Structure of Bone Morphogenetic Protein 9 Procomplex

Structure of bone morphogenetic protein 9 procomplex

Li-Zhi Mia,b,c, Christopher T. Brownd, Yijie Gaod, Yuan Tiana,b, Viet Q. Lea,b, Thomas Walze,f, and Timothy A. Springera,b,1 aDepartment of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, Boston, MA 02115; bProgram in Cellular and Molecular Medicine, Children’s Hospital Boston, Boston, MA 02115; cDepartment of Molecular and Structural Biology, School of Life Sciences, Tianjin University, Tianjin 300072, China; dGlobal BioTherapeutics Technologies, Pfizer Worldwide Research and Development, Cambridge, MA 02139; and eHoward Hughes Medical Institute and fDepartment of Cell Biology, Harvard Medical School, Boston, MA 02115

Contributed by Timothy A. Springer, January 27, 2015 (sent for review January 2, 2015; reviewed by Daniel Rifkin and Lynn Y. Sakai) Bone morphogenetic proteins (BMPs) belong to the TGF-β family, the structure of pro-BMP9 and compare it to the previously de- whose 33 members regulate multiple aspects of morphogenesis. scribed, cross-armed conformation of pro-TGF-β1 (10). Although TGF-β family members are secreted as procomplexes containing a a member of the BMP subfamily and possessing chondrogenic small growth factor dimer associated with two larger prodomains. and osteogenic activity, BMP9 is expressed in liver and is required As isolated procomplexes, some members are latent, whereas most for properly organized blood and lymphatic vascular development are active; what determines these differences is unknown. Here, (11, 12). Mutations in the prodomain of BMP9, in its receptor studies on pro-BMP structures and binding to receptors lead to β Alk1, and in its coreceptor endoglin cause phenotypically over- insights into mechanisms that regulate latency in the TGF- family lapping hereditary hemorrhagic telangiectasias (13–15). and into the functions of their highly divergent prodomains. The Here, we reveal surprising open-armed conformations of pro- observed open-armed, nonlatent conformation of pro-BMP9 and BMPs 7 and 9. We propose that binding to interactors in the pro-BMP7 contrasts with the cross-armed, latent conformation of β matrix may regulate transition between open-armed and cross- pro-TGF- 1. Despite markedly different arm orientations in pro- β BMP and pro-TGF-β, the arm domain of the prodomain can simi- armed conformations in the TGF- family and that these con- larly associate with the growth factor, whereas prodomain ele- formations regulate GF latency. ments N- and C-terminal to the arm associate differently with Results the growth factor and may compete with one another to regulate Structures of BMP Procomplexes. latency and stepwise displacement by type I and II receptors. Se- In marked contrast to the cross- quence conservation suggests that pro-BMP9 can adopt both cross- armed, ring-like conformation of pro-TGF-β1 (10), crystal armed and open-armed conformations. We propose that interactors structures of natively glycosylated pro-BMP9 reveal an un- in the matrix stabilize a cross-armed pro-BMP conformation and expected, open-armed conformation (Fig. 1 A and B and Table regulate transition between cross-armed, latent and open-armed, S1). All negative stain EM class averages show an open-armed nonlatent pro-BMP conformations. conformation for pro-BMP9 (Fig. 1C and Fig. S1) and a similar, although less homogenous, open-armed conformation for pro- embers of the TGF-β family including bone morphoge- BMP7 (Fig. 1D and Fig. S2). Crystal structure experimental Mnetic proteins (BMPs) are biosynthesized and processed electron density is excellent (Fig. S3) and allows us to trace the into complexes between large prodomains and smaller, C-ter- complete structure of each pro-BMP9 arm domain (residues 63– minal mature growth factor (GF) domains that are separated by 258; Fig. 1E). As in pro-TGF-β1, the arm domain has two β-sheets proprotein convertase (PC) (furin) cleavage sites. In the original that only partially overlap. Hydrophobic, nonoverlapping portions isolation of proteins responsible for BMP activity, bone was first of the β-sheets are covered by meandering loops and the α4-helix demineralized with 0.5 M HCl. The resulting residual matrix was (Fig. 1 E and F). Comparison of pro-BMP9 and pro-TGF-β1arm – extracted with 6 M urea or 4 M guanidine HCl (1 3). During domains defines a conserved core containing two four-stranded subsequent purification under largely denaturing conditions, the β-sheets and the α4-helix (labeled in black in Fig. 1 E and F). GF domains were separated from their prodomains. Therefore, One of the BMP9 arm domain β-sheets joins a finger-like little attention was paid to the potential existence of BMP pro- β-sheet in the GF to form a super β-sheet (Fig. 1 A and G). Each complexes. However, evidence exists that BMP prodomains GF monomer has a hand-like shape. The two BMP9 GF hands contribute to maintaining BMP GF domains inactive or latent in vivo. For example, early studies showed a 60-fold increase in total BMP activity during the first two purification steps fol- Significance lowing extraction of the BMP, which was interpreted as purification of BMP away from an inhibitor (2). This finding is Bone morphogenetic protein (BMP) activity is regulated by consistent with the presence of largely latent complexes between prodomains. Here, structures of BMP procomplexes reveal an BMPs, their prodomains, and extracellular matrix components in open-armed conformation. In contrast, the evolutionarily re- the insoluble residual matrix from which BMPs were purified. In lated, latent TGF-β1 procomplex is cross-armed. We propose agreement with a regulatory role for the prodomain, mutations that in the TGF-β and BMP family, conversion between cross- of secondary PC sites within the prodomain perturb embryonic armed and open-armed conformations may regulate release development in insects and vertebrates, suggesting that prodo- and activity of the growth factor. mains of several BMPs remain associated with GFs after secretion and regulate the distance over which BMPs signal (4–7). An Author contributions: T.A.S. designed research; L.-Z.M., C.T.B., Y.G., Y.T., V.Q.L., T.W., and T.A.S. performed research; L.-Z.M., C.T.B., Y.G., Y.T., V.Q.L., T.W., and T.A.S. analyzed important role for the prodomain in development is also illustrated data; and L.-Z.M., C.T.B., Y.T., V.Q.L., and T.A.S. wrote the paper. by prodomain mutations, including in secondary PC cleavage sites, Reviewers: D.R., NYU Langone Medical Center; and L.Y.S., Shriners Hospitals for Children. that cause human diseases (5, 7). The authors declare no conflict of interest. Pro-TGF-β is latent; however, when overexpressed as recom- Data deposition: The atomic coordinates and structure factors have been deposited in the binant proteins, most BMPs are active. Although noncovalently Protein Data Bank, www.pdb.org (PDB ID codes 4YCG and 4YCI). associated with their GF after secretion, the prodomains of most 1To whom correspondence should be addressed. Email: [email protected]. BMPs do not bind strongly enough to prevent GF from binding edu. to receptors and signaling (8, 9). To better understand such This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. differences among members of the TGF-β family, we examine 1073/pnas.1501303112/-/DCSupplemental.

3710–3715 | PNAS | March 24, 2015 | vol. 112 | no. 12 www.pnas.org/cgi/doi/10.1073/pnas.1501303112 A B Pro- Bowtie Pro- domain domain Prodomain Prodomain RGD RGD Arm Arm Arm domain domain domain Arm Pro- Pro- domain β1 BMP9 α2 TGF-β1 α2 α2 β1 β7 α5 α5 α2 β-finger β7 β6 β6 β-finger α5 Growth Growth Strait- Latency factor factor jacket lasso Latency lasso α1 α1 C D Growth Growth Pro-BMP9 Pro-BMP7 factor Cys factor linkage E248 I K393 α5 F BMP9 0.88 0.88 0.86 TGF-β1 K350 Bowtie Y396 E BMP9 β9 β8 C196 M252 C194 C214 W322 RGD β4 C133 β9’ H255 β4 β3 W319 β5 β2 β5 Arm β7 β2 β7 β1' domain β6 Arm β6 α4 α4 domain β3 J β3 α3 β10 β10 TGF-β1 β1 α2 β1 α2 Fastener L28 α5 Y339 α1 Latency W281 lasso α1 I24 I20 α5 G H I17 BMP9 TGF-β1 W279 R21 Pro- β10 Pro- β10 domain 1 β1 domain 1 K β1 β-finger α2 α5 α2 α5 β7 β7 BMP9 α1 α5 Cross- β6 β6 β-finger armed Y383 L47 model Latency F43 lasso M39 Growth α3 W322 factor 2 α1 Growth Growth Growth factor 2 L36 factor 1 factor 1 W319 K40

Fig. 1. Structures. (A and B) Cartoon diagrams of pro-BMP9 (A) and pro-TGF-β1 (10) (B) with superimposition on GF dimers. Disulfides (yellow) are shown in stick. (C and D) Representative negative-stain EM class averages of pro-BMP9 (C) and pro-BMP7 (D). Best correlating projections of the pro-BMP9 crystal structure with their normalized cross-correlation coefficients are shown below class averages. (Scale bars, 100 Å.) (E and F)BMP9andTGF-β1 prodomains shown in cartoon after superimposition. Core arm domain secondary structural elements are labeled in black and others in red. Helices that vary in position between cross- and open-armed conformations are color-coded. Spheres show Cys S atoms. (G–K) Prodomain–GF interactions in pro-BMP9 (G and I), pro-TGF-β1(H and J), and a model of cross-armed BIOPHYSICS AND

pro-BMP9 (K). Structures are superimposed on the GF monomer. Colors are as in A, B, E,andF. Key residues are shown in stick. COMPUTATIONAL BIOLOGY

Mi et al. PNAS | March 24, 2015 | vol. 112 | no. 12 | 3711 A α2 B BMP9 --KPLQNWEQASPGENAHSSLGLSGA-GEEGVFDLQMFLENMKVDFLRSLNLSGIPSQDKTRAE-PPQYMIDLYNRYTTDKSST------80 BMP9 NLKMFLENVKVDFLRSLNLSGP V 54 TGFβ1 ------LSTCKTIDMELVKRKRIEAIRGQILSKLRLASPPSQGDVPPGPlpEAVLALYNSTRDRVAgeSVEPEPEP70 BMP10 DFNTLLQ SMKDEFLKTLNLSDIP 51 β1 α1 Latency lasso TGFβ1 VKRKRIEAIRG QILSKLRLASPP 34 β1’ β2 β3 α3 β4 GDF8 TKSSRIEAIKIQILSKLRLETAP 47 BMP9 ---PASNIVRSFSVEDAI------STAATEDFPFQKHILIFNIS-----IPRHEQITRAELRLYVSCQNDVDSTHGLEGSMVVYDVLEDS 156 Nodal ------TVATALLRTRG QPP SS 16 TGFβ1 EADyyAKEVTRVLMVESgNQIYDKFKgTPH------SLYMLFNTsELREAVPEpVLLSRAELRLLRLKl------KVEQHVELYQKYSQ- 147 BMP2 PPSDEVLSEFELRLLSMFG LKQR 44 α3 BMP7 LRSQERREMQREILSILG LPP HR 39 β5 β6 α4 β7 β9’ GDF9 LLQHIDERDRAG LLP ALFKVLSV 47 BMP9 ETWDQATGTKTFLVSQDIRD---EGWETLEVSSAVKRWVRADSTTNKNKLEVTVQS------HRESCDTLDIS-VPPGSKNLPF 230 BMP3 RPPPP K F ELRKAV PG DRTA GGG P24 GDF5 APP KAG SV P SSFLLKKAREPG PP 128 TGFβ1 ------DSWRYLSNRLLApSDSPEWLS FDVTGVVRQWLTRREa--IEGFRLSAHcSCDSKDNTLHVEINgFNSGRRGdLATIHg--mNRPF 228 GDF3 ------QEYVFLQFLG LDKAP 15 β8 β9 β10 α5 β1 β2 Inh.α ARELVLAKVRALFLDAL---GPP 26 Inh.βA SQPP EMVEAVKKHILNMLHLKKR 52 BMP9 FVVFSNDRSNGTKETRLELKEMIGHEQETMLVKTAKNAYQVAGESQEEEGLDGYTAVGPLLARRKR------SAGAGSHCQKTSLR VNFE- 315 LEFTY1 ---LTGG EQLL SLLRQLQLKEVP 20 TGFβ1 LLLMATPleRAQh------LHSSRHRRALDTNYCfssteKNCCVRQLY IDFRK 275 α5 α1 α2 α2 β3 β4 α3’ α3 β5 β6 β7 β8’ β8 BMP9 DIGWDSWIIAP KEYEAYECKG GCFFPLADDVTPT KHAIVQTLVHLKFPTKVGKACCVPTKLSPISVLYKDDMGVPTLKYHYEGMSVAECGCR 407 TGFβ1 DLGW-KWIHEPKGYHANFCLGPCPYIwsld---tqyskvlalynqhnpGASAAPCCVPQALEPLPIVYYVG--RKPKVEQLSNMIVRSCKCS 361

