Bone Morphogenetic Proteins (BMPs)

First printed in R&D Systems' 1999 Catalog.

Overview

Bone morphogenetic proteins (BMPs) are members of the TGF-beta superfamily. The discovery of BMPs parallels that of the tumor necrosis factor (TNF) superfamily, where clinicians noted that infection, accompanied by acute inflammation, was correlated with necrosis and regression of pre-existing tumors.1 The first hint of the existence of BMPs came from the observation that bone formed in fascia that had been surgically used to bridge large gaps in bladder.2 An association was made between the approximation of connective tissue and “osteoinductive” tissues (or substances). Both transplanted, demineralized bone and transitional (or urinary) epithelium induced ectopic bone formation in connective tissue.3, 4 Certain tissues seemed to possess an “osteogenic protein” that could induce new bone formation.4, 5 A number of osteogenic proteins, or BMPs, have been discovered and, based on sequence homology, most of the GDFs (growth/differentiation factors) have been added to the BMP family, raising the number of known BMPs to around 20.6 Although the name BMP is descriptive of one particular function, it is somewhat misleading. While BMPs may induce ectopic bone or cartilage formation, they also play important roles in the development of the viscera. This includes roles in cell proliferation, apoptosis, differentiation, and morphogenesis.7, 8

Structural Information

The BMPs/GDFs are 30-38 kDa homodimers that are synthesized as prepropeptides of approximately 400-525 amino acids (aa).6, 7, 9 Cleavage of the variable length pro-segment occurs prior to secretion.7, 10 Secretion of the 100-140 aa C-terminal mature segment forms a dimer, sometimes non-disulfide-linked (GDF-3, GDF-9, and BMP-15).11 Although homodimers are considered the standard form, there are natural heterodimers with equal, if not increased, bioactivity.12, 13 Unlike transforming growth factor beta (TGF-beta), secreted BMP proforms apparently do not form latent complexes with their mature counterparts.7 Mature BMP segments are all assumed to form a cysteine knot with six conserved cysteine residues,7, 8, 14 and there may be an additional one to three cysteines. The presence of N-linked glycosylation sites is variable. There is considerable cross-species bioactivity for the BMPs.15-18 The BMPs/GDFs have been grouped into subsets based on aa sequence homology. The groupings are suggested to be 1) BMP-2 and BMP-4; 2) BMP-3 and BMP-3b; 3) BMP-5, BMP-6, BMP-7, and BMP-8; 4) BMP-9 and BMP-10; 5) BMP-12, BMP-13, and BMP-14; and 6) BMP-11 and GDF-8.6, 7, 12, 14

 

A short summary highlighting the structural features of the human BMPs/GDFs follows. MIS is also provided, and TGF-beta is discussed because it is the prototype for the superfamily.

TGF-beta1:

Human TGF-beta is a 55 kDa, 319 aa preproprotein that consists of a 23 aa signal sequence, a 256 aa pro-region, and a 112 aa mature segment.10, 19 The pro-region is characterized by the presence of three potential N-linked glycosylation sites, while the mature segment contains nine cysteines and no N-lniked glycosylation sites. Prior to secretion, the pro-region is cleaved at an RxxR site with a furin-like protease. This generates a non-glycosylated, 25 kDa, disulfide-linked mature dimer that non-covalently associates with its previously attached disulfide-linked pro-regions to form a "latent complex".20, 21 This complex is secreted. Activation occurs extracellularly under a variety of conditions. This is significant because BMP family members have no known latent form. Human and mouse mature TGF-beta1 are 99% identical at the aa level.22

TGF-beta2:

TGF-beta2 is a 414 aa preproprotein with a mature region of 112 aa. As with TGF-beta1, TGF-beta2 contains nine cysteines and no potential N-linked glycosylation sites. At the aa level, mature human TGF-beta2 is 71% identical to human TGF-beta1, while human TGF-beta2 is 97% identical to mouse TGF-beta2 in the mature segment.23

TGF-beta3:

TGF-beta3 is synthesized as a 410 aa preproprotein. As with TGF-beta1 and TGF-beta2, mature TGF-beta3 is 112 aa long with nine conserved cysteines and no potential N-linked glycosylation sites. In the mature segment, TGF-beta3 is 76% identical to TGF-beta1 and 80% identical to TGF-beta2.24

