Bone Morphogenetic Protein-1 Processes Probiglycan*

Bone morphogenetic protein-1 (BMP-1) is a metalloprotease that plays important roles in regulating the deposition of fibrous extracellular matrix in vertebrates, including provision of the procollagen C-proteinase activity that processes the major fibrillar collagens I–III. Biglycan, a small leucine-rich proteoglycan, is a nonfibrillar extracellular matrix component with functions that include the positive regulation of bone formation. Biglycan is synthesized as a precursor with an NH2-terminal propeptide that is cleaved to yield the mature form found in vertebrate tissues. Here, we show that BMP-1 cleaves probiglycan at a single site, removing the propeptide and producing a biglycan molecule with an NH2 terminus identical to that of the mature form found in tissues. BMP-1-related proteases mammalian Tolloid and mammalian Tolloid-like 1 (mTLL-1) are shown to have low but detectable levels of probiglycan-cleaving activity. Comparison shows that wild type mouse embryo fibroblasts (MEFs) produce only fully processed biglycan, whereas MEFs derived from embryos homozygous null for theBmp1 gene, which encodes both BMP-1 and mammalian Tolloid, produce predominantly unprocessed probiglycan, and MEFs homozygous null for both the Bmp1 gene and the mTLL-1 gene Tll1produce only unprocessed probiglycan. Thus, all detectable probiglycan-processing activity in MEFs is accounted for by the products of these two genes.

Bone morphogenetic protein-1 (BMP-1) 1 is the prototype of a family of metalloproteases involved in morphogenesis in a broad range of species (1). BMP-1 and mammalian Tolloid (mTLD), a somewhat larger protein encoded by alternatively spliced RNAs of the Bmp1 gene (2), affect morphogenesis, at least in part, by providing the procollagen C-proteinase (PCP) activity that cleaves the C-propeptides of procollagens I-III to yield the major fibrous components of extracellular matrix (ECM) (3)(4)(5)(6). These two proteases also contribute to the net deposition of insoluble ECM through proteolytic activation of lysyl oxidase (7), 2 an enzyme necessary to the formation of covalent cross-links in collagen and elastic fibers. Two additional, genetically distinct, BMP-1/mTLD-related mammalian proteases have been described and designated mammalian Tolloid-like 1 (mTLL-1) and mTLL-2, due to domain structures identical to that of mTLD (5,8). Although mTLL-1 has some PCP activity in in vitro assays (5), the significance of this activity is unclear, as procollagen processing appears unaffected in mTLL-1-deficient mice (9). PCP activity was not detected in in vitro assays of mTLL-2 (5).
Recently, BMP-1/mTLD-related proteases Xenopus Xolloid (10) and zebrafish Tolloid (11) were shown to exert ventralizing effects during vertebrate embryogenesis by cleaving the secreted protein Chordin, which forms latent complexes with ventralizing TGF-␤-like molecules, such as BMPs 2 and 4 (12). BMP-1 and mTLL-1 are also capable of affecting dorsal-ventral patterning through cleavage of Chordin, whereas mTLD and mTLL-2 do not have detectable levels of this activity (5). In later development, BMP-1 and Chordin have been found to have similar expression patterns in pre-and postnatal endochondral bone formation (5), whereas BMP-1 copurifies with TGF-␤-like BMPs from osteogenic extracts of bone (13). Thus, BMP-1 and related proteases may serve to coordinate the deposition of ECM with the activation of certain TGF-␤-like BMPs in early development and later in the development of bone and other tissues. Determining the extent of involvement of BMP-1 and related proteases in morphogenetic events will require identifying the range of substrates processed by each in vivo.
Vertebrate ECM largely comprises insoluble collagenous fibers and a hydrated interfibrillar network, predominantly composed of proteoglycans. Both large, aggregating and small, nonaggregating proteoglycans are thought to affect development and homeostasis through interactions with macromolecular structures of the ECM, growth factors, and cell surfaces (14). Biglycan and decorin are small, nonaggregating proteoglycans that contain either chondroitin sulfate or dermatan sulfate side chains and belong to the family of small leucinerich proteoglycans (SLRPs) of the ECM. There are at least nine SLRPs, including lumican and fibromodulin, all of which possess a core protein with leucine rich repeat motifs flanked by cysteine-clusters (14). Biglycan and decorin, which show greater homology to each other than to other SLRPs, are widely distributed with overlapping but divergent patterns of expres-* This work was supported by National Institutes of Health Grants AR43621 and GM46846 (to D. S. G.), by a grant from FibroGen Inc. (South San Francisco, CA) (to D. S. G.), and by grants from the Medical Research Council of Canada (to P. J. R.) and the Shriners of North America (to P. J. R. and A. D. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
** To whom correspondence should be addressed. Tel.: 608-262-4676; Fax: 608-262-6691; E-mail: dsgreens@facstaff.wisc.edu. 1 The abbreviations used are: BMP-1, bone morphogenetic protein-1; mTLD, mammalian Tolloid; mTLL, mammalian Tolloid-like; PCP, procollagen C-proteinase; C-propeptide, carboxyl-terminal propeptide; Npropeptide, amino-terminal propeptide; ECM, extracellular matrix; TGF-␤, transforming growth factor-␤; SLRP, small leucine-rich proteoglycan; MEF, mouse embryo fibroblast; PCR, polymerase chain reaction; RT, reverse transcription; PAGE, polyacrylamide gel electrophoresis; P2, P3, P1Ј, cleavage site residues amino-terminal to (nonprimed) and carboxyl-terminal to (primed) the cleaved bond; dpc, day(s) postconception. sion in vertebrate connective tissues (15,16). Creation of mice homozygous null for biglycan or decorin have shown the former to be a positive regulator of bone growth (17) and the latter to play a role in regulating type I collagen fibrillogenesis in skin and tendon (18). High levels of expression in preosteogenic cells and a pericellular distribution are consistent with a role for biglycan in osteoblast differentiation, whereas an association of decorin expression with tissues rich in fibrillar collagens is consistent with a role in fibrillogenesis (16). Although the molecular bases for the biological roles of biglycan and decorin are unclear, they may involve the demonstrated abilities of the two to interact with various collagens, other ECM proteins, and transforming growth factor-␤ (19 -25).
