Crystal structure of BMP-9 and functional interactions with pro-region and receptors.

Bone morphogenetic proteins (BMPs), a subset of the transforming growth factor (TGF)-beta superfamily, regulate a diverse array of cellular functions during development and in the adult. BMP-9 (also known as growth and differentiation factor (GDF)-2) potently induces osteogenesis and chondrogenesis, has been implicated in the differentiation of cholinergic neurons, and may help regulate glucose metabolism. We have determined the structure of BMP-9 to 2.3 A and examined the differences between our model and existing crystal structures of other BMPs, both in isolation and in complex with their receptors. TGF-beta ligands are translated as precursors, with pro-regions that generally dissociate after cleavage from the ligand, but in some cases (including GDF-8 and TGF-beta1, -2, and -3), the pro-region remains associated after secretion from the cell and inhibits binding of the ligand to its receptor. Although the proregion of BMP-9 remains tightly associated after secretion, we find, in several cell-based assays, that the activities of BMP-9 and BMP-9.pro-region complex were equivalent. Activin receptor-like kinase 1 (ALK-1), an orphan receptor in the TGF-beta family, was also identified as a potential receptor for BMP-9 based on surface plasmon resonance studies (BIAcore) and the ability of soluble ALK-1 to block the activity of BMP-9.pro-region complex in cell-based assays.

Transforming growth factor ␤ (TGF-␤) 1 signaling controls a wide variety of processes over the lifetime of an organism. A subset of this large and well conserved family is the group of bone morphogenetic proteins (BMPs) and growth and differentiation factors (GDFs), which regulate a diverse array of cellu-lar functions, including differentiation, proliferation, organogenesis, axon guidance, apoptosis, and the establishment of left-right asymmetry (1)(2)(3). BMPs and GDFs are highly conserved throughout the animal kingdom, with examples ranging from Drosophila to humans. They have frequently been implicated in the treatment of bone disorders and injury, in accordance with their robust ability to generate de novo bone formation.
All TGF-␤ ligands are translated as precursor proteins, consisting of an amino-terminal pro-region and a carboxyl-terminal ligand. This precursor forms a disulfide-linked homodimer in the cytoplasm, and the pro-region is then cleaved from the ligand. In most cases, the pro-region disassociates, and the mature ligand is secreted from the cell, but the pro-regions of GDF-8 (also known as myostatin) and TGF-␤1, -2, and -3 remain noncovalently associated with the ligand after secretion and inhibit binding of their ligands to their respective receptors (4 -6). Transgenic mice overexpressing the pro-region of GDF-8 show dramatic increases in muscle mass, further indicating that the pro-region functionally inhibits GDF-8 (7). The proregion of BMP-9 also remains tightly associated after secretion from the cell.
BMP signaling is induced when a dimeric ligand binds to the extracellular domains of two type I and two type II receptors (8). This assembly brings the intracellular domain of both receptor types into close proximity, permitting the constitutively active intracellular kinase domain of the type II receptor to cross-phosphorylate the intracellular Gly-Ser (GS) domain of the type I receptor (9). Receptor-regulated Smad proteins are phosphorylated by the activated type I receptor kinase and associate with Smad4. This complex translocates to the nucleus, interacting with various cofactors to modify gene expression (10).
In the adult rat, BMP-9 is expressed predominantly in the liver and has been shown to induce proliferation of cultured liver cells (18). BMP-9 mRNA was also found in the septum and spinal cord of E13 mice. In vitro and in vivo, BMP-9 was found to promote cholinergic differentiation and the synthesis of acetylcholine and effectively maintained the cholinergic phenotype of differentiated cells (19). Other studies indicate that BMP-9 produces ectopic bone growth and potently directs the differentiation of mesenchymal cells into cartilage (20 -22).
