Inhibition of Bone Morphogenetic Protein 1 by Native and Altered Forms of α2-Macroglobulin*

The four mammalian bone morphogenetic protein 1 (BMP1)-like proteinases act to proteolytically convert procollagens to the major fibrous components of the extracellular matrix. They also activate lysyl oxidase, an enzyme necessary to the covalent cross-linking that gives collagen fibrils much of their tensile strength. Thus, these four proteinases are attractive targets for interventions designed to limit the excess formation of fibrous collagenous matrix that characterizes fibrosis. Although it has previously been reported that the serum protein α2-macroglobulin is unable to inhibit the astacin-like proteinases meprin α and meprin β, we herein demonstrate α2-macroglobulin to be a potent inhibitor of the similar BMP1-like proteinases. BMP1 is shown to cleave the α2-macroglobulin “bait” region, at a single specific site, which resembles the sites at which BMP1-like proteinases cleave the C-propeptides of procollagens I–III. α2-Macroglobulin is an irreversible inhibitor that is shown to bind bone morphogenetic protein 1 in a covalent complex. It is also demonstrated that genetically modified α2-macroglobulin, in which the native bait region is replaced by sequences flanking the probiglycan BMP1 cleavage site, is enhanced ∼24-fold in its ability to inhibit BMP1, and is capable of inhibiting the biosynthetic processing of procollagen I by cells. These findings suggest possible therapeutic interventions involving ectopic expression of modified versions of α2-macroglobulin in the treatment of fibrotic conditions.

Bone morphogenetic protein 1 (BMP1) 3 is the prototype of a subgroup of structurally similar, secreted metalloproteinases, each member of which has an astacin-like protease domain, complement-uegf-BMP1 protein-protein interaction domains, and epidermal growth factor motifs (1,2). Mammalian members of this subgroup proteolytically convert a variety of precursor molecules into mature, functional proteins involved in formation of the extracellular matrix (ECM) (1,2). Members of this class of proteinases also activate a subset of the transforming growth factor-␤ superfamily of proteins in a broad range of species, through cleavage of extracellular protein antagonists (2). Thus, the four mammalian BMP1-like proteinases are likely to be key regulators and orchestrators of ECM formation and signaling by certain transforming growth factor-␤-like molecules, in morphogenetic events and homeostasis. Surprisingly, studies regarding regulation of the expression and activities of these key proteinases have been limited. Nevertheless, one such study showed that, whereas transcription of the BMP1 gene and secretion of BMP1 are both up-regulated ϳ8-fold by treatment of cells with transforming growth factor-␤1, induction of detectable procollagen C-proteinase activity is only ϳ2-fold, suggesting the existence of an endogenous inhibitor(s) (3). In fact, a number of studies have detected unexpectedly low levels of procollagen C-proteinase (pCP) activity in tissues and cell cultures, also leading to suggestions of endogenous inhibitors (4). ␣ 2 -Macroglobulin (␣ 2 M) is a member of the ␣-macroglobulin family of proteins found in the circulation and egg whites of a broad range of species (5). Human ␣ 2 M is found at relatively high levels (2-4 mg/ml) in plasma and is produced by hepatocytes, but is also produced by a number of other cell types that include lung fibroblasts, macrophages, astrocytes, and tumor cells (6,7). Human ␣ 2 M is a tetramer of four identical 185-kDa subunits, each of which has an exposed 39-amino acid "bait region" that contains cleavage sites for a variety of proteinases (6,8). Cleavage within the bait region results in exposure of a highly reactive ␣ 2 M thioester that can covalently bind the cleaving proteinase (6,8). Cleavage within the bait region also results in a conformational change that entraps the proteinase within the interior of the ␣ 2 M molecule, thus inhibiting further proteinase activity by steric hindrance (6,8). The conformational change also gives rise to what can be considered an "activated" form of ␣ 2 M, with exposed sites for binding of ␣ 2 M to its cognate cell surface receptor, the low-density lipoprotein receptor-related protein, and for binding to a number of cytokines, including transforming growth factor-␤, plateletderived growth factor, interleukin-1␤, basic fibroblast growth factor, and nerve growth factor (6). Binding to lipoprotein receptor-related protein results in rapid clearance of ␣ 2 M-proteinase complexes from the extracellular space and catabolism, although activated ␣ 2 M appears able to bind cytokines in a reversible manner that allows it to serve as a carrier and targeting protein involved in modulating the biological responses of various cell types (6).