Biglycan and decorin are unique among SLRPs in that they are synthesized as pro-forms containing N-propeptides of 21 and 14 residues, respectively, that are completely removed in most, but not all, connective tissues (26). Although the proteinase(s) responsible for these cleavage events has not been identified, the sites at which probiglycan and prodecorin are processed in vivo have been determined (26 -29), and the residues M(M/L)N-DEE and M(L/I)E-DE(A/G) found at the probiglycan and prodecorin sites, respectively, are conserved in various species (15, 30 -38). The similarity of these sequences suggests that the same proteinase(s) may be responsible for processing of both proteins, and interestingly, these cleavage sites show similarities to the cleavage sites of the C-propeptides of procollagens I-III (3,39,40).
In the present study, we demonstrate that BMP-1 cleaves probiglycan at a single site, thus removing the N-propeptide to produce biglycan with an NH 2 terminus identical to that of mature biglycan isolated from tissues. Consistent with a physiological role for the processing of probiglycan by BMP-1, expression domains of the two gene products are shown to be coincident in most developing tissues. A notable exception is postnatal articular cartilage, in which high levels of biglycan expression are not matched by detectable BMP-1 expression, and in which persistence of high levels of unprocessed probiglycan has previously been noted. The enzymes mTLD and mTLL-1 are also shown to have low levels of probiglycanprocessing activity. Moreover, whereas wild type mouse embryo fibroblasts (MEFs) produce only mature biglycan, MEFs deficient for BMP-1 and mTLD are shown to produce predominantly unprocessed probiglycan, and MEFs deficient for the three enzymes BMP-1, mTLD, and mTLL-1 are totally devoid of probiglycan-processing activity. Evidence that these proteases are responsible for processing probiglycan in at least some tissues in vivo is discussed, as are the implications for ECM deposition and morphogenesis.

EXPERIMENTAL PROCEDURES
Peptide Substrate Synthesis and Cleavage Assays-Peptides incorporating the propeptide cleavage sites of decorin and biglycan (Fig. 1B) were synthesized by Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry using an ABI 431A synthesizer. The peptides contained a central region of sequences spanning the biglycan or decorin propeptide cleavage site and basic terminal regions to ensure solubility. A cysteine residue was added to carboxyl-terminals for conjugating peptides to ovalbumin. The peptides were coupled to lysine residues of ovalbumin using N-hydroxysuccinimidyl bromoacetate to yield solutions of conjugate in phosphatebuffered saline (about 4 mg/ml). Peptide-ovalbumin conjugates were then dialyzed in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM CaCl 2 , prior to analysis in cleavage assays.
Peptide cleavage assays employed recombinant BMP-1, prepared using a baculovirus expression system, as described (3). Only the peptide assays employed the baculovirus-generated material, whereas all subsequent cleavage assays with recombinant probiglycan (see below) employed affinity-purified proteases produced in a mammalian expression system. Five l of recombinant BMP-1 and 5 l of peptide-ovalbumin conjugate were combined in a 40-l volume of 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM CaCl 2 and incubated at 37°C for 16 h. Control incubations contained culture medium derived from native Sf21 cells or cells infected with the wild type virus. Reactions were stopped by the addition of SDS-PAGE loading dye containing 2% ␤-mercaptoethanol and boiled for 5 min. Samples were subjected to SDS-PAGE on 10% acrylamide gels and electrotransferred to Sequi-Blot polyvinylidene difluoride membranes (Bio-Rad). Peptide-ovalbumin conjugates were identified by staining with Coomassie Blue, and NH 2 -terminal amino acid sequences were determined by automated Edman degradation on an ABI 473A protein sequencer.