More recently, BMP-9 was identified as a regulator of glucose metabolism by modulating the transcription of several genes that are involved in glucose and fatty acid metabolism, decreasing glucose production in cultured cells, and reducing glycemia in diabetic mice (23). Activin-like kinase-1 (ALK-1) is an orphan receptor in the TGF-␤ family. It has been implicated as an inhibitor of lateral TGF-␤/ALK-5 signaling (24), correlated with vasculogenesis and angiogenesis (25), and may be a factor in hereditary hemorrhagic telangiectasia (26).
The stereotypical scaffold of BMPs consists of a cysteine knot characterized by three pairs of highly conserved disulfide bonds. This fold has been previously described as a "hand" with a concave "palm" side and two parallel ␤-sheets forming the "fingers." In the mature, dimeric ligand, these two ␤-sheet fingers extend from the cysteine core of the protein like butterfly wings. Binding to type I receptors occurs near the ␣-helix on the concave side at the junction between the two subunits (27), whereas binding to the type II receptors is located on the convex side of the hand near the "fingertips" (28,29).
We report here the structure of BMP-9 at 2.3 Å and examine structural differences between BMP-9 and BMP-2, BMP-7, the BMP-2⅐BMPR-IA complex, and the BMP-7⅐ActR-IIB complex. Results of in vitro cell-based assays indicate that both BMP-9 and BMP-9⅐pro-region complex are equally active in signaling and cell-growth stimulation, whereas the pro-region alone is inactive. Based on surface plasmon resonance and functional inhibition in several cell-based assays, we suggest activin receptor-like kinase-1 (ALK-1) as a functional receptor for BMP-9.

EXPERIMENTAL PROCEDURES
Materials-All recombinant receptor:Fc chimeras were purchased from R & D Systems and were reconstituted as instructed by the manufacturer. BIAcore instrumentation, software, and CM5 sensor chips were purchased from BIACORE TM (Uppsala, Sweden). Fetal bovine serum and all other cell culture media were purchased from Invitrogen unless otherwise stated.
Cell Lines-Rat hepatoma cell line H4IIe, mouse myoblast cell line C2C12, and mouse pre-adipocyte cell line 3T3-L1 were all obtained from the American Type Culture Collection (ATCC, Manassas, VA). These cell lines were propagated in media recommended by ATCC. The dihydrofolate reductase-deficient Chinese hamster ovary cell line DG44 (30) was propagated in minimal essential ␣ ϩ medium, 5% dialyzed fetal bovine serum, and 4 mM L-glutamine. Selections were performed in minimal essential ␣ Ϫ medium lacking ribonucleotide and deoxyribonucleotide supplements, supplemented with 5% dialyzed fetal bovine serum and 4 mM glutamine. Methotrexate (Sigma) was used at the concentrations indicated.
Protein Expression and Purification-The coding sequence for fulllength BMP-9 with pro-region was cloned into pC4, a proprietary mammalian expression vector, and the construct was transfected into dihydrofolate reductase-deficient Chinese hamster ovary cell line DG44 using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. Typically, ϳ1 ϫ 10 6 cells were transfected with 5 l of Lipofectamine 2000, 5 g of pC4.BMP-9, and 0.5 g of pSV2.NEO plasmids. The transfected cells were put into selection after ϳ36 h in minimal essential ␣ Ϫ medium containing 25 nM methotrexate and 1 mg/ml geneticin. After ϳ2-3 weeks, single clones were isolated and seeded into 24-well plates. When confluent, the supernatants from these clones were screened for BMP-9 activity using cell-based reporter assays. The high expressing clones were selected and amplified stepwise up to 0.5 mM methotrexate. For large scale production of BMP-9, a high expressing clone was grown in Chinese hamster ovary-5 medium (a Human Genome Sciences proprietary serum-free medium without insulin) for 5 days, and the conditioned medium was harvested.
