| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Departments of Pathology (H.C., M.M.M.), Molecular and Human Genetics (C.W.B., M.M.M.), Molecular and Cellular Biology (M.M.M.), and Pediatrics (C.W.B.), and Program in Developmental Biology (H.C., M.M.M.), Baylor College of Medicine, One Baylor Plaza, Houston, Texas 77030
Correspondence: Address all correspondence and requests for reprints to: Martin M. Matzuk, M.D., Ph.D., The Stuart A. Wallace Chair and Professor, Department of Pathology, Baylor College of Medicine, One Baylor Plaza, Houston, Texas 77030. E-mail: mmatzuk{at}bcm.tmc.edu
 |
Abstract |
|---|
 Top Abstract I. IntroductionII. Components of the...III. TGF-ß Superfamily...IV. TGF-ß Superfamily...V. Perspectives and Future...References |
|---|
|
I. Introduction |
|---|
|
TopAbstractI. IntroductionII. Components of the...III. TGF-ß Superfamily...IV. TGF-ß Superfamily...V. Perspectives and Future...References |
|---|
|
II. Components of the TGF-ß Superfamily Signal Transduction Pathway |
|---|
|
TopAbstractI. IntroductionII. Components of the...III. TGF-ß Superfamily...IV. TGF-ß Superfamily...V. Perspectives and Future...References |
|---|
). The TGF-ß family ligands are translated as prepropeptide precursors with an N-terminal signal peptide followed by the prodomain and the mature domain. Six to nine conserved cysteine residues in the mature domain form intra- and intermolecular disulfide bonds characteristic of this family of proteins (3, 10, 11). Several members of the family (i.e., GDF-9, BMP-15, GDF-3, lefty-1, and lefty-2) have a substitution of a serine for the cysteine normally involved in intermolecular disulfide bond formation. Therefore, although dimers of most family members are held together covalently, the proteins with this substitution would be expected to be noncovalently associated and possibly more labile.
|
vß6 (17), also mediates TGF-ß activation under physiological conditions. Other mechanisms, such as proteolysis (16, 17), may also be involved in the activation of TGF-ß ligands in vivo (18).
B. Receptors
TGF-ß superfamily ligands signal through a family of transmembrane serine/threonine kinases known as the receptors for the TGF-ß superfamily. On the basis of their structural and functional properties, the TGF-ß receptors are divided into two subfamilies: type I and type II receptors. Type I and type II receptors are glycoproteins of approximately 55 kDa and 70 kDa, respectively, which interact upon ligand binding. The extracellular regions of these receptors contain about 150 amino acids with 10 or more cysteines that determine the folding of this region. A unique feature of the type I receptors is a highly conserved 30 amino acid intracellular region immediately preceding the kinase domain (Fig. 2); this 30-amino acid stretch is called the GS domain because of the SGSGSG sequence it contains (19). Ligand-induced phosphorylation of the GS domain in the type I receptor by the type II receptor is required for the activation of signaling (19, 20, 21).
|
. Because of the promiscuity of the type I receptors for different TGF-ß superfamily ligands in most cases, the ALK nomenclature is probably most appropriate.
|
Another distinct type of cell surface receptors for the TGF-ß ligands was discovered by ligand cross-linking methods. These receptors were called type III receptors because of the higher molecular weight of these proteins compared with type I and type II receptors (37). Betaglycan and endoglin are two examples of type III receptors. Endoglin is a cell surface protein expressed at high levels in endothelial cells and at lower levels in monocytes, erythroid precursors, and other cell types (38, 39). Betaglycan is a membrane-anchored proteoglycan that can bind TGF-ßs and facilitate their interaction with the type II receptor (40, 41). Betaglycan also can bind inhibin and promote its binding to the activin type II receptor, thereby competing with activins at this receptor and antagonizing activin signaling (42). A distinct inhibin receptor [called inhibin-binding protein (InhBP) or p120] has also been identified (43, 44, 45). InhBP is a transmembrane member of the Ig superfamily expressed from an X-linked gene. InhBP is expressed in gonadotrophs of the pituitary and Leydig cells of the testes. InhBP levels closely correlate with serum inhibin B levels in female rats. This finding is interesting because InhBP specifically interacts with ALK4, and inhibin B (but not inhibin A) disrupts the association of InhBP, ALK4, and ActRIIA, thereby antagonizing activin signaling in vitro. In contrast, inhibin A appears to bind a complex of betaglycan and ActRIIA to disrupt activin signaling through an ActRIIA-ALK4 complex. The significance of these complexes in activin and inhibin actions in vivo will have to await gene knockout studies.
C. SMAD proteins
SMAD proteins are intracellular components of the TGF-ß superfamily signal transduction pathways (Fig. 2
). The first member of this family is MAD [mothers against dpp (decapentaplegic)], which was identified from genetic screens in Drosophila melanogaster as an enhancer of the weak alleles of dpp (46). Other members of this family were identified on the basis of their sequence homology with MAD. Three Caenorhabditis elegans homologs of MAD have been named as sma-2, sma-3 and sma-4, because mutation of their respective genes causes developmentally arrested, third-stage dauer larva with smaller body sizes than wild-type larva (47). Vertebrate homologs of sma and MAD are called SMAD, as a combination of sma and MAD (48). At least 10 vertebrate SMAD proteins have been identified to date (49). Mutations in SMAD2, SMAD3, and SMAD4 have been found in human tumors, suggesting that these genes function as tumor suppressors in vivo (3, 50, 51). For example, human SMAD4, which is also called DPC4 (deleted in pancreatic carcinoma locus 4), is frequently deleted or mutated in human pancreatic cancers (Ref. 52 ; Table 2).
|
Members of the SMAD family play different roles in TGF-ß superfamily signaling (Fig. 2). SMAD1, SMAD2, SMAD3, SMAD5, and SMAD8 can be phosphorylated directly by type I receptors after ligand-induced dimerization of type I and type II receptors, and they are called receptor-regulated SMADs (R-SMADs; Refs. 3 , 6 , and 49). The phosphorylation of these R-SMADs triggers their localization from the cytoplasm to the nucleus where they can interact with the common SMAD and function to regulate gene transcription. An initial indication of a functional specialization among different SMADs came from frog animal cap assays. SMAD1 and SMAD5 induce ventral mesoderm formation like BMPs (54, 55, 56), whereas SMAD2 induces dorsal mesoderm formation similar to activins or Vg-1 (54). Further studies revealed that different Smad proteins are coupled to different receptors. SMAD2 and SMAD3 are phosphorylated and translocated to the nucleus upon stimulation by activin and TGF-ß (57, 58, 59). SMAD1 and SMAD5 are phosphorylated and translocated to the nucleus upon stimulation by BMPs (e.g., BMP-2, BMP-4, and BMP-7; Refs. 60, 61, 62). SMAD proteins are classed into subgroups on the basis of sequence homology. These structural similarities may provide the basis for functional redundancy between different SMAD proteins in one subgroup. SMAD8 (previously known as SMAD9) is structurally very similar to SMAD1 and SMAD5 and may also mediate BMP signals (63).