Probiglycan Expression Construct-A cDNA encoding human probiglycan, except for the signal peptide, was generated by PCR using fulllength probiglycan cDNA clone P16 (15) as template, and primers 5Ј-ACTGTCAGCTAGCACTGCCCTTTGAGCAGAGAGGC-3Ј (forward) and 5Ј-ACTGTCACTCGAGCTCCCCATCAGGATGTCTGGC-3Ј (reverse), corresponding to sequences 169 -189 and 1278 -1298, respectively, of the published human probiglycan sequence (15), plus an NheI or XhoI site, respectively, for cloning. The PCR employed Advantage cDNA polymerase mix (CLONTECH) and denaturation at 94°C for 30 s, followed by 25 cycles of 94°C for 10 s, 65°C for 30 s, and 72°C for 2 min and final extension at 72°C/10 min. After digestion with NheI and XhoI, the 1129-base pair PCR product was inserted between the NheI and XhoI sites of expression vector pCEP-Pu/BM40s (41) downstream of, and in the same reading frame as, sequences encoding the BM40 signal peptide. Fidelity of the pCEP-Pu/BM40s-probiglycan expression vector was confirmed by DNA sequencing of the PCR insert and junctions on both strands.
Expression of Recombinant Probiglycan-293-EBNA human embryonic kidney cells (Invitrogen) were maintained in growth medium consisting of Dulbecco's modified Eagle's medium supplemented with 1 mM L-glutamine, 0.1 mM nonessential amino acids, and 10% fetal bovine serum (HyClone). Cells at 90% confluence were transfected with 10 g of pCEP-Pu/BM40s-probiglycan vector or empty pCEP-Pu/BM40s vector per 100-mm tissue culture dish, using LipofectAMINE according to the manufacturer's instructions (Life Technologies, Inc.). After 48 h, cells were selected in growth medium containing 5 g/ml puromycin (Sigma), and surviving cells were allowed to grow to confluent mass cultures in growth medium containing 5 g/ml puromycin.
Medium was removed from cell cultures, which were then washed three times with phosphate-buffered saline and switched to serum-free Dulbecco's modified Eagle's medium containing 40 g/ml soybean trypsin inhibitor (Sigma). After 24 h, conditioned medium was harvested, and protease inhibitors were added. Fresh serum-free medium was applied to the cells and similarly collected after an additional 24 h, and medium samples from 24 and 48 h harvest were pooled and centrifuged to remove cell debris. For samples in which unpurified material was to be analyzed by Western blotting (i.e. Fig. 2), inhibitors were 0.4 mM phenylmethylsulfonyl fluoride and 10 g/ml leupeptin. For samples from which probiglycan was to be purified, inhibitors were 0.4 mM phenylmethylsulfonyl fluoride, 1 mM N-ethylmaleimide, 1 mM p-aminobenzoic acid, 10 g/ml leupeptin, and 5 mM EDTA.
MEFs-Fibroblasts were isolated from 13.5-days postconception embryos, as described (6), and were passaged no more than five times in Dulbecco's modified Eagle's medium supplemented with 1 mM L-glutamine, 0.1 mM nonessential amino acids, and 10% fetal bovine serum.
Western Blots-For analysis of unpurified materials, conditioned media were dialyzed against 20 mM Tris-HCl, pH 8.0, 40 mM sodium acetate, 5 mM EDTA, and then either treated or not treated with 0.4 units/ml protease-free chondroitinase ABC (Seikaguku Corp.) for 4 h at 37°C. Subsequently, samples were precipitated by addition of trichloroacetic acid to a final concentration of 10% and incubation for 1 h at 4°C. Precipitates were collected by centrifugation, washed three times with 12.5 mM Tris-HCl, pH 7.5, 75% ethanol at 4°C, dissolved in 50 l of SDS-PAGE loading dye containing 2% ␤-mercaptoethanol, and boiled for 5 min, and 5 l aliquots were electrophoresed on 4 -15% acrylamide gradient gels. Western blot analyses of these materials and of the purified samples described below involved transfers to polyvinylidene difluoride membranes, incubations of blots with antibodies, and washes, as described previously (42), using antibodies raised against a peptide (LPFEQRGFWGGC) within the probiglycan N-propeptide (26), for both murine and human samples, or antibodies LF-51, for human samples, and LF106, for murine samples, raised against peptides within the mature form of human (GVLDPDSVTPTYSA) or murine (VPDLDSVTPTFSAMC) biglycan, respectively (15,43). N-propeptide and LF-51 antibodies were diluted 1:1000, whereas LF-106 and secondary antibodies of peroxidase-linked anti-rabbit Ig antisera were diluted 1:5000 (Amersham Pharmacia Biotech). Immunoreactive proteins were detected using SuperSignal peroxidase substrate (Pierce).
Purification of Recombinant Probiglycan-Conditioned medium samples were applied to a 1.5 ϫ 3-cm DEAE-cellulose column (DE52, Whatman) pre-equilibrated with Buffer A (50 mM Tris-HCl, pH 7.5, 250 mM NaCl, 1 mM N-ethylmaleimide, 1 mM p-aminobenzoic acid, and 0.1 mM phenylmethylsulfonyl fluoride). The column was washed in Buffer A, and bound proteins were eluted in a linear gradient of Buffer A made 250 -1200 mM NaCl. The proteoglycan form of probiglycan eluted at ϳ350 mM NaCl and fractions containing this form were dialyzed into 50 mM Tris-HCl, pH 7.5, 150 mM NaCl.