To purify the BMP-9⅐pro-region complex, 2-5 liters of BMP-9 culture supernatant was adjusted to pH 6.4 with pH 5.0, 1 M MES, and diluted to a conductivity of Ͻ7-8 ms. The sample was loaded on a 50-ml POROS 50 HS (Perseptive Biosystems) column that had been preequilibrated with pH 6.4, 20 mM MES, and 50 mM NaCl. BMP-9 complex was eluted with a 20-column-volume pH gradient from pH 6.4 to 8.0, with pH 8.0 buffer containing 20 mM Tris-HCl and 50 mM NaCl. The fractions of partially purified BMP-9⅐pro-region complex were pooled and adjusted to pH 7.4 with pH 9.0, 1 M Tris, and diluted to the conductivity Ͻ5 ms. The sample was loaded onto an 8-ml MonoQ column (Amersham Biosciences) equilibrated with pH 7.4 and 20 mM Tris. BMP-9⅐pro-region complex was eluted by a 15-column-volume salt gradient of 0 -0.5 M NaCl.
For cell-based assays, BMP-9 dimer was separated from its proregion using 50% cold ethanol precipitation. After cold precipitation, the pro-region and remaining BMP-9⅐pro-region complex were spun down, leaving isolated, dimeric BMP-9 in the supernatant. The precipitation was repeated once after dissolving the pro-region and BMP-9⅐pro-region complex in pH 7.4, 20 mM Tris, 200 mM NaCl. The proregion was further separated from BMP-9⅐pro-region complex on a MonoQ column using the same procedure as used to purify the complex. Soluble BMP-9 in the supernatant was concentrated using a CentriPrep (Amicon), and the buffer was changed to pH 7.4, 20 mM Tris, 200 mM NaCl, 30% ethanol.
For native and SDS gels, BMP-9 dimer and pro-region were isolated by reverse phase chromatography, using a C4 high pressure liquid chromatography column (Vydac) and eluted with a 27-42% acetonitrile gradient. BMP-9 dimer eluted at 31% acetonitrile, and the pro-region eluted at 35% acetonitrile. Fractions were frozen in liquid nitrogen, lyophilized, and reconstituted in water.
Crystallization, Data Collection, and Refinement-Dimeric BMP-9 was crystallized from a solution containing BMP-9 dimer, BMP-9 monomer, pro-region, and other secreted proteins in a condition identified from sparse matrix (Hampton) screening. Total protein concentrations ranged from 3.8 to 9 mg/ml, with BMP-9 dimer constituting ϳ15-20% of the total protein. Crystals of BMP-9 were obtained by hanging drop at 23°C with a well solution of 1-1.2 M NaCl, 7-10 mM hexadecyltrimethylammonium bromide, and 10 mM MgCl 2 .
Drops containing crystals of BMP-9 were flooded with mother liquor supplemented with 12-15% glycerol as a cryoprotectant. Crystals were flash frozen in liquid nitrogen, and diffraction data were obtained at Stanford Synchrotron Radiation Laboratory on beam line 9.2. All integration and scaling were performed with HKL2000. The space group was found to be I4 1 A molecular replacement solution was found using MolRep (31) with the structure of BMP-2 monomer as the search model. The asymmetric unit contains one copy of the BMP-9 monomer, leaving a solvent content of 67.6%. The protein was manually rebuilt in O (32), and the structure was refined using Refmac5 (Collaborative Computational Project 4) and crystallography NMR software (33). Four translation, libration, and screw rotation groups were used, dividing the protein into regions from Ala 4 to Ala 34 , Tyr 35 to His 54 , Ala 55 to Lys 78 , and Leu 79 to Gly 108 .
Size Exclusion Chromatography-Approximately 100 g of total protein from Chinese hamster ovary cell media was run in 10 mM Tris, 100 mM NaCl over an S-200 size exclusion chromatography column.
Native Gels and SDS-PAGE-Native gels were 6% polyacrylamide, pH 8.8, and run at 27 mA. Proteins were mixed to final concentrations of 1-3 mg/ml and allowed to equilibrate for at least 15 min at room temperature or overnight at 4°C before loading. Non-reducing SDS gels were 12% polyacrylamide and run at 200 mV.