Fig. 2. Sequence alignments. (A) Sequences aligned structurally with SSM (26). Structurally aligned residues are in uppercase; lowercase residues are not structurally aligned. Residues with missing density are in italics. Decadal residues are marked with black dots. Residues with more than 50% solvent accessible surface area buried in prodomain–GF interfaces are marked with gold dots. Secondary structural elements or named regions are shown with lines. Long black arrow marks cleavage sites between the prodomain and GF; short black arrows mark putative PC cleavage sites after the α1-helix and in the α5-helix. Stars mark HHT-like mutation sites. (B) Sequence alignment of representative TGF-β family members in the prodomain α1-helix (overlined for TGF-β1). dimerize in their palm regions (Fig. 1A and Fig. S4)inanin- position of the α1-helix in the cross-armed pro-TGF-β1 confor- terface that buries 1,280 Å2 of solvent-accessible surface. mation (Fig. 1 A, B, G,andH). The differing twists between the The interface between each BMP9 prodomain and GF mono- arm domain and GF domains in open-armed and cross-armed 2 mer buries 1,440 Å . The larger size of the prodomain–GF in- conformations relate to the distinct ways in which the prodomain terface than the GF–GF interface emphasizes its significance, as α5-helix in pro-BMP9 and the α1-helix in pro-TGF-β1 bind to the β does the super -sheet interface between the prodomain and GF GF (Fig. 1 A and B). The strong sequence signature for the and the burial of hydrophobic residues by this interface and by the α1-helix in pro-BMP9, which is essential for the cross-armed prodomain α2-helix (Fig. 1A). A specialization in pro-BMP9 not β β α A B E F conformation in pro-TGF- , suggests that pro-BMP9 can also present in pro-TGF- 1 is a long 5-helix (Fig. 1 , , ,and ) adopt a cross-armed conformation (Discussion). that is a C-terminal appendage to the arm domain and that sep- α 2 A In absence of interaction with a prodomain 1-helix, the GF arately interacts with the GF dimer to bury 750 Å (Fig. 1 ). dimer in pro-BMP9 is much more like the mature GF (1.6-Å Despite markedly different arm domain orientations, topo- RMSD for all Cα atoms) than in pro-TGF-β1 (6.6-Å RMSD; logically identical secondary structure elements form the interface β Fig. S4). Moreover, burial between the GF and prodomain between the prodomain and GF in pro-BMP9 and pro-TGF- 1: the 2 β α β β dimers is less in pro-BMP9 (2,870 Å )thaninpro-TGF-β1 1-strand and 2-helix in the prodomain and the 6- and 7-strands 2 in the GF (Fig. 1 A, B, G,andH). The outward-pointing, open arms (4,320 Å ). In the language of allostery, GF conformation is β of pro-BMP9 have no contacts with one another, which results in tensed in cross-armed pro-TGF- 1 and relaxed in open-armed a monomeric prodomain–GF interaction. In contrast, the in- pro-BMP9. ward pointing arms of pro-TGF-β1 dimerize through disulfides in their bowtie motif, resulting in a dimeric, and more α2 avid, prodomain-GF interaction (Fig. 1 A and B). ABα2 Pro-BMP9 arm Twists at two different regions of the interface result in the Pro-TGFβ1 arm remarkable difference in arm orientation between BMP9 and BMP9 TGF-β1 procomplexes. The arm domain β1-strand is much more TGFβ1 twisted in pro-TGF-β1 than in pro-BMP9, enabling the β1-β10- β3-β6 sheets to orient vertically in pro-TGF-β and horizontally in pro-BMP9 in the view of Fig. 1 A and B. In addition, if we imagine the GF β7- and β6-strands as forefinger and middle α2 α2 finger, respectively, in BMP9, the two fingers bend inward to- β ward the palm, with the 7 forefinger bent more, resulting in CDBMP9 TGFβ1 G H R128 cupping of the fingers (Fig. 1 and and Fig. S4). In contrast, Pro- W179 in TGF-β1, the palm is pushed open by the prodomain amphi- domain Y65 β3 β10 Pro- α F234 β3 β1 domain pathic 1-helix, which has an extensive hydrophobic interface β10 with the GF fingers and inserts between the two GF monomers F230 B M66 α2 V48 (Fig. 1 ) in a region that is remodeled in the mature GF dimer α2 β1 and replaced by GF monomer–monomer interactions (10). V347 P345 L392 Y52 β7 β7 Y70 P390 Role of Elements N and C Terminal to the Arm Domain in Cross- and L382 V338 Open-Armed Conformations. A straitjacket in pro-TGF-β1 com- β6 β6 posed of the prodomain α1-helix and latency lasso encircles the GF GF GF on the side opposite the arm domain (Fig. 1B). Sequence for P326 P285 putative α1-helix and latency lasso regions is present in pro- A Fig. 3. The prodomain α2-helix. (A and B) Superimpositions based on the BMP9 (Fig. 2 ); however, we do not observe electron density arm domain (A) or on the GF (B), which are shown as Cα traces with the corresponding to this sequence in the open-armed pro-BMP9 α2-helix in cartoon. (C and D) Prodomain α2-helix environments in pro-BMP9 map. Furthermore, in the open-armed pro-BMP9 conformation, (C) and pro-TGF-β1(D) shown after superimposition on the prodomain the prodomain α5-helix occupies a position that overlaps with the α2-helix and GF β6andβ7-strands. Key interacting residues are shown in stick.

3712 | www.pnas.org/cgi/doi/10.1073/pnas.1501303112 Mi et al. Kd = 1.0 ± 0.3 μM Kd = 0.8± 0.1 μM BMP9 Pro-BMP9 A ΔH = -15.9 ± 0.8 kcal/mol B ΔH = -16.2 ± 0.4 kcal/mol E F GG ΔS = -25.8 cal/mol·K ΔS = -26.8 cal/mol·K EC50 (nM) EC50 (nM) BMP9 Pro-BMP9 BMP9 prodomain Pro/BMP9 dimer=1.94 ± 0.06 Pro/BMP9 dimer=2.23 ± 0.06 ALK1 0.023±0.009 0.039±0.01 EC50 (nM) EC50 (nM) 0.00 0.00 ALK2 70±40 >100 ALK3 >100 >100 ActRIIA 0.5±0.2 >100 ALK1 -0.02 ALK4 >100 >100 ActRIIB 0.04±0.02 0.2±0.1 α5 ALK5 110±60 >100 BMPRII 0.04±0.02 1.6±0.7 BMP9 -0.04 AMHRII 17±5 70±20

μcal/sec ALK6 >100 >100 μcal/sec -0.04 ALK7 33±8 60±30 TBRII >100 >100 α3 -0.06 1 BMP9 1 BMP9 0306090(min) -0.08 0 30(min) 60 90 -2.0 0 0.8 0.8 0.6 0.6 0.4 -6.0 -10.0 0.4 ActRIIB 0.2 0.2 BMP9 -10.0 -20.0 0 0 α2 1 1 Binding Binding Pro-BMP9 Pro-BMP9 -14.0 -30.0 0.8 0.8

kcal mol-1 of injectant kcal mol-1 of injectant 0.0 1.0 2.0 3.0 0.0 0.5 1.0 0.6 0.6 HH BMP9 prodomain (Fraction of maximum) (Fraction [Prodomain]/[BMP9 dimer] [BMP9 dimer]/[Prodomain] of maximum) (Fraction 0.4 0.4 0.2 0.2 BMP9 BMP9 EC50=0.9 nM Prodomain IC50=79 nM α5 Clip C Pro-BMP9 EC50=1.3 nM D GCN4-pro IC50=23 nM 0 0 10-12 10-11 10-10 10-9 10-8 10-7 10-12 10-11 10-10 10-9 10-8 10-7 1.5 [Receptor ECD-Fc] (M) [Receptor ECD-Fc] (M) BMP9 1.0 1.0 Pro-BMP9 0.5 (A405) (A405) 0.5 [BMP9] = 1 nM (of maximum) 0.0 (of maximum) 0.0 CV2 Alkaline phosp. Alkaline phosp. Alkaline phosph. Alkaline phosph. α2 10-10 10-9 10-8 10-7 10-10 10-9 10-8 10-7 VWC [BMP9] (M) [Prodomain] (M) BMP9

Fig. 4. Binding of BMP9 to the prodomain and receptors. (A and B) Representative ITC data at 15 °C adding material either to the BMP9 dimer (A)or prodomain (B) in the calorimetry cell. (Upper) Baseline-corrected raw data. (Lower) Integrated heats fit to the independent-binding site model. (C and D) Differentiation of C2C12 cells measured by alkaline phosphatase production. (C) Comparison of BMP9 and pro-BMP9. (D) Inhibition of 1 nM BMP9 by GCN4- linked, oligomeric BMP9 prodomain and native prodomain. (E and F)EC50 values of BMP9 and pro-BMP9 for Fc-fused type I (E) or type II (F) receptor ectodomains measured using quantitative ELISA. Data are plotted as the fraction of maximal bound in each experiment and fit to the equation of fractional saturation. Graphs show average of triplicates in one experiment; numerical values show mean EC50 and sd from three such experiments. (G and H) Superimpositions on pro-BMP9 of the BMP9–receptor complex (17) (G) and BMP2-crossveinless2 complex (H) in identical orientations. For clarity, BMP in receptor and inhibitor complexes is omitted, and the α5-helix is transparent in H.

The prodomain α2-helix covers the interface between the arm that the two arm domain cysteines, Cys-133 and Cys-214, form an and GF domains and acts as a buffer between the open-armed intrachain disulfide that links the β3strandtotheβ7-β9’ loop (Fig. and cross-armed conformations. With superimposition based on 1E). The disulfide helps stabilize an extension of the β3-strand in arm domains, α2-helix orientation differs by ∼90° in pro-BMP9 BMP9 and the formation of the β1’- and β9’-strands unique to and pro-TGF-β1 (Figs. 1 E and F and 3A), whereas with super- pro-BMP9 that add onto the β2-β7-β5-β4 sheet (Fig. 1 E and F). imposition based on GF monomers, α2-helix orientation is sim- The α5-helix in pro-BMP9 is its most surprising specialization. ilar in pro-BMP9 and pro-TGF-β1 (Figs. 1 G and H and 3B). It is much longer than in pro-TGF-β1, orients differently (Fig. 1 Thus, if we imagine shape-shifting between open- and cross- E and F), and binds to a similar region of the GF domain as armed conformations of pro-BMP9 (Discussion), the α2-helix the α1-helix in pro-TGF-β1. However, the prodomain α1 and moves in harmony with the GF rather than the arm domain. α5-helices orient differently on the GF domain (Fig. 1 A, B, G, Prodomain α2-helix association with the GF is stabilized by and H). The BMP9 prodomain α5-helix inserts into the hydro- conserved interactions. Homologous α2-helix residues Tyr-70 in phobic groove formed by the fingers of one GF monomer and pro-BMP9 and Tyr-52 in pro-TGF-β1 stack against homologous the α3-helix of the other monomer (Fig. 1A). This association is Pro-326 and Pro-285 residues in their GFs (Fig. 3 C and D). stabilized by a cluster of specific interactions (Fig. 1I). Glu-248, Moreover, homologous α2-helix residues Met-66 in pro-BMP9 at the N terminus of the α5-helix, forms salt bridges with GF and Val-48 in pro-TGFβ1 are buried in a hydrophobic cavity on residues Lys-393 and Lys-350. In the middle of the α5-helix, their GFs (Fig. 3 C and D). Met-252 plunges into a hydrophobic cavity. At the C terminus, The α2-helix in open-armed pro-BMP9 interacts with the arm His-255 stacks against GF residue Trp-322 (Fig. 1I). However, domain in a way not seen in cross-armed pro-TGF-β1. Tyr-65 GF burial by the pro-BMP9 α5-helix (750 Å2) is less than by the from the α2-helix together with Trp-179 and Phe-230 from the arm pro-TGF-β1 α1-helix (1,120 Å2)orα1-helix plus latency lasso domain form an aromatic cage (Fig. 3C). Arm residue Arg-128 at (1,490 Å2). Furthermore, when crystals were cryo-protected with the center of this cage forms π–cation interactions with Tyr-65 and a 10% higher concentration of ethanol (3.25-Å dataset; Table Trp-179 (Fig. 3C). Residues for the π-cation cage are well con- S1), density for the α5-helix was present in one monomer but served in BMP4, 5, 6, 7, 8, and 10, GDF5, 6, and 7, and GDF15 not the other (Fig. S6). (Fig. S5). However, in BMP2 and BMP15, Arg-128 is replaced by Gln, potentially weakening association of the prodomain with the Prodomain Functions. We next asked if interactions of the two GF in the open-armed conformation. BMP9 prodomains with the GF dimer are independent or co- The similar arm domain cores and α2-helices in the prodomains operative. Isothermal calorimetry (ITC) showed that, irrespective of BMP9 and TGF-β1 are remarkable, given that the prodomains of whether increasing amounts of prodomain were added to GF have only 11% identity in sequence and have 12 insertions/ or vice versa, heat production showed a single sigmoidal profile deletions (Fig. 2A). This contrasts with the 25% identity between (Fig. 4 A and B). Curves fit well to a model in which the two their GF domains (Fig. 2A). Among notable differences, pro- binding sites are independent, and yielded KD values of 0.8–1.0 μM BMP9 lacks the 14-residue bowtie in pro-TGF-β1 that disulfide at pH 4.5, which maintains BMP9 solubility. links the two arm domains together and has in its place a β7-β9’ A critical question concerning BMP prodomains is whether the loop (Fig. 2A). The two cysteine residues in the TGF-β1 arm do- BMP9 prodomain inhibits GF signaling and whether making the main, Cys-194 and Cys-196 (Fig. 1F), form reciprocal interchain BMP9 prodomain dimeric as in pro-TGF-β1 would provide suf- BIOPHYSICS AND

disulfide bonds (10). In contrast, our pro-BMP9 structure shows ficient avidity to keep the GF latent. Consistent with previous COMPUTATIONAL BIOLOGY

