We have approached the study of growth factors affecting cartilage and bone development by investigating those factors present in bone which are able to initiate new cartilage and bone formation in vivo. This has led to the identification and molecular cloning of seven novel human factors which we have named BMP-1 through BMP-7. Six of these molecules are related to each other, and are also distantly related to TGF-β. The presence of one of these molecules, recombinant human BMP-2 (rhBMP-2) is sufficient to produce the complex developmental system of cartilage and bone formation when implanted subcutaneously in a rat assay system. In this model, administration of rhBMP-2 ultimately results in the formation of a piece of trabecular bone, which is filled with mature bone marrow. While our studies demonstrate that rhBMP-2 by itself has the ability to induce cartilage and bone formation in vivo, we find other BMP molecules present along with BMP-2 in our highly purified nonrecombinant bone-inductive material. These results suggest that the bone inductive capacity of bone-derived proteins may reside in the combinatorial or synergistic activities of this set of BMP-2 related molecules.

The growth and maintenance of bone tissue are complex processes influenced by many systemic hormones and locally produced growth factors. Parathyroid hormone, calcitonin, and the vitamin D metabolites are known effectors of bone remodeling. Growth factors such as TGF-β1, TGF-β2, bFGF, aFGF, IGF-I and IGF-II all are present in significant quantities in bone matrix and have been demonstrated to have various effects on the growth properties and functions of bone cells in vitro (Hauschka et al. 1988; Canalis et al. 1988; Seyedin et al. 1985). In spite of these studies, the knowledge of which factors initiate and control the formation of cartilage and bone, both in the adult and during embryogenesis, is minimal. For example, morphological investigations of bone development during embryogenesis have indicated that bone can form in two distinct sequences: through direct condensation of mesenchyme (intramembranous bone formation) and through the formation and removal of a cartilage model (endochondral bone formation). The end result of each of these pathways is a functional bone with bone marrow. Whether or not the cell types and initiating signalling systems are the same for both of these pathways is unknown.

In the adult, protein extracts of bone are able to induce bone formation through the endochondral process. This activity, often referred to as bone morphogenetic protein, or BMP, initiates and possibly maintains a sequence of events in vivo which culminates in bone formation (Urist et al. 1973). The cellular events caused by BMP have been described by histological examination of BMP implanted subcutaneously in a rat assay system (Reddi, 1981). These include chemotactic events (infiltration of the implant site with cartilage and bone cell precursors), proliferative events, differentiation of precursor cells into chondrocytes, induction of vascularization, maturation of the chondrocytes and differentiation of cells into osteoblasts. This complex process results in a new piece of bone tissue, complete with osteoblasts, osteocytes, osteoclasts, and bone marrow elements. Through our investigations, we have discovered growth factors responsible for this series of events, and have begun examining which of the many specific events in this process they induce.

Identification and molecular cloning of growth factors responsible for BMP activity

The extensive (over 300 000-fold) purification of BMP activity from bovine bone has been described previously (Wang et al. 1988). Biochemical analysis of the proteins present in the most active bone-derived BMP indicated the presence of several polypeptides of approximately 30 × 103Mr, the majority of which are found as polypeptides of 16 and 18 × 103Mr after reduction. Further purification of these 30 ×103Mr species resulted in substantial loss of activity, and as disulfide reduction destroys BMP activity, the individual polypeptides could not be assayed. Trypsin digestion of the 30 × 103Mr nonreduced material, as well as the 16 and 18 ×103Mr reduced components, produced a set of short peptides whose amino acid sequences were determined. These sequences were then used to design and synthesize oligonucleotide probes, which were used to screen bovine genomic or cDNA libraries. Recombinant clones were identified which accounted for the various sequences derived from bovine BMP. These clones were then used to screen human cDNA libraries to obtain recombinant clones encoding the complete human equivalents of the bovine proteins. Using this strategy, we have obtained the sequences of seven novel human proteins, which we call BMP-1 through BMP-7 (Wozney et al. 1988; Wozney, 1989; Celeste et al. unpublished).