In Vitro Enzyme Assays-Proteases used for cleavage assays of recombinant probiglycan were recombinant human BMP-1, mTLD, mTLL-1, and mTLL-2, with COOH-terminal Flag-tags, produced in 293-EBNA cells, purified on anti-Flag affinity columns, and quantitated as described (5). To optimize comparison of activities, the various proteases were prepared and purified using identical conditions, and each enzyme preparation was pure to the extent that only a single band of appropriate size was detectable on zinc-stained SDS-PAGE gels (not shown). Approximately 500 ng of purified probiglycan and 15 ng of purified Flag-tagged BMP-1, mTLD, mTLL-1, or mTLL-2 were combined in a 50-l volume of 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM CaCl 2 and incubated 15 h at 37°C. Subsequently, 10 l of a solution containing 100 mM Tris-HCl, pH 8.0, 240 mM sodium acetate, 25 mM EDTA, and 0.02 units of protease-free chondroitinase ABC (Seikaguku Corp.) was added to each assay, and samples were incubated an additional 4 h at 37°C. Reactions were stopped by the addition of 10ϫ SDS-PAGE loading dye containing 2% ␤-mercaptoethanol and by boiling for 5 min. Samples were subjected to SDS-PAGE on 10% acrylamide gels and analyzed by Western blotting, as described above.
Amino Acid Sequence Analysis-Approximately 3 g of probiglycan was incubated with 50 ng of BMP-1 at 37°C for 15 h, and the sample was then treated with 0.1 unit of chondroitinase ABC under conditions similar to those described above. Proteins were concentrated by adding trichloroacetic acid to a final concentration of 10%, incubated at 4°C for 1 h, and centrifuged, and precipitates were dissolved in SDS-PAGE sample buffer. Proteins were resolved by SDS-PAGE on a 10% acrylamide gel and electrotransferred to a Sequi-Blot polyvinylidene difluoride membrane (Bio-Rad). Proteins were identified by staining with 0.1% Amido Black in 10% acetic acid, and NH 2 -terminal amino acid sequences were determined by automated Edman degradation on an Applied Biosystems Procise 494 HT protein sequencing system at the Harvard University Microchemistry Facility.
In Situ Hybridization-Mouse biglycan riboprobes were prepared from a 2.4-kb full-length cDNA clone in the Shlox vector (Novagen) (Ref. 43, Clone-3, GenBank TM accession number L20276). This construct was linearized with HindIII and transcribed with polymerase T7 (sense) or linearized with EcoRI and transcribed with polymerase SP6 (antisense). Mouse decorin riboprobes were prepared from a 1.3-kb cDNA insert in pBluescript SK-(Stratagene) (Ref. 43, plasmid mDCN-5, GenBank TM accession number X53929), linearized with HindIII and transcribed with polymerase T3 (antisense) or linearized with EcoRI and transcribed with T7 (sense). Biglycan and decorin clones were kindly provided by Dr. Larry Fisher (NIDCR, National Institutes of Health). Preparation of riboprobes for BMP-1, uniform labeling of riboprobes with [ 35 S]UTP, tissue preparation, and in situ hybridization were all performed as described previously (2,5). Bone sections were fixed overnight in 4% paraformaldehyde, washed for 3 min in deionized water, demineralized in Immunocal (Decal Chemical Corp.) overnight, washed for 3 min in running deionized water, and then dehydrated and embedded as for nonmineralized tissues. For histological analysis, sections were deparaffinized in xylenes, rehydrated, stained for 20 min in 0.3% Alcian blue, pH 2.5, immersed for 10 min in 0.3% sodium carbon-ate, and stained with hematoxylin and eosin, but without a final acidalcohol wash. Slides were analyzed using bright-and dark-field optics of a Zeiss Axiophot 2 microscope.

BMP-1 as a Candidate Enzyme for the Proteolytic Processing
of Probiglycan and Prodecorin-The reported cleavage sites at which probiglycan and prodecorin are proteolytically processed in vivo (26 -29) are similar to the cleavage sites at which BMP-1 has previously been shown to process procollagens I-III, Chordin, and prolysyl oxidase, particularly in that each site contains an aspartate residue at the P1Ј position and tyrosine and/or methionine residues NH 2 -terminal to the cleavage site, usually in the P3 or P2 position (Fig. 1A). In an initial assay to determine whether BMP-1 was capable of recognizing and cleaving the probiglycan and prodecorin sites, synthetic peptides spanning the two cleavage sites were prepared (Fig.  1B). The peptides were coupled to ovalbumin via their carboxylterminal residues, so that following internal cleavage the product bearing the new amino terminus would remain coupled and could be readily purified by SDS-PAGE. Analysis of these products by NH 2 -terminal amino acid sequencing revealed that in both cases, baculovirus-generated recombinant BMP-1 (3) cleaved at the predicted sites, without the occurrence of cleavages at any additional sites.