* This work was supported by National Institutes of Health Grants AR47746
and GM71679 (to D. S. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 Both authors contributed equally to this work. 2  BMP1-like proteinases have astacin-like protease domains (1,2). It has previously been reported that ␣ 2 M does not inhibit vertebrate proteinases with astacin-like protease domains (7,9), such as ␣ meprin and ␤ meprin (9), although bovine ␣ 2 M has been shown to have inhibitory activity toward the crayfish digestive protease astacin (10). Here we identify a potential site for cleavage of ␣ 2 M by BMP1-like proteinases. We demonstrate cleavage at this site by BMP1 and the potent inhibition of BMP1 proteolytic activities via the formation of covalent complexes with cleaved ␣ 2 M. A genetically altered version of ␣ 2 M, in which the bait region has been replaced by the BMP1 cleavage site of the small leucine-rich proteoglycan precursor probiglycan is shown to have a greatly enhanced ability to inhibit BMP1, and is able to inhibit procollagen I processing by cells. Implications of the data are discussed.

EXPERIMENTAL PROCEDURES
Cleavage of ␣ 2 M by BMP1-200 nM FLAG-tagged recombinant BMP1, prepared, and purified as previously described (11), was incubated overnight at 37°C with 200 nM human ␣ 2 M (Sigma) in 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, and 10 mM CaCl 2 . Subsequently, reaction proteins were analyzed by SDS-PAGE under reducing conditions on a 7.5% gel and staining with Coomassie Brilliant Blue R-250. For Western blot analysis, 320 nM BMP1 was incubated with 200 nM ␣ 2 M at 37°C overnight, and samples were loaded on 7.5% gels and separated by SDS-PAGE, under both nonreducing and reducing conditions. Separated proteins were transferred to polyvinylidene difluoride membranes and probed with a 1:5000 dilution of anti-FLAG antibody (Sigma). Subsequently, membranes were incubated with a 1:25000 dilution of goat anti-mouse IgG horseradish peroxidase conjugate as the secondary antibody.
Inhibition of BMP1 pCP Activity by ␣ 2 M-For the dose dependence study, ␣ 2 M was preincubated with BMP1 for 2 h at 37°C. The study of time-dependent cleavage of ␣ 2 M by BMP1 indicates that cleavage of ␣ 2 M by BMP1 is complete after a 2-h preincubation, under the conditions used. Thus, the rate of procollagen processing is proportional to the amount of active BMP1 remaining uncomplexed to ␣ 2 M. Relative processing of procollagen was plotted against ␣ 2 M concentrations and nonlinear regression was performed to obtain IC 50 values.
Time-dependent Inhibition of BMP1 Cleavage of Probiglcyan by ␣ 2 M-15 ng of BMP1 (9.4 nM) was preincubated with/without 5 times the amount of ␣ 2 M (47.0 nM) in 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, and 10 mM CaCl 2 , 2 h at 37°C. Subsequently, 450 ng of probiglycan, prepared as previously described (12), was added to a final volume of 20 l and incubated overnight. Cleavage reactions were quenched by adding 4 l of chondroitinase ABC (a mixture of 10 l of 0.01 units/l protease-free chondroitinase ABC (Seikaguku Corp.), 40 l of 6ϫ chondroitinase buffer (100 mM Tris-HCl, pH 8.0, 240 mM NaAc, 0.25 mM EDTA), and 10 l of 500 mM EDTA), followed by incubation at 37°C for 4 h. Samples were subjected to SDS-PAGE on a 10% gel and Western blot analysis was performed, using anti-probiglycan antibody LF51 (13) (the kind gift of Dr. Larry W. Fisher) at a 1:5000 dilution, and a 1:25000 dilution of goat antirabbit IgG alkaline phosphatase conjugate, as the secondary antibody.