Six of these seven proteins (BMP-2 through BMP-7) are closely related to each other and are distantly related to TGF-β. A schematic representation of their structural features is given in Fig. 1. Like all molecules in the TGF-β family, the BMPs are synthesized as precursor proteins. Their primary translation products, as derived from the cDNA sequences, contain hydrophobic secretory leader sequences as well as substantial propeptides. The mature portion of each molecule, as determined by homology to other members of the TGF-β family, constitutes the carboxy-terminal portion of the precursor peptide. Furthermore, all of the peptide sequences found in bovine BMP have their counterparts in these regions of the molecules. Each of the molecules contains seven cysteine residues in their carboxy-terminal portions, in positions corresponding to those present in all members of the TGF-β superfamily. Unlike TGF-β, all the mature BMP proteins contain potential N-linked glycosylation sites. Indeed, the BMP proteins found in bovine bone do contain 2–3 ×103Mr of carbohydrate per chain, as evidenced by reduction of their molecular weights with N-glycanase (Wang et al. 1988). The propeptide portions also contain potential N-linked glycosylation sites, similar to TGF-β No Arg-Gly-Asp potential cell recognition sequences are present as they are in the TGF-β1 and TGF-β3 propeptide sequences. Three of the BMP family members, BMP-2, BMP-4, and BMP-7 do not contain cysteine residues in their propeptides. The presence of cysteine residues required for dimerization of the propeptide have been implicated in the formation of a latent (inactive) complex between TGF-β and its precursor (Brunner et al. 1989). This suggests that these three BMPs are secreted in active forms.

Fig. 1.

Schematic representation of the human TGF-β-like BMPs. The amino acid sequence of each derived from the full-length cDNA is represented by a bar. Each precursor peptide contains a hydrophobic secretory leader sequence, a substantial propeptide sequence (1ight shading), and the mature region (filled shading) at the carboxy terminus. Potential N-linked glycosylation sites (Asn-X-Ser/Thr) are represented by triangles (▾) above the bars. Cysteine residues (exclusive of the leader sequences) are indicated by (c) below the bars. The molecule originally designated BMP-2A (Wozney et al. 1988) is now called BMP-2; BMP-2B has been renamed BMP-4.

Fig. 1.

Schematic representation of the human TGF-β-like BMPs. The amino acid sequence of each derived from the full-length cDNA is represented by a bar. Each precursor peptide contains a hydrophobic secretory leader sequence, a substantial propeptide sequence (1ight shading), and the mature region (filled shading) at the carboxy terminus. Potential N-linked glycosylation sites (Asn-X-Ser/Thr) are represented by triangles (▾) above the bars. Cysteine residues (exclusive of the leader sequences) are indicated by (c) below the bars. The molecule originally designated BMP-2A (Wozney et al. 1988) is now called BMP-2; BMP-2B has been renamed BMP-4.

Interrelationships of the BMP proteins and homologies with other TGF-β superfamily members

Examination of the amino acid sequences of the human BMP proteins demonstrates that they are closely related to each other, and that they form a subgroup of the TGF-β superfamily. When the mature regions of the BMPs are aligned, it is evident that, in addition to the seven cysteine residues present in all members of the TGF-β superfamily, substantial regions of sequence identity exist. Analysis of the amino acid identities between all pairs of the BMPs indicates that these molecules can be divided into three subfamilies. BMP-2 and BMP-4 are quite closely related (86% amino acid identity in this region, see Fig. 2A) and form one group. BMP-5, BMP-6, and BMP-7 are also closely related to one another (an average of 78%), and form a second group. BMP-3, while more closely related to the other BMP proteins than to other TGF-β family members, is the most different and by itself forms the third group. The relationships of these groups of BMP proteins to representative members of the TGF-β family are given in Fig. 2B, comparing the sequences from the regions containing the seven cysteine residues. While distantly related to TGF-β1, TGF-β2, and TGF-β3 (35% sequence identity), they are more closely related to the inhibin β family of molecules (41%). All of the BMP proteins are even more closely related to the developmentally implicated proteins Vgl from Xenopus (Weeks and Melton, 1987) and dpp from Drosophila (Padgett et al. 1987). From the substantial homology between dpp and the BMP-2 and BMP-4 subfamily (75%), as well as substantial regions of sequence identity in the propeptide regions, we infer that either (or both) are the human equivalents of the dpp gene product. Human BMP-6 shows 91% sequence identity across the entire precursor molecule with Vgr-1, a polypeptide defined by the derived amino acid sequence of a cDNA isolated from a mouse embryo cDNA library by cross-hybridization with Vgl (Lyons et al. 1989). For this reason, we presume that Vgr-1 is the murine homolog of BMP-6.

