Bone Morphogenetic Protein 1 Prodomain Specifically Binds and Regulates Signaling by Bone Morphogenetic Proteins 2 and 4*

Highly purified fractions of bone extracts capable of inducing ectopic bone formation have been reported to contain peptides corresponding to the mature active regions of the TGF-β-like bone morphogenetic proteins (BMPs) 2–7, and to the prodomain region of the metalloproteinase BMP1. Co-purification of BMPs 2–7 with BMP1 prodomain sequences through the multiple biochemical steps used in these previous reports has suggested the possibility of interactions between the BMP1 prodomain and BMPs 2–7. Here we demonstrate that the BMP1 prodomain binds BMPs 2 and 4 with high specificity and with a KD of ∼11 nm, in the physiological range. It is further demonstrated that the BMP1 prodomain is capable of modulating signaling by BMPs 2 and 4 in vitro and in vivo, and that endogenous BMP1 prodomain-BMP4 complexes exist in cell culture media and in tissues.

Highly purified fractions of bone morphogenetic activity isolated from bone extracts contain peptides corresponding to bone morphogenetic proteins (BMPs) 2 1-7 (1, 2). BMPs 2-7 are all TGF-␤ superfamily members and are involved not only in formation of bone, but in the formation of various tissues, including the nervous system, heart, and kidneys (3). BMPs 2 and 4 in particular have also been shown to play central roles in determining the dorsal-ventral axis in early embryogenesis (3,4). In contrast to the TGF-␤-like BMPs, BMP1 is the prototype of a subgroup of structurally similar secreted metalloproteinases with astacin-like protease domains (5). Mammalian BMP1-like proteinases play key roles in formation of the extracellular matrix via biosynthetic processing of precursors into mature, functional proteins that include collagen types I-III, V, VII, and XI; laminin 5; proteoglycans biglycan and osteoglycin; and the enzyme lysyl oxidase (5). In addition, BMP1-like proteinases are involved in activating some TGF-␤ family members; including growth differentiation factors (GDFs) 8 and 11, TGF-␤1, and BMPs 2 and 4, via cleavage of extracellular antagonists (5,6). BMPs 2 and 4 are activated via cleavage of the extracellular antagonist Chordin by BMP1-like proteinases (5).
TGF-␤-like factors and BMP1-like proteinases are both synthesized as precursors with NH 2 -terminal prodomains that are excised by subtilisin-like proprotein convertases to, in each case, yield the COOH-terminal biologically active portion of the molecule (3,5,7). Thus, it is not surprising that all BMP 2-7 tryptic peptides obtained from bone morphogenetic fractions correspond to sequences within the COOH-terminal biologically active portion of these proteins, downstream of the proprotein convertase cleavage site (1,2,8). In contrast, all BMP1 tryptic peptides obtained from bone morphogenetic fractions of bone extracts have corresponded to sequences within the NH 2 -terminal prodomain region of this protein (2,8). Thus, the BMP1 prodomain, which is thought to be cleaved upon secretion of BMP1 by fibrogenic cells (9), perhaps within the trans-Golgi network (10), persists in bone and co-purifies through a large number of varied biochemical purification steps (8) with TGF-␤-like BMPs 2-7. Persistence of the BMP1 prodomain in bone suggests a role in bone biology, whereas co-purification through multiple protein separation steps with TGF-␤-like BMPs in morphogenetic fractions of bone extracts suggests that the BMP1 prodomain might interact with the TGF-␤-like proteins.
In the present study, it is demonstrated that the BMP1 prodomain binds active BMP2 and BMP4 with high specificity and high avidity (K D ϭ 10.9 nM); that it is capable of modulating signaling by these BMPs in vitro and in vivo; and that BMP1 prodomain-BMP4 complexes are found in cell culture media and within human tissues.

EXPERIMENTAL PROCEDURES
Expression Constructs-A construct for expression of mutagenized proBMP1, in which the prodomain is impervious to SPC activity, was generated via PCR using reverse primer 5Ј-TCGGGACGTCGCCGCCTGCTGGCTACTGGATCTA-CCTCTCCA-3Ј, which encodes proBMP1 sequences in which sequences encoding the subtilisin-like proprotein convertase recognition site RSRR were converted to sequences encoding * This work was supported by National Institutes of Health Grants AR47746 and GM71679 (to D. S. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  residues SSQQ. ProBMP1 sequences within vector pBluescript II KSϩ (Stratagene) were used as template, and T3 primer was used as the forward primer. The resulting amplimer was digested with AflII and AatII and used to replace the corresponding wild-type fragment in the original proBMP1/pBluescript clone. Subsequently, an AflII-ApaI fragment containing the mutation was excised from the pBluescript construct and exchanged for the corresponding wild-type fragment in fulllength proBMP1 sequences previously placed within vector pcDNA4/TO-His (Invitrogen), such that mutant proBMP1 produced by the resultant construct would have a COOH-terminal His tag.