Inactivation of ␣ 2 M by Methylamine-1 mg/ml ␣ 2 M was incubated with 50 mM Tris-HCl, pH 7.1, in the presence or absence of 20 mM methylamine for 20 h, followed by dialysis against 50 mM Tris-HCl, pH 7.5, 100 mM NaCl. Both methylamine-treated and control ␣ 2 M samples were tested for the ability to be cleaved by BMP1, and ability to inhibit BMP1 cleavage of probiglycan.
293 T-Rex cells (Invitrogen) were maintained in Dulbecco's modified Eagle's medium (DMEM) with 5 g/ml blasticidin and 10% fetal bovine serum. Cells at 80% confluence were transfected with 1 g of expression plasmid/35-mm culture dish using Lipofectamine (Invitrogen). After 48 h, cells were selected in the same type of medium supplemented with 200 g/ml Zeocin. Production of secreted ␣ 2 M, upon induction with 1 g/ml tetracycline, was detected via Western blot.
Confluent cells were washed twice with phosphate-buffered saline (PBS), and incubated 15 min in serum-free DMEM at 37°C. Cells were then washed once with PBS, followed by addition of serum-free DMEM containing 1 g/ml tetracycline, to induce protein expression, and 40 g/ml soybean trypsin inhibitor. Conditioned medium was harvested every 24 h, and protease inhibitors were added to final concentrations of 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 mM N-ethylmaleimide, and 1 mM p-aminobenzonic acid. Conditioned medium was centrifuged to remove debris and supernatants were stored at Ϫ70°C. FLAG-tagged ␣ 2 M and b-␣ 2 M were affinity-purified from conditioned medium using an anti-FLAG M2 column (Sigma), following the manufacturer's instructions.
Cross-linking-6.9 nM recombinant ␣ 2 M or b-␣ 2 M was incubated 5 min with 80 mM glutaraldehyde (Sigma) in PBS at room temperature. Cross-linking was terminated by adding glycine to a final concentration of 0.2 M. Samples were separated by SDS-PAGE on 4 -15% acrylamide gradient gels and subjected to immunoblotting.
Determination of K i Values and Second-order Rate Constants for BMP1-␣ 2 M Interactions-Confluent 293 T-Rex cells producing recombinant BMP1 were washed twice with PBS, and incubated 15 min in serum-free, Cys/Met-free DMEM at 37°C. Cells were then washed once with PBS, followed by addition of serum-free DMEM containing 1 g/ml tetracycline, to induce protein expression, 40 g/ml soybean trypsin inhibitor, and 60 Ci/ml Pro-Mix 35 S-cell labeling mixture (Amersham Biosciences). Conditioned medium was harvested every 24 h, and protease inhibitors were added to final concentrations of 1 mM phenylmethylsulfonyl fluoride, 1 mM N-ethylmaleimide, and 1 mM p-aminobenzonic acid. Metabolically 35 S-radiolabeled BMP1 was affinity purified from conditioned medium using an anti-FLAG M2 column (Sigma), following the manufacturer's instructions.
Inhibition of Procollagen Processing by Cells-2 ϫ 10 5 MC-3T3-E1 cells were plated in a 24-well plate, allowed to attach overnight, and then treated with 50 g/ml ascorbate in DMEM, 10% FBS for 24 h. Cells were then washed twice with PBS, and incubated 15 min in serum-free DMEM at 37°C. Cells were then washed once with PBS, followed by addition of serum-free DMEM containing 50 g/ml ascorbate, 40 g/ml soybean trypsin inhibitor, and 20 nM ␣ 2 M or b-␣ 2 M, or an equivalent volume of buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl). Conditioned media were harvested after 24 h as described above. Cell layers were washed twice with ice-cold PBS, and scraped into hot SDS-sample buffer. Medium and cell layer samples were subjected to SDS-PAGE on acrylamide gels, transferred to nitrocellulose membranes, and probed with anti-collagen ␣1(I) C-telopeptide polyclonal antibody LF67 (13) (a generous gift from Larry Fisher), as described.