Fig. 2.

Inter-relationships of the BMPs and relationship to TGF-β superfamily members. (A) Amino acid identities between members of a subgroup of the BMPs to each other in either the propeptide (PRO) or mature regions. (B) Amino acid identities to other members of the TGF-β family. Numbers are an average of percent amino acid identities of pairwise comparisons aligned using the UWGCG sequence analysis system.

Fig. 2.

Inter-relationships of the BMPs and relationship to TGF-β superfamily members. (A) Amino acid identities between members of a subgroup of the BMPs to each other in either the propeptide (PRO) or mature regions. (B) Amino acid identities to other members of the TGF-β family. Numbers are an average of percent amino acid identities of pairwise comparisons aligned using the UWGCG sequence analysis system.

Distribution of mRNAs for the BMP proteins

As part of an effort to understand the roles of the BMP proteins in cartilage and bone formation, as well as their possible growth factor activities in other tissues, we have examined the tissue distribution of mRNAs for BMP-1 through BMP-4 in various bovine tissues. These values are compared with the levels of mRNAs in primary subperiosteal bone cell cultures (a generous gift from Dr Marian Young, N.I.H.) in Table 1. Using the sensitive technique of mRNA protection, low levels of BMP-1, BMP-2, and BMP-4 mRNA can be found in all tissues. With BMP-1, substantially higher levels of mRNA are present in cerebellum and liver; with approximately 3-fold higher levels in bone cells than in cerebellum. BMP-2 mRNA was present at about 40-fold higher levels in bone cells than in any other tissue examined. The relative mRNA distribution of BMP-4 is distinct from that of BMP-2, though the two molecules are extremely closely related. The levels of BMP-4 mRNA found in bone cells are only about 2-fold higher than in lung and kidney. The tissue distribution of BMP-3 mRNA is the most restricted. BMP-3 mRNA is undetectable in most tissues, with some small amount present in cerebrum and significantly higher levels in lung. Interestingly, the mRNAs for all four of these proteins can be found in brain. Many growth factors, including aFGF (Gimenez-Gallego et al. 1985), bFGF (Esch et al. 1985; Emoto et al. 1989), inhibin a, βA, and βB (Sawchenko et al. 1988) have been found in brain, though their roles there are unclear.

Table 1.

Relative distribution of the mRNAs for BMPs

Relative distribution of the mRNAs for BMPs
Relative distribution of the mRNAs for BMPs

Though low levels of BMP-1, BMP-2, and BMP-4 mRNAs are found in all tissues examined, it is difficult to interpret this result. The technique used to detect mRNA in these studies is quite sensitive, and several reports indicate there is a basal expression of any mRNA in any tissue (Chelly et al. 1988; Chelly et al. 1989; Sarkar and Sommer, 1989). Furthermore, the presence of small amounts of mRNA does not necessarily indicate that it is translated into protein. With these caveats in mind, it is interesting that the highest levels of BMP-1 and BMP-2 expression are found in the subperiosteal bone cells, consistent with a role fbr these molecules in bone formation. The levels of BMP-1 mRNA are 3-fold higher in these bone cells than in the next most highly expressing tissue, cerebellum. Almost 40-fold higher levels of BMP-2 mRNA are found in bone cells than in any other tissue examined. While BMP-3 mRNA was undetectable in these bone cells, it should be pointed out that these cells represent a distinct subset of cells of the osteoblastic lineage and other bone cell populations may synthesize BMP-3 mRNA.

