikebukuro, Toshima-ku, Tokyo, Japan
    Search for more papers by this author


The transforming growth factor-β (TGF-β) superfamily includes more than 30 members which have a broad array of biological activities. TGF-β superfamily ligands bind to type II and type I serine/threonine kinase receptors and transduce signals via Smad proteins. Receptor-regulated Smads (R-Smads) can be classified into two subclasses, i.e. those activated by activin and TGF-β signaling pathways (AR-Smads), and those activated by bone morphogenetic protein (BMP) pathways (BR-Smads). The numbers of type II and type I receptors and Smad proteins are limited. Thus, signaling of the TGF-β superfamily converges at the receptor and Smad levels. In the intracellular signaling pathways, Smads interact with various partner proteins and thereby exhibit a wide variety of biological activities. Moreover, signaling by Smads is modulated by various other signaling pathways allowing TGF-β superfamily ligands to elicit diverse effects on target cells. Perturbations of the TGF-β/BMP signaling pathways result in various clinical disorders including cancers, vascular diseases, and bone disorders. © 2001 Wiley-Liss, Inc.

Transforming growth factor-β (TGF-β) is a potent growth inhibitor for a wide variety of cells including epithelial cells, vascular endothelial cells, hematopoietic cells, and immune lymphocytes (Roberts and Sporn, 1990; Miyazono et al., 1994; Blobe et al., 2000). Perturbations of the TGF-β signaling pathways result in loss of cell growth regulation which is one of the most crucial steps in oncogenesis. TGF-β is also a potent inducer of tissue fibrosis, which may provide a microenvironment suitable for growth of transformed cells.

Three isoforms of TGF-β, i.e. TGF-β1, TGF-β2, and TGF-β3, with similar structures and in vitro biological activities have been identified in mammals (Roberts and Sporn, 1990). Many other proteins also have structures essentially similar to TGF-β and are collectively referred to as the TGF-β superfamily. The TGF-β superfamily includes more than 30 proteins in mammals, e.g. activins, bone morphogenetic proteins (BMPs), and anti-Müllerian hormone (AMH, also termed Müllerian inhibiting substance or MIS) (Kawabata and Miyazono, 2000). Growth/differentiation factors (GDFs) also belong to the TGF-β superfamily. Some of them, including GDF-5, GDF-6, and GDF-7, are BMP-like proteins, but others have been only partially characterized.

In this review article, we discuss some recent progress in research on signal transduction by the TGF-β superfamily proteins, focusing on how a wide variety of ligands bind to limited number of receptors, and how the receptors and their signal transducer Smads transmit diverse signals in target cells. We also discuss how these molecules are linked to the pathogeneses of various clinical disorders.


As described above, the TGF-β superfamily includes more than 30 members in mammals. Two questions arise here; first is that why are there so many members of the TGF-β superfamily? and second that is it possible to classify these proteins into several groups?

Proteins of the TGF-β superfamily bind to two different types of signaling receptors termed as type II and type I receptors. Although some other cell surface proteins have been shown to bind TGF-β, type II and type I receptors are most important for TGF-β superfamily signaling (Massagué, 1998). Both type II and type I receptors contain serine/threonine kinase domains in their intracellular portions and exist on cell surface in various oligomeric forms, e.g. type II homomers, type I homomers, and type II-type I heteromers (Gilboa et al., 2000). The type II receptor kinases are constitutively active without ligand stimulation. Upon ligand binding and formation of type II and type I receptor complexes, followed by possible receptor conformational changes, type I receptors are phosphorylated and activated by type II receptor kinases. Type I receptor kinases then transmit intracellular signals by phosphorylating Smad proteins.