Alkaline Phosphatase Assay-Mouse pluripotent C2C12 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 4 mM L-glutamine, 1.5 g/liter sodium bicarbonate, 4.5 g/liter glucose, 100 IU/ml penicillin, and 100 g/ml streptomycin. For the alkaline phosphatase assay, C2C12 cells were seeded in 96-well tissue culture plates at 1,000 cells/well in 100 l of medium. The following day, the medium was removed and replaced with treatments in Dulbecco's modified Eagle's medium containing 0.1% fetal bovine serum. The cells were cultured for 4 days and conditioned media were collected for measurement of alkaline phosphatase activity. In inhibitor studies, soluble receptors were added at the indicated concentrations with BMP-9⅐pro-region complex (5 nM) or BMP-4 (4.7 nM). Alkaline phosphatase activity was measured using the Phospha-Light TM System (Applied Biosystems) according to the manufacturer's directions. Briefly, the cells were rinsed with phosphate-buffered saline, lysed in buffer containing 0.2% Triton X-100, incubated 5 min in assay buffer and 20 min in reaction buffer containing CSPD (disodium 3-(4-methoxy-spirol1,2-dioxetane-3,2Ј-(5Ј-chloro)tricyclo[3.3.1.1(3,7)]decan-4-yl)phenyl phosphate) substrate. CSPD substrate produces a luminescent signal when dephosphorylated by alkaline phosphatase in an alkaline, hydrophobic environment, resulting in anion production, which emits light upon decomposition. The luminescent signal was read using a luminescent plate reader (Applied Biosystems). All treatments were performed in triplicate. The average and S.D. were determined and data plotted using Prism software (GraphPad Software, Inc., San Diego, CA).
Cell Proliferation Assay-Mouse pre-adipocyte 3T3-L1 cells were plated at 4,000 cells/well in white 96-well tissue culture plates (Costar) in 100 l of Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 IU/ml penicillin, and 100 g/ml streptomycin. The cells were cultured overnight. To arrest cell growth, the cells were washed once with low serum medium (0.1% fetal bovine serum), 100 l of low serum medium was added, and the cells were cultured overnight. BMP-9, BMP-9⅐pro-region complex, pro-region, or controls were added to cells at indicated concentrations. In inhibitor studies, soluble receptors were also added at the indicated concentrations with BMP-9 proteins (4 nM) or BMP-4 (3.7 nM). The cells were incubated for 4 days, and cell number assayed using the Cell Titer Glo assay (Promega, Madison, WI) according to the manufacturer's instructions. In brief, 100 l of assay reagent was added to the cells at room temperature, mixed, and incubated for ϳ30 min. The luminescent signal was read using a luminescent plate reader. All treatments were performed in triplicate and analyzed with Prism software.
Malic Enzyme-SEAP Reporter Assay-Promoter region (base pairs Ϫ1183 to Ϫ77) of rat malic enzyme (ME) was cloned into the promoterless pSEAP2-neomycin vector and fused in-frame with the SEAP gene. The construct was transfected into the rat hepatoma cell line H4IIe, and stable clones were selected for neomycin resistance. To perform the reporter assay, ϳ75,000 cells/well were plated in 96-well plates. The growth medium was Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 15 mM HEPES, 100 IU/ml penicillin, and 100 g/ml streptomycin. The cells were cultured for 24 h. Reporter cells were then serum-deprived for 18 -24 h before being treated with BMP-9, BMP-9⅐pro-region complex, pro-region, or controls. For the experiments using soluble BMP receptors, the cells were treated with various concentrations of soluble receptors prior to the addition of BMP-9⅐pro-region complex (2 nM) or BMP-4 (0.4 nM). After a 48-h incubation period, conditioned media were removed, and SEAP activity was determined using the Phospha-Light TM System according to the manufacturer's recommended protocol. Light emission from the wells was measured in a MLX Microtiter Plate Luminometer (Dynex Technologies). All treatments were performed in triplicate and analyzed with Prism software.
