Title: Identification of Amino Acid Residues in Bone Morphogenetic Protein-1 Important for Procollagen C-proteinase Activity
Abstract: Bone morphogenetic protein (BMP)-1, which belongs to the tolloid subgroup of astacin-like zinc metalloproteinases, cleaves the C-propeptides of procollagen at the physiologic site and is, therefore, a procollagen C-proteinase (PCP). Cleavage occurs between a specific alanine or glycine residue (depending on the procollagen chain) and an invariant aspartic acid residue in each of the three chains of procollagen. To learn more about how BMP-1 exhibits PCP activity we mapped the primary structure of BMP-1 onto the x-ray crystal structure of astacin and identified residues in the metalloproteinase domain of BMP-1 for subsequent site-directed mutagenesis studies. Recombinant wild-type and mutant BMP-1 were expressed in COS-7 cells and assayed for PCP activity using type I procollagen as the substrate. We showed that substitution of alanine for Glu94, which occurs in the HEXXH zinc-binding motif of BMP-1, abolishes PCP activity. Furthermore, mutation of residues Lys87 and Lys176, which are located in the S1′ pocket of the enzyme and are therefore adjacent to the P1′ residue in the substrate, reduced the proteolytic activity of BMP-1 by ∼50%. A surprising observation was that mutation of Cys66 reduced the activity to 20%, suggesting that this residue is crucial for activity. Further experiments showed that Cys66 and Cys63, which are located in the tolloid-specific sequence Cys63-Gly64-Cys65-Cys66in the active site, most likely form a disulfide bridge. Bone morphogenetic protein (BMP)-1, which belongs to the tolloid subgroup of astacin-like zinc metalloproteinases, cleaves the C-propeptides of procollagen at the physiologic site and is, therefore, a procollagen C-proteinase (PCP). Cleavage occurs between a specific alanine or glycine residue (depending on the procollagen chain) and an invariant aspartic acid residue in each of the three chains of procollagen. To learn more about how BMP-1 exhibits PCP activity we mapped the primary structure of BMP-1 onto the x-ray crystal structure of astacin and identified residues in the metalloproteinase domain of BMP-1 for subsequent site-directed mutagenesis studies. Recombinant wild-type and mutant BMP-1 were expressed in COS-7 cells and assayed for PCP activity using type I procollagen as the substrate. We showed that substitution of alanine for Glu94, which occurs in the HEXXH zinc-binding motif of BMP-1, abolishes PCP activity. Furthermore, mutation of residues Lys87 and Lys176, which are located in the S1′ pocket of the enzyme and are therefore adjacent to the P1′ residue in the substrate, reduced the proteolytic activity of BMP-1 by ∼50%. A surprising observation was that mutation of Cys66 reduced the activity to 20%, suggesting that this residue is crucial for activity. Further experiments showed that Cys66 and Cys63, which are located in the tolloid-specific sequence Cys63-Gly64-Cys65-Cys66in the active site, most likely form a disulfide bridge. bone morphogenetic protein BMP-1 containing the c-Myc peptide at its C terminus procollagen C-proteinase polyacrylamide gel electrophoresis Dulbecco's modified Eagle's medium Bone morphogenetic protein (BMP1)-1 is a zinc metalloproteinase whose metalloproteinase domain shares 39% sequence identity with that of astacin, the digestive proteinase from crayfish (1Titani K. Torff H.J. Hormel S. Kumar S. Walsh K.A. Rodl J. Neurath H. Zwilling R. Biochemistry. 1987; 26: 222-226Crossref PubMed Scopus (105) Google Scholar). BMP-1, also known as procollagen C-proteinase (PCP) (2Kessler E. Takahara K. Biniaminov L. Brusel M. Greenspan D.S. Science. 1996; 271: 360-362Crossref PubMed Scopus (462) Google Scholar), is fundamental to the synthesis of the extracellular matrix, because it cleaves the C-propeptides of type I, II, and III procollagen (3Hojima Y. van der Rest M. Prockop D.J. J. Biol. Chem. 1985; 260: 15996-16003Abstract Full Text PDF PubMed Google Scholar, 4Kessler E. Adar R. Goldberg B. Niece R. Collagen Relat. Res. 1986; 6: 249-266Crossref PubMed Scopus (48) Google Scholar) and presumably the C-propeptides of the other major fibrillar collagen precursors, type V and XI procollagen. The importance of BMP-1 in tissue assembly is exemplified in the BMP-1 knockout mouse, which dies soon after birth from failure of ventral body wall closure associated with abnormal collagen fibrillogenesis (5Suzuki N. Labosky P.A. Furuta Y. Hargett L. Dunn R. Fogo A.B. Takahara K. Peters D.M. Greenspan D.S. Hogan B.L. Development. 