In mammals, only five type II receptors and seven type I receptors have been identified (Miyazono et al., 2000). The type II receptors include activin type II and type IIB receptors (ActR-II and ActR-IIB), TGF-β type II receptor (TβR-II), BMP type II receptor (BMPR-II), and AMH type II receptor (AMHR-II). Type I receptors are termed activin receptor-like kinases (ALKs) 1 through 7. It is theoretically possible to form more than 30 different combinations of type II and type I receptors. However, certain type II receptors tend to interact with certain type I receptors. Thus, the combinations of type II and type I receptors appear to be limited under physiological conditions and the variety of ligands converge at the receptor level.

The intracellular substrates, Smads, can be classified into three groups, i.e. receptor-regulated Smads (R-Smads), common-mediator Smads (Co-Smads), and inhibitory Smads (I-Smads) (Heldin et al., 1997). Specificity of the intracellular signals is determined by R-Smads, whereas Co-Smad serves as an adaptor molecule for R-Smads. In contrast, I-Smads antagonize signaling by R-Smads and Co-Smads. Five different R-Smads have been identified in mammals. Smad2 and Smad3 are activated by TGF-βs and activins (they are collectively referred to as type A R-Smads or AR-Smads in this review) whereas Smad1, Smad5 and Smad8 are activated by BMPs (referred to as type B R-Smads or BR-Smads). Thus, the variety of R-Smad signaling is also limited.

R-Smads are either AR-Smads or BR-Smads, but R-Smads in each subgroup are not equally activated by their cognate receptors. ALK-4 (also termed activin type IB receptor or ActR-IB), and ALK-5 (also known as TGF-β type I receptor or TβR-I) are the functional type I receptors for activins and TGF-βs, respectively. Although, Smad2 and Smad3 are structurally very similar to each other, they transmit distinct signals because of the differences in their ability to bind to specific DNA sequences, and possibly in other functions (Yagi et al., 1999). Smad2 is a signal transducer for activins during early embryogenesis (Nomura and Li, 1998; Waldrip et al., 1998), while Smad3 appears to be more important for TGF-β signaling in cell growth, movement and immune function in adult tissues (Ashcroft et al., 1999). In cultured keratinocytes, however, both Smad2 and Smad3 are activated by TGF-β, whereas only Smad3 is strongly phosphorylated by activin (Shimizu et al., 1998). Since Smad2 is more important than Smad3 in activin signaling during early embryogenesis, why Smad3 is preferentially activated by activin in keratinocytes remains to be elucidated.

Receptors and downstream signaling pathways for BMPs are also complex. ALK-3 and ALK-6, which are also termed BMP type IA and type IB receptors (BMPR-IA and BMPR-IB), respectively, are structurally similar to each other and function as BMP type I receptors by activating BR-Smads. In addition, ALK-2 binds certain BMPs, e.g. BMP-6 and BMP-7 (ten Dijke et al., 1994; Macías-Silva et al., 1998; Ebisawa et al., 1999). Specificity of the interaction of type I receptors with R-Smads is determined by a short region located between the kinase subdomains IV and V, the L45 loop, of type I receptor kinases (Feng and Derynck, 1997). The L45 loop of ALK-2 is less similar to those of ALK-3 and ALK-6 than those of ALK-4 and ALK-5 are. However, ALK-2 also activates BR-Smads, and transmits signals similar to those of ALK-3 and ALK-6 (Chen and Massagué, 1999). Interestingly, ALK-3 and ALK-6 activate all three BR-Smads, whereas ALK-2 can activate only Smad1 and Smad5, but not Smad8 (Ebisawa et al., 1999; Aoki et al., 2001). Functional differences between Smad8 and Smad1/5 have not been determined, but these BR-Smads have different motifs in their structures. For example, Smad1 and Smad5 have a PY motif in their linker regions, which is responsible for interaction with an E3 ubiquitin ligase, Smurf1 (Zhu et al., 1999), but Smad8 lacks this motif. Thus, Smad8 may be degraded by a mechanism distinct from that for Smad1 or Smad5. Differences in the activation patterns of BR-Smads by ALK-3/6 and ALK-2 may therefore generate different biological responses in target cells.