Biosensor Surface Preparation for BIAcore Analysis-To determine receptor specificity for BMP-9, three high density CM5 chips were prepared. Each chip contained four flow cells, with one flow cell/chip immobilized with an unrelated receptor:Fc as a negative Fc control (trail-receptor:Fc). The remaining flow cells contained one of the following BMP receptors (each fused to Fc): BMPR-IA, BMPR-IB, BMPR-II, ALK-1, ALK-7, activin receptor IA (ActR-IA), ActR-IB, ActR-IIA, and ActR-IIB. The immobilization level, defined in relative units, ranged from 7,110 to 17,300. Purified BMP-9⅐pro-region complex was applied at 50 g/ml (1 M) to each individual flow cell for 2 min. Measured on-rates and off-rates after global fits to binding curves were used to derive binding affinity.

RESULTS
BMP-9 was crystallized and solved with molecular replacement at a resolution of 2.32 Å (Table I). BMP-9 shows the characteristic cysteine knot scaffold and overall butterflylike conformation, with an ␣-helix ("knuckle," ␣3) epitope and two ␤-stranded sheets (fingers F1 and F2) extending from the core of the molecule (Fig. 1A), but it also deviates from known BMP structures in regions known to be important for receptor binding.
Previous structural studies of BMP-2 in complex with BMPR-IA ectodomain revealed the residues likely to be in- FIG. 1. Structure of BMP-9. A, Ribbon diagram of BMP-9, showing finger regions 1 and 2 (F1 and F2) and ␣-helix 3 (␣3). The pre-helix loop, which likely encodes specificity determinants, is colored red. C, C terminus; N, N terminus. B, overlay of BMP-9 (blue) and BMP-2 (pink) with BMPR-IA ectodomain. Residues of BMPR-IA within 4.0 Å of BMP-2 are shaded green (36). The carboxyl-terminal residue of BMPR-IA ectodomain is shaded yellow. C, overlay of BMP-9 (blue) and BMP-7 (green) with ActR-II ectodomain. Residues of ActR-II within 4.0 Å of BMP-2 are shaded red (28). The carboxyl-terminal residues of ActR-II ectodomain are shaded yellow. B and C were produced with DINO software (www.dino3d.org). volved in determining binding affinity and specificity (defined by a 4.0-Å cutoff and colored green in Fig. 1B). To examine which structural differences between BMP-2 and BMP-9 might influence receptor binding, BMP-2/BMPR-IA (Protein Data Bank (PDB) entry 1ES7) was aligned with BMP-9 using least squares minimization. This comparison reveals the regions of the ligands that are likely to be significantly different, particularly the pre-helix loop.
To quantify differences between BMP-9 and BMP-2, C-␣ r.m.s. deviations between BMP-9 and free (uncomplexed) BMP-2 (PDB entry 3BMP) were calculated. The average r.m.s. deviation for the monomer is 1.50 Å. Of the amino acids likely to be involved in binding a type I receptor, C-␣ r.m.s. deviations Ͼ1 S.D. above the mean occur between residues Gly 27 in BMP-2 and Gly 21 in BMP-9 and the segments Phe 49 -Asp 53 in BMP-2 and Phe 43 -Asp 47 in BMP-9. One residue pair (Pro 50 in BMP-2 and Pro 44 in BMP-9) at the type I interface has an r.m.s. deviation Ͼ1.5 S.D. from the mean. All of these residues are located in the pre-helix loop, or "wrist" epitope (colored red in Fig. 1A).
The crystal structure of BMP-7 in complex with activin receptor IIA (ActR-IIA, PDB entry 1LX5) has previously been solved, and the residues likely to be involved in binding (defined by a 4.0-Å cutoff) are colored red in Fig. 1C. Root mean square deviations were also calculated between BMP-9 and free BMP-7 (PDB entry 1LXI). One position at the type II binding interface, Ala 58 in BMP-7 and Ala 28 in BMP-9, was found to have the most pronounced difference from the mean r.m.s. deviation (1.76 Å).