To create recombinant baculovirus for expression of the BMP1 prodomain, PCR amplification was performed, using primers 5Ј-GAGTAGATCTAGACCTAGCCGACTACACC-TATGACCT-3Ј (forward) and 5Ј-GAGTGCTAGCCGCCGG-CTACGGGATCTACCTCTCC-3Ј (reverse), and using wildtype proBMP1 sequences in pcDNA4/TO as the template. The resulting amplimer was restricted with BglII and NheI, and then inserted between the BamHI and XbaI sites of the pAcGP67.coco baculovirus transfer vector (11) (kind gift of Deane F. Mosher, University of Wisconsin-Madison). The pAcGP67.coco vector is a modification of vector pAcGP67A (BD Biosciences Pharmingen), with sequences encoding the thrombin cleavage site (LVPRGS) and a six-histidine tag (His 6 tag) added immediately 3Ј to the multiple cloning site. Recombinant baculovirus was generated by co-transfection with Baculogold-linearized AcNPV viral DNA (BD Biosciences Pharmingen) into SF9 cells (InVitrogen). Recombinant baculovirus was propagated, and protein was expressed and purified as described by Mosher et al. (11).
To create a construct for expression of a BMP1 prodomain-Fc fusion protein, a KpnI-NgoMI cDNA fragment encoding BMP1 prodomain sequences fused to BM40 signal peptide sequences (12) was inserted between a KpnI site and a BamHI site that had been blunt-ended with Klenow fragment in a construct that contained enterokinase cleavage site sequences, followed by sequences encoding the human IgG 1 Fc domain, in a pcDNA3 backbone (13). The enterokinase site/Fc domain construct was the kind gift of Carl P. Blobel (Hospital for Special Surgery, Cornell University).
To generate mRNA for injection into zebrafish embryos, BMP1 prodomain sequences were PCR-amplified with primers 5Ј-CTGAGAATTCCCAGCACCATGAGGGCC-3Ј (forward) and 5Ј-GAGTTCTAGACTACCGCCGGCTACGGGATCTA-CCT-3Ј (reverse) and inserted between the EcoRI and XbaI sites of vector pCS2ϩ. To generate mRNA for expression of a BMP1 prodomain/CD-2 fusion protein, prodomain sequences were PCR-amplified with primers 5Ј-CTGAGGATCCAGCA-CCATGAGGGCCTGGAT-3Ј (forward) and 5Ј-GACTGA-ATTCCCTAGGCGGCTACGGGATCTACC-3Ј(reverse). The resulting amplimer, which contained AvrII and EcoRI, sites in reverse primer sequences, was digested with BamHI and EcoRI and was inserted between corresponding sites of pCS2ϩ. The resulting construct was restricted with AvrII and EcoRI, for placement of an AvrII-EcoRI cDNA fragment encoding rat CD-2 transdomain and partial exodomain sequences (kind gift from Gary Struhl, Columbia University).
Protein Biochemistry-For expression of recombinant proteins, constructs were transfected into T-REx 293 cells using Lipofectamine TM (Invitrogen). Recombinant proteins with His tags or Fc domains were purified from conditioned serum-free medium using Ni-NTA resin (Qiagen) or protein A-Sepharose (Amersham Biosciences), respectively. FLAG-tagged murine Chordin and wild-type human BMP1 were produced, purified, and analyzed for purity on zinc-stained SDS-PAGE gels as previously described (12). Concentrations of purified proteins were calculated by comparing intensities of Coomassie Bluestained bands from serial dilutions of each sample to those of serially diluted bovine serum albumin of known concentrations. Procollagen and Chordin cleavage assays were performed as previously described (12). Immunoprecipitations of recombinant proteins were performed as previously described (14), using equimolar amounts of purified proteins. For the competition experiment, 5 nM proBMP1 SSQQ was incubated 30 min at 37°C with 5 nM BMP4 in the presence of excess BMP2, BMP5, EGF, or TGF-␤1 (with the extent of excess noted in the text) and immunoprecipitated. Similarly, mBMPR-1A/Fc was incubated with BMP4 under conditions similar to those described above, was competed with equimolar amounts of Chordin or BMP1 Prodomain, and complexes were immunoprecipitated with protein A-Sepharose. For immunoprecipitations from MG-63 cultures, 80% confluent MG-63 cells were washed twice with phosphate-buffered saline (PBS), incubated 15 min with serum-free Dulbecco's modified Eagle's medium at 37°C, followed by two PBS washes. Cells were then incubated 48 h in serum-free Dulbecco's modified Eagle's medium, 40 g/ml soybean trypsin inhibitor, 2 ng/ml TGF-␤1 (R&D systems). Conditioned medium was harvested and cells were washed twice with ice-cold PBS, followed by lysing in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EGTA, 1 mM ZnCl 2 , 1 mM CaCl 2 , 1 mM MgCl 2 , 1% Triton X-100, 1% CHAPS, 10% glycerol; with clearing of lysates by centrifugation. Cell lysate or conditioned medium, with Triton X-100 and CHAPS added to 0.1% final concentrations, was precleared with protein A-Sepharose at 4°C. Anti-BMP1 prodomain antibodies or preimmune serum was then added, followed by incubation with protein A-Sepharose. Immunoprecipitates were subjected to six 10-min washes with PBS/0.5% Triton X-100, followed by elution of samples with SDS sample buffer containing 100 mM dithiothreitol, and separation of samples on 4 -20% acrylamide SDS-polyacrylamide electrophoresis gels (Bio-Rad). Samples were transferred to nitrocellulose membranes and probed with anti-BMP1 prodomain antibodies or monoclonal anti-BMP4 antibody (R&D systems).