MIS:

Full length MIS (MüllerianInhibiting Substance) is a 70-74 kDa, 560 aa glycosylated preproprotein that contains a 21 aa signal sequence, a 4 aa pro-region, and a 535 aa mature segment. The mature segment is synthesized as a disulfide-linked homodimer of 140 kDa. Although the 140 kDa unit is functional, up to 20% of the secreted molecule is further processed into a 115 kDa N-terminal region and a 25 kDa, TGF-beta-like C-terminal segment. Each of the chains that make up this short segment dimer are 12 kDa, 109 aa long and contain seven cysteines. As with TGF-beta, there can be either two or three N-terminal chains for every 25 kDa dimer.25, 26 Rat MIS is 92% identical to human MIS over its C-terminal 108 aa residues.27

BMP-1:

BMP-1 is not a member of the TGF-beta superfamily, and thus is not technically a BMP. Its name is due to a misreading of the original bioassay for osteogenesis that yielded BMP-2, -3, and -4.28-30 Structurally, BMP-1 is a 730 aa, cysteine-rich zinc-peptidase that cleaves pro-collagens I, II, and III, yielding fragments that subsequently self-associate into mature collagen fibrils.29, 31 It has been suggested that BMP-1 may activate latent TGF-beta1, but this has not been proven.31

BMP-2/BMP-2alpha:

Full length BMP-2 is a 396 aa glycosylated polypeptide that has a 19 aa signal sequence, a 263 aa pro-region, and a 114 aa mature segment. The mature region has seven cysteines and one N-linked glycosylation site. With a predicted mass of 14 kDa, the mature segment is actually 18 kDa and is assumed to be glycosylated. The functional form of the molecule consists of two disulfide-linked mature chains.28 There is 100% aa identity in the mature region of human, mouse, and rat protein.28, 32-34

BMP-3/Osteogenin:

BMP-3 is synthesized as a 472 aa precursor with a 362 aa prepropeptide and a 110 aa mature segment. The mature segment has a 13 kDa predicted, and a 16 kDa native molecular weight, indicating extensive glycosylation. As with other BMPs, the functional form of the protein is a disulfide-linked homodimer.28 The mature segments of human and rat are 98% identical at the aa level.28, 35, 36

BMP-3b/GDF-10:

Human BMP-3b is a 478 aa prepropolypeptide with a predicted MW of 52 kDa. Its mature segment is 110 aa long and contains seven conserved cysteines plus one potential N-linked glycosylation site. The mature segments for BMP-3b and BMP-3 are 82% identical at the aa level. The mature segments of human and mouse BMP-3b differ by 3 aa (97% identity) while human and rat BMP-3b differ by 2 aa (98% identity). There is only one aa difference between mouse and rat BMP-3b mature segments (99% identity).37-39

BMP-4/BMP-2b:

BMP-4 is a 408 aa prepropeptide with a 19 aa signal sequence, a 273 aa pro-region, and a 116 aa mature segment. Both the pro-region and mature segment contain two potential N-linked glycosylation sites, with the mature region also containing seven cysteines.28, 40 Mouse and rat BMP-4 are 100% identical in the mature segment.41-44 Comparison of the mature regions of human, mouse and rat reveals 98% aa identity.

BMP-5:

BMP-5 is synthesized as a 454 aa preproprotein that is cleaved to yield a 16 kDa, 138 aa mature polypeptide. The mature molecule contains seven cysteines plus three potential N-linked glycosylation sites. The degree to which it is glycosylated is uncertain.45 Mature human and mouse BMP-5 are 96% identical at the aa level.46

BMP-6/Vgr-1:

Mature BMP-6 is formed from a rather large precursor of 513 aa. When cleaved, it creates a 46 kDa precursor and an 18-23 kDa mature segment. The 139 aa mature segment contains seven cysteines plus three potential N-linked glycosylation sites, and it is suggested that the molecule is glycosylated (the predicted MW is 16 kDa).45, 47 In mice, BMP-6 is not always cleaved prior to secretion. It is estimated that 50% of the precursor remains intact.47 Mature mouse and human BMP-6 are 96% identical (6 aa differences over 139 aa). The name Vgr-1 is derived from the similarity of the BMP-6 sequence to frog Vg-1 protein (Vg-1-related, or Vgr-1).48