1996; 122: 3587-3595Crossref PubMed Google Scholar). The occurrence of abnormal fibrils in affected tissues of this mouse is consistent with the persistence of partially processed procollagen molecules. In addition to its roles in cleaving procollagen, BMP-1 cleaves other extracellular matrix macromolecules including prolysyl oxidase (6Panchenko M.V. Stetler-Stevenson W.G. Trubetskoy O.V. Gacheru S.N. Kagan H.M. J. Biol. Chem. 1996; 271: 7113-7119Abstract Full Text Full Text PDF PubMed Scopus (174) Google Scholar), probiglycan (7Scott I.C. Imamura Y. Pappano W.N. Troedel J.M. Recklies A.D. Roughley P.J. Greenspan D.S. J. Biol. Chem. 2000; 275: 30504-30511Abstract Full Text Full Text PDF PubMed Scopus (128) Google Scholar), and prolaminin-5 (8Amano S. Scott I.C. Takahara K. Koch M. Champliaud M.F. Gerecke D.R. Keene D.R. Hudson D.L. Nishiyama T. Lee S. Greenspan D.S. Burgeson R.E. J. Biol. Chem. 2000; 275: 22728-22735Abstract Full Text Full Text PDF PubMed Scopus (198) Google Scholar). BMP-1 and its larger splice variant mammalian tolloid are important in development (9Wozney J.M. Rosen V. Celeste A.J. Mitsock L.M. Whitters M.J. Kriz R.W. Hewick R.M. Wang E.A. Science. 1988; 242: 1528-1534Crossref PubMed Scopus (3441) Google Scholar), as well as embryo patterning in Drosophila (10Shimell M.J. Ferguson E.L. Childs S.R. O'Connor M.B. Cell. 1991; 67: 469-481Abstract Full Text PDF PubMed Scopus (273) Google Scholar), Xenopus(11Maeno M. Xue Y. Wood T.I. Ong R.C. Kung H.F. Gene. 1993; 134: 257-261Crossref PubMed Scopus (38) Google Scholar), and sea urchin (12Reynolds S.D. Angerer L.M. Palis J. Nasir A. Angerer R.C. Development. 1992; 114: 769-786PubMed Google Scholar, 13Lepage T. Ghiglione C. Gache C. Development. 1992; 114: 147-163Crossref PubMed Google Scholar). For example, in vertebrates, BMP-1 cleaves chordin (an antagonist of BMP-4) during normal dorsal-ventral patterning (for review see Ref. 14Mullins M.C. Trends Genet. 1998; 14: 127-129Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar). Therefore, understanding the catalytic mechanism of this enzyme is relevant to studies of animal development and tissue organization. Furthermore, because collagen is expressed in soft tissues such as liver and lung during progressive fibro-proliferative diseases (e.g. cirrhosis, pulmonary fibrosis, and scleroderma), which are characterized by excessive deposition of connective tissue (for review see Ref. 15Trojanowska M. LeRoy E.C. Eckes B. Krieg T. J. Mol. Med. 1998; 76: 266-274Crossref PubMed Scopus (154) Google Scholar), structure-function studies of BMP-1 are a step toward the rational design of anti-fibrotic drugs. In all the known substrates of BMP-1 the scissile bond resides between a small side-chained residue and an aspartic acid. For example, the P1 residue in chordin is serine or alanine (16Scott I.C. Blitz I.L. Pappano W.N. Imamura Y. Clark T.G. Steiglitz B.M. Thomas C.L. Maas S.A. Takahara K. Cho K.W. Greenspan D.S. Dev. Biol. 1999; 213: 283-300Crossref PubMed Scopus (236) Google Scholar), in procollagen it is alanine or glycine (2Kessler E. Takahara K. Biniaminov L. Brusel M. Greenspan D.S. Science. 1996; 271: 360-362Crossref PubMed Scopus (462) Google Scholar), and in prolysyl oxidase it is glycine (6Panchenko M.V. Stetler-Stevenson W.G. Trubetskoy O.V. Gacheru S.N. Kagan H.M. J. Biol. Chem. 1996; 271: 7113-7119Abstract Full Text Full Text PDF PubMed Scopus (174) Google Scholar). Furthermore, the P1′ residue in these substrates is always aspartic acid. Presumably, the shape and size of the active site of BMP-1 accommodates only small residue side chains in the P1 position and a large acidic side chain in the P1′ position. An understanding of the mechanism of cleavage by BMP-1 requires a detailed knowledge of how the P1 and P1′ residues dock into the active site. In the absence of structural information of the metalloproteinase domain of BMP-1 we reasoned that a valid approach was to examine the function of individual residues by site-directed mutagenesis. This approach is made possible by the fact that the metalloproteinase domain of BMP-1 shares high sequence homology with astacin, whose x-ray crystal structure is known (17Bond J.S. Beynon R.J. Protein Sci. 1995; 4: 1247-1261Crossref PubMed Scopus (360) Google Scholar). The metalloproteinase domain of astacin is kidney-shaped and has two domains (at the N and C termini) that are separated by an active site cleft (18Gomis-Ruth F.X. Stöcker W. Huber R. Zwilling R. Bode W. J. Mol. Biol. 1993; 229: 945-968Crossref PubMed Scopus (135) Google Scholar). A zinc ion sits at the bottom of the cleft and is coordinated in a trigonal-bipyramidal geometry by three histidine residues, a tyrosine residue (Tyr149), and a water molecule, which is also bound to the carboxylic acid side chain of Glu93. It has been shown for astacin that Glu93and Tyr149 are essential for catalytic activity (19Yiallouros I. Berkhoff E.G. Stöcker W. FEBS Lett. 2000; 484: 224-228Crossref PubMed Scopus (62) Google Scholar). Although the metalloproteinase domains of astacin and BMP-1 are homologous and presumably have a similar tertiary structure, it is not obvious why BMP-1 cleaves scissile bonds between a small side chained residue and an aspartic acid. In this study we used information from the three-dimension structure of astacin, the x-ray crystal structure of astacin in complex with a transition state inhibitor (20Grams F. Dive V. Yiotakis A. Yiallouros I. Vassiliou S. Zwilling R. Bode W. Stöcker W. Nat. Struct. Biol. 1996; 3: 671-675Crossref PubMed Scopus (150) Google Scholar), and the primary structures of different members of the astacin family to identify residues in BMP-1 that might account for the procollagen C-proteinase activity of BMP-1. We subsequently used site-directed mutagenesis and assay of recombinant enzyme to identify specific lysyl and cysteine residues in the active site of BMP-1 that are important for PCP activity. Polymerase chain reaction products were purified with a Qiaquick kit (Qiagen). Plasmids were extracted with a Qiaprep spin miniprep kit (Qiagen). Kaleidoscope protein molecular weight standards (high range) were from Bio-Rad. A full-length cDNA for BMP-1 (accession number P13497) was cloned from a human placental cDNA library. The cDNA was inserted at theKpnI/XbaI sites of the expression vector pcDNA3 (Invitrogen), thereby placing it under the transcriptional control of a cytomegalovirus promoter. The c-Myc amino acid sequence (EQKLISEEDL), which is recognized by the 9E10 monoclonal antibody (21Evan G.I. Lewis G.K. Ramsay G. Bishop J.M. Mol. Cell. Biol. 1985; 5: 3610-3616Crossref PubMed Scopus (2269) Google Scholar), was introduced immediately 5′ of the stop codon of the BMP-1 clone. The polyclonal 1210 antibody was raised, by Sigma, in rabbits using conventional procedures. In brief, a peptide corresponding to the 10 N-terminal residues of the mature BMP-1 protein (after removal of the prodomain) was conjugated to keyhole limpet hemocyanin and subsequently used to immunize two separate rabbits. Non-immune serum was collected prior to injections. Mutants were generated by site-directed mutagenesis of the XcmI/BlpI fragment of the BMP-1 cDNA clone. The XcmI site is located at nucleotide 383. The BlpI site is located at nucleotide 913. The XcmI/BlpI fragment corresponds to residue numbers Ser124 to Gly308in the BMP-1 protein. The mutations were made by standard procedures using the strand overlap polymerase chain reaction (22Horton R.M. White B.A. PCR Protocols: Current Methods and Applications. Humana Press Inc., Totowa, NJ1993: 251-261Google Scholar) and usingPwo polymerase (Roche Molecular Biochemicals), a forward primer containing an XcmI site (underlined) (5′-GTCCCGACCAGAGCGTGTGTGGCC-3′; XcmForward primer), a reverse primer containing a BlpI site (underlined) (5′-CCCTTGCTGAGCCGTGTCCTT-3′;BlpReverse primer), and oligonucleotides in both orientations containing the desired modification (in bold): E94A, 5′-ATTGTGGTCCACGCTCTGGGCCACG-3′; K87A, 5′-AACTGTGACGCGTTCGGCATT-3′; K176A, 5′-GAACGGGGTGGCACCTCCCATTG-3′; C63G, 5′-CCTATCGACCTGGCGGGTGCTGCTC-3′; C65A, 5′-GACCTTGCGGGGCCTGCTCCTACGT-3′; C66G, 5′-TGCGGGTGCGGCTCCTACGTG-3′; C85A, 5′-GGCAAGAACGCTGACA AGTT-3′. Pwo DNA polymerase was used to minimize base misincorporation during the polymerase chain reactions. Briefly, a DNA fragment was amplified using the XcmForward primer and the antisense mutant primer, and an overlapping fragment was amplified using the sense mutant primer and the downstream BlpReverse primer. Both fragments were gel-purified (Macherey-Nagel), mixed, and reamplified with thePwo enzyme with the XcmForward andBlpReverse primers. The product was digested using appropriate restriction enzymes, gel-purified, and introduced in place of the corresponding wild-type fragment in BMP-1myc. DNA sequencing was used to verify the mutations and to ensure that the cDNA clones were error-free. COS-7 cells (European Collection of Cell Cultures number 87021302) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Life Technologies, Inc.) (complete DMEM) in a 37 °C incubator with 5% CO2. The recombinant wild-type and mutant BMP-1 proteins were expressed in transiently transfected COS-7 cells. Transfections were made with Fugene 6 reagent (Roche Molecular Biochemicals) and 10 µg of plasmid/100-mm tissue culture plate. Cells were grown to ∼80% confluency by overnight incubation in complete DMEM. After three rinses with phosphate-buffered saline (Life Technologies, Inc.), cells were transfected in serum-free DMEM/Iscove's/Ham's F12 (1:1:1) (Life Technologies, Inc.) following the manufacturer's instructions and returned to the incubator. Media were removed after 48 h and replaced by DMEM without serum and conditioned for another 24 h. The tissue culture media were collected at 72 h post-transfection and cleared of cell debris by centrifugation at 1600 ×g for 10 min and concentrated to 100 µl using Centriprep-30 and Microcon-10 concentrators (Amicon, Inc.). The samples were used immediately or stored at −80 °C. Recombinant BMP-1 was assayed for procollagen C-proteinase activity using human14C-labeled type I procollagen substrate and analysis of the cleavage products on SDS gels as described (3Hojima Y. van der Rest M. Prockop D.J. J. Biol. Chem. 1985; 260: 15996-16003Abstract Full Text PDF PubMed Google Scholar). A minor change was that laser densitometry of film exposed to dried gels was replaced by image plate quantitation of the 14C-labeled proteins. In brief, 14C-labeled type I procollagen was obtained from the medium of human skin fibroblasts that had been cultured in DMEM supplemented with ascorbic acid (25 µg/ml), l-glutamine, and a mixture of uniformly labeled 14C-l-amino acids (1 µCi/ml). The procollagen was purified by ammonium sulfate precipitation (176 mg/ml) and ion exchange chromatography and was then concentrated by ultrafiltration on Amicon YM100 membranes. The procollagen content was determined by the sensitive hydroxyproline assay method of Terlink (23Terlink T. Tavenier P. Netelenbos J.C. Clin. Chim. Acta. 1989; 183: 309-316Crossref PubMed Scopus (70) Google Scholar). BMP-1 was purified from the culture medium of MG63 cells by liquid chromatography on sequential columns of matrix Green A, concanavalin A-Sepharose, heparin-Sepharose, and Superose S-300 and S-200 as described (24Ovens A. Joule J.A. Kadler K.E. J. Pept. Sci. 2000; 6: 489-495Crossref PubMed Scopus (21) Google Scholar). Cells were rinsed once with phosphate-buffered saline and incubated on ice for 15 min with 500 µl of RIPA buffer (150 mm NaCl, 1% deoxycholate, 0.1% SDS, 10 mm Tris, pH 7.6) containing 10 mm EDTA and protease inhibitor mixture (Roche Molecular Biochemicals). Cells in RIPA buffer were scraped on ice and sonicated. Lysates were subjected to a 5-min centrifugation at 14,000 × g at 4 °C. Supernatants were retained and stored at −80 °C for further analysis. Proteins were concentrated on Centriprep YM30 membranes (Amicon) and separated by discontinuous SDS-PAGE. BMP-1myc was examined by Western blot analysis in which the primary antibody was either the mouse monoclonal anti-c-myc peptide antibody, 9E10, or the rabbit 1210 neoepitope polyclonal antibody. Secondary antibodies were either horseradish peroxidase conjugated to anti-mouse or anti-rabbit IgG and were detected by the enhanced chemiluminescence method (Supersignal West Dura Extended Duration; Pierce). The levels of BMP-1myc were quantitated by laser densitometry of enhanced chemiluminescence fluorograms exposed to pre-flashed film. In preliminary studies we performed a multiple sequence alignment of 31 members of the astacin family of metalloproteinases using MultAlin (25Corpet F. Nucleic Acids Res. 1988; 16: 10881-10890Crossref PubMed Scopus (4429) Google Scholar) (data not shown). This included astacin, BMP-1, mammalian tolloid, mammalian tolloid-like-1, mammalian tolloid-like-2, and meprins. Fig. 1 contains a trimmed down version of this alignment in which only the sequences of astacin and human BMP-1 are shown. The high degree of sequence homology between the metalloproteinase domains of astacin and BMP-1, and the fact that the metalloproteinase domains are similar in size, suggests that the structure of the metalloproteinase domain of BMP-1 is similar to that of astacin (26Stöcker W. Gomis-Ruth F.X. Bode W. Zwilling R. Eur. J. Biochem. 