Common SMADs in vertebrates include SMAD4 or SMAD4ß (found only in Xenopus laevis; Refs. 64 and 65), sma-4 in C. elegans (47), and Medea in Drosophila (66, 67, 68). Compared with other SMAD proteins, common SMADs have a characteristic insertion in the MH2 domain and lack the C-terminal SSXS motif, the site of phosphorylation by the type I receptors (59). Because of this sequence alteration, common SMADs are not substrates of the type I receptors. Common SMADs form hetero-oligomers with receptor-specific SMAD proteins and translocate with the receptor-specific SMAD proteins into the nucleus upon activation of the signaling pathways (55, 59, 61, 69). SMAD4 forms a complex with SMAD1, SMAD5, or SMAD8 when BMP pathways are activated and forms a complex with SMAD2 and SMAD3 when activin or TGF-ß pathways are activated. Consequently, overexpression of Smad4 in frog animal caps induces both ventral and dorsal mesoderm as a result of the activation of both BMP and activin pathways, respectively (70, 71).
The inhibitory Smad proteins include SMAD6 and SMAD7 in vertebrates, Dad in Drosophila, and possibly Daf-3 in C. elegans (6, 49). These Smad proteins contain a characteristic C-terminal MH2 domain, but their N terminus has little similarity with typical MH1 domains (49). The only known function of this group of Smad proteins is to inhibit the signaling activity of R-SMADs. SMAD6 preferentially inhibits BMP signaling, whereas SMAD7 can inhibit both TGF-ß and BMP signaling (72, 73, 74). SMAD7 inhibits phosphorylation of R-SMADs by occupying the type I receptors for BMPs, activins, and TGF-ßs. SMAD6 preferentially inhibits BMP signaling by competing with SMAD4 for binding to receptor-activated SMAD1 and forms an inactive SMAD1-SMAD6 complex. SMAD6 and SMAD7 levels are increased in response to BMP, activin, or TGF-ß signaling, suggesting that these SMADs function as negative feedback controls for different pathways.
D. Other components in the TGF-ß superfamily signaling cascade
Signaling of the TGF-ß superfamily members is highly regulated at multiple levels. Extracellular proteins such as follistatin, noggin, and chordin function as antagonists of many TGF-ß ligands to alter the signaling process. Follistatin can bind to activin and prevent the binding of activin to its cell surface receptors (75, 76). Follistatin can also bind to BMP-7 at lower affinity than activin and may antagonize BMP signals in vivo (77). Noggin and chordin bind to BMP-4 and prevent its interaction with receptors (78, 79).
TSP-1 is a large homotrimeric protein secreted by many cell types (80, 81). In cell-free systems, TSP-1 binds to and activates both small and large latent forms of TGF-ß1 (15, 82, 83). TSP-1 is likely a major activator of TGF-ß1 because mice lacking TSP-1 highly resemble the mutant phenotypes observed in young TGF-ß1 null mice with regard to lung, pancreas, liver, kidney, stomach, testes, skin, bone, and heart development (84). Furthermore, systemic treatment with a peptide that blocks the activation of TGF-ß1 by TSP-1 can induce similar lung and pancreas phenotypes. Lung and pancreatic abnormalities of TSP-1 null mice reverted toward wild type when treated with a peptide derived from TSP-1 that could activate TGF-ß1 (84).
Intracellularly, FKBP12 can interact with the cytoplasmic domain of type I receptors (32, 85, 86). FKBP12 was initially thought to be a negative regulator of TGF-ß signaling (86, 87). However, no detectable effects were observed on TGF-ß superfamily signaling or cell growth in Fkbp12 null fibroblasts (on processes such as Müllerian duct regression or FSH regulation by activins) or by disrupting FKBP12/TßRI (TGF-ß type I receptor) binding with FK506 or rapamycin (88, 89, 90). However, another group surprisingly showed that fibroblasts that they isolated from the same Fkbp12 null mice grew more slowly than controls (91). Presently, it is unclear why such apparently conflicting data have been generated.
TRIP-1 was identified as a TßRII (TGF-ß type II receptor)-interacting protein, and its interaction with TßRII is ligand-independent (92). TRIP-1 functions as a modulator of TGF-ß signaling by receptor-dependent and -independent mechanisms (93). Lastly, SARA (SMAD anchor for receptor activation) interacts with SMAD2 and SMAD3 and appears to function as a linker to recruit SMAD2 to the TGF-ß receptor (94). The functions of TRIP-1 and SARA in vivo are unknown.
TAK1 (TGF-ß-activating kinase 1), a member of the MAPK kinase kinase family (95), is another protein found to be involved in TGF-ß superfamily signaling. It participates in regulating TGF-ß-induced gene expression, as well as BMP-induced mesoderm formation and embryonic patterning (95, 96, 97). TAB1 (TAK1-binding protein 1) is an activator of TAK1 so that an association between the kinase domain of TAK1 and the C-terminal portion of TAB1 triggers activation of phosphorylation-dependent TAK1 (98, 99, 100). XIAP (human X-chromosome-linked inhibitor of apoptosis protein) was isolated as a TAB1-binding protein that also binds to BMP receptors in mammalian cells; XIAP may function as a positive regulator of the BMP signaling pathways by linking the BMP receptors and TAB1-TAK1 (101).
Although SMAD proteins bind directly to DNA, work from several groups suggests that high-affinity binding requires interactions with other DNA binding proteins such as coactivators (for review, see Ref. 102). Two important coactivators are the paralogous proteins, cAMP-response element binding protein (CREB) binding protein (CBP) and p300, which function in multiple transcriptional pathways by acting as scaffolding proteins and promoting histone acetylation-dependent chromatin relaxation and acetylation of transcription factors. In TGF-ß signaling, CBP and p300 cooperate with SMAD2-SMAD4 and SMAD3-SMAD4 in activating TGF-ß/activin-mediated gene transcription (103, 104). Humans with heterozygous mutations at the CBP locus have Rubinstein-Taybi syndrome (105), which is characterized by multiple craniofacial, skeletal, and cardiac defects, as well as growth retardation, severe mental retardation, and increased risk of cancer. Biallelic mutations of the p300 gene in humans have been found in colorectal and gastric carcinomas (106). Mice homozygous for a truncated CBP protein (amino acids 1–1084 out of 2441 amino acids) died between embryonic day (E)9.5 and E10.5 due to defects in neural tube closure, hematopoiesis, and yolk sac vascular development (107, 108); in a few cases, hematological malignancies including lymphocytic and myelogenous leukemia and histiocytic sarcomas occur (109). Homozygous mutant mice lacking p300 also die at midgestation due to defects in neural tube closure, enlarged heart cavities, poor vascular development within the yolk sac, and overall developmental retardation, possibly due to decreased cell proliferation (110). Mice heterozygous for both Cbp and p300 (i.e., Cbp+/- p300+/-) also died at midgestation with defects similar to the single homozygous mutants (110). This suggests important redundancy between these two related proteins, as might be expected. Some of these developmental defects in the Cbp and p300 knockouts could be attributed to abnormalities in TGF-ß superfamily signaling cascades (see Section III.B–D).