In vivo activity of BMP-2

The clear definition of the in vivo and in vitro activities of any growth factor require its purification to absolute biochemical homogeneity and/or expression in a recombinant system. The derivation of mammalian cell lines which stably express BMP-2 has allowed us to examine its activity in the rat ectopic bone formation assay used to identify BMP activity (Wang et al. 1990). In this system, recombinant human BMP-2 (rhBMP-2) is reconstituted with a bone-derived collagenous matrix. Histological examination of the implants removed at various times after implantation of moderate doses (2-20μg) of rhBMP-2 show that rhBMP-2 alone can induce cartilage and bone formation in a subcutaneous site in the rat. The cellular events observed are similar if not identical to those observed with crude or purified nonrecombinant bone-derived BMP. Undifferentiated mesenchymal cells infiltrate the implant site, proliferate, and differentiate into chondroblasts by day 5. These cartilage-forming cells mature, hypertrophy, and mineralize; vascularization of the implant area can be seen at about this time. Osteoblasts appear in the implant site and bone matrix (osteoid) is deposited while the cartilage is removed. The bone matrix is then mineralized, and by day 14 only the newly formed trabecular bone tissue remains, containing osteoblasts, osteoclasts, and bone marrow elements. The normal bone remodeling sequence continues, with osteoblasts depositing bone and osteoclasts resorbing bone. Interestingly, the dose of rhBMP-2 administered affects the time at which these events occur. This process can be accelerated such that bone is observable at day 14 with 0.5 μg, at day 7 with 12 μg, and at day 5 with 115 ug of rhBMP-2. When the latter amount of rhBMP-2 is administered, large amounts of both cartilage and bone formation are observable at day 5. This may suggest that BMP-2 acts on both cartilage and bone induction directly, and that the sequence of cartilage induction followed by bone formation observed with lower amounts of BMP in this assay system is not truly a cascade. While reconstitution of rhBMP-2 with collagenous bone matrix decreases the amount of BMP needed to see cartilage and bone induction, it is not necessary for BMP activity. Implantation of rhBMP-2 without any matrix carrier induces both cartilage and bone formation, as seen in Fig. 3. Therefore our working hypothesis is that the matrix contributes slow release and/or immobilization characteristics to the BMP-2, but does not contribute any additional factors necessary for bone induction. The dissection of which particular events in this complex system of in vivo cartilage and bone induction are affected by BMP-2 will require further studies on the in vivo and in vitro activities of BMP-2.

Fig. 3.

Induction of cartilage and bone by rhBMP-2 without addition of carrier matrix. BMP-2 was implanted subcutaneously in a rat for 9 days. Von Kossa stain, c, calcified cartilage; b, bone; ob, osteoblasts.

Fig. 3.

Induction of cartilage and bone by rhBMP-2 without addition of carrier matrix. BMP-2 was implanted subcutaneously in a rat for 9 days. Von Kossa stain, c, calcified cartilage; b, bone; ob, osteoblasts.

Bone-derived BMP activity

While rhBMP-2 appears to have the complete cartilage and bone inductive activities of BMP derived from bone, there are several lines of evidence to suggest that BMP-2 is not the sole growth factor responsible for this activity. In our work, we have found that the most active bone-derived material contains a mixture of molecules, BMP-1 through BMP-7. Other researchers have also purified bone-inductive substances from bovine bone (Luyten et al. 1989). The reported amino acid sequence of this material, called osteogenin, is the same as that of BMP-3. This may indicate that BMP-3 also has bone-inductive activity by itself, as does BMP-2, but one cannot exclude the possibility that other BMP molecules are also contained in this nonrecombinant material. In addition, the amount of rhBMP-2 necessary to produce in vivo bone induction is an order of magnitude higher than that of highly purified bone-derived BMP. Though it is possible that the recombinant BMP-2 made in our mammalian cell expression system is somehow different from bone-derived BMP-2 and therefore is less active, these facts suggest that natural BMP activity is some combination of the activities of multiple BMP molecules or represents synergistic activity between them. The fact that mRNA for the BMP proteins can be found in many non-bony tissues also supports the argument that multiple factors may normally have to work together to reach a critical concentration to induce bone formation in the animal. As a further complication, since all the TGF-/)-like BMPs form dimers, it is possible that heterodimers of these molecules exist. By analogy to other members of the family, these heterodimers could have increased or even opposite activities to the homodimers themselves. Again, the possibility of heterodimer formation is supported by the finding that many tissues express several of the BMP mRNAs. Production of the rest of the BMP proteins through recombinant means will allow further clarification of their roles in the cartilage and bone induction process.