Antibodies-Antibody was raised against recombinant BMP1 prodomain produced by the recombinant baculovirus described above. The His-tagged protein was purified with Ni-NTA resin; followed by removal of the His tag by thrombin cleavage, SDS-polyacrylamide electrophoresis on a preparatory gel, and excision of a single band, visualized with Coomassie Blue. The band was equilibrated with PBS, emulsified with an equal volume of Freund's complete adjuvant, and injected subcutaneously into a New Zealand White rabbit. The rabbit was boosted twice at 4-week intervals with 250 g of His-tagged BMP1 prodomain in Freund's incomplete adjuvant per boost. Antibodies were purified on a prodomain affinity column of purified prodomain-His, coupled in 0.1 M NaHCO 3 , 0.5 M NaCl, pH 8.5 to CNBr-activated agarose (Sigma). After washing the column with 10 bed volumes of 500 mM NaCl, 10 mM Tris, pH 7.5, bound antibody was eluted with 10 bed volumes of 100 mM glycine, pH 2.5. Antibody-containing pools were dialyzed against PBS, 0.2% sodium azide and then stored at Ϫ70°C. Specificity of the affinity-purified antibody was ascertained by immunoblotting of conditioned medium of TGF-␤-treated MG-63 cells. Although the latter produce both 104-kDa proBMP1 and 79-kDa mature BMP1 (9), only the former was detected by the antibody (Fig. 4D). For immunoblots, prodomain antibody was used at 1:5000 dilution; anti-BMP4, anti-TGF-␤, and anti-BMPR1A antibodies (R&D) at 1:5000; and anti-BMP2, anti-EGF, and anti-BMP-5 (R&D) at 1:1000.
Surface Plasmon Resonance Analysis-SPR measurements were performed in a BIAcore 2000 system. Recombinant BMP2, the generous gift of Elisabeth Schwarz (16), was dissolved at 2.5 g/ml in 50 mM sodium acetate (pH 4.7) and immobilized on a CM5 sensor chip, using the amine coupling method, to a level of ϳ887 response units. BMP1 Prodomain-Fc fusion protein was purified, and the Fc portion was cleaved off with Enterokinase (Novagen). Enterokinase and Fc domains were removed using Ekapture agarose (Novagen) and protein G-Sepharose, respectively. The BMP2 used for SPR was produced, purified, and analyzed for purity on Coomassie Bluestained SDS-PAGE gels, as previously described (16).
Control flow cells were prepared by executing the coupling reaction in the presence of coupling buffer alone. Control cells were then used to obtain control sensorgrams showing nonspecific binding to the surface as well as refractive index changes resulting from changes in bulk properties of the solution. Control sensorgrams were subtracted from sensorgrams obtained with immobilized ligand to yield accurate binding responses.
Binding and washes were performed in 20 mM Hepes, pH 7.5, 150 mM NaCl, 0.005% Tween-20. Each experimental cycle consisted of an initial 4-min flow of various concentrations of BMP1 prodomain or Chordin into the respective flow cells at a flow rate of 20 l/min, followed by a 4-min flow of buffer. After each cycle, chip surfaces were regenerated by a 5-s flow of 0.1 M HAc, 1 M NaCl, 5 M guanidine-HCl. Data were analyzed with BIAevaluation 4.1 software and curve-fitting was done with the assumption of one-to-one binding. SPR experiments were conducted by G. G. at the Biophysics Instrumentation Facility (Department of Biochemistry), established by funding from NSF (BIR-9512577), NIH (S10 RR13790), and the University of Wisconsin.