Another important SMAD DNA-binding protein is FoxH1 (forkhead or winged helix DNA-binding protein 1, also known as FAST2). FoxH1 was initially isolated in X. laevis as a protein that bound to an activin response element in the Mix.2 gene (111). FoxH1 can bind to complexes of SMAD4/phosphorylated SMAD2 or SMAD3, suggesting that these have an important role in activin signaling pathways in mammals. Knockout of the FoxH1 gene results in early embryonic lethality due to defects in anterior primitive streak and node formation (112), consistent with a function in the activin/Nodal-Smad signaling pathway (see Section III.A).
In addition to binding to DNA with transcriptional coactivators such as FoxH1, SMADs also associate with a number of transcriptional corepressors. This group of corepressors includes SnoN and Ski, members of the Ski family of oncogenes, as well as SMAD nuclear interacting protein 1 (SNIP1; Ref. 113), 5' transforming growth 3' interacting factor (TGIF; Ref. 114), TGIF2 (115), and SMAD-interacting protein 1 (SMADIP1; Ref. 116). SNIP1, which interacts with SMAD1, SMAD2, SMAD4, and CBP/p300, appears to modulate TGF-ß signaling mainly through preventing the association of SMAD4 and CBP/p300. TGIF and TGIF2 interact with TGF-ß-activated SMADs to repress TGF-ß-responsive transcription. TGIF and TGIF2 also recruit histone deacetylase, and TGIF (but not TGIF2) binds CBP and competes with p300 for binding to a SMAD2 complex. Both TGIF and TGIF2 bind to a specific TGIF binding site via their unique DNA-binding homeodomains. Interestingly, mutations in the human TGIF gene cause holoprosencephaly (117), presumably by interfering with NODAL/TGF-ß signaling through SMAD2 at specific loci, thereby affecting the development of midline structures in mammals (see Section III.E.4). Lastly, SMADIP1, a member of the 
EF1/ZFH-1 family of two-handed zinc finger/homeodomain proteins, interacts with SMADs after receptor-mediated activation and also binds to 5'-CACCT sequences in several promoters. A knockout of the mouse Smadip1 gene has not been created. However, heterozygous mutations in the human SMADIP1 gene cause syndromic Hirschsprung disease with a number of other findings, including mental retardation, hypospadias, and agenesis of the corpus callosum (118, 119, 120, 121). Thus, altering the dosage of SMADIP1 in humans is sufficient to cause a complex developmental disorder. Other genes involved in TGF-ß superfamily signaling and mutated in humans are presented in Table 2.
SnoN and Ski have also been shown to act as repressors by recruiting histone deacetylases. However, they appear to function in other ways as well, and their regulation in response to TGF-ß is very interesting. In particular, SnoN has been shown to antagonize TGF-ß signaling by binding to SMAD2, SMAD3, and SMAD4 (122, 123). SnoN is rapidly degraded in response to TGF-ß. SMAD2 and SMAD3 mediate this SnoN degradation using different pathways. TGF-ß induces a complex of SMAD2 and the E3 ubiquitin ligase, which stimulates ubiquitin-mediated degradation of SnoN (124). TGF-ß also promotes the formation of a complex containing SMAD3, the anaphase-promoting complex, and SnoN, which leads to SnoN ubiquination and degradation (125). Thus, TGF-ß stimulation results via both transcription activation and degradation of a corepressor. In Drosophila, DSmurf can bind MAD (the SMAD1/5 homolog) and promote its proteolysis, thereby restricting the BMP signaling pathway (126). Likewise, in mammals SMURF2 can target SMAD1 for ubiquination and proteosome-mediated degradation (127), thus favoring a TGF-ß/activin signaling pathway over the BMP signaling pathway. Interestingly, SMURF2, which is normally a nuclear protein, can complex with SMAD7, be exported into the cytoplasm, complex with the TGF-ß receptor, and cause ubiquitin-mediated degradation (down-regulation) of the receptor complex (128). In this way, ubiquitin-mediated proteolysis acts bifunctionally first to rapidly stimulate TGF-ß superfamily signaling and subsequently to halt further TGF-ß signaling. Similar to SMURF2, SMURF1 is also an E3 ubiquitin ligase that has been shown in X. laevis to specifically interact with BMP pathway-specific SMADs to trigger their ubiquination and degradation (129). In mammalian cells and parallel to SMURF2, SMURF1 interacts with SMAD7 to cause translocation of the complex to the cytoplasm, in which the complex will bind to the TßRI where it enhances ubiquination of both TßRI and SMAD7, thereby inducing turnover of these two proteins (130). On the basis of this data, there might be functional redundancy of SMURF1 and SMURF2 in vivo. The in vivo functions of the TGF-ß superfamily transcriptional corepressors SnoN, SNIP1, and TGIF2 and the E3 ubiquitin ligases, SMURF1 and SMURF2, are not yet known. However, significant in vivo functional data have been accumulated on SKI. Mice lacking the Ski proto-oncogene have multiple craniofacial defects, including exencephaly, midline facial defects, eye abnormalities (e.g., iris defects that range from aniridia to coloboma), decreased muscle mass, and extra digits (postaxial polydactyly; Refs. 131 and 132). When challenged with carcinogens, Ski heterozygous mutant mice show an increased incidence of tumors (133). In humans, absence of SKI may be in part causal for the craniofacial and muscle findings observed in individuals with the 1p36 deletion syndrome. These findings suggest that SKI may function in multiple TGF-ß signaling pathways. In fact, the neural tube defects (exencephaly) may be secondary to elevated BMP-4 levels. Thus, multiple ligands, receptors, binding proteins, and downstream proteins may modulate TGF-ß superfamily signaling cascades.
|
III. TGF-ß Superfamily Signaling and Development |
|---|
|
TopAbstractI. IntroductionII. Components of the...III. TGF-ß Superfamily...IV. TGF-ß Superfamily...V. Perspectives and Future...References |
|---|
Genetic evidence indicates that TGF-ß family members regulate both embryonic and extraembryonic development of an early postimplantation mouse embryo. Mice null for Bmp4, Nodal, lefty2, Alk2, Alk3 (Bmpr1a), Alk4 (ActRIB), Bmpr2, Smad2, or Smad4 either fail to initiate gastrulation or have defects in mesoderm differentiation. Other mutant mice that have defects in components of the TGF-ß superfamily signaling cascade are described in Table 3.
|
|
|
The defects in Alk2-mutated embryos are less severe than those of mouse embryos lacking either ALK3 (BMPR1A) or ALK4 (ActRIB) receptors. Embryos homozygous for an Alk2 null mutation are arrested at early gastrulation with abnormal visceral endoderm morphology and severe disruption of mesoderm formation (138, 139). In Alk3 null or Alk4 null embryos, the formation of the primitive streak and mesodermal cells is completely blocked. Although ALK3 probably functions as the type I receptor for BMP-4 to regulate epiblast cell proliferation and mesoderm formation, ALK4 appears to mediate a Nodal or an activin-like signal that is essential for egg cylinder organization and primitive streak formation.