RESULTS
BMP1 Cleaves ␣ 2 M-Perusal of the amino acid sequence of the ␣ 2 M bait region found a site comprising residues Ser 687 , Asp 688 , and surrounding residues that resembled the majority of known BMP1 cleavage sites (Fig. 1). The resemblance resided primarily in the placement of Phe 684 and Tyr 686 3 and 4 residues, respectively, NH 2 -terminal to Asp 688 , because residues with aromatic side chains are frequently found in positions P2-P5, and an Asp is almost always found in the P1Ј position of previously identified substrates of BMP-like proteinases (2) (Fig. 1). Moreover, the majority of previously characterized cleavage sites of BMP1-like proteinases have residues with small side chains in the P1 positions (Ref. 2 and Fig. 1), such that the placement of ␣ 2 M Ser 687 in relationship to Asp 688 , Phe 684 , and Tyr 686 is also reminiscent of a BMP1 cleavage site.  (27), procollagens I-III, and the pro-␣2(V) and pro-␣1(VII) procollagen chains (15,28,29), the ␥2 chain of laminin 5 (30,31), probiglycan (12), and the NH 2 -terminal cleavage site of Chordin (11). Immediately beneath the ␣ 2 M sequence is the probiglycan sequence that substitutes for the native bait region in the mutant b-␣ 2 M protein. To test whether ␣ 2 M can be cleaved by BMP1-like proteinases, the two proteins were co-incubated and ␣ 2 M examined for processing. Although ␣ 2 M incubated alone at 37°C overnight was stable ( Fig. 2A), 185-kDa ␣ 2 M co-incubated at 37°C overnight with BMP1 was cleaved to produce two bands of ϳ100 and ϳ85 kDa. NH 2 -terminal amino acid sequencing of the 85-kDa band yielded the sequence SVSGKPQYMV, corresponding to the NH 2 terminus of secreted ␣ 2 M. NH 2 -terminal amino acid sequencing of the 100-kDa band yielded the sequence DVMGRGHR, thus demonstrating cleavage of ␣ 2 M by BMP1 at the predicted site between Ser 687 and Asp 688 in the bait region.
Cleaved ␣ 2 M Forms a Complex with BMP1-Because a common feature in the inhibition of proteinases by ␣ 2 M is formation of a covalent complex between ␣ 2 M and the proteinase, we next sought to determine the capability of ␣ 2 M to covalently bind BMP1. BMP1 is normally detected on SDS-PAGE gels as a ϳ90-kDa monomer (3,11), but it can also be detected as a ϳ270-kDa form on a reduced gel, subsequent to incubation with ␣ 2 M (Fig. 2B). Under non-reducing conditions (Fig. 2C), BMP1 can be detected as even higher molecular weight forms. The 85-kDa NH 2 -terminal and 100-kDa COOH-terminal fragments of cleaved ␣ 2 M can remain covalently bound, via disulfide linkage, and this form can be linked to other 85-and 185-kDa forms via disulfide bonds (8). Thus, the high molecular weight forms observed under non-reducing conditions likely represent BMP1 covalently bound, via reaction with the thioester, to 185-, 270-, and 375-kDa disulfide-bonded oligomers, although the exact identity of each band on the non-reducing gel remains somewhat speculative. The above interpretations are consistent with Western blots probed with anti-␣ 2 M antibodies, which show ␣ 2 M to co-localize to the same high molecular weight forms as BMP1 under both reducing (Fig. 2D) and non-reducing (Fig. 2E) conditions. The observation here of covalent binding of BMP1 to the 100-kDa ␣ 2 M cleavage product is consistent with the mechanism whereby ␣ 2 M has been found to covalently bind other proteinases that it inhibits (8).