Alkaline Phosphatase (AP) BMP Signaling Assay-AP assays were performed as previously described (21). Briefly, M2-10B4 cells were plated in triplicate in 96-well plates at 10,000 cells per well. After 16 h, cells were washed once with PBS and then placed in 200 l of Dulbecco's modified Eagle's medium, 10% fetal bovine serum containing either 6 nM BMP4 (dissolved in 4 mM HCl, 0.1% bovine serum albumin) or 6 nM BMP4 that had been preincubated with proBMP1 or BMP1 prodomain at 37°C. AP activity was determined 48-h later. Cells were washed once with PBS and lysed by freeze-thawing twice in 50 l of 0.2% Nonidet P-40, 1 mM MgCl 2 . 150 l of AP buffer (5 mg of p-nitrophenyl phosphate dissolved in 2.5 ml of Sigma 221 alkaline buffer solution diluted 1:2 with water) was added to each well, and plates were incubated 1 h at 37°C. AP activity was then measured by absorbance readings at 405 nm using a Universal Microplate Reader (Bio-Tek Instruments, Winooski VT).

RESULTS
ProBMP1, but Not Mature BMP1, Specifically Binds BMPs 2 and 4-To begin characterizing possible functional roles of the BMP1 prodomain, site-directed mutagenesis was employed to alter the proBMP1 furin recognition site RSRR to SSQQ (Fig.  1A), to enable comparison of the properties of pro and mature BMP1. When expressed in transfected T-Rex-293 cells, only BMP1 Prodomain Binds BMP2/4 MARCH 23, 2007 • VOLUME 282 • NUMBER 12 full-length proBMP1 SSQQ was detected in conditioned medium, with no evidence of processing to mature BMP1 (Fig.  1B). The proBMP1 SSQQ displayed no detectable activity for processing BMP1 substrates type I procollagen (Fig. 1C) and Chordin (Fig. 1D). Moreover, proBMP1 SSQQ was found to interfere with cleavage of procollagen and Chordin by mature/ active wild-type BMP1 (Fig. 1, C and D). There is evidence that BMP1-like proteinases do not form oligomers (22). Thus, the apparent dominant negative effect of proBMP1 SSQQ likely results from binding to, and sequestering of substrate. The abil-ity of proBMP1 SSQQ to bind substrate, suggests that it is correctly folded.
In an initial attempt to ascertain the ability of BMP1 prodomain sequences to interact with TGF-␤-like BMPs, mature BMP1, and proBMP1 SSQQ were separately incubated with BMP4, and then immunoprecipitated with anti-BMP1 antibody. Immunoprecipitation of proBMP1 SSQQ , but not of mature BMP1, was found to co-precipitate BMP4 (Fig. 2A); consistent with the probability that BMP1 prodomain sequences bind BMP4. Moreover, this binding is highly specific, as closely related BMP2 (93% sequence identity) readily displaced BMP4 from binding to proBMP1 SSQQ , and was itself pulled down by proBMP1 SSQQ from which BMP4 had been displaced (Fig. 2B). Competition of BMPs 2 and 4 for binding to the BMP1 prodomain is not surprising, given their close In the right panel, purified His-tagged proBMP1 SSQQ and BMP1 were visualized by immunoblotting using antibody directed against mature BMP1 (26). C, an autofluorogram is shown of purified human type I procollagen incubated in the absence (Ϫ) or presence (ϩ) of BMP1 and/or proBMP1. pN␣1(I) and pN␣2(I) are processing intermediates of the pro␣1(I) and pro␣2(I) chains of type I procollagen, in which the N-, but not C-propeptides are retained. D, an immunoblot is shown of purified FLAGtagged Chordin incubated in the absence (Ϫ) or presence (ϩ) of BMP1 and/or proBMP1 SSQQ . Chordin bands were detected with antibodies directed against the NH 2 terminus of mature Chordin (40). FIGURE 2. ProBMP1 SSQQ binds BMP4 and BMP2 in a highly specific manner. A, an immunoblot is shown of samples containing (ϩ) or lacking (Ϫ) 5 nM proBMP1 SSQQ , BMP1, and/or BMP4; which were immunoprecipitated with anti-BMP1 antibodies (26). Blots were cut and the separate pieces were incubated with anti-BMP1 or anti-BMP4 antibodies. B, proBMP1 SSQQ was incubated with a fixed amount of BMP4 (5 nM) alone, or in the presence of increasing amounts of BMP2 (molar ratios of BMP4:BMP2 of 1:1, 1:2, 1:5, and 1:10) followed by immunoprecipitation with anti-BMP1 antibody (26) and immunoblotting with anti-BMP4 or anti-BMP2 monoclonal antibodies. As a control for pull-down efficiency, the anti-BMP4 blot was stripped and re-probed with anti-BMP1 antibodies to detect immunoprecipitated proBMP1 SSQQ . C, purified BMP1 prodomain, produced in a baculovirus system, was visualized by staining with Coomassie Brilliant Blue R-250 (lane 2). Numbers to the left of gel correspond to the approximate sizes in kDa of molecular mass markers (lane 1). D, Western blot is used for characterization of anti-BMP prodomain antibodies, which at a 1:20,000 dilution detect 6 ng of BMP1-Fc fusion protein (lane 1), but do not detect 6 ng of Fc domain (lane 2). sequence homology and evidence of similar binding of the two proteins to previously described BMP-binding proteins (23,24). In contrast, 10-fold molar excesses of the unrelated growth factor EGF, or even other TGF-␤-like growth factors such as TGF-␤1 (36% sequence identity to BMP4) or BMP5 (54% sequence identity to BMP4) did not compete with BMP4 for binding to proBMP1 SSQQ (data not shown).