BMPRII transduces signals for BMPs through heterodimerization with ALK2, ALK3, or ALK6. Bmpr2 null embryos are arrested at the egg cylinder stage before gastrulation (140), very similar to Alk3 null embryos (141). This suggests that the essential BMP signaling pathway in this early developmental period involves heterodimerization of BMPRII and ALK3.
Several other key TGF-ß family signal mediators, including ALK4 (ActRIB), SMAD2, and SMAD4, have been shown to function in extraembryonic cells during early mouse development. Although ALK4 (ActRIB) and SMAD2 function in both epiblast and extraembryonic cells, their roles in the extraembryonic cells are required for the formation of the embryonic rather than the extraembryonic mesoderm. Similar to ALK2/ActRIA, SMAD4 is required for the differentiation and function of the visceral endoderm, and inactivation of SMAD4 in the extraembryonic cells blocks embryonic growth and mesoderm formation. It is possible that ALK2/ActRIA and SMAD4 function in the same signaling pathway during early mouse development.
Genetic studies have shown that embryos lacking either Acvr2 (142) or Acvr2b (143) develop to term with no gross defects in mesoderm formation. However, mutant embryos homozygous for both Acvr2 and Acvr2b are arrested at the egg cylinder stage and do not form mesoderm (144), demonstrating functional compensation between these two receptors. Furthermore, Acvr2-/- Acvr2b+/- embryos fail to form an elongated primitive streak, causing disruption of the mesoderm formation. These gastrulation defects are similar to the defects in Nodal+/- Acvr2-/- (144) and Nodal+/- Smad2+/- (145) double mutants, suggesting that Nodal, the type II activin receptors, and Smad2 function in the same genetic pathway to regulate mouse gastrulation (144). Other craniofacial findings in these mutant mice will be discussed below.
In addition to the above defects, BMP signaling is also important for extraembryonic development. In mice lacking Bmp2, Bmp4, or Bmp8b, there are extraembryonic defects. Bmp2 null embryos demonstrate shortened and delayed allantois development, respectively (146, 147). About 50% of the Bmp2 null embryos fail to undergo normal chorioallantoic fusion. In contrast, Bmp4 null embryos completely lack an allantois. Lastly, Bmp5/Bmp7 double mutant embryos demonstrate developmentally delayed and smaller allantois and also demonstrate failure of the allantois to fuse with the chorion in most cases (148). Knockouts of Smad1 and Smad5, which encode proteins downstream of BMP stimulation, also have defects in extraembryonic tissue development. Both Smad1 and Smad5 knockout mice die at midgestation before E10.5 (149, 150, 151, 152). Both models have defects in allantois formation, this being more dramatic in the Smad1 knockout mice. These findings suggest that there is redundancy of these two SMADs in allantois development (see Section IV.B). In Smad5 knockout mice, there is mislocation of allantois tissue that is also found in the amnion (149). In the Smad1 knockout, there are defects in fusion of the allantois and chorion (similar to the Bmp2 null embryos and the Bmp5/Bmp7 double mutant embryos) and overproliferation of the chorion (151, 152). Thus, BMP-2, BMP-4, BMP-5, BMP-7, and BMP-8b signaling through SMAD1 and SMAD5 appear to play important functions in the integrity of the allantois and chorion.
In summary, the TGF-ß superfamily is essential for several important steps in early postimplantation development, including organization and growth of the egg cylinder, development of the extraembryonic membranes, formation of the primitive streak, and differentiation of the mesoderm. One recent report suggests BMP-2 signal is important for the spacing of embryo implantation (153). Functional importance of TGF-ß superfamily signaling in preimplantation embryo development is less clear.
B. Heart development
During mouse embryogenesis, the heart is the first organ to differentiate and function (154). The heart initially appears as two cardiac primordia that then fuse together to form a linear heart tube consisting of an inner endothelium surrounded by an outer layer of myocardium. Between these two layers lies a complex layer of extracellular matrix known as cardiac jelly secreted mainly by the myocardium. At later developmental stages, loops and turns break the symmetry of the linear heart tube. In addition, endocardial cells respond to signals from the overlying myocardium and undergo an epithelial-to-mesenchymal transformation to invade the intervening extracellular matrix and form the cardiac cushion. The mature heart valves and septa derived from the cardiac cushions ultimately divide the heart into four functional chambers (154). For more details on heart development and congenital heart diseases, please refer to a recent review by Harvey (155).
Genetic studies have shown that TGF-ß superfamily signaling is essential for heart development. BMP-2 is required for the initial formation of cardiac primordium, because Bmp2 null mice either do not have a heart or develop a very retarded and malformed heart (156). BMPs are also likely to be involved in later septa and valve formations as suggested by their expression patterns. In the midgestation mouse heart, BMP-2 and BMP-4 are expressed in the atrial-ventricular (AV) canal and AV cushion, respectively, whereas BMP-5, -6, and -7 are expressed more homogeneously in the myocardium, and BMP-10 is expressed exclusively in the trabeculae. Gene inactivation analysis indicates functional redundancy among 60A subgroup of BMPs, namely BMP-5, BMP-6, and BMP-7. Heart development is normal in embryos lacking BMP-5 (157), BMP-6 (158), or BMP-7 (159, 160). However, mice carrying mutations in both Bmp5 and Bmp7 genes die around E10.5 with multiple defects in heart development, involving the AV cushion, septum, free wall, and trabeculae (148). Mice deficient for both BMP-6 and BMP-7 have delayed cardiac cushion formation in the outflow tract, which results in subsequent valve and septation defects. These embryos die due to cardiac insufficiency (161).
In addition, several TGF-ß isoforms may control epithelial-mesenchymal transformation in the AV canal of the heart (162, 163, 164). During murine endocardial cushion formation, TGF-ß1 is expressed in endothelial/mesenchymal cells, whereas TGF-ß2 and TGF-ß3 are expressed in the myocardium (162, 165, 166, 167, 168). Although TGF-ß1 knockout mice do not seem to have any heart abnormalities, TGF-ß2 knockout mice have specific defects in the development of valves and septa of the heart (169). In addition, both TGF-ß type II and type III (betaglycan) receptors are expressed in AV endothelial cells. At least in avian explants, antibodies against either TGF-ß type II or type III receptors inhibit epithelial-mesenchymal transformation and mesenchymal cell migration after transformation (162, 163, 164). Further studies on TGF-ß ligand signaling can be expected to provide important information for understanding normal heart development and congenital heart diseases.