1993; 214: 215-231Crossref PubMed Scopus (104) Google Scholar). With this information, we identified the following residues for site-directed mutagenesis studies. In astacin the active site zinc is pentacoordinated by three histidines, a unique tyrosine residue (in the Met turn), and a water molecule. The zinc-bound water is thought to be polarized for nucleophilic attack of the scissile bond by the glutamic acid residue in the consensus sequence HEXXH. In BMP-1, this glutamic acid is at amino acid position 94. To test the hypothesis that Glu94 is important for the PCP activity of BMP-1, this residue was replaced by alanine. It has been suggested that Lys87 and the Lys176-containing S1′ loop define the S1′ pocket of BMP-1 (26Stöcker W. Gomis-Ruth F.X. Bode W. Zwilling R. Eur. J. Biochem. 1993; 214: 215-231Crossref PubMed Scopus (104) Google Scholar). Furthermore, the Lys176-containing S1′ loop of BMP-1 is absent from astacin and non-tolloid members of the astacin family. It was an attractive idea that the positively charged side chains of these lysyl residues might bind to the acidic side chain of aspartic acid in the P1′ position of the procollagen chains. Of particular interest was the anti-parallel β-strand IV (edge strand), which forms the upper edge of the active site cleft. Noteworthy, when astacin binds its substrate, the residues in the substrate N-terminal to the scissile-bound (i.e. the non-primed side) align with the β-strand IV of the enzyme, forming two or three hydrogen bonds (27Bode W. Reinemer P. Huber R. Kleine T. Schnierer S. Tschesche H. EMBO J. 1994; 13: 1263-1269Crossref PubMed Scopus (326) Google Scholar). It has been pointed out that the β-strand IV in BMP-1 contains three cysteine residues. Two of these, Cys63 and Cys66, might form a disulfide bond with each other (26Stöcker W. Gomis-Ruth F.X. Bode W. Zwilling R. Eur. J. Biochem. 1993; 214: 215-231Crossref PubMed Scopus (104) Google Scholar). These cysteine residues straddle Cys65, which is conserved across the astacin family. In astacin, the equivalent cysteine (Cys64) forms a vital disulfide bond with a non-active site cysteine (Cys84). It has been suggested that the equivalent disulfide bond in BMP-1 is formed between Cys65 and Cys85 (26Stöcker W. Gomis-Ruth F.X. Bode W. Zwilling R. Eur. J. Biochem. 1993; 214: 215-231Crossref PubMed Scopus (104) Google Scholar). The strategic position of Cys63, Cys65, and Cys66 in the active site β-strand IV of BMP-1 and the proposed disulfide bond between Cys65and Cys85 made these residues good targets for site-directed mutagenesis studies of the structure and function of BMP-1. The pcDNA3 vector containing the cDNA for BMP-1myc was transfected into COS-7 cells, and the conditioned culture medium and the cell lysate were analyzed by Western blotting using the 9E10 antibody (which detects the c-myc tag at the C terminus of the molecule) and the 1210 neoepitope antibody (which was raised to a peptide corresponding to the 10 residues at the N terminus of mature BMP-1). Fig. 2 shows that the 9E10 antibody detected the latent BMP-1myc in cell lysates and the mature BMP-1myc in the culture medium. The 1210 neoepitope anti-peptide antibody detected only the mature enzyme, which occurred in the culture medium and not in the cell lysate. Furthermore, the culture medium from COS-7 cells transfected with the empty vector contained no immunoreactive proteins, which shows that the endogenous levels of BMP-1 were very low. Furthermore, Western blots using the preimmune rabbit serum were blank (data not shown). Previous studies had shown that the addition of the FLAG peptide at the C terminus of BMP-1 has negligible effect on the ability of the molecule to cleave procollagen (16Scott I.C. Blitz I.L. Pappano W.N. Imamura Y. Clark T.G. Steiglitz B.M. Thomas C.L. Maas S.A. Takahara K. Cho K.W. Greenspan D.S. Dev. Biol. 1999; 213: 283-300Crossref PubMed Scopus (236) Google Scholar). Therefore, we anticipated that the c-myc