BMP-7/Op-1:

Human BMP-7, or osteogenic protein-1 (Op-1), is a 49 kDa, 431 aa preproprotein that is cleaved into a 292 aa preproregion and a 139 aa mature segment. The mature segment contains three potential N-linked glycosylation sites plus seven cysteine residues.45, 49 There is 98% aa identity in the mature polypeptides of mouse and human BMP-7.50 There is 75% aa identity within the mature regions of human BMP-5, -6, and -7.45

BMP-8/BMP-8a/Op-2:

BMP-8 is synthesized as a 402 aa prepropeptide that gives rise to a 19 aa signal sequence, a 244 aa pro-region and a 139 aa mature segment. Like most BMP mature segments, the BMP-8 mature polypeptide contains one potential N-linked glycosylation site. However, BMP-8 contains 8 cysteines vs. the seven cysteines typical of other BMPs. There are 13 aa differences over 139 aa, for an aa identity of 93.5% between mouse and human BMP-8. The aa sequences of the mature segments of human Op-2 and human Op-1 are 57% identical.51

BMP-8b/Op-3:

BMP-8b apparently occurs only in mice. The BMP-8b gene is suggested to have arisen by BMP-8 gene duplication. The 139 aa mature segments for BMP-8 and 8b are 75% aa identical, whereas the 241 aa pro-regions for each molecule have 87% aa identity.52

BMP-9/GDF-2:

Currently, only the mature BMP-9 sequence has been reported. The mature region is 110 aa long and contains seven cysteines. There are 6 aa differences over the 110 aa mature region between the mouse and human BMP-9 sequence (95% identity). Based on mouse BMP-9, the full length human BMP-9 molecule will span 428 aa.53, 54 BMP-9 may not be involved in bone morphogenesis thus suggesting a novel function for this molecule.55

BMP-10:

BMP-10 is a 424 aa preproprotein that contains a 316 aa preprosequence and a 108 aa mature segment. There are seven cysteines in the mature segment. Human and bovine BMP-10 are 100% aa identical in their mature regions.56, 57

BMP-11/GDF-11:

Mature BMP-11 is 12 kDa and 109 aa long. Somewhat analogous to the TGF-betas, the mature segment contains 9 cysteine residues. Mature human and bovine BMP-11 sequences are identical.58

BMP-12/GDF-7/CDMP-3:

The mature segment for BMP-12 is an 11 kDa, 104 aa polypeptide that contains seven cysteines. Mouse and human BMP-12 mature segments show 98% aa identity.59 CDMP is an acronym for cartilage-derived morphogenetic protein.

BMP-13/GDF-6/CDMP-2:

Only the mature form of BMP-13 has been reported in its entirety (120 aa long with seven cysteines). The mature segments of human and mouse BMP-13 differ by only one aa over their entire length, yielding an aa identity greater than 99%.59

BMP-14/GDF-5/CDMP-1:

BMP-14 is a 501 aa prepropeptide with a predicted MW of 55 kDa. It consists of a 19 aa signal sequence, a 362 aa pro-region, and a 120 aa mature peptide. The mature region contains seven cysteines and shows no potential for N-linked glycosylation. Thus, the disulfide-linked dimer is approximately 27 kDa. The mature BMP-14 segments of mouse and human show 98% aa identity.60, 61

BMP-15:

The mature form of BMP-15 is 125 aa long and contains six cysteines.62, 63 The mature mouse BMP-15 molecule contains eight cysteines and has 71% overall aa identity with human BMP-15.64 Although the human molecule lacks the seventh cysteine residue, which is believed to participate in intermolecular disulfide linkages, it may form homodimers through noncovalent linkages.