␣ 2 M Inhibits the pCP Activity of BMP1-Because the first identified and best characterized activity of BMP1 is as a pCP (15), we next sought to determine whether this activity is inhibited in the presence of ␣ 2 M. Control experiments demonstrated that BMP1 achieves maximum cleavage of plasma ␣ 2 M by 2 h (Fig. 3A). Thus, to gauge the effect of ␣ 2 M-BMP1 interaction on BMP1 pCP activity, a constant amount of BMP1 (9.4 nM) was preincubated 2 h with increasing concentrations of ␣ 2 M (0, 4.7, 9.4, 18.7, 37.5, 56.2, 75.0, or 93.7 nM), prior to incubation overnight with 3 H-radiolabeled type I procollagen. Reaction mixtures were subjected to SDS-PAGE and cleavage of procollagen was measured by densitometric analysis of autofluorogams. As can be seen (Fig. 3B), prior incubation with ␣ 2 M led to potent inhibition of BMP1 pCP activity, with a calculated IC 50 of 4.8 nM.
␣ 2 M Inhibits the Cleavage of Probiglycan by BMP1-To determine whether ␣ 2 M can inhibit BMP1 cleavage of substrates other than procollagens, we tested the ability of ␣ 2 M to inhibit BMP1 cleavage of probiglycan. As can be seen (Fig. 4), BMP1 was able to completely convert probiglycan to biglycan after 30 min under assay conditions, whereas preincubation with a 5-fold molar excess of ␣ 2 M for 2 h prior to incubation with probiglycan resulted in inhibition of a majority of probiglycan processing. Thus, ␣ 2 M appears to be a general inhibitor of BMP1 activity against various substrates.
Mechanism of ␣ 2 M Inhibition of BMP1-Binding and inhibition of proteinases by ␣ 2 M is thought to follow cleavage of the

␣ 2 -Macroglobulin Complexes and Inhibits BMP1
bait region. The latter results in activation of a highly reactive ␣ 2 M thioester bond, which can form an amide bond between lysyl amino groups of the proteinase and the glutamyl residue of the thioester (8). In addition, cleavage of the bait region triggers a conformational change in ␣ 2 M, such that it "collapses" around the proteinase, thus trapping it and inhibiting its ability to interact with large protein substrates via steric hindrance (8). Treatment of ␣ 2 M with methylamine, which reacts with the thioester, also causes conformational changes in ␣ 2 M, and removes its ability to bind and inhibit proteinases by the mechanism described above. To determine whether binding and inhibition of BMP1 by ␣ 2 M is likely to be via the mechanism described above, ␣ 2 M was treated with methylamine prior to incubation with BMP1. BMP1 was unable to cleave methylamine-treated ␣ 2 M (Fig. 5A), ␣ 2 M pretreated with methylamine did not form complexes with BMP1 (Fig. 5B), and ␣ 2 M pretreated with methylamine was unable to inhibit BMP1 processing of probiglycan to biglycan (Fig. 5C). Inability of BMP1 to cleave methylamine-treated ␣ 2 M was probably the consequence of the conformational change induced in ␣ 2 M by methylamine interaction with the ␣ 2 M thioester. Subsequent to this conformational change, the bait region is presumably not available for cleavage. Together, results thus bolster the conclusion that binding of BMP1 by ␣ 2 M can involve formation of an amide bond between BMP1 lysyl amino groups and the glutamyl residue of the ␣ 2 M thioester, and that inhibition involves ␣ 2 M conformational changes consequent to cleavage of the bait region.