In a sequence alignment of BMP-9 and BMP-2 (Fig. 2), sequence differences in binding regions identified by structural and mutational (27) studies are most notable in residues Asn 59 in BMP-2 and Lys 53 in BMP-9 (acidic to basic) and in residues His 54 in BMP-2 and Asp 58 in BMP-9 (basic to acidic). A comparison between the amino acid sequences of BMP-9 and BMP-7 (Fig. 2) also reveals significant changes, including Tyr 44 in BMP-7 to Arg 14 in BMP-9 (aromatic to basic), Glu 60 in BMP-7 to Lys 30 in BMP-9 (acidic to basic), and Gly 61 in BMP-7 to Glu 31 in BMP-9 (aliphatic to acidic) at the type II binding interface. Based on the BMP-7/ActR-IIB structure (28), Tyr 44 and Gly 61 both interact with the acidic Asn 65 , and Glu 60 interacts with basic Lys 76 . These data indicate that differences between BMP-7 and BMP-9 in binding affinity for type II receptors are most likely because of amino acid changes at the binding interface, although type II receptors are generally quite promiscuous (37).
BMP-9, when purified from cell medium, remains noncovalently associated with its pro-region. This complex runs as a single peak by size exclusion chromatography, with a retention volume consistent with a complex of one BMP-9 dimer and two pro-regions (Fig. 3A). An SDS gel shows clean separation after reverse-phase chromatography (Fig. 3B). When separated, BMP-9 and pro-region are recombined in equimolar quantities and run on a native gel, a band similar to what is seen in the original protein solution appears, and neither the isolated dimeric BMP-9 band nor the isolated pro-region band is observed, indicating that the complex can be reformed after separation (Fig. 3C). Biological activities of isolated BMP-9, BMP-9⅐pro-region complex, and pro-region alone were tested by three different in vitro assays using distinct cell types. Isolated BMP-9⅐proregion complex, BMP-9, and pro-region were tested for the ability to stimulate secretion of alkaline phosphatase from C2C12 cells (Fig. 4A), to activate transcription of H4IIe/ME-SEAP gene reporter (Fig. 4B), and to induce proliferation of the 3T3-L1 pre-adipocyte cell line (Fig. 4C). Both purified BMP-9 and BMP-9⅐pro-region complex were highly active in all three assays, and the degree of activation was dose-dependent. The presence of the pro-region did not alter the activity of BMP-9, as the activities of BMP-9 and BMP-9⅐pro-region complex were not significantly different. The pro-region alone was devoid of biological activity in these assays.
In an effort to identify the receptor(s) mediating the effects of BMP-9, five type I and four type II commercially available soluble BMPs or activin receptor:Fc chimeras were tested for the ability to block activity of BMP-9⅐pro-region complex in the C2C12 alkaline phosphatase assay. BMP-4, a known ligand of BMPR-IA (38), was used as the control (Table II, columns 2 and  3). In this experiment, BMP-9⅐pro-region complex (5 nM) or BMP-4 (4.7 nM) was added to C2C12 cells either alone or in the presence of a 10-fold molar excess of soluble receptor. After an incubation period, the conditioned medium was collected and alkaline phosphatase activity was measured as described above. As presented in Table II, ALK-1:Fc completely inhibited BMP-9⅐pro-region complex activity but had no effect on BMP-4. In contrast, BMPR-IA:Fc completely abolished BMP-4 activity but had no effect on BMP-9⅐pro-region complex. Of the other soluble receptor:Fc chimeras tested, only BMPR-II:Fc partially blocked BMP-9⅐pro-region complex activity, whereas the rest did not exhibit any inhibitory activity.