Brunner
,
A. M.
,
Marquardt
,
H.
,
Malacko
,
A. R.
,
Lioubin
,
M. N.
and
Purchio
,
A. F.
(
1989
).
Site-directed mutagenesis of cysteine residues in the pro region of the transforming growth factor /31 precursor
.
J. biol. Chem.
264
,
13 660
13 664
.
Canalis
,
E.
,
McCarthy
,
T.
and
Centrella
,
M.
(
1988
).
Isolation of growth factors from adult bovine bone
.
Calc. Tiss. Int.
43
,
346
351
.
Chelly
,
J.
,
Concordet
,
J.-P.
,
Kaplan
,
J.-C.
and
Kahn
,
A.
(
1989
).
Illegitimate transcription: transcription of any gene in any cell type
.
Proc. natn. Acad. Sci. U.S.A.
86
,
2617
2621
.
Chelly
,
J.
,
Kaplan
,
J. C.
,
Maire
,
P.
,
Gautron
,
S.
and
Kahn
,
A.
(
1988
).
Transcription of the dystrophin gene in human muscle and non-muscle tissues
.
Nature
333
,
858
860
.
Emoto
,
N.
,
Gonzalez
,
A.-M.
,
Walicke
,
P. A.
,
Wada
,
E.
,
Simmons
,
D. M.
,
Shimasaki
,
S.
and
Baird
,
A.
(
1989
).
Basic fibroblast growth factor (FGF) in the central nervous system: identification of specific loci of basic FGF expression in the rat brain
.
Growth Factors
2
,
21
29
.
Esch
,
F.
,
Baird
,
A.
,
Ling
,
N.
,
Ueno
,
N.
,
Hill
,
F.
,
Denoroy
,
L.
,
Klepper
,
R.
,
Gospodarowicz
,
D.
,
Bohlen
,
P.
and
Guillemin
,
R.
(
1985
).
Primary structure of bovine pituitary basic fibroblast growth factor (FGF) and comparison with the amino-terminal sequence of bovine brain acidic FGF
.
Proc. natn. Acad. Sci. U.S.A.
82
,
6507
6511
.
Gimenez-Gallego
,
G.
,
Rodkey
,
J.
,
Bennett
,
C.
,
Rios-Candelore
,
M.
,
Disalvo
,
J.
and
Thomas
,
K.
(
1985
).
Brain-derived acidic fibroblast growth factor: complete amino acid sequence and homologies
.
Science
230
,
1385
1388
.
Hauschka
,
P. V.
,
Chen
,
T. L.
and
Mavrakos
,
A. E.
(
1988
).
Polypeptide growth factors in bone matrix
. In
Cell and Molecular Biology of Vertebrate Hard Tissues
(ed.
Evered
,
D.
and
Harnett
,
S.
), pp.
207
225
.
Chichester
:
Wiley
.
Luyten
,
F. P.
,
Cunningham
,
N. S.
,
Ma
,
S.
,
Muthukumaran
,
N.
,
Hammonds
,
R. G.
,
Nevins
,
W. B.
,
Wood
,
W. I.
and
Reddi
,
A. H.
(
1989
).
Purification and partial amino acid sequence of osteogenin, a protein initiating bone differentiation
.
J. biol. Chem.
264
,
13 377
13380
.
Lyons
,
K.
,
Graycar
,
J. L.
,
Lee
,
A.
,
Hashmi
,
S.
,
Lindquist
,
P. B.
,
Chen
,
E. Y.
,
Hogan
,
B. L. M.
and
Derynck
,
R.
(
1989
).
Vgr-1, a mammalian gene related to Xenopus Vg-1, is a member of the transforming growth factor f gene superfamily
.
Proc. natn. Acad. Sci. U.S.A.
86
,
4554
4558
.
Padgett
,
R. W.
,
St Johnston