Modified ␣ 2 M Has Enhanced Ability to Inhibit BMP1-We have previously noted wide differences in the efficiency with which different substrates are cleaved by BMP1. 4 One of the substrates processed most efficiently by BMP1 is probiglycan (12). We therefore sought to determine whether we could enhance the ability of ␣ 2 M to inhibit BMP1 by replacing the native bait region with sequences surrounding the probiglycan scissile bone (see Fig. 1). It was found that the mutant recombinant ␣ 2 M (b-␣ 2 M) forms complexes with and is cleaved by BMP1 more readily than plasma ␣ 2 M or recombinant wild type ␣ 2 M, prepared under the same conditions as b-␣ 2 M (Fig. 6A). When the pCP inhibitory activities of varying concentrations of recombinant wild type ␣ 2 M and b-␣ 2 M were measured (Fig. 6, B and C) (see "Experimental Procedures"), they led to calculated IC 50 values of 133 and 1.88 nM, respectively. The IC 50 value of 133 nM for recombinant wild type ␣ 2 M suggests considerably less effectiveness in BMP1 inhibition than the 4.82 value obtained for plasma ␣ 2 M (Fig. 3B), whereas the b-␣ 2 M IC 50 value of 1.88 is consistent with increased inhibitory effectiveness. In fact, because 9.4 nM BMP1 would be expected to be 50% inhibited by 4.7 nM ␣ 2 M under assay conditions, if one molecule of BMP1 is inhibited by one molecule of ␣ 2 M, the IC 50 value of 1.88 nM suggests that molecules of b-␣ 2 M are capable of inhibiting more than one molecule of BMP1, with an approximate stoichiometry of 2 molecules of BMP1 inhibited by one molecule of b-␣ 2 M. This in turn suggests that b-␣ 2 M is cleaved very rapidly by BMP1, because only in such cases are molecules of ␣ 2 M known to inhibit proteases at a greater than 1:1 ratio    6D), some small difference in folding and/or post-translational modification may render the activity of the recombinant wild type protein less stable over the course of the overnight pCP assay than that of the corresponding protein from plasma. Importantly, however, b-␣ 2 M is shown to have markedly improved efficiency in inhibiting BMP1 compared with wild type ␣ 2 M prepared under identical conditions, or compared with plasma ␣ 2 M. The improved efficiency of interaction of b-␣ 2 M with BMP1, compared with wild type ␣ 2 M, is further illustrated by comparing the rapidity with which b-␣ 2 M is cleaved by BMP1 compared with cleavage of wild type recombinant or plasma ␣ 2 M (Fig. 3A).
To obtain a quantitative comparison of the rates of interaction of BMP1 with the various wild type and mutant forms of ␣ 2 M, we employed the methodology of Enghild et al. (14), which involves quantitation of covalent proteinase-␣ 2 M complex formation, subsequent to incubation of radiolabled proteinase with ␣ 2 M. In Fig.  6E it can be seen that, subsequent to a 2-h co-incubation of 40 nM  (Fig. 7). Second-order rate constants obtained from the same data showed b-␣ 2 M to be 24-fold more effective in interacting with ␣ 2 M than was recombinant ␣ 2 M prepared under identical conditions and 16-fold more effective than wild type ␣ 2 M from plasma (Table 1). ␣ 2 M Inhibition of Procollagen Processing by Cells-As ␣ 2 M is capable of inhibiting the pCP activity of BMP1, we attempted to determine whether it might be able to inhibit the processing of procollagen by cells. Toward this end, MC-3T3-E1 murine osteoblastic cells were incubated either alone or in the presence of recombinant wild type ␣ 2 M or b-␣ 2 M, and levels of processing of procollagen and insertion into the cell layer were compared. As can be seen (Fig. 8), MC-3T3 processing of procollagen was inhibited by both wild type and mutant ␣ 2 M. However, in the case of media from wild type ␣ 2 M-treated cells, most detectable collagenous material was in the form of processing intermediate pN␣1(I) (in which the N-, but not the C-propeptide is retained), or mature ␣1(I) chains, whereas in media from b-␣ 2 M-treated cells most detectable collagenous material was in the form of unprocessed pro-␣1(I) chains (Fig.  8, A and B). These results show efficient inhibition of cellular BMP1-like proteins by b-␣ 2 M, and less efficient inhibition by wild type ␣ 2 M. The appearance of pN␣1(I) chains in the wild type ␣ 2 Mtreated sample and procollagen in the b-␣ 2 M-treated sample indicate that both forms of ␣ 2 M are able to inhibit N-propeptide cleavage in