Expression and Characterization of Recombinant Probiglycan-The experiments with synthetic peptides indicated that BMP-1 can recognize and cleave both the probiglycan and prodecorin sites. To examine the ability of BMP-1 to process such sites in the context of a full-length protein precursor, a human probiglycan cDNA was inserted into the episomal expression vector pCEP-Pu/BM40s (41) for production of recombinant probiglycan, differing from native preprobiglycan only in substitution of the native signal peptide with the BM40 signal peptide, for optimization of secretion (41). The conditioned media of confluent, puromycin-resistant mass cultures of 293-EBNA human embryonic kidney cells transfected either FIG. 1. BMP-1 as a candidate enzyme for the proteolytic removal of probiglycan and prodecorin N-propeptides. A, alignment of the sites at which BMP-1 cleaves the C-propeptides of procollagens I-III (3,39,40), Chordin (5), and prolysyl oxidase 2 (7), with the sites at which the N-propeptides of probiglycan and prodecorin are removed in vivo (26 -29). Conserved Asp, Tyr, and Met residues are in boldface. B, sequences of synthetic peptides used for cleavage assays with BMP-1. Synthetic peptides of 22 amino acid residues each were prepared with central 10 amino acid regions comprising the cleavage sites of probiglycan and prodecorin. Basic amino acid residues were added to ensure solubility, and a cysteine residue was added to the carboxyl terminus to allow coupling to ovalbumin. with the expression construct or with an empty pCEP-Pu/ BM40s vector were examined by Western blot for secreted probiglycan. As can be seen ( Fig. 2A), antibodies specific for probiglycan N-propeptide sequences detected a heterogeneous smear centered around 100 kDa and a discrete band of approximately 50 kDa. These two forms are larger than the M r 42,510 predicted for preprobiglycan by the cDNA sequence (15) and are similar in size to ϳ100and ϳ49-kDa forms of recombinant biglycan previously produced in a vaccinia virus expression system (44). In the previous study, in which a truncated form of recombinant biglycan was produced that lacked the N-propeptide, the ϳ100-kDa species was shown to be the proteoglycan form, and the ϳ49-kDa species was shown to be a nonglycanated but Asn-glycosylated form of the biglycan core protein (44). In the present study, treatment of medium samples with chondroitinase ABC resulted in disappearance of the ϳ100-kDa form and a concomitant increase in levels of the ϳ50-kDa form ( Fig. 2A), indicating that the ϳ100and ϳ50-kDa species represent the proteoglycan and nonglycanated forms, respectively, of probiglycan. The similarity in the sizes of probiglycan produced in the present study and the truncated forms of biglycan produced previously (44) is due to the relatively small size (ϳ2 kDa) of the propeptide.
Analysis of conditioned medium samples with antibodies specific for sequences within mature biglycan also detected the ϳ100and ϳ50-kDa-forms (Fig. 2B), although binding of these antibodies to the proteoglycan form was weak. The latter effect can probably be ascribed to steric hindrance by glycosaminoglycan chains, because the sequence GVLDPDSVTPTYSA, against which the antibody was made (15), begins with a Gly contained in one of the two biglycan Ser-Gly glycosaminoglycan attachment sites. In addition, these antibodies detected bands of ϳ47 and ϳ35 kDa that were not detected by antibodies directed against propeptide sequences (compare Fig. 2A to Fig.  2B). These latter bands were also found in the media of cells transfected with an empty pCEP-Pu vector (Fig. 2B) and in the media of untransfected 293-EBNA cells (not shown). Thus, the ϳ47-kDa band, which approximates the size of mature biglycan, may represent endogenous biglycan produced by 293-EBNA cells, all of which is processed to the mature form, whereas the 36-kDa band may represent a proteolytic fragment of the endogenous biglycan. However, levels of neither the ϳ47-kDa nor the 36-kDa form seem appreciably increased in cultures of transfected 293-EBNA cells producing large amounts of recombinant probiglycan (Fig. 2B). Thus, either the cells lack the capacity to process additional probiglycan into the ϳ47and ϳ35-kDa forms, or these forms represent nonbiglycan proteins that cross-react with the biglycan antibodies. In either case, neither the ϳ47-kDa nor the ϳ35-kDa species co-purified in detectable quantities with the recombinant probiglycan isolated by DEAE-chromatography and used in the enzyme cleavage assays described below.

BMP-1 Efficiently Processes the N-propeptide of Recombinant
Probiglycan at the Physiological Site-To determine whether the N-propeptide of probiglycan might be processed by BMP-1 and/or related mammalian enzymes, recombinant probiglycan was purified from 293-EBNA cell conditioned media by DEAEcellulose (DE52) cation exchange chromatography (Fig. 2C) and incubated in separate reactions with purified recombinant human BMP-1, mTLD, mTLL-1, or mTLL-2, each of which was prepared in a mammalian expression system and affinity-purified, under identical conditions, which have been previously described (5). As can be seen (Fig. 3A), Western blot analysis using antibodies specific for the probiglycan N-propeptide showed a total loss of detectable probiglycan in material that had been incubated with BMP-1, whereas the intensities of probiglycan bands were not detectably diminished in samples that had been incubated with mTLD, mTLL-1, or mTLL-2. Western blot analysis using antibodies against sequences within mature biglycan (Fig. 3B) showed that the probiglycan incubated with BMP-1 had been totally converted to a ϳ48-kDa form. Because the latter form corresponds in size to mature biglycan and does not react with the propeptide antibodies (Fig.  3A), these results were consistent with the possibility that BMP-1 is capable of cleaving the probiglycan N-propeptide. Fig. 3B also shows that both mTLD and mTLL-1 had converted a small fraction of probiglycan to the 48-kDa form, consistent with the presence of low levels of probiglycan-processing activity in these enzymes, whereas no such processing was observed in the sample that had been incubated with mTLL-2.