Mi et al. PNAS | March 24, 2015 | vol. 112 | no. 12 | 3713 (2) bloodstream at physiologically relevant concentrations (11), Pro-BMP+ ? our finding that the BMP9 prodomain blocks binding to ActRIIA Cell Matrix ? α5 and alters selectivity for ActRIIB compared with BMPRII has (b) important implications for BMP9 signaling in vivo. α1 The pro-BMP9 structure explains the selective effect of BMP prodomains on type II receptor binding (18) (Fig. 4G). The prodo- Pro-BMP main arm domain and α2-helix occupy the type II receptor binding α (c) Prodomains site. In contrast, only the prodomain 5-helix blocks the type I re- (a) ceptor binding site (Fig. 4G). Selective displacement of the α5-helix (1) by the type I receptor with retention of arm domain association is α Pro-BMP consistent with the relatively weak 5-helix interface described above. α1 In contrast, arm domain binding to the type II receptor site is strong enough to alter receptor affinity and selectivity. α5 Antagonists including noggin, gremlins, and chordins bind to (3) (4) BMP GFs and regulate activity. However, little attention has α1 GF been paid to whether BMP prodomains might prevent antagonist α5 binding. Interestingly, the BMP-inhibiting fragment of the chor- + din family member crossveinless-2 binds to interfaces on BMP2 (20) similar to those on BMP9 to which the prodomain binds (Fig. 4H). The von Willebrand factor C (VWC) domain binds P P to a similar site on the GF fingers as the arm domain, whereas an N-terminal appendage called clip binds to the same site as Type I the prodomain C-terminal appendage, the α5-helix (Fig. 4H). Type II Whether prodomains can protect GF from inhibitors, as well as prevent GF binding to receptors, deserves study. Fig. 5. Models for pro-BMP9 structures and binding to receptors. Models The crystal structure of pro-BMP9 begins to reveal how prodo- for the open-armed, nonlatent pro-BMP9 conformation characterized here β – mains contribute to the tremendous functional diversity among the (1), the proposed (dashed lines) pro-TGF- 1 like cross-armed conformation β of pro-BMP9 (2), and stepwise binding (18) to type I (3) and type II receptors (4). 33 members of the TGF- family. Many of these members have prodomains that differ even more than BMP9 and TGF-β,which have only 11% sequence identity. Prodomain divergence may in- studies (16), pro-BMP9 and BMP9 showed little difference in crease the specificity of GF signaling in vivo by regulating pro- inducing C2C12 cell differentiation, with EC50 values of 1.3 and complex localization, movement, release, and activation in the 0.8 nM, respectively (Fig. 4C). However, by doing the experiment extracellular environment. a different way, i.e., by titrating the prodomains into assays with 1 The open-armed pro-BMP7 and 9 and cross-armed pro-TGF-β1 nM BMP9, we found that the prodomain is inhibitory, with IC50 = conformations differ greatly. Overall learnings from protein families 79 nM (Fig. 4D). Thus, a large excess of the BMP9 prodomain that can adopt multiple conformations, such as tyrosine kinases, over the GF is required to inhibit signaling, whereas at equimolar integrins, G protein-coupled receptors, membrane channels, and 1-nM concentrations of prodomain and GF monomers, the membrane transporters, show that when markedly distinct con- BMP9 prodomain has little effect on signaling. We added a cys- formations are glimpsed for individual members, most family teine-bearing GCN4 coil at the N terminus of the BMP9 pro- members can visit each state, often in a manner that is regulated domain to disulfide-link BMP9 prodomains into dimers or by other interactors. Thus, we hypothesize that most members of the TGF-β family can visit both cross-armed and open-armed oligomers (Fig. S7). Multimerization modestly lowered the IC50 to β 23 nM (Fig. 4D) but did not induce latency at equimolar conformations. TGF- is a later evolving family member; whereas BMPs and activins are found in all metazoans, TGF-β is found concentrations. only in deuterostomes. Furthermore, TGF-β is the only known To determine the basis for the partial inhibition of BMP9 member with disulfide-linked arm domains. Thus, trapping pro- signaling by the prodomain, we compared binding of receptor Fc TGF-β in a solely cross-armed conformation with disulfides may fusions to pro-BMP9 and BMP9 in quantitative ELISA (Fig. 4 E F be a later evolutionary adaptation. and ). Among the seven type I receptors, ALK1 bound by far The amino acid sequence of a protein is constrained by its the strongest, and there was little difference in affinity and no E structure, and sequence conservation in evolution is a powerful difference in selectivity between BMP9 and pro-BMP9 (Fig. 4 ). predictor of protein structure and conformation. The prodomain In contrast, the prodomain altered both EC50 values and selec- α1-helix has an important function in stabilizing the cross-armed F tivity of type II receptors (Fig. 4 ). Whereas BMP9 bound conformation but has no function in the open-armed conformation, ActRIIB and BMPRII similarly and 10-fold better than ActRIIA, as shown by lack of electron density and presence of the prodomain pro-BMP9 bound ActRIIB 10-fold better than BMPRII and did not α5-helix in a position that prevents α1-helix binding. In support of F detectably bind ActRIIA (Fig. 4 ). The EC50 values for type I and II the hypothesis that pro-BMP9 can adopt a cross-armed confor- receptors agree with previous affinity measurements with BMP9 mation, the amino acid sequence corresponding to the α1-helix is (16, 17). Interestingly, the BMP7 prodomain also competes binding highly conserved (44–79% identity at residues 29–47) among of BMP7 to type II but not type I receptors (18), consistent with the human, mouse, zebrafish, and chicken BMP9s. Indeed, the se- open-armed conformation of pro-BMP7 (Fig. 1D). quence of the α1-helix is more conserved than the remainder of – Discussion the prodomain (33 74% identity). Furthermore, the prodomain α1-helix sequence and its amphipathic signature are also con- BMPRII and ActRIIA are coexpressed with ALK1 and mediate served among diverse representatives of the 33-member TGF-β BMP9 signaling in endothelial cells (12, 19). Furthermore, family including BMP7 (Fig. 2B). Importantly, the α1-helix and its ALK1, which primarily functions as a receptor for BMP9 and amphipathic signature are highly conserved between pro-TGF-β1 BMP10, has recently been implicated as an important target and pro-BMP9 (Fig. 2). These results support the hypothesis that of antiangiogenic tumor therapy (17). Together with the finding pro-BMP9 and other TGF-β family members can adopt an α1-helix- that BMP9 complexed with its prodomain circulates in the bound, cross-armed conformation similar to that of TGF-β1.

3714 | www.pnas.org/cgi/doi/10.1073/pnas.1501303112 Mi et al. To more directly test evolutionary support for a cross-armed competent, open-armed conformations (Fig. 5, pathway a and BMP9 conformation, we made a pro-TGF-β1–like model of pro- structure 1). Other family members may be secreted together with BMP9 that uses the BMP9 conformation of the arm domains, or immediately form complexes after secretion with extracellular superimposed on the cross-armed orientation of the arm do- matrix components including heparin, proteoglycans, fibrillin, and mains in pro-TGF-β1, and pro-TGF-β1–like conformations of latent TGF-β binding proteins (8, 9, 21–23). These interactors may prodomain α1- and α2-helices and GF domains (Movie S1). In stabilize the cross-armed conformation (Fig. 5, pathway b and addition to the α1-helix, pro-BMP9 also includes a latency lasso- structure 2), and enable the GF domain, which is very short lived like sequence, including an identical PSQ sequence (Fig. 2A). in vivo, to remain latent and reach storage concentrations as high There are no clashes between the two pro-BMP9 arm domains in as 100 ng/g in demineralized bone (24). Release from storage in the crossed-arm conformation; notably, the arm domains come vivo may then yield the open-armed conformation, which is close together at their β4 and β5-strands, which are on the side of ready for receptor or inhibitor binding (Fig. 5, pathway c). the arm domain conserved between pro-BMP9 and pro-TGF-β1 TGF-β family members with long sequences at the ends of (Movie S1). The extensive, amphipathic α1-helix–GF interface in their prodomains that may have α5-helix–like functions include pro-TGF-β1 is recapitulated well in the cross-armed pro-BMP9 BMP3, BMP10, BMP15, GDF5, 6, 7, and 9, anti-Müllerian model, and the long α5-helix can adopt a conformation similar to hormone, and nodal (Fig. S5). A number of these, including BMP9 the shorter α5-helix in pro-TGF-β1 without clashes (Fig. 1K). and BMP10, have basic sequences resembling PC cleavage sites These results compellingly support a cross-armed conformation (25) in or before the α5-helix (Fig. 2A and Fig. S5). Moreover, for pro-BMP9. A plausible pathway for structural interconversion many TGF-β family members have PC or tolloid cleavage sites in between open-armed and cross-armed conformations of BMP9 or after the prodomain α1-helix that regulate activation or sig- can be described in which crossing of the arms is accompanied by naling (6, 7, 9, 10, 25) (Fig. S5). Indeed, recombinant pro-BMP9 dissociation of the α5-helix from the GF and its replacement by preparations contain a minor component cleaved at a putative the α1-helix and latency lasso (Movie S1). PC site in this region (Fig. 2A and Methods). Thus, potential The strong evolutionary and 3D structural support for a cross- mechanisms for regulating the switching between open-armed and armed conformation of BMP9 (and also BMP7; Fig. 2B) con- cross-armed procomplex conformations include removal of the trasts with our lack of observation of cross-armed BMP7 and α1- or α5-helix by proteases in addition to binding to extracellular BMP9 conformations in EM (Fig. 1 C and D). However, this is matrix components. Our results suggest that the open-armed easily explicable, because it is compatible with a lower energy of the conformation of pro-BMP9 can readily bind to type I receptors, open-armed conformation for the isolated procomplex, and on the with displacement of the α5-helix (Fig. 5, structure 3). The other hand, with a lower energy of the cross-armed conformation final step in signaling could then be binding to type II receptors, for the procomplex bound to an interactor. For BMPs in bone, such with complete prodomain dissociation, consistent with a pre- interactors may be present in the residual matrix, and release from vious model of stepwise receptor binding and prodomain dis- interactors may in part be responsible for the increase in BMP placement (18) (Fig. 5, structure 4). activity found after extraction by denaturants and purification (2). We hypothesize that cross-armed and open-armed conformations Methods β of TGF- family members correspond to latent and nonlatent Pro-BMP9 and pro-BMP7 were purified from supernatants of CHO and HEK293 states, respectively, and propose a model for conformational regu- cell transfectants, respectively. Crystals formed in 0.15 M zinc acetate, 0.1 M lation of release from storage and latency (Fig. 5). Some family sodium cacodylate, pH 5.8, and 4% (vol/vol) isopropanol. Phases were solved members may be secreted as isolated procomplexes in signaling- using Zn anomalous diffraction. Complete methods are described in SI Methods.