was unlikely to influence the PCP activity of BMP-1. Nevertheless, we assayed c-myc-tagged BMP-1 for PCP activity and compared it to that of native BMP-1. Figs.3 and 4show that type I procollagen is readily cleaved by BMP-1myc. We wanted to know whether BMP-1myc cleaved procollagen at the physiological site. We used a neoepitope antibody that recognized the N-terminal 10 residues of the α1(I) chain C-propeptide. Western blot analysis showed that the antibody recognized the C-propeptides after cleavage of procollagen with recombinant BMP-1 and BMP-1myc (data not shown).Figure 4Cleavage of type I procollagen by recombinant BMP-1myc (analyzed under reducing conditions). Human type I procollagen (0.3 µg) was incubated at 37 °C for 24 h with and without EDTA. pcDNA3, without enzyme;BMP-1, recombinant untagged BMP-1; BMP-1myc, concentrated medium (4 µl) from COS-7 cells transfected with BMP-1myc. Products of the reactions were separated on a 7% SDS gel under reducing conditions. Arrows indicate the position of the proα1(I) and proα2(I) chains of procollagen and the pNα1(I) and pNα2(I) chains of a normal intermediate in the conversion of procollagen to collagen containing the N-propeptides but not the C-propeptides. The proα2(I) and pNα1(I) chains migrated as a single band.View Large Image Figure ViewerDownload Hi-res image Download (PPT) COS-7 cells were transfected with pcDNA3 containing cDNAs encoding for BMP-1myc and Glu94, Lys87, Lys176, and K87A/K176A mutants. The culture media of the cells were examined by Western blot analysis using the 9E10 monoclonal antibody. Fig. 5shows that BMP-1myc and the mutants were secreted as mature enzymes. In some experiments small levels of latent BMP-1 were detectable in the culture medium, which showed that cleavage of the prodomain of BMP-1 was not a prerequisite for secretion. Medium harvested from COS-7 cells transfected with the empty vector contained no immunoreactive proteins. Constructs encoding BMP-1myc and the mutants C63G, C65A, C66G, and C85A were transfected into COS-7 cells, and the conditioned culture media were subjected to Western blot analysis using the 9E10 antibody. A typical result is shown in Fig. 6. The results showed that BMP-1myc was secreted into the culture medium mostly as the mature enzyme. The Cys63 and Cys65 mutants were efficiently secreted into the culture medium as both mature and latent enzymes. C66G was also secreted efficiently into the culture medium. In contrast to the other mutants studied, the C85A mutant was poorly secreted and could only be detected in the culture medium when the gel was overloaded. Cell lysates contained only the latent form of BMP-1. The low level of secretion of the C85A mutant suggested to us that Cys85 is crucial for the stability of the protein, presumably because it participates in the formation of an important structural disulfide bond with Cys65 in the β-strand IV. This was in agreement with the suggestion that Cys85 bonds with Cys65. However, the fact that the C65A mutant was well secreted raised the possibility that another cysteine residue could substitute for Cys65 in bonding to Cys85. The only candidates were Cys63 and Cys66. We examined the migration of the C63G and C65A mutants in SDS gels under non-reducing conditions. Fig. 7 shows that wild-type BMP-1 migrated mostly as a single band (M-SS). In comparison, the Cys63 and Cys65 mutants migrated as a slower band (M-SH). This indicated that the Cys63 and Cys65 mutant molecules have a more open conformation than the wild-type molecule. Culture media from COS-7 cells transfected with cDNAs encoding BMP-1myc and the mutants described above was concentrated on YM30 membranes, and the levels of BMP-1myc were determined by Western blot analysis using the 9E10 antibody. The concentration of BMP-1 molecules was normalized. The preparations were then examined in assays of procollagen C-proteinase. The results are summarized in Fig.8. All the mutants exhibited reduced PCP activity. Most notably, the E94A mutant exhibited no PCP activity. This showed that Glu94 is essential for PCP activity of BMP-1. Furthermore, this result showed that endogenous levels of BMP-1 in the culture medium of COS-7 cells were below the detection limit of the assay and that nonspecific proteinases that might occur in the preparations did not interfere with the PCP assays. The Lys87 and Lys176 mutants