To measure relative affinities between BMP-9⅐pro-region complex and BMP/activin receptors, BIAcore analysis was performed using purified BMP-9⅐pro-region complex and receptor:Fc chimera immobilized on a BIAcore chip surface (Table II, column 1). BMP-9⅐pro-region complex did not bind the control flow cell. However, it did bind strongly to ALK-1 and BMPR-II. In addition, BMP-9 also binds weakly to ActR-IA, ActR-IIA, and ActR-IIB at different relative levels. Binding constants cannot be accurately derived from this study, as only one ligand concentration was applied to a high density chip, and it has been shown in other studies that dissociation constants vary based on the density of the fixed receptor and whether the ligand or the receptor was immobilized due to cooperative binding by the receptors (36,37). However, by comparative analysis, it is clear that BMPR-II and ALK-1 have the highest affinity.
To further examine interactions between BMP-9 and ALK-1, BMP-9⅐pro-region complex was used in the absence or presence of various concentrations of soluble ALK-1 or BMPR-IA receptors in the C2C12 alkaline phosphatase and H4IIe/ME-SEAP reporter assays (Fig. 5). BMP-4 was included in the experiment as a control. ALK-1:Fc completely inhibited BMP-9⅐pro-region complex activity in both assays when given at Ͼ6-fold molar excess but showed no effect on BMP-4. As expected, BMPRIA:Fc inhibited BMP-4 activity in a dose-dependent manner, ranging from 1-to 6-fold molar excess but did not block activity of BMP-9⅐pro-region complex. Combining the results, our data demonstrate that ALK-1 can bind to purified BMP-9⅐pro-region complex and neutralize its biological activity in cell-based assays. DISCUSSION In the TGF-␤ superfamily, which consists of over 30 known ligands but only 7 type I and 5 type II receptors, overlapping specificities among the ligands and receptors are common. The characteristic structural scaffold of TGF-␤ family ligands is provided by the amino acids (boxed in gray in Fig. 2) forming the protein core or palm of the hand and in the receptors by the highly conserved set of disulfide bond-forming cysteines (39). This common structural framework likely provides the basis for overlapping specificities between ligands and receptors. Recent structural studies on the extracellular fragments of TGF-␤ receptors in complex with several related ligands have provided important insights on the roles of stoichiometry and affinity in determining the overall outcome of a ligand receptor interaction (40 -42). These studies highlighted the importance of the membrane in restricting the orientations of the two receptor subtypes, suggesting that the relative orientations and distances of the pair of high affinity receptors may influence the affinity of the ligand for the lower affinity receptors. Furthermore, as the first high affinity receptor binds to a ligand, it restricts the entropic freedom of the ligand, resulting in enhancement of binding to the second high affinity receptor to the complex (42,43).
Crystallization of BMP-2 in complex with BMPR-IA helped to identify the points of contact between a BMP ligand and type I receptor, and further studies identified the hydrogen bonds that determined binding affinity (44). In BMP-2, these are the amide and carbonyl of Leu 51 and the amide of Asp 53 , and these residues are conserved in BMP-9. BMP-9, however, shows no affinity for BMPR-IA in cell-based or BIAcore studies. Significant C-␣ r.m.s. deviations do occur between these residues, however, suggesting that main chain position in the pre-helix loop may be an important determinant of binding affinity for type I receptors.
We have determined from the crystal structure of BMP-9 that there are few significant conformational differences between BMP-9 and BMP-7 at the binding sites for the type II receptors. These results are consistent with known promiscuity among type II receptors and our BIAcore data (indicating some binding affinity between BMP-9 and ActR-IIA) but do not exclude the possibility that the different splicing of exons encoding the ϳ15-amino-acid linker between the extracellular domain and the transmembrane domain of each receptor may influence signaling efficiency.
The pro-regions of TGF-␤s and GDF-8 remain associated with their ligands after secretion from the cell and have been found to be functionally inhibitory in vitro and in vivo (4 -7). The mechanism for this inhibition is not known, although particular regions of the pro-region relevant for inhibition have been identified (5). The pro-region of BMP-9 also remains associated after secretion, but we have shown that both BMP-9 and BMP-9⅐pro-region complex were equally active in three cell-based assays covering a range of reported BMP-9 activities, including osteoinduction, proliferation, and gluconeogenesis (20,21,23). These assays demonstrated that the pro-region of BMP-9, unlike those of TGF-␤s and GDF-8, does not appear to functionally inhibit BMP-9. It is not clear whether the BMP-9 pro-region does not block the binding sites for either receptor subtype or whether the pro-region is effectively competed off by one or both of the receptor subtypes. It is instead possible that the pro-region of BMP-9 may act to protect and stabilize BMP-9 in vivo.