GDF-1:

GDF-1 is a 39 kDa, 372 aa molecule with a 253 aa preproregion and a 119 aa mature segment. The mature region contains seven cysteines. The mature segments of mouse and human GDF-1 show only 82% aa identity.65, 66 The gene encoding the GDF-1 sequence is bicistronic. In the region 5’ to the GDF-1 open reading frame, lies a second open reading frame that codes for a 7-transmembrane receptor with unknown function.65

GDF-3/Vgr-2:

Human GDF-3 is a 313 aa prepropeptide with a predicted MW of 35 kDa. Following cleavage, a 114 aa mature segment is released. Although the mature region contains seven cysteines, the cysteine that is believed to participate in intermolecular bonding is absent. This suggests that GDF-3 homodimers are non-disulfide-linked.67 In mice, GDF-3 is 366 aa long with a 114 aa mature segment.68 The mature polypeptides of human and mouse GDF-3 are 83% identical at the aa level.

GDF-8:

GDF-8, otherwise known as myostatin, is a 375 aa, 50 kDa prepropeptide that is cleaved into a 267 aa, 35 kDa prepropeptide and a 12.5 kDa, 109 aa mature segment. The mature region of GDF-8 has nine cysteines and shows 100% aa identity between human, mouse, rat and cow. It is suggested that BMP-11 and GDF-8 may represent a new BMP subgrouping.69-71

GDF-9:

Full-sized human GDF-9 is 56 kDa and 454 aa long. When cleaved, it generates a glycosylated 37 kDa pro-region and a 16 kDa mature segment. The mature segment is 135 aa long and contains only six cysteines. Thus, it is assumed that GDF-9 dimers are noncovalently linked.72 The mature segments of mouse and human GDF-9 are 90% identical at the aa level.68

Receptors

There are at least three types of TGF-beta receptors; a 53 kDa type I receptor, a 70-85 kDa type II receptor and a 200-400 kDa type III receptor.73 Only type I and type II receptors appear to play significant roles in BMP binding and signaling. Although each receptor type seems to play a distinct role in BMP biology, there are a number of similarities between these two receptors in the BMP system. For instance, type I and type II receptors are both considered to be serine/threonine kinase receptors, have relatively short extracellular regions containing a cysteine “box”, can bind select BMPs and exist constitutively as homodimers in the absence of ligand.73, 74 Differences between the receptors are notable, however. Type II receptors have a 21 aa cytoplasmic C-terminal extension which, although capable of being phosphorylated, is not necessary for activity. Type II receptors also constitutively autophosphorylate on various serines, which is required for their subsequent association with type I receptors. Finally, type II receptors may be the primary ligand-binding subunit in the functional receptor-binding complex.

Type I receptors have no cytoplasmic C-terminal extension, but do possess a 20-30 aa glycine-serine (GS) rich domain between their transmembrane region and kinase domain. The GS region is extremely important for type II-type I receptor interaction. Following ligand binding to the type II receptor homodimer, the type II receptor cross-phosphorylates the type I receptor in the GS region. This phosphorylation activates the type I receptor kinase domain and initiates downstream signaling. Current theory suggests that type I and type II receptors are either constitutively homodimerized or constitutively heterotetramerized (i.e., a complex of two homodimers). Heterotetramerization may occur independently of ligand interaction because a constitutively autophosphorylated type II receptor naturally recruits type I receptors. Upon ligand binding, “functional” or signaling heterotetramerization occurs, with the type II receptor phosphorylating serines and threonines in the GS domain of the type I receptor. This phosphorylation activates the type I receptor kinase domain initiating phosphorylation of cytoplasmic Smad proteins and signal transduction.74, 75

Type I Receptors

ALK-1:

ALK-1 (Activin-like kinase-1) is a 62-72 kDa, 503 aa type I transmembrane glycoprotein that contains a 21 aa signal sequence, a 97 aa extracellular region, a 23 aa transmembrane domain and a 362 aa cytoplasmic segment. The extracellular region contains 10 cysteines plus one potential N-linked glycosylation site.75 Mouse and human full-length ALK-1 are 95% identical at the aa level.77 Mouse and rat full-length ALK-1 are 98% identical at the aa level.77, 78 Although the function of ALK-1 is not well understood, some evidence suggests it may be involved in TGF-beta1 binding.79

ALK-2/ActRI:

Human ALK-2 (ActRI/activin receptor type I) is a 53-55 kDa, 509 aa type I transmembrane glycoprotein that contains a 108 aa extracellular region and a 363 aa cytoplasmic domain.76, 79 The extracellular region contains 10 cysteines and one potential N-linked glycosylation site. The extracellular domains of mouse and human ALK-2 show 96% aa identity.76, 80 The extracellular domains of mouse and rat ALK-2 show 93% aa identity.78, 80 ALK-2 is the only known type I receptor currently suggested to interact with the type II receptor for MIS.81

ALK-3/BMPRIA:

BMPRIA is a 64 kDa, 532 aa type I transmembrane glycoprotein that contains a 23 aa signal sequence, a 129 aa extracellular region, a 24 aa transmembrane domain and a 356 aa cytoplasmic segment.76 The extracellular region contains 10 cysteines and one potential N-linked glycosylation site. There is 98% aa identity between mouse and human ALK-3.76, 82, 83 This receptor is known to bind both BMP-2 and BMP-4 with Kds in the range of 250-500 pM.83

ALK-4/ActRIB:

ALK-4 is a 505 aa type I transmembrane protein with a 22 aa signal sequence, a 104 aa extracellular segment, a 23 aa transmembrane domain and a 356 aa cytoplasmic region. The extracellular segment contains 10 conserved cysteines plus one potential N-linked glycosylation site..84, 85 There appears to be three alternately spliced isoforms of the receptor that occur in endothelial cells.84, 85 Activin A binding to an ActRII-ActRIB complex exhibits a Kd of approximately 100 pM.84 There is 98% aa identity over the entire length of the receptor between mouse and human ALK-4, with 93% aa identity noted in the extracellular segment.84, 86 The extracellular segments of human and rat ALK-4 differ by 5 aa (aa identity of 95%).78, 84

ALK-5/TGF-beta RI:

ALK-5 is a 53-57 kDa, 503 aa type I transmembrane glycoprotein that contains a 503 aa signal sequence, a 101 aa extracellular region, a 22 aa transmembrane domain and a 355 aa cytoplasmic segment. Consistent with other type I TGF-beta superfamily receptors, the extracellular region contains 10 cysteines and one potential N-linked glycosylation site.87 The extracellular segment of human and mouse ALK-5 is 90% identical at the aa level. Between human and rat ALK-5, there is 93% aa identity in the extracellular region.78, 87, 88 The aa sequences of mouse and rat extracellular regions of ALK-5 are 97% identical.78, 88 In the mouse, there is apparently alternate splicing in the extracellular domain.88

ALK-6/BMPRIB:

Human ALK-6 is a 502 aa type I transmembrane protein that contains a 113 aa extracellular region and a 355 aa cytoplasmic domain.89 Mouse and human ALK-6 differ by only four aa in the extracellular region, providing an overall aa identity of 96%.89, 90

ALK-7:

ALK-7 has only been reported for rat. It is a 55 kDa type I transmembrane glycoprotein that has conserved 10 cysteines plus two potential N-linked glycosylation sites in the extracellular region.91 Unlike the other type I TGF-beta receptors, ALK-7 has a relatively short extracellular region (about 90 aa). When compared to other type I receptors, this molecule seems to be the most closely related to ALK-4 and ALK-5. It is not known what ligand(s) may bind to ALK-7.

Type II Receptors

TGF-beta RII:

Human TGF-beta RII is an 80 kDa, 565 aa type I transmembrane glycoprotein that contains a 23 aa signal sequence, a 136 aa extracellular region, a 30 aa transmembrane domain, and a 376 aa cytoplasmic segment. The extracellular region contains 12 cysteines plus three potential N-linked glycosylation sites.92 Between mouse and human TGF-beta RII sequences, there is 83% aa identity in the extracellular region and 96% aa identity in the cytoplasmic segment.92, 93 When the extracellular regions are aligned, TGF-beta RI/ALK-5 and TGF-beta RII demonstrate only 19% aa identity, with conservation of eight cysteines.

ActRII:

Human ActRII is a 70 kDa type I transmembrane glycoprotein that has 99% aa identity with mouse ActRII.94, 95 Approximately 515 aa long, ActRII contains a 116 aa extracellular region and a 352 aa cytoplasmic domain. Embedded in the extracellular region are 10 cysteines and two potential N-linked glycosylation sites. Activin A is known to bind to ActRII with a Kd equal to 180-360 pM. The extracellular region of ActRII shows only 10% aa identity to the extracellular regions in ActRI and ActRIB. For the same region, ActRI to ActRIB, show 19% aa identity.