BMP1 Prodomain Sequences Specifically Bind BMP2/4 with High Affinity-To more closely explore the ability of BMP1 prodomain sequences to interact with TGF-␤-like BMPs, BMP1 prodomain, free of other sequences other than a COOHterminal His tag, was produced in a baculovirus system (Fig.  2C). In addition, for such studies antibodies were raised against the recombinant prodomain, capable of detecting 6 ng of prodomain at a 1:20,000 antibody dilution (Fig. 2D). Upon incubation of the His-tagged BMP1 prodomain with BMP4, pull down of the His-tagged prodomain with nickel-charged resin was found to co-precipitate BMP4 (Fig. 3A). Similarly, pull down of the prodomain with the anti-BMP1 prodomain anti-bodies also co-precipitated BMP2 (Fig. 3B). Thus, BMPs-2 and 4 are bound by the BMP1 prodomain. The specificity of this interaction is demonstrated in Fig. 3C, in which a 10-fold molar excess of BMP2, but not of TGF-␤1, EGF, or BMP5, effectively competed with BMP4 for binding to the BMP1 prodomain.
To compare the affinity of the BMP1 prodomain for BMP4 to that of Chordin, a known in vivo BMP4-binding partner (23), an equilibrium binding experiment was performed. This experiment entailed incubating BMP4 with an equimolar amount of soluble BMP receptor IA (ALK-3), fused to an Fc domain, in the presence of 1:1, 2:1, or 5:1 molar ratios of either BMP1 prodomain or Chordin. As can be seen (Fig. 3D), BMP1 prodomain, like Chordin, was able to effectively compete with the BMPR-IA-Fc fusion protein for BMP4 binding, and both proteins were able to essentially block BMP4 binding to the BMPR-1A-Fc fusion protein at a molar ratio of 5:1. Thus, affinity of the BMP1 prodomain for BMP4 is shown to be of the same order of magnitude as the affinity of Chordin for this ligand.
BMP1 Prodomain Sequences Bind BMP2 with an Apparent K D of 11 nM-To quantitate the strength of the interaction between the BMP1 prodomain and BMP2, binding affinity was determined via surface plasmon resonance (BIAcore) analysis. Toward this end, BMP2 was immobilized on the surface of a sensor chip, and the strength of binding was separately determined for different concentrations of BMP1 prodomain. Toward this end, a BMP1 prodomain-Fc fusion protein was produced and purified (Fig. 4A), the Fc domain was removed, and the prodomain region was repurified (Fig. 4B). For comparison, the strength of binding to immobilized BMP2 was also determined for different concentrations of Chordin. From the kinetic association and dissociation rates, the equilibrium dissociation constants (apparent K D ) for binding of BMP1 prodomain (Fig. 4C) and Chordin (Fig. 4D) to BMP2 were calculated to be ϳ11 and 7 nM, respectively. The apparent K D for murine Chordin calculated here, is comparable to the K D of ϳ12 nM previously calculated for murine Chordin via surface plasmon resonance (25). Furthermore, the K D calculated here for BMP1 prodomain binding of BMP2 is clearly in the same range as the K D values of proteins like Chordin and Crossveinless 2 (25), which are thought to bind BMPs 2 and 4 in vivo and modulate their signaling under physiological conditions. Endogenous BMP1 Prodomain-BMP4 Complexes Are Found in Skin-The strength and specificity of BMP1 prodomain-BMP4 interactions, and the co-purification of BMP1 and TGF-␤-like BMPs from bone (2) prompted an attempt to identify BMP1 prodomain-BMP4 complexes in tissues. Because BMP1 is expressed at relatively high levels in skin (26,27), and BMP signaling is important to the development and homeostasis of skin and hair follicles (28,29), skin was assayed for the possible presence of BMP1 prodomain-BMP complexes. Toward this end, immunoprecipitations with the anti-BMP1 prodomain antibodies were performed on skin extracts followed by immunoblot analysis for detection of the BMP1 prodomain and for detection of co-precipitated TGF-␤-like molecules. As can be seen (Fig. 5A), an immunoblot probed with anti-BMP1 prodomain antibody shows that BMP1 prodomain sequences persist in dermis primarily as isolated prodomains, but also, in lesser amounts, as uncleaved proBMP1 FIGURE 