,
R. D.
and
Gelbart
,
W. M.
(
1987
).
A transcript from a Drosophila pattern gene predicts a protein homologous to the transforming growth factor-β family
.
Nature, Lond.
325
,
81
84
.
Reddi
,
A. H.
(
1981
).
Cell biology and biochemistry of endochondral bone development
.
Collagen Rel. Res.
1
,
209
226
.
Sarkar
,
G.
and
Sommer
,
S. S.
(
1989
).
Access to a messenger RNA sequence or its protein product is not limited by tissue or species specificity
.
Science
244
,
331
334
.
Sawchenko
,
P. E.
,
Plotsky
,
P. M.
,
Pfeiffer
,
S. W.
,
Cunningham
,
E. T”
Jr
,
Vaughan
,
J.
,
Rivier
,
J.
and
Vale
,
W.
(
1988
).
Inhibin βin central neural pathways involved in the control of oxytocin secretion
.
Nature
334
,
615
617
.
Seyedin
,
S. M.
,
Thomas
,
T. C.
,
Thompson
,
A. Y.
,
Rosen
,
D. M.
and
Piez
,
K. A.
(
1985
).
Purification and characterization of two cartilage-inducing factors from bovine demineralized bone
.
Proc. natn. Acad. Sci. U.S.A.
82
,
2267
2271
.
Urist
,
M. R.
,
Iwata
,
H.
,
Ceccotti
,
P. L.
,
Dorfman
,
R. L.
,
Boyd
,
S. D.
,
McDowell
,
R. M.
and
Chien
,
C.
(
1973
).
Bone morphogenesis in implants of insoluble bone gelatin
.
Proc. natn. Acad. Sci. U.S.A.
70
,
3511
3515
.
Wang
,
E. A.
,
Rosen
,
V.
,
Cordes
,
P.
,
Hewick
,
R. M.
,
Kriz
,
M. J.
,
Luxenberg
,
D. P.
,
Sibley
,
B. S.
and
Wozney
,
J. M.
(
1988
).
Purification and characterization of other distinct bone-inducing factors
.
Proc. natn. Acad. Sci. U.S.A.
85
,
9484
9488
.
Wang
,
E. A.
,
Rosen
,
V.
,
D’Alessandro
,
J. S.
,
Bauduy
,
M.
,
Cordes
,
P.
,
Harada
,
T.
,
Israel
,
D.
,
Hewick
,
R. M.
,
Kerns
,
K.
,
Lapan
,
P.
,
Luxenberg
,
D. P.
,
McQuaid
,
D.
,
Moutsatsos
,
I.
,
Nove
,
J.
and
Wozney
,
J. M.
(
1990
).
Recombinant human bone morphogenetic protein induces bone formation
.
Proc. natn. Acad. Sci. U.S.A.
87
,
2220
2224
.
Weeks
,
D. L.
and
Melton
,
D. A.
(
1987
).
A maternal mRNA localized to the vegetal hemisphere in Xenopus eggs codes for a growth factor related to TGF-/1
.
Cell
51
,
861
867
.
Wozney
,
J. M.
(
1989
).
Bone morphogenetic proteins
.
Prog. Growth Factor Res.
1
,
267
280
.
Wozney
,
J. M.
,
Rosen
,
V.
,
Celeste
,
A. J.
,
Mitsock
,
L. M.
,
Whitters
,
M. J.
,
Kriz
,
R. W.
,
Hewick
,
R. M.
and
Wang
,
E. A.
(
1988
).
Novel regulators of bone formation: molecular clones and activities
.
Science
242
,
1528
1534
.
Zinn
,
K.
,
Dimaio
,
D.
and
Maniatis
,
T.
(
1983
).
Identification of two distinct regulatory regions adjacent to the human β-interferon gene
.
Cell
34
,
865
879
.