ActRIIB:

ActRIIB is a 65-70 kDa type I transmembrane glycoprotein that is 512 aa long and contains a 116 aa extracellular region plus a 352 aa cytoplasmic domain. Between human, mouse and rat ActRIIB sequences, there is 98%-99% aa identity with only two aa differences noted in the mature segments.96 The mouse receptor has four splice variants.97 The first two variants (B1 and B2) are long forms and bind activin A with a Kd equal to 100 pM; the last two variants (B3 and B4) are short forms and bind activin A with a Kd equal to 400 pM.97 The single human transcript is analogous to mouse ActRIIB2. There is 51% aa identity in the extracellular region and 75% aa identity in the cytoplasmic segment between ActRIIB2 and ActRII.97

BMPRII:

BMPRII is an approximately 80 kDa, 1038 aa type I transmembrane glycoprotein that contains a very large 866 aa cytoplasmic domain.98-100 Mouse and human BMPRII differ by one aa in the extracellular region and 30 aa in the cytoplasmic segment.98, 101 In both mouse and human, there is an alternatively spliced short form (530 aa) of unknown significance.101 The extracellular regions of human BMPRII and BMPRIA show 16% aa identity and BMPRII and BMPRIB show 18% aa identity. The extracellular regions of BMPRIA/ALK-3 and BMPRIB/ALK-6 are approximately 50% identical.

MISRII:

The type II receptor for MIS is 573 aa long with a 17 aa signal sequence, a 127 aa extracellular region, a 26 aa transmembrane domain and a 403 aa cytoplasmic segment. The extracellular region has 10 cysteines and two potential N-linked glycosylation sites.102 There is approximately 80% aa identity in the extracellular domain between human, mouse and rat MISRII. Mouse and rat MISRII aa sequences are 95% identical.103, 104

Biological Activity

BMPs have been implicated in a variety of functions. BMPs induce the formation of both cartilage and bone.8, 105 During the process of bone formation, BMPs create a rudimentary environment that is conducive to the development of a functional bone marrow.106 In addition, BMPs play a role in a number of non-osteogenic developmental processes. BMP-2 can direct the development of neural crest cells into neuronal phenotypes, while BMP-4 and BMP-7 specifically induce a sympathetic adrenergic phenotype.105, 107 BMP-4 (and possibly GDF-8) gives direction to somite development by inhibiting the process of myogenesis.70, 108 BMPs also appear to be responsible for normal dorsal/ventral patterning. BMP-4 specifies the development of ventral structures (e.g., skin from ectoderm and connective tissue/blood from mesoderm). Dorsal structures (nervous system and muscle) apparently appear when BMP-4 signals are interrupted through the activities of binding proteins.109 In the limb bud, and as part of the FGF-4 and SHH interaction, BMP-2 apparently inhibits limb bud expansion and induces the formation of chondrocyte and osteoblast precursors.110, 111 Finally, it has been reported that a heterodimer composed of BMP-4 and BMP-7 is a potent inducer of mesoderm. Thus, it is possible that natural BMP heterodimers may play specific roles in BMP biology.112