BMP-9 shows a tissue expression profile largely restricted to the liver and has been shown to stimulate proliferation in hepatic cells (45), but its receptor has not been identified. ALK-1 is highly expressed in endothelial cells, and inactivating mutations of ALK-1 have been associated with hereditary hemorrhagic telangiectasia (26), a vascular pathology affecting FIG. 4. Activities of BMP-9 and BMP-9⅐pro-region complex are similar in cell-based functional assays, whereas pro-region alone has no effect. A, C2C12 alkaline phosphatase assay: C2C12 cells were growth-arrested for 24 h in low serum medium and treated with either buffer or BMP-9, BMP-9⅐pro-region complex, or pro-region at the indicated concentrations (nM). After 4 days, the cells were lysed and alkaline phosphatase activity determined. B, H4IIe/ME-SEAP reporter assay. Serum-deprived reporter cells were treated with either buffer or BMP-9, BMP-9⅐pro-region complex, or pro-region at the indicated concentrations. SEAP activity in the conditioned medium was measured after a 48-hour incubation period. C, 3T3-L1 cell proliferation assay. 3T3-L1 cells were growth-arrested for 24 h in low serum medium and treated with either buffer or BMP-9, BMP-9⅐pro-region complex, or pro-region at the indicated concentrations. After 4 days, cell numbers were determined by Cell-Titer Glo assay. For all assays, the cells were treated in duplicate or triplicate wells and average Ϯ S.D. for each treatment is shown. Data is normalized to vehicle-treated samples and presented as the signal from the BMP-9-treated sample divided by the signal from the vehicle-treated sample. AP, alkaline phosphatase. multiple organs. Transfection of constitutively active forms of ALK-1 has been shown to stimulate Smad-1 (46) and induce alkaline phosphatase expression in C2C12 cells (47). However, to date the ligand for the ALK1 receptor has not been clearly identified. In this report, we have provided the first evidence that ALK-1 is a strong candidate for a functional receptor for BMP-9. In cell-based assays employing two distinct cell types, a soluble ALK-1:Fc protein was able to potently and selectively block BMP-9⅐pro-region complex activity. BIAcore analysis indicates that BMP-9⅐pro-region complex binds to additional receptors to some degree, including BMPR-II, ActR-IIA, and ActR-IIB. Of these, only BMPR-II inhibited BMP-9⅐pro-region complex activity to any significant extent. The exact nature of the interactions of BMP-9 with ALK-1 and other receptors in FIG. 5. Inhibition of BMP-9 complex activity by soluble ALK-1:Fc. A, C2C12 alkaline phosphatase assay. C2C12 cells were growtharrested for 24 h in low serum medium. The cells were treated with either BMP-9⅐pro-region complex (5 nM) or BMP-4 (4.7 nM,) alone or in the presence of the indicated -fold excess molar concentration of soluble ALK-1:Fc or BMP-RIA:Fc proteins. After 4 days, the cells were lysed and alkaline phosphatase activity determined. AP, alkaline phosphatase. B, H4IIe/ME-SEAP reporter assay. Serum-deprived reporter cells were treated with either BMP-9⅐pro-region complex (2 nM) or BMP-4 (0.1 nM,) alone or in the presence of the indicated -fold excess molar concentration of soluble ALK-1:Fc or BMP-RIA:Fc proteins. SEAP activity in the conditioned medium was measured after a 48-h incubation period. All assays used chemiluminescent detection methods, and data were presented in relative light units (RLU). Experiments were performed with duplicate or triplicate wells, and average Ϯ S.D. for each treatment is shown.