Non-Smad pathways may also be important in understanding the diversity of signals generated by the TGF-β superfamily proteins (Massagué 2000; Mulder, et al., 2000). MAP kinases, including ERK, JNK, and p38 MAP kinases, are activated by BMPs and TGF-βs in various cells, and might play important roles in TGF-β signaling in certain cell types. For example, the p38 MAP kinase plays crucial roles in the induction of chondrogenic phenotypes in chondroblastic cells and of apoptosis of lymphocytes (Nakamura et al., 1999; Kimura et al., 2000). However, activation of the MAP kinases by the TGF-β superfamily proteins depends on cell type and culture conditions, and these kinases are not always specifically activated by the TGF-β superfamily ligands (Massagué et al., 2000).


Since R-Smads are classified into two groups and are specifically activated by cognate type I receptors, ligands of the TGF-β superfamily can also be classified into two subgroups depending on whether they activate AR-Smads or BR-Smads (Fig. 1).

Figure 1.

Divergent members of the TGF -β superfamily, their receptors, and R-Smads. Factors activating AR-Smads bind to type II receptors including TβR-II, ActR-II, and ActR-IIB, and type I receptors, including ALK-4/ActR-IB, ALK-5/TβR-I, and possibly ALK-7. Factors activating BR-Smads bind to BMP-II, ActR-II, ActR-IIB, or AMHRII, and ALK2, ALK3/BMPR-IA or ALK-6/BMPR-IB. As an exception of this scheme, TGF-βs may bind to ALK-1 as well as ALK-5/TβR-I in endothelial cells, and ALK-1 activates BR-Smads.

All three TGF-β isoforms activate AR-Smads, since they bind TβR-II and TβR-I/ALK-5. Of four different isoforms of activins, biological activities of activin βC and activin βE have been poorly defined. They are specifically expressed in liver, but gene targeting of activin βC, activin βE or both did not result in any functional defects in liver development and function (Lau et al., 2000). Thus, the signaling activities of activins βC and βE remain to be determined, but they are thought to activate AR-Smads because of their structural similarity to activins βA and βB. Nodal plays important roles in mesoderm formation, anterior patterning, and left-right axis specification during early embryogenesis. Biochemical analyses and studies on crossing of different knock-out mice revealed that Nodal binds to ActR-IB and activates AR-Smads (Gu et al., 1999; Nomura and Li, 1998; Kumar et al., 2001). In addition to these ligands, certain GDFs may also activate AR-Smads. GDF-8, also termed myostatin, inhibits proliferation of skeletal muscle cells (Thomas et al., 2000). Null mutations of the GDF-8/myostatin gene result in dramatic increase in skeletal muscle mass in mice and cattle (Grobet et al., 1997; McPherron et al., 1997; McPherron and Lee, 1997). GDF-8/myostatin activates p3TP-Lux promoter-reporter construct (Paralkar V., personal communication), which preferentially responds to AR-Smads, suggesting that GDF-8/myostatin transmits signals similar to those of TGF-βs and activins.

BMP-2 and BMP-4, the prototype of BMPs, induce bone and cartilage tissues in vivo (Reddi, 1998). They bind to ALK-3/BMPR-IA and activate BR-Smads. BMP-6 and BMP-7, which are structurally similar to each other, bind ALK-2 and also activate BR-Smads (Ebisawa et al., 1999; Miyazono, 2001). GDF-5, an inducer of cartilage-like structures in vivo, preferentially binds to ALK-6/BMPR-IB in the presence of different type II receptors, and also activates BR-Smads (Nishitoh et al., 1996). Thus, BMPs and their related molecules activate BR-Smads. In addition to those BMP-like molecules, AMH also activates BR-Smads through binding to ALK-6/BMPR-IB in the presence of AMHR-II (Gouédard et al., 2000).