1. Wang EA, et al. (1988) Purification and characterization of other distinct bone- 14. David L, Feige JJ, Bailly S (2009) Emerging role of bone morphogenetic proteins in inducing factors. Proc Natl Acad Sci USA 85(24):9484–9488. angiogenesis. Cytokine Growth Factor Rev 20(3):203–212. 2. Luyten FP, et al. (1989) Purification and partial amino acid sequence of osteogenin, 15. Wooderchak-Donahue WL, et al. (2013) BMP9 mutations cause a vascular-anomaly a protein initiating bone differentiation. J Biol Chem 264(23):13377–13380. syndrome with phenotypic overlap with hereditary hemorrhagic telangiectasia. Am J 3. Celeste AJ, et al. (1990) Identification of transforming growth factor beta family Hum Genet 93(3):530–537. members present in bone-inductive protein purified from bovine bone. Proc Natl 16. Brown MA, et al. (2005) Crystal structure of BMP-9 and functional interactions with Acad Sci USA 87(24):9843–9847. pro-region and receptors. J Biol Chem 280(26):25111–25118. 4. Cui Y, et al. (2001) The activity and signaling range of mature BMP-4 is regulated by 17. Townson SA, et al. (2012) Specificity and structure of a high affinity activin receptor- sequential cleavage at two sites within the prodomain of the precursor. Genes Dev like kinase 1 (ALK1) signaling complex. J Biol Chem 287(33):27313–27325. 15(21):2797–2802. 18. Sengle G, Ono RN, Lyons KM, Bächinger HP, Sakai LY (2008) A new model for growth 5. Harrison CA, Al-Musawi SL, Walton KL (2011) Prodomains regulate the synthesis, factor activation: Type II receptors compete with the prodomain for BMP-7. J Mol Biol extracellular localisation and activity of TGF-β superfamily ligands. Growth Factors 381(4):1025–1039. 29(5):174–186. 19. Scharpfenecker M, et al. (2007) BMP-9 signals via ALK1 and inhibits bFGF-induced 6. Constam DB (2014) Regulation of TGFβ and related signals by precursor processing. endothelial cell proliferation and VEGF-stimulated angiogenesis. J Cell Sci 120(Pt 6): Semin Cell Dev Biol 32:85–97. 964–972. 7. Akiyama T, Marqués G, Wharton KA (2012) A large bioactive BMP ligand with distinct 20. Zhang JL, et al. (2008) Crystal structure analysis reveals how the Chordin family signaling properties is produced by alternative proconvertase processing. Sci Signal member crossveinless 2 blocks BMP-2 receptor binding. Dev Cell 14(5):739–750. 5(218):ra28. 21. Sengle G, et al. (2008) Targeting of bone morphogenetic protein growth factor 8. Gregory KE, et al. (2005) The prodomain of BMP-7 targets the BMP-7 complex to the complexes to fibrillin. J Biol Chem 283(20):13874–13888. extracellular matrix. J Biol Chem 280(30):27970–27980. 22. Li S, et al. (2010) Activin A binds to perlecan through its pro-region that has heparin/ 9. Sengle G, Ono RN, Sasaki T, Sakai LY (2011) Prodomains of transforming growth heparan sulfate binding activity. J Biol Chem 285(47):36645–36655. factor β (TGFβ) superfamily members specify different functions: Extracellular matrix 23. Anderson SB, Goldberg AL, Whitman M (2008) Identification of a novel pool of ex- interactions and growth factor bioavailability. J Biol Chem 286(7):5087–5099. tracellular pro-myostatin in skeletal muscle. J Biol Chem 283(11):7027–7035. 10. Shi M, et al. (2011) Latent TGF-β structure and activation. Nature 474(7351):343–349. 24. Pietrzak WS, Woodell-May J, McDonald N (2006) Assay of bone morphogenetic 11. Bidart M, et al. (2012) BMP9 is produced by hepatocytes and circulates mainly in an protein-2, -4, and -7 in human demineralized bone matrix. J Craniofac Surg 17(1): active mature form complexed to its prodomain. Cell Mol Life Sci 69(2):313–324. 84–90. 12. Chen H, et al. (2013) Context-dependent signaling defines roles of BMP9 and BMP10 25. Seidah NG, Prat A (2012) The biology and therapeutic targeting of the proprotein in embryonic and postnatal development. Proc Natl Acad Sci USA 110(29):11887–11892. convertases. Nat Rev Drug Discov 11(5):367–383. 13. Castonguay R, et al. (2011) Soluble endoglin specifically binds bone morphogenetic 26. Krissinel E, Henrick K (2004) Secondary-structure matching (SSM), a new tool for fast proteins 9 and 10 via its orphan domain, inhibits blood vessel formation, and sup- protein structure alignment in three dimensions. Acta Crystallogr D Biol Crystallogr presses tumor growth. J Biol Chem 286(34):30034–30046. 60(Pt 12 Pt 1):2256–2268. BIOPHYSICS AND COMPUTATIONAL BIOLOGY

Mi et al. PNAS | March 24, 2015 | vol. 112 | no. 12 | 3715 Supporting Information

Mi et al. 10.1073/pnas.1501303112 SI Methods identified by SDS/PAGE and pooled. Purified protein (11.4 mg) Expression and Purification of Mouse–Human Chimeric BMP9. CHO was concentrated on Amicon-ultra-15 centrifugal concentration cells stably expressed full-length pro-BMP9 containing the mouse devices and dialyzed into 5 mM NaCl, 10% (vol/vol) glycerol, prodomain and human GF domain. Cells were seeded in roller 2.5% (wt/vol) glycine, and 5 mM glutamic acid, pH 4.5 centri- bottles in a 50:50 (vol/vol) mix of F12 and DMEM supplemented fuged at 1,400 × g, sterilized with a 0.22-μm syringe filter, sup- with nonessential amino acids, biotin, vitamin B12, 0.2 μM plemented with 0.01% Tween-80, and stored at 4 °C. The final methotrexate, and 10% (vol/vol) FBS and grown for 3–4dto yield was 9.1 mg (0.46 mg/L CM). near confluence. Serum-containing growth medium was dis- For purification of the prodomain, pro-BMP9 in PBS was mixed carded, roller bottles were rinsed with PBS, and a serum-free with an equal volume of a 9:1 (vol/vol) solution of ethanol:1 M production medium was added containing nonessential amino sodium acetate, pH 4.4. The mixture was incubated on ice for 1 h. acids, 5 μg/mL insulin, 12.9 μM putrescine, 0.2 μM hydrocorti- Precipitated BMP9 prodomain was collected by centrifugation at sone, 29 nM selenium, and 0.6 g/L polyvinyl alcohol. Condi- 20,000 × g for 10 min and dissolved in 20 mM Tris, pH 8.0. tioned medium (CM) was collected at 48 h, and roller bottles BMP9 prodomain was further purified by ion exchange chroma- were refed with fresh serum-free medium. Two sequential 48-h tography in the same buffer with a gradient of 0–1 M NaCl and harvests were collected. CM was clarified by passing through a S200 SEC. 5-μm pore filter (pass Profile) and a 0.22-μm pore Duropore filter (Millipore) (Patent: BMP9 Compositions WO 1995033830 A). Crystallization, Data Collection, and Structure Determination. For In a typical purification, the BMP9 procomplex was captured crystallization, pro-BMP9 in PBS was further purified by Superdex from ∼20 L of CHO CM by loading onto a 265-mL, 5.0 × 13.5-cm 200 HR SEC in 20 mM Bis-Tris-propane, pH 8.0, and 50 mM Q-Sepharose column equilibrated with 20 mM Tris, pH 8.6 NaCl, and concentrated to 6 mg/mL. Screening using sparse- (buffer A). After washing with 10 column volumes (CVs) of matrix, hanging drop vapor diffusion yielded a hit in 0.2 M zinc buffer A, BMP9 procomplex was eluted in a single step with acetate, 0.1 M Mes, pH 6.0, and 15% (vol/vol) ethanol. Fine- buffer A containing 1.0 M NaCl. matrix and additive screening found a final crystallization con- To purify pro-BMP9, the main peak was exchanged into buffer dition of 0.15 M zinc acetate, 0.1 M sodium cacodylate, pH 5.8, × A plus 30 mM NaCl with a 735-mL, 5.0 37.4-cm Sephadex G25 4% (vol/vol) isopropanol, and 0.15 M nondetergent sulfobetaine size-exclusion chromatography (SEC) column. The peak (720 mg (NDSB-211). Crystals were cryo-protected by rapid transfer to total protein) was divided into thirds and loaded onto an 8-mL, either 30% (vol/vol) ethanol (3.25-Å dataset) or 20% ethanol/ × 1.0 10.2-cm Source 15Q column equilibrated with 50 mM Mes, 10% glycerol (vol/vol) (3.3-Å dataset) in mother liquor and pH 6.1. After a 10-CV wash with equilibration buffer, the BMP9 – flash-frozen in liquid nitrogen. procomplex was eluted with a 12-CV gradient of 0.03 1 M NaCl X-ray diffraction data were collected at the Northeastern in 50 mM Mes, pH 6.1. SDS/PAGE and analytical SEC showed Collaborative Access Team beamline of Advanced Photon Source, that pro-BMP9 eluted between 0.15 and 0.32 M NaCl. Pooled Argonne, IL, and processed with X-ray Detector Software (1). For fractions from the Source15Q step (240 mg total protein from refinement of the 3.3-Å structure, eight datasets collected from three runs) were further purified with three SECs on two tandem eight isomorphic crystals were merged together. Initial phases 2.15 × 60-cm TOSOH G3000SWL columns equilibrated with were determined, and models were built using molecular re- PBS (total CV = 435 mL). Pro-BMP9 and free BMP9 prodomain placement with single wavelength anomalous scattering, (2) using eluted in separate peaks as shown by SDS/PAGE and analytical SEC and were pooled separately, sterilized with a 0.22-μm sy- the BMP9 dimer (3) as a search model and weak zinc anomalous ringe filter, and stored at 4 °C. Final yield was 125 mg (6.25 mg/L diffraction at 0.97918 Å in our 3.3-Å dataset. Subsequently, more CM). Material was characterized by SDS/PAGE and N-terminal accurate phases were solved using SAD in Phenix (2) with anom- sequencing of individual Coomassie blue-stained bands. Present alous data collected at a zinc absorption peak at 1.2823 Å in in the preparation was a putatively uncleaved prodomain GF a thin-sliced (0.2° oscillation per frame over 360° oscillation precursor that yielded the N-terminal prodomain sequence, the range), low-dose (1% transmission) mode using a PILATUS prodomain that yielded a major sequence beginning with the 6M detector. A total of 11 zinc binding sites were found by the N-terminal sequence and a minor sequence beginning at prodo- program, and the electron density was well resolved, giving main residue 46 (SLN), and a GF that yielded the GF sequence a figure of merit of 0.75. As the data for phasing and refinement of SAG beginning at residue 298. are not isomorphic (Table S1), phases were transferred to the 3.3-Å To purify mature BMP9, the main peak from the Q-Sepharose dataset using density_cut in Phenix (2) and Molrep in CCP4 (4). step of a separate purification run (720 mg total protein) was Phases were extended to higher-resolution shells by solvent diluted with 19 volumes of 6 M urea, 25 mM Mes, and 25 mM flattening and multicrystal averaging using the 3.3- and 3.6-Å Hepes, pH 6.5 and loaded onto a 265-mL, 5.0 × 13.5-cm datasets in Phenix (2). The 3.25-Å dataset was solved by mo- Q-Sepharose column in tandem with a 15 mL, 1.6 × 7.5-cm lecular replacement using the structure from the 3.3-Å dataset. SP-Sepharose column, both equilibrated with the same buffer. Model building with Coot (5) and refinement with Phenix After washing with equilibration buffer, the columns were sepa- including phenix.rosetta_refine (2) were iterated many times. rated, and BMP9 was eluted from the SP-Sepharose column with The resolution of the data was determined using paired re- a 12 CV gradient of 0–1 M NaCl in 6 M urea, 25 mM Mes, and finement (6). In each asymmetric unit, there are two pro-BMP9 25 mM Hepes, pH 6.5. BMP9 eluted between 0.25 and 0.4 M protomers, which form a functional dimer. NaCl. Fractions containing mature BMP9 were identified by SDS/ PAGE and pooled. Protein (13.3 mg total) was further purified Superimposition. Arm domain superimposition was with pro-BMP9 with two SECs on two tandem 2.15 × 60-cm TOSOH G3000SWL residues 84–92, 107–115, 116–132, 143–155, 164–175, 177–197, columns equilibrated with 6 M urea, 0.3 M NaCl, 25 mM Mes, and and 226–237; pro-TGF-β1 residues 77–85, 101–109, 115–131, 135– 25 mM Hepes, pH 6.5. Fractions containing mature BMP9 were 147, 148–159, 164–184, and 224–235; and “super” in PyMol.