had reduced PCP activity, and the PCP activity of the double lysine mutant was 48% of control values. The results also showed that the Cys63 and Cys65 mutants exhibited a marked reduction in PCP activity (38 and 33%, respectively). A surprising result was that the C66G mutant exhibited very low levels of PCP activity (∼20% of control). In this study we have used site-directed mutagenesis to identify residues in the metalloproteinase domain of BMP-1 that are important for its PCP activity. It was first necessary to establish and validate a system capable of expressing recombinant BMP-1 that exhibited PCP activity. We found that COS-7 cells efficiently expressed a cDNA clone encoding the full-length latent BMP-1 (driven by a cytomegalovirus promoter) to produce active procollagen C-proteinase. Furthermore, if these cells synthesized endogenous BMP-1, it was undetectable in assays of procollagen C-proteinase and in Western blotting analyses using a neoepitope antibody that recognizes the N terminus of mature BMP-1. The amino acid sequence within the N-terminal 10 residues of mature BMP-1 that is recognized by the 1210 antibody has not been characterized. However, because the 1210 antibody does not recognize latent BMP-1 and only recognizes the mature BMP-1, it is likely that COS-7 cells convert latent BMP-1 to mature BMP-1 at, or close to, the physiologic site. The study also showed that the introduction of the c-myc tag at the C terminus of BMP-1 did not prohibit subsequent assays of PCP activity. The results show that mature BMP-1 (lacking the prodomain) was not observed in lysates of transfected COS-7 cells. This is consistent with processing of latent BMP-1 by COS-7 cells occurring after, or during, secretion from the cells. Latent BMP-1 contains the dibasic RSRR peptide at the junction of the prodomain and the mature protein. Furin and furin-like subtilisin-related proprotein convertases are known to process a variety of protein precursors at dibasic sequences including growth factors, receptors, and matrix metalloproteinases (28Steiner D.F. Curr. Opin. Chem. Biol. 1998; 2: 31-39Crossref PubMed Scopus (584) Google Scholar, 29Zhou A. Webb G. Zhu X. Steiner D.F. J. Biol. Chem. 1999; 274: 20745-20748Abstract Full Text Full Text PDF PubMed Scopus (417) Google Scholar) and most probably BMP-1/mammalian tolloid-like proteases (17Bond J.S. Beynon R.J. Protein Sci. 1995; 4: 1247-1261Crossref PubMed Scopus (360) Google Scholar). In some experiments we noticed that latent BMP-1 was detected in the culture medium. This might have occurred when the rate of synthesis of latent BMP-1 exceeded the capacity of the BMP-1 convertase to convert all the latent BMP-1 to mature BMP-1. Nevertheless, these experiments demonstrated that, in COS-7 cells, removal of the prodomain is not a prerequisite for secretion of the protein. One of our targets in site-directed mutagenesis experiments was Glu94, which occurs in the zinc-binding consensus motif HEXXHXXGXXH. We showed that substitution of alanine for Glu94 eliminated the PCP activity of BMP-1 and confirmed that this glutamic acid residue is essential for catalytic activity of this enzyme. The observation that the culture medium of cells transfected with the E94A mutant did not contain detectable PCP activity also validated the use of COS-7 cells as a good model cell system in which to express recombinant BMP-1. By mapping the primary structure of BMP-1 onto the secondary structure of astacin, Stöcker et al. (30Stöcker W. Ng M. Auld D.S. Biochemistry. 1990; 29: 10418-10425Crossref PubMed Scopus (61) Google Scholar) showed that one side of the S1′ pocket of BMP-1 is formed from an S1′ loop containing residues 170–184. This loop contains Lys176. When this residue was mutated to alanine the PCP activity of BMP-1 was reduced by 25%, possibly because the positive charge of its side chain stabilizes the carboxylic side chain of the aspartic acid in the P1′ position of procollagen. The S1′ pocket of BMP-1 is supposedly bigger than the corresponding pocket in astacin, which explains how BMP-1 can accommodate the large polar side chain of aspartic acid in procollagen. Of special interest, the S1′ pocket of meprin A is thought to be larger than that in astacin (17Bond J.S. Beynon R.J. Protein Sci. 1995; 4: 1247-1261Crossref PubMed Scopus (360) Google Scholar) and accounts for how bulky and charged side chains in the P1′ position of substrates can be accommodated in the S1′ pocket (31Wolz R.L. Bond J.S. Methods Enzymol. 