cell culture by the proteinase ADAMTS-2. This is consistent with a previous report that ␣ 2 M is capable of inhibiting ADAMTS-2 in vitro (17). b-␣ 2 M may retain the ability to inhibit ADAMTS-2, due to a potential ADAMTS-2 cleavage site at the N terminus of the native bait region that is retained in the b-␣ 2 M sequence (Fig. 1). Collagen was not detected in the media of untreated cells in Fig. 8A, and was only detected upon longer exposure of the Western blot (not shown), presumably because of efficient processing and insertion of mature collagen into the cell layer. Untreated cell layers contained only fully processed mature ␣1(I) chains, whereas cell layers of cultures treated with either wild type ␣ 2 M or b-␣ 2 M contained both mature ␣1(I) chains and pN␣1(I) forms (Fig. 8A). pC␣1(I) forms (in which the C-, but not the N-propeptide is retained) were detected only in the media of b-␣ 2 M-treated cultures (Fig. 8A), and were not found in the cell layers of any of the cultures, presumably because pC␣1(I) chains are not inserted into ECM under normal circumstances (3). Treatment of MC-3T3-E1 cells with plasma ␣ 2 M (data not shown), yielded effects on procollagen processing similar to those obtained from treatment of cells with recombinant wild type ␣ 2 M.

DISCUSSION
BMP1-like proteinases are important regulators of ECM deposition in vertebrates. In particular, they are of central importance to the formation of collagen fibrils because 1) they process types I-III procollagen C-propeptides to yield the major fibrous components of ECM; 2) they cleave a zymogen to produce active lysyl oxidase, the enzyme that catalyzes the formation of covalent cross-links in collagen fibers; and 3) they process the N-propeptides, and in some cases C-propeptides, of procollagen chains of the minor fibrillar collagen types V and XI (2). The latter are incorporated into growing fibrils of collagen types I and II, respectively, and appear to control the geometries of the resulting heterotypic fibrils (18,19).
Cleavage of procollagen Cpropeptides appears to be the essential step that determines whether or not collagen fibrillogenesis will occur, and it has been demonstrated in vitro that, whereas type I collagen monomers that retain N-propeptides are incorporated into growing collagen fibrils as efficiently as mature monomers, monomers which retain C-propeptides are excluded from fibril incorporation (20). Our finding of inclusion of pN␣1(I) forms, but not pC␣1(I) forms in ECM associated with cell layers (Fig. 8A), is consistent with the previous in vitro fibrillogenesis results. The inhibitory effects of uncleaved C-propeptides on fibrillogenesis may be 2-fold: 1) via steric hindrance by the relatively bulky C-propeptide of the highly ordered packing of monomers necessary for fibrillogenesis, and 2) because C-propeptides confer ϳ1000-fold increased solubility to collagen monomers, which would be expected to interfere with monomer-to-monomer association (20). Incorporation of monomers with retained N-propeptides results in fibrils with aberrant morphologies and, in vivo, results in the heritable tissue disorder Ehlers-Danlos syndrome type VII (21). In contrast, no genetic disease has yet been associated with inability to cleave C-propeptides, perhaps because such inability would be incompatible with fibrillogenesis, and thus with life itself.
Because BMP1-like proteinases have been demonstrated to provide most, if not all procollagen C-proteinase activity in vivo (15,22), and because removal of the C-propeptide is essential for collagen fibrillogenesis, the BMP1-like proteinases are attractive targets for therapeutic interventions in situations where inhibition of collagen fibrillogenesis is desirable. Although the formation of collagen fibrils is essential to morphogenesis and to the healing of wounds and bone fractures in the adult, formation of excessive amounts of fibrous collagenous ECM is the cause of much morbidity in the general population. These conditions range from keloids (excessive scarring of the skin), to the formation of surgical adhesions, to deep-seated fibroses of organs such as the lungs, liver, and kidneys. The deep-seated fibroses are particularly ominous, as the replacement of parenchmal tissue by scar tissue, composed essentially of fibrous collagenous ECM, destroys organ function.