Mi et al. www.pnas.org/cgi/content/short/1501303112 1of9 ELISA. BMP9 or pro-BMP9 was coated overnight at 4°C in PBS at 3 d in Freestyle 293 medium, medium was concentrated, and 1 μg/mL on Nunc MediSorb 96-well plates. Plates were washed buffer was exchanged with 20 mM Tris, pH 8.0, and 200 mM with PBST (PBS with the addition of 0.05% Tween 20) and NaCl. Protein was purified by Strep-tactin affinity chromatography. blocked with 2% (wt/vol) BSA in PBST for 5–8 h at 20 °C. Blocking solutions were decanted, and receptor-Fc fusion pro- C2C12 Differentiation Assay. The C2C12 cell line was from teins (R&D Systems, except ActRIIB-Fc, which was made in- American Type Culture Collection. p-Nitrophenyl phosphate was house) were applied at 2 μg/mL and 1:3 serial dilutions in PBST from Sigma-Aldrich. Activity of BMP9 was measured as pre- with 2% (wt/vol) BSA and incubated at 4 °C overnight. Plates viously described (10, 11) using induction of alkaline phospha- were washed three times with PBST, followed by the addition tase, a biomarker of osteogenesis, in myoblast C2C12 cells. of JDC-10 mouse anti-human Fc conjugated with HRP (Southern C2C12 cells seeded at 104 cells per well in 96-well plates 1 d Biotech) at 1:5,000 in PBST + 1% BSA. After overnight incubation previously in DMEM/10% (vol/vol) FBS were washed two times at 4 °C, plates were washed with PBST (0.05% Tween 20), and color and suspended in 100 μL DMEM. Test proteins in 100 μL was developed with a standard 3,3′,5,5′ tetramethylbenzidine DMEM and 1 mg/mL BSA were added in triplicate. After 2 d, μ reagent. After stopping with 1 M H2SO4, plates were read at 450 nm. wells were washed with PBS, 50 L distilled water was added, Binding data were plotted as the percentage of maximal receptor Fc and plates were frozen and thawed three times. Freshly made boundineachexperimentandfitto an equation for fractional sat- 0.1 M glycine, pH 10.3, 0.1% Triton X-100, 19 mM p-nitrophenyl μ uration. Reported EC50 values represent the average and SD from phosphate, and 17 mM MgCl2 (50 L) was added. After 30 min triplicate experiments. at 37 °C, 100 μL 0.2 N NaOH was added, and alkaline phosphatase activity was measured at 405 nm. Pro-BMP7 Expression and Purification. HEK293 cells that stably express WT pro-BMP7 were kindly provided by Lynn Sakai, ITC. Proteins were dialyzed overnight against 0.5% sucrose, 5 mM Shriners Hospital for Children, Portland, OR (7). Cells were NaCl, 0.01% Tween 80, 2.5% (wt/vol) glycine, and 5 mM glutamic cultured in DMEM with 10% (vol/vol) FBS and 0.5 mg/mL acid, pH 4.5, degassed, and centrifuged at 20,000 × g for 10 min. G418. Supernatants (2 L) were clarified by centrifugation, con- ITC experiments used MicroCal iTC200 (GE Healthcare Life centrated 10-fold with tangential flow filtration (Vivaflow 200; Sciences). Either 94.8 μM BMP9 prodomain was titrated into Sartorius Stedim), diluted 5-fold with Tris-buffered saline, pH 8.0 5 μM BMP9 dimer in the cell or 47.4 μM BMP9 dimer was titrated (TBS), and then concentrated 5-fold. The material was loaded into 10 μM BMP9 prodomain in the cell. A priming injection of onto 20 mL Ni-NTA agarose (Qiagen), washed with 200 mL of 0.4 μL (not included in data analysis) was followed by 2-μLin- 1 M NaCl, 25 mM imidazole, and 20 mM Tris·HCl, pH 8.0, and jections every 300 s. Data averaged over 2-s windows were an- eluted by 0.5 M NaCl and 1 M imidazole, pH 8.0. The Ni-NTA alyzed using Origin 7. Fits using cooperative binding or two eluent containing pro-BMP7 was purified by SEC using a HiLoad sequential binding sites did not yield markedly lower χ2 values 26/60 Superdex-200 prep grade column (GE Healthcare) in TBS. than a single-binding-site model, which was used to fit N (binding The pro-BMP7–containing fractions were pooled and further sites), Ka (association constant), and ΔH (enthalpy). purified with a heparin Hitrap column (GE Healthcare), followed by Superdex S200 SEC in TBS. The peak fractions containing Negative-Stain EM. Pro-BMP9 and pro-BMP7 were purified by pure pro-BMP7 were pooled and concentrated using a 50-kDa SEC using Superdex 200 HR preequilibrated with 20 mM Tris, cutoff centrifugal concentrator (Vivaspin) to 20 mg/mL, frozen in pH 8.0, and 150 mM NaCl. The peak fraction was loaded onto liquid nitrogen, and stored at −80 °C. glow-discharged carbon-coated grids, buffer was wicked off, and grids were immediately stained with 0.75% (wt/vol) uranyl for- BMP9 Prodomain with a Coiled-Coil. A signal sequence, Strep-tag, mate. EM was as previously described (12). Images were re- His-tag, 3C protease cleavage site, and GCN4 coiled-coil were corded with a CCD camera at 52,000× magnification with a fused to the N terminus of the human BMP9 prodomain. The defocus of −1.5 μm. The pixel size was 2.88 Å at specimen level. C-terminal Val-30 residue of GCN4 [numbering of Harbury et al. Particles were interactively picked with BOXER in EMAN (13). (8)] was mutated to cysteine. The construct in pEF1-puro was Class averages were calculated with SPIDER (14) as previously transiently transfected into HEK 293T cells (9). After culture for described (12).

1. Kabsch WF (2001) Crystallography of biological macromolecules. International Tables 8. Harbury PB, Zhang T, Kim PS, Alber T (1993) A switch between two-, three-, and four- for Crystallography, eds Rossmann MG, Arnold E (Kluwer Academic Publishers, stranded coiled coils in GCN4 leucine zipper mutants. Science 262(5138):1401–1407. Dordrecht), pp 730–734. 9. Mi LZ, et al. (2008) Functional and structural stability of the epidermal growth factor 2. Adams PD, et al. (2010) PHENIX: a comprehensive Python-based system for macro- receptor in detergent micelles and phospholipid nanodiscs. Biochemistry 47(39): molecular structure solution. Acta Crystallogr D Biol Crystallogr 66(Pt 2):213–221. 10314–10323. 3. Brown MA, et al. (2005) Crystal structure of BMP-9 and functional interactions with 10. Partridge NC, et al. (1981) Functional properties of hormonally responsive cultured pro-region and receptors. J Biol Chem 280(26):25111–25118. normal and malignant rat osteoblastic cells. Endocrinology 108(1):213–219. 4. Bailey S; Collaborative Computational Project, Number 4 (1994) The CCP4 suite: Pro- 11. Rosen V, et al. (1994) Responsiveness of clonal limb bud cell lines to bone morpho- grams for protein crystallography. Acta Crystallogr D Biol Crystallogr 50(Pt 5): genetic protein 2 reveals a sequential relationship between cartilage and bone cell 760–763. phenotypes. J Bone Miner Res 9(11):1759–1768. 5. Emsley P, Cowtan K (2004) Coot: model-building tools for molecular graphics. Acta 12. Mi LZ, et al. (2011) Simultaneous visualization of the extracellular and cytoplasmic Crystallogr D Biol Crystallogr 60(Pt 12 Pt 1):2126–2132. domains of the epidermal growth factor receptor. Nat Struct Mol Biol 18(9):984–989. 6. Karplus PA, Diederichs K (2012) Linking crystallographic model and data quality. 13. Ludtke SJ, Baldwin PR, Chiu W (1999) EMAN: semiautomated software for high- Science 336(6084):1030–1033. resolution single-particle reconstructions. J Struct Biol 128(1):82–97. 7. Gregory KE, et al. (2005) The prodomain of BMP-7 targets the BMP-7 complex to the 14. Frank J, et al. (1996) SPIDER and WEB: processing and visualization of images in 3D extracellular matrix. J Biol Chem 280(30):27970–27980. electron microscopy and related fields. J Struct Biol 116(1):190–199.

Mi et al. www.pnas.org/cgi/content/short/1501303112 2of9 Fig. S1. Pro-BMP9 EM structure. Representative electron micrograph (A) and all 50 class averages of 7,098 particles (B). Averages are ranked (left to right in each row, from top to bottom row) by numbers of particles in each class.

Mi et al. www.pnas.org/cgi/content/short/1501303112 3of9 Fig. S2. Pro-BMP7 EM structure. Representative electron micrograph (A) and all 50 class averages of 4,427 particles (B). Averages are ranked (left to right in each row, from top to bottom row) by numbers of particles in each class.

Mi et al. www.pnas.org/cgi/content/short/1501303112 4of9 A

V188 W191 F231

F234 V232 Y65 L69

F230

B α5 α5 β5

β8

C406 β8 C406

C370 C338 C370 C338 α3 α3

Fig. S3. Electron density map of pro-BMP9. (A) Representative region of experimentally phased electron density map of pro-BMP9 shown with mesh at 1σ contour level. (B) The free cysteines at the dimer interface and nearby disulfides shown with stick and S atom spheres, with experimentally phased electron density at 1σ level carved within 3 Å of cysteine residues. The structure appears to contain a mixture of disulfide and nondisulfide bonded Cys-370 residues at the GF monomer-monomer interface, consistent with presence of both GF monomers and dimers in nonreducing SDS/PAGE (Fig. S7 right, lane 2). In the refined molecular model, the disulfide is not built.

BMP9 TGF-β1 A C β7 β7 β5 β6 β8 β3 β6 β3 α3 β2 β2 α2 α2 α3 B β7 D β7 β5 β5 β6 β8 β6 β8 β3 β3

β2 α2 α2 α3 α3

Fig. S4. Comparison of BMP9 and TGF-β1 structures in different complexes. (A and B) Comparison of prodomain-associated BMP9 (A) with receptor-bound BMP9 [Protein Data Bank (PDB) ID 4FAO] (B). (C and D) Comparison of prodomain-associated TGF-β1(PDBID3RJR)(C) with receptor-bound TGF-β1 (PDB ID 3KFD) (D). The structures are represented in cartoon with one growth factor in pink and the other in gold.

Mi et al. www.pnas.org/cgi/content/short/1501303112 5of9 α1

TGFβ1po ------LSTCKTIDMELVKRKRIEAIRGQILSKL 28 TGFβ1hu ------LSTCKTIDMELVKRKRIEAIRGQILSKL 28 TGFβ2 ------SLSTCSTLDMDQFMRKRIEAIRGQILSKL 29 TGFβ3 ------SLSLSTCTTLDFGHIKKKRVEAIRGQILSKL 31 Myostatin ------NENSEQKENVEKEGLCNACTWRQNTKSSRIEAIKIQILSKL 41 GDF11 ------AEGPAAAAAAAAAAAAAGVGGERSSRPAPSVAPEPDGCPVCVWRQHSRELRLESIKSQILSKL 63 Inhibin βA ------SPTPGSEGHSAAPDCPSCALAALPKDVPNSQPEMVEAVKKHILNML 46 Inhibin βB ------SPTPPPTPAAPPPPPPPGSPGGSQDTCTSCGGFRRPEELGRVDGDFLEAVKRHILSRL 58 Inhibin βC ------TPRAGGQCPACGG--PTLELESQRELLLDLAKRSILDKL 37 Inhibin βE ------QGTGSVCPSCGG--SKLAPQAERALVLELAKQQILDGL 36 Inhibin α ------CQGLELARELVLAKVRALFLDAL 23 BMP3 ------ERPKPPFPELRKAVPGDR 18 GDF10 ------SHRAPAWSALPAAADGLQGDR 21 BMP15 ------MEHRAQMAEGGQSSIALLAEAPTLPLIEEL 30 GDF9 ------SQASGGEAQIAASAELESGAMPWSLLQHIDERDRAGLLPAL 41 GDF1 ------PVPPGPAAALLQAL 14 GDF3 ------QEYVFLQFL 9 BMP9mo ------KPLQNWEQASPGENAHSSLGLSGA-GEEGVFDLQMFLENMKVDFLRSL 47 BMP9hu ------KPLQSWGRGSAGGNAHSPLGVPGGGLPEHTFNLKMFLENVKVDFLRSL 48 BMP10 ------SPIMNLEQSPLEEDMSLFGDVFSEQDGVDFNTLLQSMKDEFLKTL 45 Nodal ------TVATALLRTR 10 BMP2 ------LVPELGRRKFAAASSGRPSSQPSDEVLSEFELRLLSMF 38 BMP4 ------GASHASLIPETGKKKVAEIQGHAGGRRSGQ-SHELLRDFEATLLQMF 46 BMP5 ------DNHVHSSFIYRRLRNHERREIQREILSIL 29 BMP7 ------DFSLDNEVHSSFIHRRLRSQERREMQREILSIL 33 BMP6 ------CCGPPPLRPPLPAAAAAAAGGQLLGDGGSPGRTEQPPPSPQSSSGFLYRRLKTQEKREMQKEILSVL 67 BMP8 ------GGGPGLRPPPGCPQRRLGARERRDVQREILAVL 33 GDF5 ------APDLGQRPQGTRPGLAKAEAKERPPLARNVFRPGGHSYGGGATNANARAKGGTGQTGGLTQPKKDEPKKLPPRPGGPEPKPGHPPQTRQATARTVTPKGQLPGGKAPPKAGSVPSSFLLKKA 122 GDF6 ------FQQASISSSSSSAELGSTKGMRSRKEGKMQRAPRDSDAGREGQEPQPRPQDE 52 GDF7 ------RDGLEAAAVLRAAGAGPVRSPGGGGGGGGGGRTLAQAA 38 GDF15 ------L 1 LEFTY1 ------LTGEQLLGSLLRQL 14 LEFTY2 ------LTEEQLLGSLLRQL 14 AMH LLGTEALRAEEPAVGTSGLIFREDLDWPPGIPQEPLCLVALGGDSNGSSSPLRVVGALSAYEQAFLGAVQRARWGPRDLATFGVCNTGDRQAALPSLRRLGAWLRDPGGQRLVVLHLEEVTWEPTPSL 128 Latency lasso α2