Back to Top

References

  1. Coley, W.B. (1891) Ann. Surg. 14:199.
  2. Neuhof, H. (1917) Surg. Gynec. Obst. 24:383.
  3. Huggins, C.B. (1931) Arch. Surg. 22:377.
  4. Urist, M.R. (1965) Science 150:893.
  5. Huggins, C. et al. (1970) J. Exp. Med. 132:1270.
  6. Yamashita, H. et al. (1996) Bone 19:569.
  7. Hogan, B.L.M. (1996) Genes Dev. 10:1580.
  8. Reddi, A.H. (1998) Nature Biotechnology 16:247.
  9. Wozney, J.M. (1989) Prog. Growth Factor Res. 1:267.
  10. Dubois, C.M. et al. (1995) J. Biol. Chem. 260:10618.
  11. Wozney, J.M. (1992) Mol. Reprod. Dev. 32:160.
  12. Mehler, M.F. et al. (1997) Trends Neurosci. 20:309.
  13. Sampath, T.K. et al. (1990) J. Biol. Chem. 265:13198.
  14. Ebendal, T. et al. (1998) J. Neurosci. Res. 51: 139.
  15. Hughes, F.J. et al. (1995) Endocrinology 136:2671.
  16. Fann, M-J. & P.H. Patterson (1994) J. Neurochem. 63:2074.
  17. Gross, R.E. et al. (1996) Neuron 17:595.
  18. Song, J.J. et al. (1995) Endocrinology 136:4293.
  19. Derynck, R. et al. (1985) Nature 316:701.
  20. Lawrence, D.A. (1996) Eur. Cytokine Netw. 7:363.
  21. Bonewald, L.F. (1991) Mol. Endocrinol. 5:741.
  22. Derynck, R. et al. (1986) J. Biol. Chem. 261:4377.
  23. de Martin, R. et al. (1987) EMBO J. 6:3673.
  24. Derynck, R. et al. (1988) EMBO J. 7:3737.
  25. Cate, R.L. et al. (1986) Cell 45:685.
  26. Pepinsky, R.B. et al. (1988) J. Biol. Chem. 263:18961.
  27. Haqq, C. et al. (1992) Genomics 12:665.
  28. Wozney, J.M. et al. (1988) Science 242:1528.
  29. Kessler, E. et al. (1996) Science 271:360.
  30. Reddi, A.H. (1996) Science 271:463.
  31. Sarras, M.P. (1996) BioEssays 18:439.
  32. Feng, J.Q. et al. (1994) Biochim. Biophys. Acta 1218:221.
  33. SWISS-PROT: Accession # P49001
  34. SWISS-PROT: Accession # P12645
  35. SWISS-PROT: Accession # P49002
  36. Chen, D. et al. (1995) DNA Cell Biol. 14:235.
  37. Hino, J. et al. (1996) Biochem. Biophys. Res. Commun. 223:304.
  38. Cunningham, N.S. et al. (1995) Growth Factors 12:99.
  39. Takao, M. et al. (1996) Biochem. Biophys. Res. Commun. 219:656.
  40. SWISS-PROT: Accession # P12644
  41. SWISS-PROT: Accession # P21275
  42. Feng, J.Q. et al. (1995) J. Biol. Chem. 270:28364.
  43. SWISS-PROT: Accession # Q06826
  44. Chen, D. et al. (1993) Biochim. Biophys. Acta 1174:289.
  45. Celeste, A.J. et al. (1990) Proc. Natl. Acad. Sci. USA 87:9843.
  46. King, J.A. et al. (1994) Dev. Biol. 166:112.
  47. Gitelman, S.E. et al. (1994) J. Cell Biol. 126:1595.
  48. Lyons, K. et al. (1989) Proc. Natl. Acad. Sci. USA 86:4554.
  49. Ozkaynak, E. et al. (1990) EMBO J. 9:2085.
  50. Ozkaynak, E. et al. (1991) Biochem. Biophys. Res. Commun. 179:116.
  51. Ozkaynak, E. et al. (1992) J. Biol. Chem. 267:25220.
  52. Zhao, G-Q.& B.L.M. Hogan (1996) Mech. Dev. 57:159.
  53. Wozney, J.M. et al. (1997) U.S. Patent # 5661007
  54. GENBANK-PROT: Accession # I63216
  55. Celeste, A.J. et al. (1994) J. Bone Mineral Res. 9 (Suppl 1):S136.
  56. Celeste, A.J. et al. (1997) U.S. Patent # 5637480
  57. GENBANK-PROT: Accession # I87652
  58. Wozney, J.M. et al. (1997) U.S. Patent # 5639638