After the cloning of ALK-7 (Kang and Reddi, 1996; Ryden et al., 1996; Tsuchida et al., 1996), efforts to isolate novel serine/threonine kinase receptors have not been successful. It is possible that there are yet unidentified serine/threonine kinase receptors which may be specifically expressed in certain tissues. However, it is more likely that there are only limited numbers of type II and type I receptors for the TGF-β superfamily. On the other hand, several molecules have been demonstrated to serve as co-receptors or co-factors for the TGF-β superfamily ligands. Betaglycan (also known as TGF-β type III receptor) has been shown to bind inhibin as well as TGF-β, but not activin (Lewis et al., 2000). Nodal requires Xenopus EGF-CFC protein or its mammalian homologues Cripto and Cryptic for efficient receptor-binding and intracellular signaling (Shen and Schier, 2000). It is thus possible that these co-receptors and co-factors modulate the binding of ligands to signaling receptors and contribute to specific receptor binding of certain ligands.

Another important point is that certain members of the TGF-β superfamily compete with other members for receptor binding, and function as antagonists of these ligands. Lefty-1 and -2 and their Xenopus homologue Antivin play important roles in the determination of left-right asymmetry (Meno et al., 1999; Thisse and Thisse, 1999). Lefties have six conserved cysteine residues, and are structurally distantly related to members of the TGF-β superfamily. Although Lefty/Antivin may bind to activin receptor complexes, they do not transmit intracellular signals, and interfere with the signaling activity of Nodal.

The α-subunit of inhibin is also distantly related to other members of the TGF-β superfamily, and forms heterodimers with β-subunits which are components of activin dimers. Inhibins have been reported to function as antagonists of activins by binding to activin type II receptors through the β-subunits (Lebrun and Vale, 1997; Lewis et al., 2000). Certain BMPs have also been suggested to act as antagonists of other BMPs. However, further analyses are required to determine whether they actually function as BMP antagonists under physiological conditions.

Interaction of SMAD proteins with various partners

Since there are limited number of serine/threonine kinase receptors and Smads it is important to elucidate how the wide variety of biological activities of the TGF-β superfamily proteins is generated. Spatial and temporal expression patterns of ligands are important for exhibition of diverse biological activities. GDF-5 is expressed in appendicular skeleton and plays an important role in chondrogenesis (Francis-West et al., 1999). GDF-8/myostatin is preferentially expressed in skeletal muscles, and regulates the proliferation of myocytes (McPherron et al., 1997; Thomas et al., 2000). Certain GDFs/BMPs are specifically expressed in female and male reproduction systems, and play important roles in oogenesis and spermatogenesis (Elvin et al., 1999; Galloway et al., 2000). However, expression patterns of ligands do not fully explain why the TGF-β superfamily proteins have different biological effects in distinct cells.

TGF-β inhibits growth of most types of cells but induces growth of fibroblasts and osteoblasts in vitro (Urano et al., 1999). TGF-β inhibits the growth and most functions of lymphocytes but induces the synthesis of IgA in B lymphocytes (Cazac and Roes, 2000). BMPs induce differentiation of myoblasts into osteoblasts, while they induce differentiation of neuroepithelial cells into astrocytes in the presence of leukemia inhibitory factor (LIF) (Nakashima et al., 1999). BMP-like molecules are present in Drosophila, including Decapentaplegic (DPP), 60A-GBB, and Screw, but never induce bone formation in invertebrates. However, the DPP protein can induce formation of bone and cartilage tissues when subcutaneously injected into vertebrates. The question is that what could be the mechanisms for these diverse biological activities induced by the TGF-β superfamily proteins?

One possible mechanism is interaction of Smads with various proteins in the cytoplasm and in the nucleus in different fashions in various cells. Most importantly, Smads associate with various transcription factors in the nucleus and exhibit a wide array of biological activities in cooperation with these proteins in cells. Activity of Smads is also modified by signaling cross-talk with other signaling pathways (Miyazono et al., 2000), which may also lead to exhibition of diverse biological responses.