3. The BMP1 prodomain directly binds BMP2 and BMP4 with high specificity and affinity. A, 5 nM BMP4 and His-tagged BMP1 prodomain were incubated separately or together, followed by precipitation with nickelcharged affinity resin (left panel) or with anti-BMP1 prodomain antibody (right panel) and immunoblot analysis with anti-prodomain or anti-BMP4 antibodies. B, 5 nM BMP2 and His-tagged BMP1 prodomain were incubated separately or together, followed by precipitation with anti-BMP1 prodomain antibody and protein A-Sepharose, and immunoblot analysis with anti-BMP1 prodomain or with anti-BMP2 antibodies. C, 5 nM BMP4 and His-tagged BMP1 prodomain were coincubated, as in A, but in the presence of 10-fold excesses of BMP2, TGF-␤1, EGF, or BMP5; followed by precipitation with nickel-charged affinity resin and immunoblot analyses. D, 5 nM BMP4 and soluble BMP receptor IA (BMPR-1A, ALK-3), fused to an Fc domain, were coincubated in the presence of 1:1, 2:1, or 5:1 molar ratios of either BMP1 prodomain or Chordin, and the samples then precipitated with protein A-Sepharose and analyzed by immunoblots employing anti-BMP4 or anti-BMPR-1A antibodies. MARCH 23, 2007 • VOLUME 282 • NUMBER 12 (lane 3). The same blot also shows small amounts of proBMP1, but no isolated BMP1 prodomain, to be detectable in epidermis (lane 4). An immunoblot probed with anti-BMP4 antibody (Fig. 5B), demonstrates that prodomain sequences in dermis pulled down, and are thus contained in complexes with, BMP4 (lane 3). In contrast, immunoblots show that BMP1 prodomain sequences did not pulldown BMP5, TGF-␤1, or the unrelated growth factor EGF. Similarly, there was no detectable pull down of BMP2 or BMP6 by BMP1 prodomain sequences in such assays (data not shown). The finding of BMP1 prodomain-BMP4 complexes, but lack of detectable endogenous BMP1 prodomain-BMP2 complexes in dermis may be explained by the fact that the two factors are expressed in different compartments of the skin: BMP2 appears to be expressed primarily in the epithelial compartment, whereas BMP4 is expressed primarily in the mesenchymal/dermal compartment (30). In fact, skin BMP2 expression is primarily associated with hair follicles (30), which are absent in foreskin samples, such as that used in Fig. 5, A and B.

BMP1 Prodomain Binds BMP2/4
Excision of the BMP1 Prodomain and Formation of BMP1 Prodomain-BMP4 Complexes Are Extracellular Events-To determine whether endogenous BMP1 prodomain-BMP4 complexes are intracellular or extracellular, immunoprecipitations with anti-BMP1 prodomain antibodies were performed on cell lysates and conditioned medium of MG-63 osteosarcoma cells, previously shown to express both BMP1 and BMP4 (9,31). As can be seen (Fig. 5C), anti-BMP1 prodomain antibody co-precipitated BMP4 from conditioned medium, but not from cell lysates of MG-63 cultures, consistent with the conclusion that BMP1 prodomain-BMP4 complexes form in the extracellular space. It should also be noted that free BMP1 prodomain was detected in conditioned medium, but not in cell extracts (Fig. 5C). This indicates that, at least in some cells, excision of the BMP1 prodomain occurs concurrent with or subsequent to secretion, and is not an intracellular event.
BMP1 Prodomain Sequences Can Modulate BMP4 Induction of Osteoblastic Differentiation-We next sought to determine the ability of BMP1 prodomain sequences to affect BMP4 signaling. Toward this end, we compared the abilities of proBMP1 SSQQ and isolated Histagged BMP1 prodomain to that of Chordin in affecting the ability of BMP4 to induce osteoblastic differentiation of M2-10B4 stromal stem cells; a process dependent upon BMP signaling (21). Both proBMP1 SSQQ and His-tagged BMP1 prodomain were found to be comparable to Chordin in ability to inhibit the BMP signaling-dependent osteoblastic differentiation of M2-10B4 cells (Fig. 6A). In a similar assay, mature active BMP1 was found to have no effect on BMP4-induced M2-10B4 cell osteoblastic differentiation (Fig. 6B). Thus, BMP1 prodomain sequences, but not other BMP1 sequences, are capable of blocking BMP4 signaling.