To ascertain the site at which BMP-1 cleaves the probiglycan molecule, the NH 2 -terminal sequence of the 48-kDa cleavage product was determined. Toward this end, a scaled-up cleavage reaction was performed in which 3 g of purified probiglycan was incubated with 50 ng of BMP-1. Subsequent to incubation, the material in the cleavage reaction was further treated with chondroitinase ABC and subjected to SDS-PAGE, and the 48-kDa cleavage product, which could be visualized by staining with Coomassie Blue (Fig. 4A), was isolated and subjected to automated Edman degradation for determination of NH 2 -terminal amino acid sequences. Ten cycles of Edman degradation produced a sequence identical to the published human biglycan sequences that occur just COOH-terminal of the N-propeptide and corresponding to the NH 2 terminus of mature biglycan isolated from tissues (15,26,28,29). Thus, BMP-1 processes probiglycan at the physiologically relevant site.
A Product, or Products, of the Bmp1 Gene Is Responsible for the Majority of Probiglycan Processing in MEFs-We have previously described mice with null alleles for the Bmp1 gene, which encodes both BMP-1 and mTLD (6). Mice homozygous null for the Bmp1 gene are perinatal lethal, with defects in fibrillogenesis and diminished procollagen processing, but with some residual PCP activity, presumably provided through functional substitution by mTLL-1 and/or other proteases with related activity. To ascertain whether MEFs from Bmp1-/-and Bmp1ϩ/-embryos have diminished ability to process probiglycan, MEFs were derived from Bmp1ϩ/ϩ, ϩ/-, and -/-13.5-days postconception (dpc) embryo littermates, as described (6), and the conditioned media of confluent cultures of each type of MEFs were examined by Western blot for secreted pro-and mature forms of biglycan. As can be seen (Fig. 5A), antibodies specific for probiglycan N-propeptide sequences detected a ϳ50-kDa probiglycan band, subsequent to treatment of medium samples with chondroitinase ABC, in the culture medium of Bmp1-/-MEFs but not in the culture media of Bmp1ϩ/ϩ or Bmp1ϩ/-MEFs. Conversely, analysis of chondroitinase ABCtreated conditioned medium samples with antibodies specific for sequences within mature biglycan detected solely mature ϳ48-kDa biglycan in the culture media of Bmp1ϩ/ϩ and Bmp1ϩ/-MEFs, but predominantly ϳ50-kDa probiglycan and lesser amounts of ϳ48-kDa biglycan in the medium of Bmp1-/-MEFs (Fig. 5A). Thus, a product, or products, of the Bmp1 gene is shown to provide the majority of in vivo probiglycan-processing activity, at least in mouse embryo fibroblasts. Because BMP-1 has far more probiglycan-processing activity than does mTLD (Fig. 3), the most straightforward interpretation of these data is that BMP-1 provides the preponderance of probiglycan-processing activity in MEFs.
Although the majority of probiglycan remains uncleaved in Bmp1-/-MEF cultures (Fig. 5A), significant residual process-FIG. 6. Expression of BMP-1, mTLD, mTLL-1, and mTLL-2 RNAs in MEFs. RT-PCR was used to analyze the expression of mTLL-1, mTLL-2, and BMP-1/mTLD RNAs in Bmp1 -/-, ϩ/-, and ϩ/ϩ MEFs. RT-PCR for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) RNA was performed for each sample to control for variations in RNA isolation, RT efficiency, and gel loading. FIG. 4. BMP-1 cleaves probiglycan at the physiological site for removal of the N-propeptide. A, electrophoretic mobilities are compared for purified probiglycan incubated for 18 h in the absence (-BMP-1) or presence (ϩBMP-1) of purified BMP-1. Samples were run under reducing conditions on a 4 -15% acrylamide SDS-PAGE gel and visualized by staining with Coomassie Blue. Molecular masses are indicated for protein standards. B, alignment of the NH 2 -terminal amino acid residues of the cleavage product resulting from BMP-1 cleavage of probiglycan, as determined by automated Edman degradation, with the published human probiglycan sequence deduced from cDNA and the physiological site at which proteolytic cleavage removes the N-propeptide in vivo (denoted by an arrow) to produce the mature tissue form of biglycan (15,26,28,29). Edman degradation assigned all residues with highest confidence except for the Ser at position 5, which, with a yield lower than expected based upon the yield of Ser at position 10, was assigned as probable and is indicated by brackets. ing remains. Because mTLL-1 was found to have some probiglycan-processing activity in in vitro assays (Fig. 3), we sought to determine whether the mTLL-1 gene, Tll1, is expressed in MEF cells isolated from 13.5-dpc embryos, such that mTLL-1 would be properly situated to supply some portion of the residual probiglycan processing activity observed in Bmp1-/-MEFs. It was also of interest to determine whether the mTLL-2 gene, Tll2, is expressed in MEF cultures. Toward these ends, a series of RT-PCR analyses were performed on RNA isolated from Bmp1 ϩ/ϩ, ϩ/-, and -/-MEFs to gauge the expression of RNA for mTLL-1, mTLL-2, and the two products of the Bmp1 gene, BMP-1 and mTLD. As can be seen (Fig. 6), expression of mTLL-1 RNA is readily detectable in Bmp1 ϩ/ϩ, ϩ/-, and -/-MEF cells, as is the expression of mTLL-2 RNA, whereas PCR products corresponding to both wild type and mutant Bmp1 alleles are detectable in the Bmp1ϩ/-MEF sample, and only products corresponding to the Bmp1 mutant allele are detectable in the Bmp1-/-MEF sample.