TGFβ1po RLASPPSQGDV------PP------GPLPEAVLALYNSTRDRVAGESVEPE------PEPEADYYAKE 78 TGFβ1hu RLASPPSQGEV------PP------GPLPEAVLALYNSTRDRVAGESAEPE------PEPEADYYAKE 78 TGFβ2 KLTSPPEDY------PE------PEEVPPEVISIYNSTRDLLQEKASRRA------AAC---ERERSDEEYYAKE 83 TGFβ3 RLTSPPE------PT------VMTHVPYQVLALYNSTRELLEEMHGERE------EGC---TQENTESEYYAKE 84 Myostatin RLETAPNISKDVIRQLL------PK------APPLRELIDQYDVQRDD-SSDGSLED------DDYHATT 92 GDF11 RLKEAPNISREVVKQLL------PK------APPLQQILDLHDFQGDALQPEDFLEE------DEYHATT 115 Inhibin βA HLKKRPDVTQPV------PK------AALLNAIRKLHVGKVGENGYVEI------EDDIGRRAEMNELMEQTSE 102 Inhibin βB QMRGRPNITHAV------PK------AAMVTALRKLHAGKVREDGRVEI------PHLDGHASPGADGQERVSE 114 Inhibin βC HLTQRPTLNRPV------SR------AALRTALQHLHGVPQGALLEDNR------EQECE 79 Inhibin βE HLTSRPRITHPP------PQ------AALTRALRRLQPGSVAPGN------GEE 72 Inhibin α ---GPPAVTREGGD------PG------VRRLPRRHALGGFTHRGSEP------EEEEDVSQ 64 BMP3 TAGGGPDSELQ------PQ------DKVSEHMLRLYDRYSTVQAARTPGSL------EGGSQPWRPRLLREGNT 74 GDF10 DLQRHPGDAAATLG------PS------AQDMVAVHMHRLYEKYSRQGARPGG------GNT 65 BMP15 -LEESPGEQPRK------PR------LLGHSLRYMLELYRRSADSHGHPRENRT------IGATM 76 GDF9 FKVLSVGRGGSPRLQ------PD------SRALHYMKKLYKTYATKEGIPKSNRS------HLYNT 89 GDF1 GLRDEPQGA------PR------LRPVPPVMWRLFRRRDPQETRSGSRRT------SPGVTLQPCHVEELGVAGNI 72 GDF3 GLDKAPS------PQ------KFQPVPYILKKIFQDREAAATTGVSRDL------CYVKELGVRGNV 58 α2 BMP9mo NLSGIPSQDKTRAE------P------PQYMIDLYNRYTTDKSST------PASNI 85 BMP9hu NLSGVPSQDKTRVE------P------PQYMIDLYNRYTSDKSTT------PASNI 86 BMP10 NLSDIPTQDSAKVD------P------PEYMLELYNKFATDRTS------MPSANI 83 Nodal GQPSSPS------PL------AYMLSLYR------DPLPRADI 35 BMP2 GLKQRPT------PS------RDAVVPPYMLDLYRRHSGQPGSPAPDHR------LERAASRANT 85 BMP4 GLRRRPQ------PS------KSAVIPDYMRDLYRLQSGEEEEEQIHST------GLEYPERPASRANT 97 BMP5 GLPHRPR------PF---SPGKQASSAPLFMLDLYNAMTN---EENPEESEYSVRASLAEETRGARKGYPASPNGYPRRIQLSRTTPLTTQSPPLASLHDTNFLNDADM 126 BMP7 GLPHRPR------PH---LQGKH-NSAPMFMLDLYNAMAV---EEGGGPGGQGFSYPYKAVFS------TQGPPLASLQDSHFLTDADM 103 BMP6 GLPHRPRPLHGLQQPQPPALRQQEEQQQQQQLPRGEPPPGRL-KSAPLFMLDLYNALSADNDEDGASEGERQQSWPHEAASSSQRRQPPPGAAHPLNRKSLLAPGSGSGGASPLTSAQDSAFLNDADM 194 BMP8 GLPGRPR------PRAPPAASRLPASAPLFMLDLYHAMAGDDDEDGAPAE------QRLGRADL 85 GDF5 REPGPPREPKEPFRPPPIT------PH------EYMLSLYRTLSDADRKGGNSSV------KLEAGLANT 174 GDF6 PRAQQPRAQEPPGRGPRVV------PH------EYMLSIYRTYSIAEKLGINASF------FQSSKSANT 104 GDF7 GAAAVPAAAVPRARAARRAAGSGFRNGSVV--PH------HFMMSLYRSLAGRAPAGAAAVS------ASGHGRADT 101 GDF15 SLAEASRASFPG------PS------ELHSEDSRFRELRKRYEDLLTRLRANQS------WEDS 47 LEFTY1 QLKEVPTLDRADMEELVI------PT------HVRAQYVALLQRSHGDRSRGKRFSQS------FREV 64 LEFTY2 QLSEVPVLDRADMEKLVI------PA------HVRAQYVVLLRRSHGDRSRGKRFSQS------FREV 64 AMH RFQEPP------PG------GAGPPELALLVLYPGPGPEVTVTRAGLP--GAQSLCPSRD------TRYLVLAVDRPAGAWRGSGL 194 β1 β2 α3 β3 β4 β5

TGFβ1po VTRVLMVESGNQIYDKFKGT------PHSLYMLFNTSELREAVPEP--VLLSRAELRLLRLK------LKVEQHVELYQKYSND------SWRYLSNRLLA---- 159 TGFβ1hu VTRVLMVETHNEIYDKFKQS------THSIYMFFNTSELREAVPEP--VLLSRAELRLLRLK------LKVEQHVELYQKYSNN------SWRYLSNRLLA---- 159 TGFβ2 VYKIDMPPFFPSENAIPPTFYR----PYFRIVRFDVSAMEKNA-----SNLVKAEFRVFRLQ------NPKARVPEQRIELYQILKSKDLT------SPTQRYIDSKVVK---- 172 TGFβ3 IHKFDMIQGLAEHNELAVCPKG----ITSKVFRFNVSSVEKNR-----TNLFRAEFRVLRVP------NPSSKRNEQRIELFQILRPDEHI------AKQRYIGGKNLP---- 172 Myostatin ETIITMPTESDFLMQVD------GKPKCCFFKFSSKIQY------NKVVKAQLWIYL------RPVETPTTVFVQILRLIKPMKDG------TRYTGIRSLKLD-M 173 GDF11 ETVISMAQETDPAVQTD------GSPLCCHFHFSPKVMF------TKVLKAQLWVYL------RPVPRPATVYLQILRLKPLTGEGTA------GGGGGGRRHIRIRSLKIE-L 204 Inhibin βA IITFAESG------TARKTLHFEISKEGSDL-----SVVERAEVWLFL------KVPKANRTRTKVTIRLFQQQKHPQGSLDTGEEAEEVGLKGERSELLLSEKVVD---- 192 Inhibin βB IISFAETDGLA------SSRVRLYFFISNEGNQN-----LFVVQASLWLYL------KLLPYVLEKGSRRKVRVKVYFQEQGH------GDRWNMVEKRVD---- 192 Inhibin βC IISFAETGLST------INQTRLDFHFSSDRTAGD----REVQQASLMFFV------QLPSNTTWTLKVRVLVLGPHN------TNLTLATQYLLE---- 153 Inhibin βE VISFATVTDSTS------AYSSLLTFHLSTPRSH------HLYHARLWLHV------LPTLPGTLCLRIFRWGPRRRRQ------GSRTLLAEHHI----- 144 Inhibin α AILFPATDASCEDKSAARGLAQEAE-EGLFRYMFRPSQHTRS------RQVTSAQLWFHT------GLDRQGTAASNSSEPLLGLLALSPG------GPVAVPMSLG---- 152 BMP3 VRSFRAAAAETL------ERKGLYIFNLTSLTKS------ENILSATLYFCIGELGNISLSCPVSGGCSHHAQRKHIQIDLSAWTLK------FSRNQSQLLGHLSVDMAKS 168 GDF10 VRSFRARLEVV------DQKAVYFFNLTSMQDS------EMILTATFHFYS-EPPRWPRALEVLCKPRAKNASGRPLPLGPPTRQHLLFR------SLSQNTATQGLLRGAMAL--A 161 BMP15 VRLVKPLTSVARPHRGTWHIQILGFPLRPNRGLYQLVRATV------VYRHHLQLTR------FNLSCHVEPWVQKNPTN------HFPSSEGDSSKP 156 GDF9 VRLFTPCTRHKQAPGDQVTGI-----LPSVELLFNLDRITTV------EHLLKSVLLYNI------NNSVSFSSAVKCVCNLMIKEPKSSSR------TLGRAPYSFTFNSQFEF---- 181 GDF1 VRHIPDRGAPTRASEPASAAGH----CPEWTVVFDLSAVEPA------ERPSRARLELRFAAA------AAAAPEGGWELSVAQAGQ------GAGADPGPVLLRQLV--P--- 159 GDF3 LRFLPDQGFFLYPKKISQASS-----CLQKLLYFNLSAIKER------EQLTLAQLGLDL------GPNSYYNLGPELELALFLVQEPHVWGQ------TTPKPGKMFVLRSVPWPQ--- 152 β1 β1’ β2 β3 α3’ β4 β5 BMP9mo VRSFSVEDAISTAATEDFP------FQKHILIFNIS-IPRH------EQITRAELRLYVSCQ------NDVDSTHGLEGSMVVYDVLEDSETW------DQATGTKTFLVSQDIRD--- 176 BMP9hu VRSFSMEDAISITATEDFP------FQKHILLFNIS-IPRH------EQITRAELRLYVSCQ------NHVDPSHDLKGSVVIYDVLDGTDAW------DSATETKTFLVSQDIQD--- 177 BMP10 IRSFKNEDLFSQPVSFNG------LRKYPLLFNVS-IPHH------EEVIMAELRLYTLVQ------RDRMIYDGVDRKITIFEVLESKGDN------EGERNMLVLVSGEIYG--- 172 Nodal IRSLQAEDVAVD------GQNWTFAFDFSFLSQQ------EDLAWAELRLQLS------SPVDLPTEGSLAIEIFHQPKPDTEQ------ASDSCLERFQMDLFTVTLSQ 121 BMP2 VRSFHHEESLEELPETSG------KTTRRFFFNLSSIPTE------EFITSAELQVFR------EQMQDALGNNSSFHHRINIYEIIKPATAN------SKFPVTRLLDTRLVN---- 175 BMP4 VRSFHHEEHLENIPGTSE------NSAFRFLFNLSSIPEN------EVISSAELRLFR------EQVDQGPDWERGF-HRINIYEVMKPPAEV------VPGHLITRLLDTRLVH---- 187 BMP5 VMSFVNLVERDKDFSHQR------RHYKEFRFDLTQIPHG------EAVTAAEFRIYK------DRSNNRFENETIKISIYQIIKEYTNR------DADLFLLDTRKAQ---- 211 BMP7 VMSFVNLVEHDKEFFHPR------YHHREFRFDLSKIPEG------EAVTAAEFRIYK------DYIRERFDNETFRISVYQVLQEHLGR------ESDLFLLDSRTLW---- 188 BMP6 VMSFVNLVEYDKEFSPRQ------RHHKEFKFNLSQIPEG------EVVTAAEFRIYK------DCVMGSFKNQTFLISIYQVLQEHQHR------DSDLFLLDTRVVW---- 279 BMP8 VMSFVNMVERDRALGHQE------PHWKEFRFDLTQIPAG------EAVTAAEFRIYK------VPSIHLLNRTLHVSMFQVVQEQSNR------ESDLFFLDLQTLR---- 169 GDF5 ITSFIDKGQDDRGPV------VRKQRYVFDISALEKD------G-LLGAELRILRKKPSDTAK-PAAPGGGRAAQLKLSSCPSGR------QPASLLDVRSVPG--- 258 GDF6 ITSFVDRGLDDLSHTP------LRRQKYLFDVSMLSDK------EELVGAELRLFR------QAPSAPWGPPAGPLHVQLFPCLSP------LLLDARTLDPQ-- 183 GDF7 ITGFTDQATQDESAA------ETGQSFLFDVSSLNDA------DEVVGAELRVLRRGSPE----SGPGSWTSPPLLLLSTCPGAA------RAPRLLYSRAAEP--- 183 GDF15 NTDLVPAPAVRILT------PEVRLGSGGHLHLRISRAALPEGLPEASRLHRALFRLS------99 LEFTY1 AGRFLALE------ASTHLLVFGMEQRLPPNS-----ELVQAVLRLFQEPVPKAALHRHGRLSPRSARARVTVEWLRVRDDGS------NRTSLIDSRLVS---- 148 LEFTY2 AGRFLASE------ASTHLLVFGMEQRLPPNS-----ELVQAVLRLFQEPVPKAALHRHGRLSPRSAQARVTVEWLRVRDDGS------NRTSLIDSRLVS---- 148 AMH ALTLQPRGEDSRLST------ARLQALLFGDDHRCFTR------MTPALLLLPR------SEPAPLPAHGQLDTVPFPPPR------PSAELEESPPS---- 268