1995; 248: 325-345Crossref PubMed Scopus (57) Google Scholar). We noted that BMP-1 has a lysine residue at position 87 that occurs in all tolloids but is a tyrosine residue in astacin (Tyr86). Interestingly, neither astacin nor meprin A, which also exhibits a tyrosine in this position (32Wolz R.L. Harris R.B. Bond J.S. Biochemistry. 1991; 30: 8488-8493Crossref PubMed Scopus (38) Google Scholar), can accept an acidic side chain in P1′. In contrast, meprin B has an arginine in place of Lys87 (an equivalent basic amino acid) and prefers an acidic side chain in P1′ (33Chestukhin A. Litovchick L. Muradov K. Batkin M. Shaltiel S. J. Biol. Chem. 1997; 272: 3153-3160Abstract Full Text Full Text PDF PubMed Scopus (45) Google Scholar). Taken together, these observations led us to hypothesize that Lys87 might be important for PCP activity. To test this hypothesis we mutated Lys87 to alanine and assayed the mutant enzyme. The results showed that the PCP activity of the Lys87 mutant was 64% of control and was, therefore, less effective than the Lys176 mutant. When both Lys87 and Lys176 were mutated, the PCP activity of BMP-1 was reduced to ∼50%. These results indicated that Lys87 and Lys176 act together to stabilize enzyme-substrate interactions involving the primed side of the substrate. As these two residues contribute to only 50% of the PCP activity of BMP-1, it was of interest to determine which other residues might be important for PCP activity. The most conspicuous difference between BMP-1 and other non-PCP astacins was the occurrence of two unique cysteine residues at the active site of the enzyme. The x-ray crystal structure of astacin shows a disulfide bond between a cysteine in the upper edge of the active site cleft and a cysteine buried in the body of the metalloproteinase domain (34Bode W. Gomis-Ruth F.X. Huber R. Zwilling R. Stöcker W. Nature. 1992; 358: 164-167Crossref PubMed Scopus (306) Google Scholar). These cysteine residues are Cys64 and Cys84, respectively, and are invariant in all astacin family members (26Stöcker W. Gomis-Ruth F.X. Bode W. Zwilling R. Eur. J. Biochem. 1993; 214: 215-231Crossref PubMed Scopus (104) Google Scholar). As this disulfide bridge clamps the β-strand IV to the loop between strand V and helix B, it is likely to be crucial for stabilizing the structure of the metalloproteinase domain. The equivalent cysteine residues in BMP-1 are Cys65 and Cys85. The importance of Cys85 in stabilizing the structure of the metalloproteinase domain of BMP-1 was confirmed when C85A was found to be poorly secreted. Surprisingly, however, the Cys65 mutant was secreted efficiently. Also, BMP-1 molecules in which the other two cysteine residues on the active site edge strand,i.e. Cys63 and Cys66, had been mutated were also well secreted. This raised the possibility that Cys85 could form a disulfide bond with a cysteine other than Cys65. The most likely candidate was Cys63, because based on structure comparison with astacin in complex with an inhibitor, its side chain is oriented toward the Cys85 residue (20Grams F. Dive V. Yiotakis A. Yiallouros I. Vassiliou S. Zwilling R. Bode W. Stöcker W. Nat. Struct. Biol. 1996; 3: 671-675Crossref PubMed Scopus (150) Google Scholar). This suggestion is supported by the observation that the C63G and C65A mutants were equivalent to each other both in PCP assays and migration position in non-reducing SDS gels. In Fig. 9 we propose a model for the PCP activity of BMP-1 that explains the results obtained in this study and takes into account published studies on the structure of astacin. We propose that the disulfide-bonded side chain of Cys66 is fundamental to the PCP activity of BMP-1, because it backs onto the P1 residue in procollagen. This is a direct analogy with what happens in astacin when Trp65 (the equivalent of Cys66 in BMP-1) backs onto the P1 position of the astacin substrate. The absence of the Cys63-Cys66 disulfide bond in 1) the Cys66 mutant, 2) the Cys63 mutant (because Cys66 has a free thiol group in this mutant), and 3) the Cys65 mutant (because Cys85 has bonded with Cys63 thus leaving a free thiol group on Cys66) alters the chemical structure of the S1 site and decreases PCP activity. We thank Walter Stöcker (Münster, Germany) for helpful discussions.