Two type I receptors, ALK3 and ALK5, have also been shown to be involved in heart development. ALK3, also known as BMP type IA receptor, is essential for mesoderm formation as revealed by conventional knockout mice (141). Conditional loss of ALK3 in midgestation mouse myocardium results in defects in AV cushion formation and subsequent abnormalities in cardiac septa and valves (170). The cardiac defects of ALK3 conditional knockout mice resemble those of Bmp5, Bmp7 double knockout mice (148), consistent with the biochemical data that ALK3 is upstream of BMPs. TGF-ß2 seems to be downstream of ALK3 during cardiac cushion formation, as suggested by the finding from ALK3 conditional knockout mice that TGF-ß2 expression in the myocardium adjacent to the AV canal is greatly reduced when ALK3 is absent from the myocardium. This finding explains the resemblance of the cardiac defects between TGF-ß2-deficient mice and ALK3 conditional knockout mice. It also suggests that a cascade of TGF-ß superfamily signaling is required for normal cardiac cushion formation. ALK5, a receptor for TGF-ß isoforms, may be required for heart looping, because directed expression of a constitutively active form of ALK5 (L193A, P194A, T204D) in the mouse myocardium arrests heart looping (171).
Both Smad5 and Smad6 knockout mice have defects in heart development. Smad5 knockout mice (172) display defects in heart looping due to abnormalities in left-right axis determination (see Section III.C). In contrast to the Smad5 knockout mice, which die during embryogenesis, the Smad6 null mice live to birth, but a majority of null pups die before weaning (173). Analysis of the expression of a LacZ reporter in the Smad6 locus reveals high specificity of the reporter in the developing outflow tract and AV cushion of the heart between E9.5 and E13.5, expansion of the expression pattern to the vascular endothelium of larger vessels during late embryogenesis, and a restriction of the expression to the cardiovascular system postnatally. Consistent with this expression pattern and a role of SMAD6 negatively regulating TGF-ß signaling during the endocardial cushion transformation, Smad6 null mice show hyperplasia of the cardiac valves and septal defects in the outflow tracts. Smad6 null mice that live to the adult stage display aortic ossification and elevated blood pressure. Thus, SMAD6 inhibitory function is key to normal development of the cardiovascular system.
In summary, heart development is regulated by both BMPs and TGF-ß isoforms at multiple developmental stages, including cardiac primordia specification, heart looping (discussed further in Section III.C), and cardiac cushion formation, and later, in septum and valve formation. The finding that TGF-ß2 is dramatically down-regulated when ALK3 is absent from the heart suggests genetic connections among signaling pathways of different TGF-ß superfamily members (170). Mutations in BMPRII have been found in the congenital heart disease, primary pulmonary hypertension, implying functional conservation of TGF-ß signaling in mammals (174). Future studies on conditional knockout mice as well as double knockout mice will provide us with more information on the roles of TGF-ß superfamily signaling in cardiac development and congenital heart diseases.
C. Left-right asymmetry
During mouse embryogenesis, morphological asymmetry of the left-right axis first occurs around E8.0, when the embryonic heart tube loops toward the right (154). This is followed by a leftward axial rotation of the embryo at the 9–10 somite stage (154). Several TGF-ß family ligands, such as Nodal, lefty-1, and lefty-2, are expressed asymmetrically before or around the appearance of the morphological asymmetry, and these ligands appear to be involved in normal left-right axis formation (175, 176, 177, 178).
1. Nodal.
Expression of Nodal is first detected at E5.5 in the primitive ectoderm (136). At gastrulation, the highest level of Nodal expression is maintained at the posterior region of the epiblast and marks the site of future primitive streak formation. Nodal is also expressed transiently in the visceral endoderm before and during early streak formation (136). The roles of Nodal in primitive streak formation and subsequent mesoderm differentiation during mouse embryogenesis are confirmed by the lack of primitive streak development in mice deficient in Nodal (136, 179, 180).
Besides its early roles in primitive streak formation, Nodal is also involved in left-right asymmetry establishment. At E7.5, Nodal is expressed in cells around the node, and this expression becomes asymmetric at E8.0, with the left side being greater. When the embryo develops three to five pairs of somites (E8.0), Nodal is expressed on the left side of the lateral plate mesoderm (LPM), but not on the right side (2). This asymmetric expression of Nodal precedes the appearance of morphological asymmetry. In mouse strains carrying mutations causing defects in left-right asymmetry, the expression pattern of Nodal is changed according to the direction of heart looping and embryonic turning. For example, in iv (inversus viscera) mice, heart looping and embryonic turning are randomized, and the expression of Nodal in the LPM is also randomized (i.e., Nodal is expressed on the left side, the right side, or bilaterally, or it is absent). In inv (inversion of embryonic turning) mice, in which situs is completely inversed, Nodal is expressed only on the right side of the LPM (2, 181). In addition, ectopic expression of Nodal on the right side of the LPM in chicken embryos randomizes the direction of heart looping (182). Furthermore, genetic evidence also reveals roles for Nodal in left-right asymmetry establishment. Mice heterozygous mutant for both Nodal and Smad2 develop defects in left-right asymmetry, including transposition of the great arteries, right pulmonary isomerism, and right-sided stomach (145).
2. Lefty-1 and lefty-2.
Lefty-1 and lefty-2 are two divergent members of the TGF-ß superfamily with highest homology to each other (183). Their names are based on their embryonic expression patterns. Lefty-1 is expressed strongly on the left side of the putative ventral floor plate (PFP) and weakly on the left side of the LPM at E8.0 to E8.5 (183). Lefty-2 is expressed strongly on the left side of the LPM and weakly on the left side of the PFP at E8.0 to E8.5 (183). Nodal and lefty-2 expression patterns in knockout mice lacking lefty-1 are normal at early somite stages (three to five pairs of somites) but become bilateral at later developmental stages (six to eight pairs of somites). In accordance with these expression patterns, earlier events in left-right asymmetry, such as heart looping and embryonic turning, are normal in lefty-1 knockout mice, whereas later events, such as the position and differentiation of visceral organs, are abnormal (184). The most common defect in lefty-1 knockout mice is pulmonary left isomerism in which both lungs of the mutant mice have only one lobe in contrast to wild-type mice in which the left lung has one lobe and the right lung has four lobes (184).
Two conclusions can be drawn from these studies. First, left-right asymmetry is established sequentially. Early events (such as embryonic turning and heart looping) are related to an early phase of Nodal and lefty-2 expression, whereas later events (such as the differentiation of the visceral organs) are related to a late phase of Nodal and lefty-2 expression. Second, lefty-1 functions as a negative regulator of late phase Nodal and lefty-2 expression, but not the early phase of Nodal and lefty-2 expression.