β6 α4 β7 β8 β9 β10 α5

TGFβ1po PSDSPEWLSFDVTGVVRQWLTRRE--AIEGFRLSAHCSCDSKD------NTLHVEINGFN------SGRRGDLATIHGMNRPFLLLMATPLERAQH------240 TGFβ1hu PSDSPEWLSFDVTGVVRQWLSRGG--EIEGFRLSAHCSCDSRD------NTLQVDINGFT------TGRRGDLATIHGMNRPFLLLMATPLERAQH------240 TGFβ2 TRAEGEWLSFDVTDAVHEWLHHKD--RNLGFKISLHCPCCTFVPSNNYIIPNKSEELEARFAGIDGTSTYTSGDQKTIKSTRKKNSGKTPHLLLMLLPSYRLES------273 TGFβ3 TRGTAEWLSFDVTDTVREWLLRRE--SNLGLEISIHCPCHTFQP-NGDILENIHEVMEIKFKGVDNED---DHGRGDLGRLKKQKDHHNPHLILMMIPPHRLDN------269 Myostatin NPGTGIWQSIDVKTVLQNWLKQPE--SNLGIEIKALDENGHDLAVTFPG------PGEDGLNPFLEVKVTDT------237 GDF11 HSRSGHWQSIDFKQVLHSWFRQPQ--SNWGIEINAFDPSGTDLAVTSLG------PGAEGLHPFMELRVLEN------268 Inhibin βA -ARKSTWHVFPVSSSIQRLLDQGK--SSLDVRIACEQCQESGASLVLLGKKKKKEE------EGEGKKKGGGEGGAGADEEKEQSHRPFLMLQARQSEDHP------284 Inhibin βB -LKRSGWHTFPLTEAIQALFERGE--RRLNLDVQCDSCQELAVVPVFV------DPGEESHRPFVVVQARLGDSR------258 Inhibin βC -VDASGWHQLPLGPEAQAACSQGH--LTLELVLEGQVAQSSVIL------GGAAHRPFVAARVRVGGK------212 Inhibin βE --TNLGWHTLTL---PSSGLRGEKS-GVLKLQLDCRPLEGNSTV------TGQPRRLLDTAGHQQPFLELKIRANEPGA------211 Inhibin α -HAPPHWAVLHLATSALSLLTHPV--LVLLLRCPLCTCSA------RPEATPFLVAHTRTRPPSG------208 BMP3 HRDIMSWLSKDITQLLRKAKENEE--FLIGFNITSK------GRQLPKRRLPFPEPYILVYANDAAISEPESVVSSLQGHRNFPTGTVPKWDS H 254 GDF10 PPPRGLWQAKDISPIVKAARRDGE--LLLSAQLDSE------ERDPGVPRPSPYAPYILVYANDLAISEPNSVAVTLQRYDPFPA---GDPEP R 244 BMP15 SLMSNAWKEMDITQLVQQRFWNNKGHRILRLRFMCQQQKD------SGGLELWHGTSSLDIAFLLLYFNDTHKSIRKAK------229 GDF9 -GKKHKWIQIDVTSLLQPLVASNK--RSIHMSINFTCMKDQLEHP------SAQNGLFNMTLVSPSLILYLNDTSAQAYHSW------254 GDF1 ----ALGPPVRAELLGAAWARNASWPRSLRLALALR------PRAPAACARLAEASLLLVTLDPRLCHP------218 GDF3 --GAVHFNLLDV---AKDWNDNPR--KNFGLFLEILVKEDRDSGVNFQPEDT------CARLRCSLHASLLVVTLNPDQCH------220 β6 α4 β7 β9’ β10 α5 BMP9mo ----EGWETLEVSSAVKRWVRADSTTNKNKLEVTVQS------HRESCDTLDISVPPGSKNLPFFVVFSNDRSNGTKETRL------247 BMP9hu ----EGWETLEVSSAVKRWVRSDSTKSKNKLEVTVES------HRKGCDTLDISVPPGSRNLPFFVVFSNDHSSGTKETRL------248 BMP10 --TNSEWETFDVTDAIRRWQKSGS--STHQLEVHIESKHDEAEDAS------SGRLEIDTSAQNKHNPLLIVFSDDQSSDKERKE------247 Nodal VTFSLGSMVLEVTRPLSKWLKR-----PGALEKQMS------RVAGECWPRPPTPPATNVLLMLYSNLSQEQRQLG------186 BMP2 -QNASRWESFDVTPAVMRWTAQGH--ANHGFVVEVAHLEEKQGVSK------RHVRISRSLHQDEHSWSQIRPLLVTFGHDGKGHP------252 BMP4 -HNVTRWETFDVSPAVLRWTREKQ--PNYGLAIEVTHLHQTRTHQG------QHVRISRSLPQGSGNWAQLRPLLVTFGHDGRGHA------264 BMP5 -ALDVGWLVFDITVTSNHWVINPQ--NNLGLQLCAETGDGRSINVKSAGL------VGRQGPQSKQPFMVAFFKASEVLLRSV------285 BMP7 -ASEEGWLVFDITATSNHWVVNPR--HNLGLQLSVETLDGQSINPKLAGL------IGRHGPQNKQPFMVAFFKATEVHFRSI------262 BMP6 -ASEEGWLEFDITATSNLWVVTPQ--HNMGLQLSVVTRDGVHVHPRAAGL------VGRDGPYDKQPFMVAFFKVSEVHVRTT------353 BMP8 -AGDEGWLVLDVTAASDCWLLKRH--KDLGLRLYVETEDGHSVDPGLAGL------LGQRAPRSQQPFVVTFFRASPSPIRTP------243 GDF5 -LDGSGWEVFDIWKLFRNFKNS------AQLCLELEAWERGRAV------DLRGLGF---DRAARQVHEKALFLVFGRTKKRDLFFN------329 GDF6 GAPPAGWEVFDVWQGLRHQPWK------QLCLELRAAWGELDAGEAEARARGPQQPPPP------DLRSLGF---GRRVRPPQERALLVVFTRSQRKNLFAEMREQL------GSAE A 280 GDF7 -LVGQRWEAFDVADAMRRHRREPR--PPRAFCLLLRAVAGPVPSPL------ALRRLGFGWPGGGGSAAEERAVLVVSSRTQRKESLFREIRAQ------A 269 GDF15 PTASRSWDVTRP--LRRQLSLARPQAPALHLRLSPPPSQSDQLL------AESSSARPQLELHLRPQAARG------162 LEFTY1 -VHESGWKAFDVTEAVNFWQQLSRPRQPLLLQVSVQREHLGPLASGAHKLV------RFASQGAPAGLGEPQLELHTLDLGD------223 LEFTY2 -VHESGWKAFDVTEAVNFWQQLSRPRQPLLLQVSVQREHLGPLASGAHKLV------RFASQGAPAGLGEPQLELHTLDLRD------223 AMH ---ADPFLETL-TRLVRALRVPPARASAPRLALDPDALAGFPQGLVNLSDPAALERLLDG------EEPLLLLLRPTAATTGDPAPLHDP------TSAPW A 354

α1

TGFβ1po ------LHSSRHRR------ALDTNYCFSSTEKNC C 265 TGFβ1hu ------LQSSRHRR------ALDTNYCFSSTEKNC C 265 TGFβ2 ------QQTNRRKKR------ALDAAYCFRNVQDNC C 299 TGFβ3 ------PGQGGQRKKR------ALDTNYCFRNLEENC C 296 Myostatin ------PKRSRR------DFGLDCDEHSTESRC C 259 GDF11 ------TKRSRR------NLGLDCDEHSSESRC C 290 Inhibin βA ------HRRRRR------GLECDGKVNIC C 302 Inhibin βB ------HRIRKR------GLECDGRTNLC C 276 Inhibin βC ------HQIHRR------GIDCQGGSRMC C 230 Inhibin βE ------GRARRR------TPTCEPATPLC C 229 Inhibin α ------GERARR----STPLMSWPWSPSALRLLQRPPEEPAAHANC H 245 BMP3 IRAALSIERRKKRSTGVLLPLQNNELPGAE------YQYKKDEVWEERKPYKTLQAQAPEKSKNKKKQRKGPHRKSQTLQFDEQTLKKARR------KQWIEPRNC A 349 GDF10 AAPNNSADPRVRRAAQATGPLQDNELPGLDERPPRAHAQHFHKHQLWP--SPFRALKPRPGRKDRRKKGQEV-FMAASQVLDFDEKTMQKARR------KQWDEPRVC S 344 BMP15 ------FLPRGMEEFMERESLLRRTRQADGISAEVTA------SSSKHSGPENNQC S 274 GDF9 ------YSLHYKRRPSQGPDQERSLSA------YPVGEEAAEDGRSSHHRHRRGQETVSSELKKPLGPASFNLSEYFRQFLLPQNEC E 330 GDF1 ------LARPRR------DAEPVLGGGPGGAC R 239 GDF3 ------PSRKRR------AAIPVPKLSCKNLC H 241

BMP9mo -ELKEMIGHEQETMLVKTAKNAYQVAGESQ------EEEGLDGYTAVGPLLARRKR------STGASSHC Q 305 BMP9hu -ELREMISHEQESVLKKLSKDGSTEAGESS------HEEDTDGHVAAGSTLARRKR------SAGAGSHC Q 306 BMP10 -ELNEMISHEQLPELDNLGLDSFSSGPGEE------ALLQMRSNIIYDSTARIRR------NAKGNYC K 303 Nodal ------GSTLLWEAE------SSWRAQEGQLSWEWGKRHRR------HHLPDRSQLC R 226 BMP2 ------LHKREKR------QAKHKQRKRLKSSC K 274 BMP4 ------LTRRRRAKR------SPKHHSQRARKKNKNC R 290 BMP5 ------R-AANKRKNQNRNK------SSSHQDSSRMSS-VGDYNTSEQKQAC K 324 BMP7 ------RSTGSKQRSQNRSK------TPKNQEALRMAN-VAENSSSDQRQAC K 302 BMP6 ------RSASSRRRQQSRNR------STQSQDVARVSS-ASDYNSSELKTAC R 393 BMP8 ------RAVRPLRRRQPK-K------SNELPQANRLPGIFDDVRGSHGRQVC R 283 GDF5 ------EIKARSGQDDKTVYEYLFSQR------RKRRAPLATRQGKRPSKNLK------ARC S 374 GDF6 AGPGAGAEGSWPPPSGAPDARPWLPSPGRR------RRRTAFASRHGKRHGKKSR------LRC S 333 GDF7 RALGAALASEPLPDPGTGTASPRAVIGGRR------RRRTALAG-TRTAQGSGGGAGRGHGRRGR------SRC S 331 GDF15 ------RRRARAR------NGDHCPLGPGRCC R 183 LEFTY1 ------YGAQGD------CDPEAPMTEGTRC C 243 LEFTY2 ------YGAQGD------CDPEAPMTEGTRC C 243 AMH TALARRVAAELQAAAAELRSLPGLPPATAP------LLARLLALCP------GGPGGLGDPLRALLLLKALQGLRVEWRGRDPRGPGRAQR------SA--GATAADGPC A 445