Proteins that interact with Smads are listed in Table 1. Smad-interacting proteins include those that regulate membrane anchoring of Smads, ubiquitin-dependent degradation, and transcription by Smads. In this section we focus on the function of Runx proteins. Runx is a transcription factor containing a Runt domain, which is important for DNA binding (Ito, 1999). Runx is also denoted polyomavirus enhancer binding protein-2 α subunit (PEBP2α), core binding factor-a (Cbfa), and acute myeloid leukemia (AML). Runx1, also known as PEBP2αB, Cbfa2, and AML1, plays an important role in definitive hematopoiesis. Abnormalities in the Runx1/AML1 gene induced by chromosomal translocation are found in approximately one-third of human acute leukemias. Germ-line mutations of the Runx1/AML1 gene are responsible for the human familial platelet disorder with predisposition to acute myelogenous leukemia (FPD/AML) (Song et al., 1999). Runx2 is also known as PEBP2αA and Cbfa1. Gene targeting of Runx2/Cbfa1 results in the loss of bone formation. Germ-line mutations of Runx2/Cbfa1 are responsible for the pathogenesis of human cleidocranial dysplasia syndrome (CCD). Runx3, also known as PEBP2αC and Cbfa3, induces IgA synthesis in B lymphocytes. These biological activities of Runx proteins are reminiscent of those induced by TGF-β and BMPs.

Permission to reproduce this table online was not granted by the copyright holder. Readers are kindly asked to refer to the printed version.

Physical interaction between Smads and Runx can be detected in mammalian cells (Hanai et al., 1999; Pardali et al., 2000; Zhang and Derynck, 2000; Zhang et al., 2000). In the immunoglobulin constant α region (Ig Cα) promoter region, Smad binding sites and Runx binding sites are located in close vicinity, therefore, the Smad-Runx complex binds to and enhances transcription from the IgCα promoter. Physical interaction and functional cooperation of Runx2 and BR-Smads have also been demonstrated (Zhang et al., 2000) indicating that these two proteins may cooperate in osteoblast differentiation. Importantly, expression of Runx1 is induced by BMPs and TGF-β in C2C12 cells. Runx1 prevents the differentiation of myoblasts into mature myocytes. However, Runx1 by itself is not sufficient for the induction of osteoblast differentiation; cooperation of BR-Smads with Runx1 is required for this process (Lee et al., 2000; Zhang et al., 2000).

TGF-β is a potent growth inhibitor of hematopoietic cells, but many leukemic cells are resistant to the growth inhibitory activity of TGF-β. Runx1/AML1 plays an important role in growth and differentiation of hematopoietic cells. A leukemia-associated fusion protein, AML1/ETO, has been shown to interact with AR-Smads, but fail to induce transcriptional responses induced by TGF-β (Jakubowiak et al., 2000). Abnormalities of the AML1 protein may thus play an important role in acquiring resistance to the effect of TGF-β on hematopoietic cells and development of leukemias.

Many other transcription factors interact with Smads. The forkhead transcription factor FAST1 specifically interacts with AR-Smads and plays an important role in formation of left-right asymmetry and mesoderm formation during early embryogenesis. Not much is known about transcription factors that interact with BR-Smads (Miyazono, 2001) but it has been shown that BR-Smads indirectly interact with STAT3 through a transcriptional coactivator, p300, leading to the differentiation of neuronal epithelial cells into astrocytes (Nakashima et al., 1999). Thus, by selecting various transcription factors as partners, Smads may be able to exhibit a wide range of biological activities in target cells.