The effect of lefty-2 on left-right axis formation cannot be evaluated because lefty-2 has an essential role in primitive streak formation, and lefty-2 knockout mice fail to develop to a stage in which defects in the left-right axis can be observed (137). Analysis of conditional knockout mice lacking lefty-2 at later developmental stages may provide direct evidence for the role of lefty-2 in left-right axis formation.
3. GDF-1.
In contrast to Nodal, lefty-1, and lefty-2, Gdf1 is expressed symmetrically around the time of left-right asymmetry establishment (185). Gdf1 is initially expressed throughout the embryo proper at E7.5 and later becomes restricted to cells around the node, cells in the ventral neural tube, and cells in the proximal and LPM (185). In late-stage embryos and adult mice, the expression of Gdf1 is restricted to the nervous system (186). A minority of knockout mice lacking GDF-1 die between E14.5 and birth. About two thirds of the mutant mice live to birth and die within 48 h of severe cardiac defects. A wide range of left-right axis defects are obvious in Gdf1 knockout mice, including visceral situs inversus, right pulmonary isomerism, and cardiac anomalies (185). Furthermore, the absence of lefty-1, Nodal, and lefty-2 expression in the PFP and LPM of Gdf1 knockout mice suggests that GDF-1 is an upstream factor required for the asymmetric expression of these genes (185). The receptors and SMADs through which GDF-1 signals are unknown at this time.
4. Other components.
Although several TGF-ß superfamily ligands have been implicated in left-right axis formation in mice, little is known about their receptors and downstream SMADs. ACVR2B is the only receptor shown to be essential for left-right asymmetry. Acvr2b knockout mice develop pulmonary right isomerism and other defects related to left-right axis formation (143), similar to Nodal/Smad2 double heterozygous mutants (145). Although activins A and B bind to ACVR2B in vitro, the physiological ligands that bind to ACVR2B and regulate left-right axis development in mammals are not known.
Direct evidence for the involvement of SMAD2 in left-right asymmetry is lacking because mice deficient in Smad2 fail to initiate gastrulation. However, mice double heterozygous for both Nodal and Smad2 mutations develop variable mutant phenotypes, including gastrulation defects, complex craniofacial abnormalities (see Section III.E), and defects in left-right patterning. These data indicate that SMAD2 may mediate Nodal signaling in these developmental processes (145).
Smad5 knockout mice have shown that a signaling pathway through SMAD5 is involved in left-right asymmetry (172). At early stages, the mutant embryos fail to initiate or complete turning properly, a nearly completely penetrant phenotype that is obvious at E9.5. In addition, there are defects in heart looping in these mutants. In wild-type embryos at E9.5, the rostral part of the heart loops toward the right, whereas the caudal part twists to the left. In the Smad5-/- embryos, the rostral part of the heart either fails to loop (6 of 17), loops randomly (9 of 17), or elongates and fails to show any looping. At the molecular level, lefty-1 was essentially undetectable, and Nodal, lefty-2, and Pitx2 were expressed bilaterally. This suggests that a SMAD5 signaling cascade is upstream of these genes and is essential for left-right axis determination.
D. Vasculogenesis and angiogenesis
Vasculogenesis and angiogenesis are two processes leading to the formation of new blood vessels (187). Vasculogenesis refers to the primary in situ differentiation of endothelial precursor cells from mesoderm and their subsequent organization into a primary capillary plexus. Angiogenesis is the formation of new vessels from the preexisting vessel plexus through splitting and sprouting (187, 188). During embryogenesis, vasculogenesis first occurs in the extraembryonic yolk sac when blood islands differentiate from the extraembryonic mesodermal cells. Blood islands contain hemangioblasts capable of forming both endothelial cells and hematopoietic cells. A primary blood plexus is formed in the yolk sac and then connects with the primary blood plexus in the embryo proper. The earliest circulatory system is established when the embryonic heart starts to beat (154). Primary embryonic vasculature is remodeled at later developmental stages through vasculogenesis and angiogenesis.
TGF-ß signaling pathways play important roles in vasculogenesis and angiogenesis (Fig. 3; Ref. 189). During mouse embryogenesis, TGF-ß1 is expressed in many tissues, including endothelial and hematopoietic precursor cells (165). Targeted disruption of TGF-ß1 in mice results in midgestation lethality in half of the homozygotes and about a quarter of the heterozygotes (167, 190, 191, 192). The primary causes of death are defects in the yolk sac vasculature and hematopoietic system. Although initial differentiation of mesodermal precursors into endothelial cells occurs, subsequent differentiation of endothelial cells into capillary-like tubules is defective, resulting in vessels with decreased wall integrity. TßRII-deficient mice have also been generated and demonstrate a similar mutant phenotype (193).
|
SMAD5, originally thought to be a signal transducer for BMPs, appears to mediate TGF-ß1 signaling through ALK1 and TßRII. ALK1 can interact with TßRII or ActRII and phosphorylate SMAD1 and SMAD5 (200, 201). Constitutively active ALK1/Q201D can phosphorylate SMAD1 and SMAD5 but not SMAD2 and SMAD3 (198), suggesting that ALK1 transduces signals through SMAD1 and/or SMAD5. ALK1, however, is not the only type I receptor for TGF-ß1 signaling during vascular development.
Consistent with the biochemical data that SMAD5 is downstream of ALK1, vascular lesions in Smad5 knockout mice highly resemble those in TGF-ß1, TßRII, and ALK1 knockout mice (149, 150). Smad5 knockout mice die between E9.5 and E11.5. After E9.0, the yolk sacs of Smad5 mutants contain red blood cells but lack well organized vasculature.
ALK1, however, is not the only type I receptor for TGF-ß1 signaling during vascular development. Mice deficient in ALK5 also show severe vascular lesions reminiscent of TGF-ß1 knockout mice (202). Endothelial cells derived from Alk5 null mice show enhanced proliferation, improper migratory behavior, and impaired fibronectin production in vitro (202). ALK5 appears to signal through Smad2 in endothelial cells and induces the expression of PAI-1, a known proteinase inhibitor, to prevent degradation of extracellular matrix proteins around nascent vessels (203). Therefore, the possible function of the TGF-ß/ALK5 pathway during angiogenesis is to inhibit cell migration and proliferation.
Therefore, in endothelial cells, TGF-ß signals through both ALK1 and ALK5, and the two receptors phosphorylate different sets of downstream Smad proteins (ALK1 phosphorylates Smad1 and Smad5, ALK5 phosphorylates Smad2 and Smad3). The TGF-ß/ALK5 pathway leads to inhibition of cell migration and proliferation, whereas the TGF-ß/ALK1 pathway induces endothelial cell migration and proliferation (203). The activation state of each pathway is determined by the dose of TGF-ß present (203). The fate of the endothelium may thus be dependent on the balance of ALK1 vs. ALK5 activation. These genetic studies solved a puzzle in the field for many years, namely, how TGF-ß can function as both an inhibitor and a promoter of angiogenesis.