β1 β2 α2 β3 β4 β5 β6 β7 β8

TGFβ1po VRQLYIDFRKDLGW--KWIHEPKGYHANFCLGPCPYIWSLD-----T----QYSKVLALYNQH--N-PGASAAPC--CVPQALEPLPIVYYVG----RKPKVEQLSNMIVRSCKCS------361 TGFβ1hu VRQLYIDFRKDLGW--KWIHEPKGYHANFCLGPCPYIWSLD-----T----QYSKVLALYNQH--N-PGASAAPC--CVPQALEPLPIVYYVG----RKPKVEQLSNMIVRSCKCS------361 TGFβ2 LRPLYIDFKRDLGW--KWIHEPKGYNANFCAGACPYLWSSD-----T----QHSRVLSLYNTI--N-PEASASPC--CVSQDLEPLTILYYIG----KTPKIEQLSNMIVKSCKCS------395 TGFβ3 VRPLYIDFRQDLGW--KWVHEPKGYYANFCSGPCPYLRSAD-----T----THSTVLGLYNTL--N-PEASASPC--CVPQDLEPLTILYYVG----RTPKVEQLSNMVVKSCKCS------392 Myostatin RYPLTVDFE-AFGW--DWIIAPKRYKANYCSGECEFVFLQK--YPHT------HLVHQA--N-PRGSAGPC--CTPTKMSPINMLYFNGK---EQIIYGKIPAMVVDRCGCS------352 GDF11 RYPLTVDFE-AFGW--DWIIAPKRYKANYCSGQCEYMFMQK--YPHT------HLVQQA--N-PRGSAGPC--CTPTKMSPINMLYFNDK---QQIIYGKIPGMVVDRCGCS------383 Inhibin βA KKQFFVSFK-DIGW-NDWIIAPSGYHANYCEGECPSHIAGTSGSSLS----FHSTVINHYRMRGHS-PFANLKSC--CVPTKLRPMSMLYYDDG---QNIIKKDIQNMIVEECGCS------406 Inhibin βB RQQFFIDFR-LIGW-NDWIIAPTGYYGNYCEGSCPAYLAGVPGSASS----FHTAVVNQYRMRGLN-P-GTVNSC--CIPTKLSTMSMLYFDDE---YNIVKRDVPNMIVEECGCA------379 Inhibin βC RQEFFVDFR-EIGW-HDWIIQPEGYAMNFCIGQCPLHIAGMPGIAAS----FHTAVLNLLKANTAA-GTTGGGSC--CVPTARRPLSLLYYDRD---SNIVKTDIPDMVVEACGCS------334 Inhibin βE RRDHYVDFQ-ELGW-RDWILQPEGYQLNYCSGQCPPHLAGSPGIAAS----FHSAVFSLLKAN--N-PWPASTSC--CVPTARRPLSLLYLDHN---GNVVKTDVPDMVVEACGCS------331 Inhibin α RVALNISFQ-ELGW-ERWIVYPPSFIFHYCHGGCGLHIPPN--LSLP----VPGAPPTPAQPY--S-LLPGAQPCCAALPGTMRPLHVRTTSDGG--YSFKYETVPNLLTQHCACI------348 BMP3 RRYLKVDFA-DIGW-SEWIISPKSFDAYYCSGACQFPMPKS--LKPS----NHATIQSIVRAV--GVVPGIPEPC--CVPEKMSSLSILFFDEN---KNVVLKVYPNMTVESCACR------450 GDF10 RRYLKVDFA-DIGW-NEWIISPKSFDAYYCAGACEFPMPKI--VRPS----NHATIQSIVRAV--GIIPGIPEPC--CVPDKMNSLGVLFLDEN---RNVVLKVYPNMSVDTCACR------445 BMP15 LHPFQISFR-QLGW-DHWIIAPPFYTPNYCKGTCLRVLRDG--LNSP----NHAIIQNLINQL--V-DQSVPRPS--CVPYKYVPISVLMIEAN---GSILYKEYEGMIAESCTCR------374 GDF9 LHDFRLSFS-QLKW-DNWIVAPHRYNPRYCKGDCPRAVGHR--YGSP----VHTMVQNIIYEK--L-DSSVPRPS--CVPAKYSPLSVLTIEPD---GSIAYKEYEDMIATKCTCR------430 GDF1 ARRLYVSFR-EVGW-HRWVIAPRGFLANYCQGQCALPVALS--GSGGPPALNHAVLRALMHAA--A-PGAADLPC--CVPARLSPISVLFFDNS---DNVVLRQYEDMVVDECGCR------343 GDF3 RHQLFINFR-DLGW-HKWIIAPKGFMANYCHGECPFSLTIS--LNSS----NYAFMQALMHAV--D-P-EIPQAV--CIPTKLSPISMLYQDNN---DNVILRHYEDMVVDECGCG------340 β1 β2 α2 β3 β4α3’ α3 β5 β6 β7β8’ β8 BMP9mo KTSLRVNFE-DIGW-DSWIIAPKEYDAYECKGGCFFPLADD--VTPT----KHAIVQTLVHLK--F-PTKVGKAC--CVPTKLSPISILYKDDM--GVPTLKYHYEGMSVAECGCR------406 BMP9hu KTSLRVNFE-DIGW-DSWIIAPKEYEAYECKGGCFFPLADD--VTPT----KHAIVQTLVHLK--F-PTKVGKAC--CVPTKLSPISVLYKDDM--GVPTLKYHYEGMSVAECGCR------407 BMP10 RTPLYIDFK-EIGW-DSWIIAPPGYEAYECRGVCNYPLAEH--LTPT----KHAIIQALVHLK--N-SQKASKAC--CVPTKLEPISILYLDKG---VVTYKFKYEGMAVSECGCR------403 Nodal KVKFQVDFN-LIGW-GSWIIYPKQYNAYRCEGECPNPVGEE--FHPT----NHAYIQSLLKRY--Q-PHRVPSTC--CAPVKTKPLSMLYVDN----GRVLLDHHKDMIVEECGCL------325 BMP2 RHPLYVDFS-DVGW-NDWIVAPPGYHAFYCHGECPFPLADH--LNST----NHAIVQTLVNSV----NSKIPKAC--CVPTELSAISMLYLDE---NEKVVLKNYQDMVVEGCGCR------373 BMP4 RHSLYVDFS-DVGW-NDWIVAPPGYQAFYCHGDCPFPLADH--LNST----NHAIVQTLVNSV----NSSIPKAC--CVPTELSAISMLYLDE---YDKVVLKNYQEMVVEGCGCR------389 BMP5 KHELYVSFR-DLGW-QDWIIAPEGYAAFYCDGECSFPLNAH--MNAT----NHAIVQTLVHLM--F-PDHVPKPC--CAPTKLNAISVLYFDD---SSNVILKKYRNMVVRSCGCH------424 BMP7 KHELYVSFR-DLGW-QDWIIAPEGYAAYYCEGECAFPLNSY--MNAT----NHAIVQTLVHFI--N-PETVPKPC--CAPTQLNAISVLYFDD---SSNVILKKYRNMVVRACGCH------402 BMP6 KHELYVSFQ-DLGW-QDWIIAPKGYAANYCDGECSFPLNAH--MNAT----NHAIVQTLVHLM--N-PEYVPKPC--CAPTKLNAISVLYFDD---NSNVILKKYRNMVVRACGCH------493 BMP8 RHELYVSFQ-DLGW-LDWVIAPQGYSAYYCEGECSFPLDSC--MNAT----NHAILQSLVHLM--K-PNAVPKAC--CAPTKLSATSVLYYDS---SNNVILRKHRNMVVKACGCH------383 GDF5 RKALHVNFK-DMGW-DDWIIAPLEYEAFHCEGLCEFPLRSH--LEPT----NHAVIQTLMNSM--D-PESTPPTC--CVPTRLSPISILFIDS---ANNVVYKQYEDMVVESCGCR------474 GDF6 KKPLHVNFK-ELGW-DDWIIAPLEYEAYHCEGVCDFPLRSH--LEPT----NHAIIQTLMNSM--D-PGSTPPSC--CVPTKLTPISILYIDA---GNNVVYKQYEDMVVESCGCR------433 GDF7 RKPLHVDFK-ELGW-DDWIIAPLDYEAYHCEGLCDFPLRSH--LEPT----NHAIIQTLLNSM--A-PDAAPASC--CVPARLSPISILYIDA---ANNVVYKQYEDMVVEACGCR------431 GDF15 LHTVRASLE-DLGW-ADWVLSPREVQVTMCIGACPSQFRAA-----N----MHAQIKTSLHRL--K-PDTVPAPC--CVPASYNPMVLIQKTD----TGVSLQTYDDLLAKDCHCI------279 LEFTY1 RQEMYIDLQ-GMKWAENWVLEPPGFLAYECVGTCRQPPEAL------AFKWPFLGPRQC---IASETDSLPMIVSIKEGGRTRPQVVSLPNMRVQKCSCASDGALVPRRLQ P 345 LEFTY2 RQEMYIDLQ-GMKWAKNWVLEPPGFLAYECVGTCQQPPEAL------AFNWPFLGPRQC---IASETASLPMIVSIKEGGRTRPQVVSLPNMRVQKCSCASDGALVPRRLQ P 345 AMH LRELSVDLR-A----ERSVLIPETYQANNCQGVCGWPQSDR-----NPRYGNHVVLLLKMQVR--G-AALARPPC--CVPTAY-AGKLLISLSE---ERISAHHVPNMVATECGCR------542

Fig. S5. Alignment of human TGF-β family sequences including structurally aligned pro-TGF-β1 and pro-BMP9 sequences. Black dots mark decadal residues. Gold dots mark residues with >50% surface accessible surface area buried in prodomain-GF interfaces. Potential proprotein convertase cleavage sites and tolloid cleavage sites are marked with green and red circles, respectively. BMP8 shows BMP8A sequence which is almost identical to BMP8B.

Mi et al. www.pnas.org/cgi/content/short/1501303112 6of9 Fig. S6. Poor α5-helix density in one monomer of pro-BMP9 in the 3.25 Å dataset. Simulated annealing composite omit map is shown in mesh contoured at 1σ. The crystal was cryo-protected with 30% ethanol plus mother liquor.

123 4 1234GCN4-Pro Tetramer Dimer GCN4-Pro Prodomain Prodomain

BMP9 dimer

BMP9 BMP9 reducing non-reducing

Fig. S7. Coomassie-stained SDS/PAGE of proteins used in crystallization and functional studies. Purified samples were loaded at equal molar amount: 1, BMP9; 2, pro-BMP9; 3, BMP9 prodomain; 4, GCN4-prodomain.

Mi et al. www.pnas.org/cgi/content/short/1501303112 7of9 Table S1. Data collection and refinement statistics Zinc anomalous peak Cryo with 20% ethanol/10% glycerol Cryo with 30% ethanol

Wavelength (Å) 1.2823 0.97918 1.0332 Resolution range (Å)* 50.0–3.6 (3.69–3.60) 50.0–3.3 (3.39–3.30) 50.0–3.25 (3.33–3.25) Space group P 32 2 1 P 32 2 1 P 32 2 1 Unit cell a, b, c (Å) 121.9, 121.9, 223.6 120.8, 120.8, 220.6 120.3, 120.3, 221.3 α, β, γ (°) 90, 90, 120 90, 90, 120 90, 90, 120 Unique reflections*,† 31,412 (3,191) 29,933 (2,973) 29,711 (2,147) Multiplicity* 14.5 (13.5) 23.3 (23.9) 4.03 (4.10) Completeness (%)* 99.9 (99.9) 99.6 (100.0) 99.5 (99.7) I/ σ(I)* 10.08 (0.43) 9.97 (0.67) 8.36 (0.53) Wilson B-factor 136.5 142.7 128.2 ,‡ Rmerge* 0.190 (4.82) 0.277 (5.26) 0.133 (3.394) CC1/2 (%)*,§ 0.999 (0.139) 0.994 (0.161) 0.995 (0.123) Phasing and refinement { Phasing FOM 0.75(0.29) Resolution range (Å)* 50.0–3.3 (3.36–3.30) 50.0–3.25 (3.35–3.25) ,jj Rwork/Rfree* 0.213/0.230 (0.335/0.314) 0.240/0.260 (0.387/0.384) Nonhydrogen atoms Protein/NAG/Zn/water 4,802/56/10/21 4,681/42/10/14 RMSD bonds (Å) 0.002 0.002 RMSD angles (°) 0.59 0.62 Ramachandran plot** 93/7/0 93/7/0 (% favored/allowed/outliers) Clashscore** 3.03 5.06 Average B-factor 172.4 147.7 Macromolecules 171.8 147.2 Solvent 130.9 104.3 Protein databank codes 4YCG 4YCI

*Statistics for the highest-resolution shell are shown in parentheses. † Friedel pairsP P are treated as separateP P reflections. ‡ = j − ð Þj= j ð Þj Rmerge hkl j Ihkl Ihkl J hkl j Ihkl J , where Ihkl(J) and Ihkl are the jth and mean measurement of the intensity of reflection hkl. §Pearson’s correlation coefficient between average intensities of random half-data sets of the measurements for each unique reflection. { The valueP in parentheses is the figure of merit determined without solvent flattening and noncrystallographic symmetry averaging. jj = jj j − j jj=j j Rwork hkl Fobs Fcalc Fobs , where Fobs and Fcalc are the observed and calculated structure factors, respectively. Rfree is the cross- validation R factor computed for the 5% test set of unique reflections. **Ramachandran and clash values were reported by Phenix.

Mi et al. www.pnas.org/cgi/content/short/1501303112 8of9 α5 α5 α5 α5

α1 α1 α1 α1 Cross-armed pro-BMP9 model Open-armed pro-BMP9 model

Movie S1. A feasible pathway for conformational change of pro-BMP9 between open-armed and cross-armed conformations. A cross-armed model of pro- BMP9 was constructed as described in Discussion. The movie, if looped, will display reversible conformational change from cross-armed to open-armed and open-armed to cross-armed. Structures are shown in cartoon with yellow disulfide bonds. BMP9 is shown with magenta and orange monomers. Prodomain monomers are shown in blue and green. Different shades are used for different elements. Thus for the prodomain monomer on the left, the arm domain and α2-helix are light blue, the α1-helix plus latency loop are marine, and the α5-helix is dark blue. The cross-armed pro-BMP9 model accommodates the longer α5- helix of pro-BMP9 in the same position as the shorter α5-helix of pro-TGF-β1. The open-armed pro-BMP9 model adds the α1-helix and the straight jacket elements to the pro-BMP9 crystal structure in positions where they do not clash. Using the open-armed and cross-armed models as starting and ending states, 88 morphing states in-between were calculated with adiabatic mapping using the script morph_dist from the Morph Server (1).

Movie S1

1. Krebs WG, Gerstein M (2000) The morph server: A standardized system for analyzing and visualizing macromolecular motions in a database framework. Nucleic Acids Res 28:1665–1675.

Mi et al. www.pnas.org/cgi/content/short/1501303112 9of9