An important unanswered question is that how growth inhibition is induced by TGF-β in target cells. BMPs and activins can inhibit the growth of certain cells but the potency and magnitude of growth inhibition they induce are less than those induced by TGF-β. TGF-β induces growth inhibition of various cells which is mediated by repression of c-myc, induction of p21 and p15 cyclin-dependent kinase (CDK) inhibitors, and suppression of tyrosine phosphatase Cdc25A (Massagué et al., 2000). However, it is still unclear why only TGF-β can induce potent growth inhibitory effects. Abnormalities of activins, activin receptors, BMPs, BMP receptors, or BR-Smads have not been identified in human cancer cells, except for a C-terminal truncation of Smad5 in certain leukemias (Jiang et al., 2000). Mutations in Smad4 have been identified in various tumors but Smad4 is shared with TGF-β/activin and BMP signaling pathways. These findings suggest a specific role of TGF-β signaling pathway in oncogenesis.

Modulation of SMAD signaling

In the Smad signaling pathways, R-Smads are activated by serine/threonine kinase receptors, form complexes with Co-Smads, and translocate into the nucleus where they regulate transcription of target genes. Smads play major roles as receptor substrates and transcription factors but their activity is modulated by multiple regulatory mechanisms (Miyazono, 2000).

Various extracellular antagonists including Noggin, Chordin, Cerberus, and Gremlin, regulate BMP signaling (Massagué and Chen, 2000; Miyazono, 2000). Follistatin is an inhibitor for activins but is also known to antagonize the effects of certain BMPs. Xenopus BAMBI (its mammalian homologue is the nma gene product) is a pseudoreceptor for the TGF-β superfamily members (Onichtchouk et al., 1999). BAMBI is structurally similar to type I serine/threonine kinase receptors in the extracellular domain but lacks the intracellular kinase domain. By forming heteromeric complexes with serine/threonine kinase receptors BAMBI/Nma antagonizes the effects of TGF-β/BMPs.

I-Smads, i.e. Smad6 and Smad7, function as antagonists of TGF-β/BMP signaling inside cells. I-Smads stably interact with activated type I receptors and compete with R-Smads for receptor activation. In addition, Smad6 has been reported to form a complex with Smad1 and to compete with Smad4 in complex formation. Smad6 was also reported to interact with a homeobox protein, Hoxc8, and to act as a transcriptional repressor in the nucleus (Bai et al., 2000).

ERK MAP kinase interferes with Smad signaling under certain conditions. ERK1/2, activated by tyrosine kinase receptors or oncogenic Ras, phosphorylates linker regions of R-Smads, resulting in interference with nuclear translocation of R-Smads (Kretzschmar et al., 1999), although the molecular mechanisms of this have not been elucidated.

In addition to those molecules which primarily act in the cytoplasm certain molecules act as transcriptional co-repressors in the nucleus (Massagué and Chen, 2000 Miyazono, 2000). c-Ski and its related protein, SnoN, and TGIF interact with AR-Smads in the nucleus. They compete with transcriptional co-activators p300 and CBP, and recruit histone deacetylases to Smad complexes, resulting in transcriptional repression. TGF-β/Smad signaling is thus regulated by multiple mechanisms at extracellular, cytoplasmic and nuclear levels.

Importantly, expression of these regulatory molecules is regulated via negative feedback mechanisms (Miyazono, 2000). For example, expression of Noggin and BAMBI/Nma is regulated by TGF-β and BMPs. I-Smads are strongly induced by ligand stimulation. In addition, Smad7 is induced by various signals other than TGF-β/BMPs including NF-κB and interferon-γ signaling. Thus, cells treated with interleukin-1 or interferon-γ become resistant to the effects of TGF-β and possibly those of BMPs.

TGF-β/BMP signaling and diseases

Perturbations of TGF-β superfamily signaling result in various clinical disorders including some cancers, bone diseases, and vascular diseases. Involvement of TGF-β receptors and Smads in tumorigenesis has been discussed in several other review articles (Blobe et al., 2000; Massagué et al., 2000). The relationship of BMP signaling with bone diseases has also been discussed elsewhere (Reddi, 1998; Luyten et al., 2000). TGF-β has also been reported to be involved in some bone diseases. Recently, mutations in the N-terminal region of TGF-β1 precursor and hyperactivation of latent TGF-β have been found to result in Camurati-Engelmann disease (Janssens et al., 2000, Kinoshita et al., 2000). In this article, we show only a summary figure of the diseases induced by abnormalities of TGF-β/BMP signaling (Fig. 2) and focus on genetically-inherited vascular diseases induced by mutations of the TGF-β superfamily receptors.