  59. Celeste, A.J. et al. (1997) U.S. Patent # 5658882
  60. Chang, S.C. et al. (1994) J. Biol. Chem. 269:28227.
  61. Storm, E.E. et al. (1994) Nature 368:639.
  62. Celeste, A.J. et al. (1998) U.S. Patent # 5728679
  63. GENBANK-PROT: Accession # I45043
  64. GENBANK-PROT: Accession # I45042
  65. Lee, S-J. (1991) Proc. Natl. Acad. Sci. USA 88:4250.
  66. Lee-S-J. (1990) Mol. Endocrinol. 4:1034.
  67. Caricasole, A.A.D. et al. (1998) Oncogene 16:95.
  68. McPherron, A.C. & S-J. Lee (1993) J. Biol. Chem. 268:3444.
  69. Jones, C.M. et al. (1992) Mol. Endocrinol. 6:1961.
  70. McPherron, A.C. & S-J. Lee (1997) Proc. Natl. Acad. Sci. USA 94:12457.
  71. McPherron, A.C. et al. (1997) Nature 387:83.
  72. McGrath, S.A. et al. (1995) Mol. Endocrinol. 9:131.
  73. Massague, J. et al. (1994) Trends Cell Biol. 4:172.
  74. Derynck, R. & X-H. Feng (1997) Biochim. Biophys. Acta 1333:F105.
  75. Cho, K.W.Y. & I.L. Blitz (1998) Curr. Opin. Genet. Dev. 8:443.
  76. ten Dijke, P. et al. (1993) Oncogene 8:2879.
  77. Wu, X. et al. (1995) Biochem. Biophys. Res. Commun. 216:78.
  78. He, W.W. et al. (1993) Dev. Dyn. 196:133.
  79. Attisano, L. et al. (1993) Cell 75:671.
  80. Ebner, R. et al. (1993) Science 260:1344.
  81. Teixeira, J. et al. (1996) Endocrinology 137:160.
  82. Dewulf, N. et al. (1995) Endocrinology 136:2652.
  83. Koenig, B.B. et al. (1994) Mol. Cell. Biol. 14:5961.
  84. Carcamo, J. et al. (1994) EMBO J. 14:3810.
  85. Xu, J. et al. (1994) Proc. Natl. Acad. Sci. USA 91:7957.
  86. EMBL-PROT: Accession # A38819.
  87. Franzen, P. et al. (1993) Cell 75:681.
  88. Tomoda, T. et al. (1994) Biochem. Biophys. Res. Commun. 198:1054.
  89. GENBANK-PROT: Accession # U89326
  90. ten Dijke, P. et al. (1994) Science 264:101.
  91. Ryden, M. et al. (1996) J. Biol. Chem. 271:30603.
  92. Lin, H.Y. et al. (1992) Cell 68:775.
  93. Lawler, S. et al. (1994) Development 120:165.
  94. Donaldson, C.J. et al. (1992) Biochem. Biophys. Res. Commun. 184:310.
  95. Mathews, L.S. & W.W. Vale (1991) Cell 65:973.
  96. Hilden, K. et al. (1994) Blood 83:2163.
  97. Attisano, L. et al. (1992) Cell 68:97.
  98. Rosenzweig, B.L. et al. (1995) Proc. Natl. Acad. Sci. USA 92:7632.
  99. Nohno, T. et al. (1995) J. Biol. Chem. 270:22522.
  100. Liu, F. et al. (1995) Mol. Cell. Biol. 15:3479.
  101. Beppu, H. et al. (1997) Biochem. Biophys. Res. Commun. 235:499.
  102. Imbeaud, S. et al. (1995) Nature Genet. 11:382.
  103. Mishina, Y. et al. (1997) Biochem. Biophys. Res. Commun. 237:741.
  104. Baarends, W.M. et al. (1994) Development 120:189.
  105. Moses, H.L. & R. Serra (1996) Curr. Opin. Genet. Dev. 6:581.
  106. An, J. et al. (1996) Exp. Hematol. 24:768.
  107. Reissman, E. et al. (1996) Development 122:2079.
  108. Tajbakhsh, S. & G. Cossu (1997) Curr. Opin. Genet. Dev. 7:634.
  109. Graff, J.M. (1997) Cell 89:171.
  110. Niswander, L. & G.R. Martin (1993) Nature 361:68.
  111. Wall, N.A. & B.L.M. Hogan (1994) Curr. Opin. Genet. Dev. 4:517.
  112. Suzuki, A. et al. (1997) Biochem. Biophys. Res. Commun. 232:153.