Tethered BMP1 Prodomain Sequences Induce Dorsalization in Developing Zebrafish Embryos-Extracellular antagonists of signaling by BMPs 2 and 4, such as Chordin, induce dorsalization when overexpressed in early zebrafish and Xenopus embryos (32,33). Thus, we overexpressed the BMP1 prodomain in zebrafish embryos to determine whether these prodomain sequences might be similarly capable of modulating BMP signaling in vivo. In initial attempts, injection of 1-2-cell embryos with 1 ng of mRNA encoding isolated BMP1 prodomain, which had effectively blocked BMP4 signaling in cell culture assays (Fig. 6A), only mildly dorsalized 6 of 56 surviving injected embryos (Table 1). In contrast, 250 pg of Chordin mRNA strongly dorsalized 24 of 32 surviving embryos. Interestingly, the observation that BMP1 prodomain overexpression does not ventralize embryos is consistent with the probability that prodomain and Chordin binding sites on BMP2/4 molecules do not coincide. It has previously been shown that the BMP4-binding protein twisted gastrulation (Tsg) is unable to induce secondary axis formation upon overexpression in Xenopus embryos, whereas a membrane-tethered form, created via fusion of Tsg to the integral membrane protein CD2, has a strong dorsalizing activity similar to that of Chordin (34). Presumably, this difference in effects is caused by the normally highly diffusible nature of Tsg, which renders it incapable of accumulating to a high enough localized concentration to establish an organizer-like activity (34). To examine whether a less diffusible, membrane-tethered form of BMP1 prodomain would have modified/enhanced dorsalizing activity, a chimeric protein composed of BMP1 prodomain fused to the CD2 transmembrane domain was overexpressed in 1-2-cell stage zebrafish embryos. As can be seen (Table 1), microinjection of 1 ng or 500 pg of mRNA for the membrane-tethered form of BMP1 prodomain yielded dorsalization of 100% of surviving embryos, and microinjection of 250 pg resulted in 78% dorsalization. Moreover, the majority of embryos dorsalized by injection with 1 ng or 500 pg of tethered BMP1 prodomain mRNA were severely dorsalized and could be rated as having class 4/5 dorsalization defects, according to the classification scheme of Mullins et al. (35). Thus, as with Tsg (34), a form of BMP1 prodomain designed to accumulate to high, localized concentrations within embryos acts like known in vivo antagonists of BMP signaling. Embryos injected at the two cell stage with 500 pg of prodomain-CD2 mRNA are shown in Fig. 7, B and D at 12-and 24-h postfertilization (hpf), respectively. At 12 hpf such embryos become ovoid, rather than round just before the bud stage, a characteristic of dorsalized gastrulating embryos (35,36). A representative 24 hpf-injected embryo is shown with a class 4 dorsalized phenotype, in which the trunk axis twists around itself, reminiscent of homozygous recessive snailhouse (snh) mutations (35). Consistent with a dorsalized phenotype, in situ hybridization analysis of injected embryos at 70% epiboly demonstrated expansion of the expression domain for the dorsal marker chd, in 22/24 embryos observed (Fig. 7F), and markedly reduced expression, in 32/32 embryos observed, of gata2 (Fig. 7H), a marker for ventral ectoderm and hematopoetic cells in the ventral mesoderm. In contrast, the expression domain for goosecoid, an early marker gene associated with formation of the organizer, was unchanged in injected embryos. These data are consistent with the interpretation that dorsalization occurred downstream of organizer induction, as a consequence of disrupted BMP signaling during gastrulation. BMP2 appears to be the primary ventralizing factor in early zebrafish dorsoventral patterning (4), in contrast to Xenopus and mouse, in which BMP4 appears to play this role (37,38). Thus, results in the zebrafish ( Fig. 7 and Table 1) and M2-10B4 (Fig. 6) systems are together consistent with the interpretation that the BMP1 prodomain is capable of binding and modulating signaling by both BMPs 2 and 4.

DISCUSSION
Here a cogent explanation is provided for the previously observed co-purification of BMP1 prodomain sequences and TGF␤-like BMPs through multiple and varied biochemical purification steps. We demonstrate that the BMP1 prodomain binds BMP2/4 with high specificity and with high affinity, with a K D similar to that of Chordin, in the physiological nM range. Indeed, the affinity and specificity of binding to BMP2/4 within organisms is such that overexpression of a tethered form of the BMP1 prodomain induces a severely dorsalized phenotype in zebrafish embryos that closely resembles the phenotype achieved by Chordin overexpression. Importantly, it is demonstrated here that endogenous free BMP1 prodomain not only persists in tissues, but that it is found bound to endogenous BMP4 in such tissues. The latter results, obtained from dermal  3) and epidermal (lanes 2 and 4) extracts were separately subjected to immunoprecipitation with preimmune serum (lanes 1 and 2) or anti-BMP1 prodomain antibodies (lanes 3 and 4). Immunoblots are shown in which the samples are probed with anti-BMP1 prodomain (A), or anti-BMP4, anti-BMP5, TGF-␤, or anti-EGF antibodies, respectively (B). Recombinant proteins run as marker controls include proBMP1 and isolated BMP1 prodomain (A, lanes 5 and 6, respectively), or BMP4, BMP5, TGF-␤1, and EGF (B, lane 5). C, conditioned medium and cell lysates from MG-63 cultures were separately immunoprecipitated with preimmune serum (PI) or anti-BMP prodomain antibodies (␣-Pro). Immunoblots were probed with either anti-BMP4 (␣ BMP4) or anti-BMP1 prodomain antibodies. The first lane (Ϫ) is not an immunoprecipitated sample, but instead contains non-precipitated input conditioned medium. samples, combined with the previous findings of BMP1 prodomain persistence in bone and co-purification with TGF-␤-like BMPs (2), build a strong case for in vivo roles for BMP1 prodomain-BMP2/4 interactions in these tissues, and perhaps in other tissues as well. Other results presented herein strongly suggest that BMP1 prodomain-BMP4 interactions occur in the extracellular space, rather than intracellularly.