mTLL-1 Provides the Residual Probiglycan-Processing Activity in Bmp-1-null MEFs-Because the Bmp1 gene and the Tll1 gene, which encodes mTLL-1, map to different chromosomes (8), we were able to generate embryos homozygous null at both loci through matings of Bmp1ϩ/-;Tll1ϩ/-double heterozygotes. Thus, a comparison could be made of probiglycan processing in MEF cultures derived from wild type, Bmp1-/-, Tll1-/-, and Bmp1-/-;Tll1-/-doubly null embryos (Fig. 5B). Despite the low but detectable levels of probiglycan processing shown by recombinant mTLL-1 in vitro (Fig. 3B), processing of probiglycan to mature biglycan appeared to be complete in cultures of Tll1-/-MEFs (Fig. 5B). This is similar to our previous finding that type I procollagen is processed to similar extents in cultures of wild type and Tll1-/-MEFs, despite evidence of mTLL-1 PCP activity (5,9). Nevertheless, analysis found probiglycan to be totally unprocessed in cultures of MEFs from Bmp1-/-;Tll1-/-doubly null embryos (Fig. 5B). The most straightforward interpretation of these results is that although the more robust probiglycan-processing activity of BMP-1, combined with that of mTLD, seem sufficient to fully compensate for loss of mTLL-1 activity in Tll1-/-MEFs, mTLL-1 provides the residual probiglycan-processing activity observed in Bmp1-/-MEFs. Thus, products of the related Bmp1 and Tll1 genes appear to provide all detectable probiglycan-processing activity in mouse embryo fibroblasts.
Overlapping Expression Domains of BMP-1/mTLD and Biglycan RNA in 15.5-dpc Mouse Embryos and 21-Day-old Mouse Femur-Because BMP-1 can correctly and efficiently process probiglycan in vitro and because products of the Bmp1 gene seem responsible for the preponderance of probiglycan processing in MEFs, it was of interest to obtain insights regarding the possible codistribution of expression of the Bmp1 gene and the probiglycan gene Bgn, in vivo. As can be seen in serial sagittal sections of a 15.5-dpc mouse embryo (Fig. 7), both genes are broadly co-expressed throughout mesenchymal tissues, with particularly high signals for both found in developing bones, as is particularly evident in cross sections of the developing clavicle, basioccipital bone, vertebrae, and bones of the hind limb. It has previously been noted (26) that unprocessed probiglycan is detectable in some tissues, and particularly in articular cartilage. To ascertain whether there might be a correlation between the distributions of expression of Bmp1 and Bgn, and the previous finding of unprocessed probiglycan in articular cartilage (26), expression patterns of the two genes were compared in serial sections of the femoral growth plate of a 21-day-old mouse (Fig. 8). As can be seen, both Bmp1 and Bgn are expressed at high levels in the epiphyseal and metaphyseal centers of ossification and in the area of ossification corresponding to the subperiosteal collar. In contrast, although Bgn is expressed at high levels in articular cartilage, Bmp1 is not. The latter observation thus shows a correlation between the relative levels of expression of the two genes and persistence of probiglycan in articular cartilage. DISCUSSION Previously, BMP-1 and mTLD have been shown to play multiple roles integral to the formation of vertebrate fibrous ECM. These have included cleavage of the C-propeptides of procollagens I-III (3,4), activation of lysyl oxidase (7), 2 and the processing of type V procollagen N-propeptides, which appear to be involved in regulating the shape and diameters of heterotypic type I/V collagen fibrils (45). BMP-1 also appears to play roles in the deposition of nonfibrillar ECM, as BMP-1 has been found to process precursors into the mature ␥2 chains of laminin 5, a major structural component of the basement membranes of skin and other stratified squamous epithelia (46). Recently, BMP-1 and the related mammalian protease mTLL-1 have been shown to be capable of cleaving Chordin, an antagonist of TGF-␤-like BMPs (5). Thus, BMP-1, and related proteases, may also serve to orchestrate the deposition of various matrix components with BMP signaling in the course of morphogenetic events. Here we further expand the range of known roles for these enzymes by showing that 1) BMP-1 cleaves full-length recombinant probiglycan at a single site, corresponding to the physiological site at which the N-propeptide is removed in vivo; 2) recombinant mTLD and mTLL-1 appear to have lower levels of the same activity; and 3) products of the Bmp1 gene are responsible for most detectable probiglycan-processing activity in MEFs, and products of the Bmp1 and Tll1 genes combined are responsible for all. Thus, the data presented herein indicate BMP-1 to be the major protease responsible for processing probiglycan to its ma-ture form in at least some tissues in vivo, with additional probiglycan-processing activity provided by mTLD and mTLL-1.