In summary, TGF-ß signaling is essential for vascular development. A genetic pathway composed of TGF-ß1, endoglin, TßRII, ALK1, ALK5, and SMAD5 (Fig. 3) determines several important aspects of angiogenesis, including maintenance of the integrity of vessel walls, recruitment of smooth muscle cells, deposition of extracellular matrix, as well as differentiation of arteries and veins.
E. Craniofacial development
1. Cleft palate.
A variety of craniofacial malformations have been associated with mutations in TGF-ß superfamily members. Isolated cleft palate is a relatively common human birth defect that is generally remedied by surgical correction within the first few months to years of life, depending on its severity. It is considered a complex trait in humans, and presumably the malformation reflects deleterious effects on multiple gene products. This is in contrast to many mouse models in which the knockout of a single gene is sufficient to cause clefting in 100% of homozygous null mutants (204, 205, 206). Palate formation is a complex process that requires the growth, elevation, and fusion of palatal shelves, followed by a loss of the midline epithelial seam and subsequent differentiation of overlying and adjacent mesenchymal cells (207). During embryogenesis, many TGF-ß superfamily members, their receptors and their associated regulatory proteins are expressed in the developing palate (Refs. 208, 209, 210, 211, 212 and Fig. 4). TGF-ß1, TGF-ß2, and TGF-ß3 are all expressed in the palate, but their spatiotemporal expression patterns differ. All are thought to contribute to palate shelf elongation and fusion. At E13.5, TGF-ß3 expression is restricted to the medial edge epithelium (MEE) as the palatal shelves move from a vertical to horizontal position. One day later, TGF-ß1 and TGF-ß3 are expressed in the MEE, whereas TGF-ß2 expression is restricted to mesenchymal cells. The type I TGF-ß receptor (Tgfbr1) is expressed in cells comprising the palate, oral, and nasal epithelia, as well as the medial edge seam (212). The type II TGF-ß receptor (Tgfbr2) is highly expressed in the MEE seam but is only weakly expressed in oral and nasal epithelia (212).
|
Activin/inhibin ßA (Inhba) is another member of the TGF-ß superfamily that is highly expressed in the developing palate, oral, and nasal mesenchyme (210). Similar to TGF-ß2 knockout mice, activin ßA null mutants have incomplete penetrance of cleft palate, ranging from 29–33%, depending on the background strain (213). The severity of the palate defect ranges from complete anteroposterior clefting to a membranous hard palate devoid of ossified tissue. Although the precise mechanism for the clefting has not been determined, the ossification failure may reflect important roles for activin ßA in facilitating palate chondrogenesis and subsequent ossification, as well as palate fusion (213).
Activin receptor type II (Acvr2) null mutants also have cleft palates. However, the clefting is likely secondary to the hypoplastic mandibles that are present in 22% of these mice, although primary effects on palatogenesis cannot be excluded. This phenotype is similar to the Pierre Robin sequence in humans, in which the development of a hypoplastic mandible results in a posterior and superior positioning of the tongue during craniofacial development. This results in a mechanical obstruction to palate fusion. Like the Acvr2 null mutant mouse, the severity of the Pierre Robin sequence ranges from relatively mild retrognathia (i.e., posteriorly placed mandible) with or without clefting, to severe micrognathia (i.e., small, hypoplastic mandible) and vestigial tongue and airway compromise (214). In contrast to the Acvr2 null mutant mouse, however, the Pierre Robin sequence occurs sporadically.
Follistatin was initially described as an activin binding protein (215); however, on the basis of studies from follistatin null mutant mice (216), transgenic mice overexpressing the protein (217), and studies in Xenopus embryos (77, 218), it is clear that follistatin modulates the effects of other TGF-ß superfamily members as well (218). Follistatin homozygous null mutant mice die in the perinatal period, and some of the craniofacial features of these mice are strikingly similar to those of the activin ßA knockout mouse (216). Twenty-one percent of follistatin null mutant mice lack a hard palate, consistent with its role in modulating the effects of activin signaling.
In addition to skeletal patterning defects, Gdf11 null mutant mice have cleft palates, consistent with Gdf11 expression pattern during embryogenesis (219, 220). Some functions of GDF-11 are hypothesized to be transmitted through the activin receptor, ACVR2B (see Section III.F). In summary, several TGF-ß superfamily members participate in the process of palatogenesis. The TGF-ßs, activins, GDF-11, TGF-ß and activin receptors, and follistatin have all been shown to play a role in mouse models. However, the failure of activin ßA to rescue the palate fusion defects of Tgfb3 null mutant mice (212) suggests that alternative signaling pathways may contribute synergistically to the complex process of palatogenesis. It is also likely that there are major players and minor players in the game, and the relative importance of each is reflected by the penetrance of the null phenotype. This model is consistent with the multifactorial nature of this complex biological process.
2. Tooth development.
The development of teeth in mammals is a complex process involving at least five well characterized developmental signaling pathways [BMP, activin A, fibroblast growth factor (FGF), Hedgehog, and WNT]. The expression of more than 200 genes has been examined in teeth, and the expression patterns of many of these are now available in a web-based graphical format (221). The process of tooth development first requires the definition of a tooth region (i.e., it must first be determined where a tooth will grow). Next, tooth identity must be established (i.e., incisor, canine, premolar, or molar in humans; incisor or molar in mice). Both processes are influenced by antagonistic effects of BMP and FGF signaling and rely on reciprocal, temporally regulated signals between the inductive epithelium and its underlying mesenchyme. Also, the same signals are used repeatedly at various stages as tooth identity is established. (For excellent reviews on mammalian tooth morphogenesis, please refer to Refs. 222 and 223 .) Within the TGF-ß superfamily, both BMP-4 and activin ßA have been shown to play important roles in tooth morphogenesis. BMP-4 has been the most extensively studied (224, 225, 226). During the determination of the tooth region, BMP-4 inhibits and FGF-8 stimulates the expression of Pax9 (which encodes a paired box transcription factor) in the mesenchyme (227). PAX9 and other transcription factors are likely required for the progression of tooth morphogenesis beyond the bud stage. BMP-4 has also been shown to play several important roles in determining tooth identity. Exposure of E9–E10 mouse mandibles to Noggin (a BMP antagonist) causes an incisor to molar transformation (225) due to the inability of BMP-4 to inhibit BARX1, a homeobox-containing transcription factor expressed in molar mesenchyme. Expression of mesenchymal BMP-4 (induced by epithelial BMP-4; Refs. 224 and 226) and activin ßA (induced by epithelial FGF-8) at E10–11 (228), precede the budding of epithelium at the sites where teeth will form. This observation makes these two TGF-ß superfamily members candidate signals for the initiation of tooth bud formation. Activin ßA also plays an important role in dental patterning, because homozygous null mutant mice lack incisors and mandibular molars but maintain normal maxillary molar development (228, 229). Comprehensive expression studies of the tooth buds of these mice using in situ hybridization show that none of the genes thought to influence the early stages of tooth development (including BMP-4) are affected in the knockout mice. However, follistatin, normally expressed in the tooth epithelium, was absent in the mutants. Interestingly, follistatin null mutants have essentially the same tooth phenotype as activin ßA null mutants (216). These observations suggest that activin ßA in the tooth mesenchyme normally induces follistatin expression in the overlying epithelium. This induction is possibly mediated through activin type II and IIB receptors, because double heterozygous mutant mice for these receptors have a similar tooth phenotype as activin ßA and follistatin knockout mice (228). It has been suggested that follistatin acts as a sink during tooth morphogenesis, removing activin from the local environment, thereby resetting the activin signaling cascade (228).