Figure 2.

Signaling by TGF-β superfamily ligands and diseases induced by mutations of the molecules involved in the signaling pathways. In addition to the diseases listed in this Figure, left-right laterality defects are induced by abnormalities of Lefty-2, Cryptic, or ActR-IIB (Kosaki K et al., 1999; Kosaki R et al., 1999; Bamford et al., 2000). References that described the diseases listed in this figure are as follows: TGF-β (Yamada et al., 1998; El-Gamel et al., 1999; Li et al., 1999; Bertoldo et al., 2000; Suthanthiran et al., 2000; Yokota et al., 2000); AMH /MIS and AMH/MIS type II receptor (Belville et al., 1999); GDF-5/CDMP-1 (Thomas et al., 1996; Polinkovsky et al., 1997; Thomas et al., 1997); Noggin (Gong et al., 1999); TβR-II (Markowitz et al., 1995; Myeroff et al., 1995; Kim et al., 1996a; Knaus et al., 1996; Togo et al., 1996; Vincent et al., 1996; de Jonge et al., 1997; Izumoto et al., 1997; McCaffrey et al., 1997; Goggins et al., 1998; Kang et al., 1998; Tanaka et al., 2000; Wang et al., 2000); TβR-I (Kim et al., 1996b; DeCoteau et al., 1997; Chen et al., 1998; Anbazhagan et al., 1999; Schiemann et al., 1999); Smad4/DPC4 (Hahn et al., 1996a; Hahn et al., 1996b; Nagatake et al., 1996; MacGrogan et al., 1997; Nishizuka et al., 1997; Howe et al., 1998; Bartsch et al., 1999; Schutte, 1999; Takakura et al., 1999; Woodford-Richens et al., 2000); Smad2 (Barrett et al., 1996; Eppert et al., 1996; Uchida et al., 1996; Yakicier et al., 1999); endoglin (McAllister et al., 1994; Pece-Barbara et al., 1999); ALK-1 (Johnson et al., 1996; Abdalla et al., 2000); BMPR-II (Deng et al., 2000; Lane et al., 2000; Thomson et al., 2000); TGIF (Gripp et al., 2000); CBP (Petrij et al., 1995; Borrow et al., 1996; Taki et al., 1997); p300 (Ida et al., 1997); Evi-1 (Pekarsky et al., 1997; Suzukawa et al., 1999); Runx (Erickson et al., 1992; Nucifora et al., 1994; Golub et al., 1995; Romana et al., 1995; Song et al., 1999; Zhou et al., 1999; Preudhomme et al., 2000).

BMPR-II is a type II receptor that specifically bind BMPs. BMPR-II is ubiquitously expressed in various tissues, and mice lacking the BMPR-II gene die during early embryogenic stages due to abnormal mesoderm formation (Beppu et al., 2000). Although the bioactivities of BMPs in vascular wall cells have not been fully determined, BMP-2 has been shown to inhibit the growth of smooth muscle cells and to prevent progression of vascular proliferative diseases (Nakaoka et al., 1997). Mutations in the human BMPR-II gene have been shown to be involved in the pathogenesis of primary pulmonary hypertension (PPH) (Deng et al., 2000; Lane et al., 2000). Familial PPH is inherited in autosomal dominant fashion with low penetrance. In PPH patients, obstruction of pre-capillary pulmonary arteries is observed due to proliferation of endothelial cells and smooth muscle cells. Lung-heart or lung transplantation is an effective treatment for the disease. Interestingly, mutations can be observed in various regions of the human BMPR-II gene including the extracellular domain and intracellular domain. BMPR-II has a unique, long C-terminal tail with approximately 530 amino acid residues which is not observed in other type II or type I receptors (Rosenzweig et al., 1995). The functional importance of the C-terminal tail of BMPR-II has not been reported (Ishikawa et al., 1995) but truncations of the C-terminal tail also lead to development of PPH. It will be interesting to examine whether certain BMP-like molecules specifically bind to BMPR-II in lung and regulate the function of pulmonary arteries.