BMP1 prodomain-BMP2/4 interactions appear to have been conserved at least over the span of evolutionary time that separates humans and zebrafish, as human BMP1 prodomain sequences are shown in the current study to be capable of dorsalizing zebrafish embryos. It should also be noted that we have recently isolated zebrafish BMP1 sequences (39), and have found CD2-tethered zebrafish BMP1 prodomain sequences to induce severe dorsalization of zebrafish embryos in a manner indistinguishable from the effects of the CD2-tethered human BMP1 prodomain sequences. 3 In regard to the specificity of interactions of BMP1 prodomain-BMP2/4 interactions, we have found the prodomains of the other mammalian BMP1-like proteinases, mammalian Tolloid-like 1 and 2 (mTLL1 and mTLL2), to lack the ability to bind BMP2/4. 3 It remains to be determined whether the prodomains of these other proteinases have specificities for binding other growth factors.
Despite the demonstrated ability of the BMP1 prodomain to bind BMP2/4 and to inhibit BMP2/4 signaling in cell culture assays and zebrafish embryos, and despite the isolation of BMP1 prodomain-BMP4 complexes from tissues, the normal developmental and physiological consequences that result from BMP1 prodomain-BMP2/4 complex formation await further elucidation. However, genetic approaches toward such elucidation are hampered by the possibility that manipulation of BMP1 prodomain sequences and/or expression levels may also affect the physiological roles of mature BMP1, thus interfering with the interpretation of results. For example, ablation of BMP1 prodomain sequences could lead to ventralization because of loss of prodomain dorsalizing activity, but such an effect might also be ascribed to an increase in the ventralizing activity of mature BMP1, which normally acts to clear the BMP antagonist Chordin from the extracellular space (40). In regard to the latter possibility, it has previously been shown that deletion of the prodomain 3 D. S. Greenspan, unpublished data. FIGURE 6. BMP1 prodomain is comparable to Chordin in ability to inhibit BMP4-dependent osteoblastic differentiation of M2-10B4 stromal stem cells. BMP4 (6 nM) was preincubated in the absence (Ϫ) or presence (ϩ) of equimolar amounts of proBMP1 SSQQ , His-tagged BMP1 prodomain, or Chordin (A), or proBMP1 SSQQ or mature BMP1 (B), and the samples were added to M2-10B4 cell cultures. AP activity in cell lysates was determined 48-h later by absorbance readings at 405 nm. sequences of the BMP1-related Drosophila proteinase Tolloid results in a superactivated form of the proteinase (41). It is likely instructive that overexpressed BMP1 prodomain, like overexpressed Tsg, only induced embryo dorsalization when in a membrane-tethered form. This suggests that the BMP1 prodomain, like Tsg, may normally be highly diffusible in tissues, and thus unable to accumulate to a high enough localized concentration to establish an organizer-like activity, upon overexpression. Consistent with the latter possibility is the finding that cleaved BMP1 prodomains are found in conditioned medium, but not associated with cell layers of cultured cells (Fig. 5C), suggesting that prodomain sequences do not bind to extracellular matrix or cell surface components. The suggestion of high diffusibility, in turn suggests the possibility that BMP1 prodomain sequences may serve a carrier function that increases the diffusibility, and perhaps affects the half-life, of BMP2/4 in tissues.
Tsg can serve a linker function, as it binds both BMP2/4 and Chordin in a tripartite complex (14,34,42). The possibility remains open that BMP1 prodomain sequences may serve to link BMP2/4 to other molecules, and it will thus be of interest to determine whether the BMP1 prodomain binds molecules other than BMP2/4. However, regardless of whether BMP1 prodomain serves a linker function, it is highly probable that BMP 2/4 bound to BMP1 prodomain sequences have different properties than do unbound BMP2/4, and that BMP1 prodomain sequences therefore modulate BMP2/4 signaling. In fact, BMP1 prodomain sequences may serve to modulate BMP signaling in a complex fashion, as do Tsg and another BMP-binding protein Crossveinless 2, both of which appear able to either enhance or inhibit BMP signaling according to varying in vivo conditions (14,25,34,42). Indeed, it seems unlikely that the BMP1 prodomain would serve only to inhibit BMP2/4 signaling, since mature BMP1 serves to activate BMP2/4 via cleavage of Chordin (5). However, the possibility that the BMP1 prodomain modulates BMP2/4 signaling in a complex way, per-haps serving to refine signaling gradients in response to extracellular levels of other proteins, may further serve to hamper elucidation of its in vivo roles.