BMP-4 is absent from the dental mesenchyme of mice with homozygous null mutations for the transcription factors LEF1, MSX1, or PAX9 (224, 230, 231), all of which show arrest of tooth development at the bud stage. However, progression can occur in tooth buds from Msx1 mutants in vitro if BMP-4 is added to the medium (232). Thus, BMP-4 is also a good candidate for a mesenchymal factor that signals the transition from the bud to cap stage. Later, the apoptosis of enamel knots (structures on the apical surface that ultimately give the tooth its rough contour) is associated with BMP-4 expression in the knot cells (233), suggesting that this protein is important at essentially all stages of tooth morphogenesis. In summary, both BMP and activin signaling play important roles in normal tooth development. Furthermore, BMP expression is controlled by a variety of transcription factors that have their effects at the earliest stages of this process.
3. Eye development.
Several TGF-ß superfamily members are involved in eye development. Mice homozygous for an activin ßA null mutation have reduced rod photoreceptor cells in the neural retina, suggesting that activin A promotes progenitors to differentiate into photoreceptors (234). The presence of ALK2 protein in facial sensory organ primordia, including the eye area, also points to a role for activin signaling in the development of eye (235). BMP-4 and BMP-7 seem to be involved in earlier stages of eye development (236, 237). Bmp4 is expressed strongly in the optic vesicles and weakly in the surrounding mesenchyme and surface ectoderm. In Bmp4 null mice, lens induction is absent. Exogenous BMP-4 added to optic vesicles in explant cultures rescues the lens induction defect of the Bmp4 null mice. Interestingly, Pax6 expression is not affected in Bmp4 null mice, and Bmp4 expression is not affected in Pax6 null mice. These observations suggest that BMP-4 is a critical factor for lens induction and functions in a pathway independent of PAX6. BMP-7 is another TGF-ß superfamily ligand required for lens induction. BMP-7 protein is present in the head ectoderm of mouse embryos at the time of lens induction, and when BMP-7 antagonists are added to in vitro cultures during the period of lens induction, lens formation is inhibited (237). Furthermore, mice deficient in BMP-7 have eye abnormalities ranging from anophthalmia to microphthalmia (159, 160, 238, 239). Tgfb2 null mice also have eye defects, including a thin corneal stroma, and hypercellularity of the posterior chamber (169). Histologically, a hamartomatous mix of melanocytes, neuronal cells, and mesenchymal cells with vascular elements is observed. In addition, there is hyperplasia of the inner and outer neuroblastic layers of the retina. It is unclear whether the hypercellularity of the posterior chamber and retina represents aberrant cell proliferation or a failure of apoptotic programs. Nevertheless, these observations suggest an important role for TGF-ß2 in normal eye remodeling.
Activin/inhibin ßB (Inhbb) homozygous null mutant mice have variable defects in eyelid closure that are evident at birth, suggesting a possible role for this activin in facilitating the proliferation and migration of the peridermal cells involved in this process (240, 241). No other craniofacial anomalies are observed in these mice. Because mutations in other genes (e.g., epidermal growth factor, TGF-, integrins; Refs. 242, 243, 244, 245) also result in open-eye phenotypes, it is likely that many genetic determinants acting in concert contribute to the process of eyelid fusion in mice. In summary, the development of most major eye structures (retina, lens, cornea, posterior chamber, and eyelid) is influenced by members of the superfamily, including the activins, BMPs, and TGF-ßs. Moreover, at least the BMPs exert their effects through pathways independent of PAX6.
4. Other craniofacial malformations.
Holoprosencephaly is the most severe of a spectrum of disorders caused by the abnormal establishment of midline craniofacial structures. Aberrant midline division of the embryonic forebrain results in poor distinction of the cerebral hemispheres, which results in severe neurological abnormalities, and, frequently, death. Anatomically, holoprosencephalies have been subclassified into lobar, semilobar, and alobar forms, in increasing order of severity. Abnormalities are not restricted to the forebrain, because the midline facial structures are also affected. The severity of the condition is highly variable and can include dramatic craniofacial features, such as midline fusion of the developing eye structures (cyclopia), union of the olfactory placodes to form a tube-like structure that projects from the forehead (proboscis), or midline clefting of the lip and palate. Defects may also be as subtle as an abnormally formed nasal septum, narrowly spaced eyes (hypotelorism), or a single central incisor (214, 246). Holoprosencephaly is genetically heterogeneous and is observed as a feature in many genetic disorders (214). Conversely, in rare familial cases, the condition does not breed true, (i.e., the same mutation within a family can give rise to the full spectrum of clinical severity).
Within the context of the TGF-ß superfamily, holoprosencephaly has been linked to deficiencies in Nodal signaling in humans and in several experimental model systems. Nodal+/-, Smad2+/- double heterozygous mice have a wide range of abnormal phenotypes that include gastrulation defects, errors in left-right axis determination, cyclopia, and truncation of the anterior head structure (145). Similar conditions are observed in Nodal+/-, Acvr2-/- mice (144). These findings strongly suggest that normal forebrain development and the formation of midline facial structures require an intact Nodal signaling pathway and that the signal is mediated at least in part by the ACVR2. This hypothesis is further supported by the zebrafish mutants, cyclops, squint, and one-eyed pinhead, which all have cyclopia; the mutated genes all encode proteins that are components of the Nodal signaling pathway. Finally, some holoprosencephaly patients have mutations in TGIF (117). This factor can act as a transcriptional corepressor with SMAD2 in vivo (114, 247). Mutations in TGIF are hypothesized to result in holoprosencephaly by disrupting the delicate balance between transcriptional activation and repression by SMAD2 complexes during forebrain development (117, 248). Because a nearly identical phenotype has been documented in patients with mutations in the human sonic hedgehog (SHH) gene, it has been suggested that compromised Nodal signaling disrupts the normal juxtaposition of SHH-expressing cells and developing cells of the forebrain during gastrulation, placing SHH downstream of the Nodal signaling pathway (117). In summary, holoprosencephaly, the extreme manifestation of disrupted midline development, can occur through perturbation of ligands (Nodal), receptors (ACVR2), and downstream components of Nodal signaling (SMAD2, TGIF). Moreover, disruption of normal Nodal signaling has adverse effects on either cells expressing SHH or SHH signaling.