Previously, we have speculated that products of the Bmp1 and Tll1 genes might be capable of functional substitution for each other in vivo, with mTLL-1 partially compensating for loss of BMP-1 and mTLD in Bmp1-/-embryos, and with BMP-1 and mTLD partially compensating for loss of mTLL-1 in Tll1-/-embryos (6,8,9). Similarly, we have speculated that mTLL-1 might supply the residual procollagen C-proteinase activity found in Bmp1-/-MEFs (6), despite the observation that MEFs cultured from Tll1-/-embryos show levels of procollagen processing activity indistinguishable from those of wild type MEFs (9). The observation in the current study that mTLL-1 provides the residual probiglycan-processing activity observed in Bmp1-/-MEF cultures is the first demonstration of functional substitution by products of the two genes in a biological system and is consistent with the possibility that the product of the Tll1 gene can compensate for a deficiency in the products of the Bmp1 gene, and vice versa, in at least some tissues. Nevertheless, observations that BMP-1 has higher levels of PCP activity (5) and probiglycan-processing activity, in vitro, than either mTLD or mTLL-1, and that Tll1-/-MEFs have levels of PCP (6) and probiglycan-processing activities indistinguishable from those of wild type MEFs, suggest that BMP-1 may normally be the major protease responsible for provision of both activities in vivo.
In a previous study, immunohistochemical analysis of aorta with an antibody specific for the probiglycan N-propeptide found intense intracellular staining but an absence of extracellular staining, suggesting that processing of the propeptide occurs prior to secretion (47). In contrast, the finding of secreted, unprocessed probiglycan in the ECM of tissues such as articular cartilage (26) has suggested that processing of the N-propeptide may occur extracellularly. It is clear that BMP-1, mTLD, and mTLL-1 are secreted proteases (3,5,39,40,42), whereas both BMP-1 and mTLD have been shown to have slightly basic pH optima (39,40), suggesting that they operate most efficiently in the extracellular milieu. Thus, the finding in the present report that all MEF probiglycan-processing activity is provided by products of the Bmp1 and Tll1 genes suggests that processing of probiglycan to its mature form occurs extracellularly. Interestingly, however, although the cleavage site for removal of the prodecorin N-propeptide is similar to a number of previously characterized sites utilized by BMP-1 and related enzymes, and although we have shown that BMP-1 is capable of recognizing and cleaving the prodecorin site in a synthetic peptide, pulse-chase experiments have previously suggested that the N-propeptide of prodecorin is removed intracellularly, in the Golgi apparatus, prior to elongation of glycosaminoglycan chains (48). Thus, it is possible that prodecorin is not processed by BMP-1 and/or related enzymes but that it is instead processed by different proteases than is probiglycan. In support of the possibility that processing of prodecorin differs from that of probiglycan, our attempts to produce recombinant prodecorin, using full-length human prodecorin cDNA and the same pCEP-Pu/BM40s vector/293-EBNA cell expression system used to efficiently produce unprocessed probiglycan, resulted in the detection of only processed mature decorin in the media of transfected 293-EBNA cultures. 3 Nevertheless, because 293-EBNA cells produce low levels of endogenous BMP-1, 3 it remains possible that prodecorin is cleaved by the same enzymes as is probiglycan but at a much higher rate. Efforts to examine processing of prodecorin in MEF cultures were inconclusive, 3 due to the small size difference between pro-and mature forms of decorin and the absence of an antibody that recognizes the murine decorin N-propeptide. If, however, prodecorin is indeed processed by the same enzymes as probiglycan, and if prodecorin is in fact processed in the Golgi, then it is implied that BMP-1 and related proteases are capable of processing some substrates within the Golgi. Relevant to the latter possibility, a number of studies have found the processing of procollagen C-propeptides in tissues to be extremely rapid (49,50), consistent with the possibility that cleavage occurs coincident with secretion and in a pericellular environment. The latter environment might well include the trans Golgi compartment, in which the processing of some portion of procollagens, probiglycan, and prodecorin by BMP-1 and related proteases may occur in vivo.
Cleavage of the N-propeptide not only changes the primary structure of probiglycan, by removing the NH 2 -terminal 21 amino acid residues, but previous antibody studies also suggest that removal of the N-propeptide induces conformational changes that affect the availability of epitopes in both the NH 2 -terminal (47) and COOH-terminal (26) portions of the mature biglycan molecule. Therefore, removal of the N-propeptide, through cleavage by BMP-1 and related enzymes, is likely to have significant effects on the properties of biglycan, which, in turn, are likely to affect its interactions with other molecules. Interestingly, biglycan has been shown to bind TGF-␤ (24) and may also interact with various collagen types (21,22,44), whereas BMP-1, mTLD, and mTLL-1 are involved in the formation of collagenous ECM and in modulating the activity of certain TGF-␤-like molecules (3)(4)(5)(6)9). Thus, the interaction of these proteases with biglycan would conform to the previously suggested roles of these molecules. In particular, biglycan has been shown to play an important role in the formation of bone (17), whereas BMP-1 and mTLD have also been implicated in bone morphogenesis (5,6,13). Thus, the interplay of these molecules may normally be of particular importance in the proper formation of this tissue.