Despite earlier reports that ␣ 2 M does not inhibit astacin-like proteinases (7,9), we herein demonstrate ␣ 2 M to be an efficient inhibitor of the BMP1-like proteinases, a subgroup of the astacin-like proteinases (1). Thus, ␣ 2 M becomes the second endogenous inhibitor of BMP1-like proteinases to be identified because, as this article was being prepared, De Robertis and colleagues (23) elegantly demonstrated the secreted Xenopus/ zebrafish protein sizzled/ogon to be an inhibitor of such proteinases. The demonstration that ␣ 2 M is a potent inhibitor of BMP1-like proteinases suggests new approaches toward therapeutic interventions in the fibroses. These could include: 1) topical application in cases of superficial fibroses, such as keloids and corneal scarring; 2) delivery via aerosol for pulmonary fibrosis; and 3) ectopic expression of ␣ 2 M via gene therapy.  Blots were probed with anti-␣1(I) C-telopeptide antibodies. B, the media sample blot was scanned and quantified, using NIH Image software, providing the graft shown in which the percentage of total signal each for the wild type ␣ 2 M lane and for the b-␣ 2 M lane is given for uncleaved pro-␣1(I) chains (pro), pC␣1(I) (pC), and pN␣1(I) (pN) processing intermediates, and completely processed mature ␣1(I) chains (col ␣1(I)). Ectopic expression via gene therapy, in addition to targeting delivery of high levels of ␣ 2 M to specific tissues, would also result in co-expression of recombinant ␣ 2 M with endogenous BMP1-like proteinases in the same cells. The latter result may additionally boost inhibitory effects, as it has been hypothesized that some portion of the cleavages affected by BMP1-like proteinases may occur intracellularly and/or near the cell surface (2,24), and it has also been previously demonstrated that recombinant ␣ 2 M is capable of inhibiting proteolysis both extracellularly and intracellulary, when co-expressed with endogenous proteinases in the same cells (25). Although the above approaches may be undertaken with wild type ␣ 2 M sequences, the exciting possibility also exists of engineering the bait region of ␣ 2 M, such that the IC 50 for inhibition of BMP1like proteinases is decreased. The validity of the latter approach has been proven in the current report, in which substitution of the native ␣ 2 M bait region with sequences flanking the probiglycan BMP1 cleavage site increased the efficiency about 24-fold ( Table 1). The desirability of the latter approach, in terms of inhibiting deposition of a collagenous ECM, is also demonstrated in Fig. 8, where it is shown that the mutant b-␣ 2 M was effective in inhibiting procollagen processing by fibrogenic cells, whereas wild type ␣ 2 M was much less effective.
Although there is no strict consensus sequence for the cleavage sites of BMP1-like proteinases, we can begin to identify preferred amino acid residues at each position flanking such sites, based on analysis of the sites of known substrates of BMP1-like proteinases (2). In the future, modified versions of ␣ 2 M may be further optimized for cleavage by BMP1-like proteinases by substituting preferred residues into positions flanking cleavage sites for BMP1-like proteinases in bait regions. In addition, removal of recognition sites for cleavage of the bait region by other proteinases could make the recombinant ␣ 2 M specific for inhibition of BMP1-like proteinases, thereby eliminating the possibility of unwanted secondary effects from inhibition of other proteinases. In regard to the treatment of fibroses, it may be particularly desirable to engineer versions of ␣ 2 M that would inhibit BMP1 proteinases, which are involved in forming collagenous ECM while not inhibiting those matrix metalloproteinases involved in the turnover of collagenous ECM. Studies akin to the approaches described above are currently underway.