Mutations of endoglin and ALK-1 are responsible for the pathogenesis of hereditary hemorrhagic telangiectasia (HHT) type I and type II, respectively (also known as Osler-Rendu-Weber syndrome) (McAllister et al., 1994; Johnson et al., 1996). HHT is characterized by arteriovenous malformations and recurrent bleeding due to vascular dysplasia. ALK-1 is a type I receptor specifically expressed in endothelial cells. ALK-1 is structurally similar to ALK-2 and has been shown to bind TGF-βs in the presence of TβR-II (Oh et al., 2000). However, it is possible that other members of the TGF-β superfamily bind to ALK-1 under certain conditions. In endothelial cells, two different type I receptors, i.e., ALK-1 and ALK-5, may serve as receptors for TGF-βs. ALK-5/TβR-I activates AR-Smads, while ALK-1 phosphorylates BR-Smads. Thus, TGF-βs activate AR-Smads in most cell types, but phosphorylate both types of R-Smads in endothelial cells.

Structurally endoglin is weakly related to betaglycan. Betaglycan and possibly endoglin may function as co-receptors for the TGF-β superfamily proteins (Pece-Barbara et al., 1999). Endoglin may regulate ligand binding to signaling receptors and is specifically expressed in endothelial cells. In human umbilical endothelial cells, endoglin physically interacts with ALK-1 (Abdalla et al., 2000). How endoglin and ALK-1 cooperate in transduction of TGF-β/BMP signaling remains to be determined.

Analyses by gene targeting have revealed that both ALK-1 and endoglin null mice die during embryogenesis due to abnormalities in vascular development which are reminiscent of those in human HHT (Li et al., 1999; Oh et al., 2000; Urness et al., 2000). Both types of mice exhibit enlarged vasculatures lacking smooth muscle cells surrounding endothelial cells. Smad5−/− mice also exhibit defects in vascular tissues as well as other tissues (Chang et al., 1999; Yang et al., 1999). The vascular abnormalities in Smad5−/− mice are similar to those in ALK-1−/− mice, suggesting that Smad5 is a downstream signaling component of ALK-1. Mice lacking TGF-β1 or TβR-II also die of vascular abnormalities and anemia. However, the phenotypes of these mice are distinct from those of ALK-1 and endoglin null mice. Thus, TGF-β and TβR-II may act upstream of the endoglin/ALK-1/Smad5 signaling pathway but other pathways, i.e., the ALK-5/TβR-I and Smad2/3 pathway also play important roles in the vascular development induced by TGF-β and TβR-II.


The large number of TGF-β superfamily proteins is notable. Despite this, however, limited numbers of serine/threonine kinase receptors and Smad proteins are present in mammals and are sufficient for transmitting diverse intracellular signaling. Importantly, the TGF-β/BMP signaling is regulated by various mechanisms at extracellular, membrane, cytoplasmic, and nuclear levels. TGF-β ligands, receptors, and Smad proteins have been reported to be involved in the pathogeneses of various clinical diseases. In addition, molecules that regulate TGF-β/BMP signaling are also involved in the pathogeneses of various diseases. Thus, in order to elucidate the roles of TGF-β/BMPs in clinical disorders it is very important to understand the signaling mechanisms of those proteins in vivo. Some recently developed technologies, e.g. DNA microarray and proteome analyses may in the future be helpful in determining the signaling mechanisms of TGF-β/BMPs in vivo.


We are grateful to Dr. V Paralkar for sharing unpublished observations. We also thank Ayako Sakai-Nishitoh for secretarial help.